From ab98af9ed0680ce00f7f2f79682b4dac460fd218 Mon Sep 17 00:00:00 2001 From: Claude Date: Sun, 5 Jul 2026 00:18:15 +0000 Subject: [PATCH 01/86] db.query() rejected statements roll back, RETURNING commits immediately Fixes two transaction bugs in db.query(): - A non-row-returning statement such as an UPDATE was executed and auto-committed by execute() before the ValueError was raised, so the write took effect despite being rejected. query() now runs the statement inside a savepoint and rolls it back before raising, so a rejected statement has no effect on the database - in every connection mode, including autocommit=True. - INSERT ... RETURNING only committed once the returned generator was fully exhausted, so calling query() without iterating - or partially iterating with next() - left the transaction open and the write could be rolled back on close. The write is now completed and committed at call time, as the documentation already promised. Plain SELECTs are still fetched lazily. Transaction control statements, VACUUM, ATTACH and DETACH never return rows, so query() now rejects them without executing them. PRAGMA statements skip the savepoint guard because some of them - such as pragma journal_mode=wal - refuse to run inside a transaction. Co-Authored-By: Claude Fable 5 Claude-Session: https://claude.ai/code/session_012U3iRfJoTZ5vd22cBSF2nJ --- docs/changelog.rst | 2 +- docs/python-api.rst | 6 +-- docs/upgrading.rst | 2 +- sqlite_utils/db.py | 77 ++++++++++++++++++++++--------- tests/test_query.py | 110 ++++++++++++++++++++++++++++++++++++++++++++ 5 files changed, 171 insertions(+), 26 deletions(-) diff --git a/docs/changelog.rst b/docs/changelog.rst index fbc58d8..f110e2e 100644 --- a/docs/changelog.rst +++ b/docs/changelog.rst @@ -12,7 +12,7 @@ Unreleased Breaking changes: - Write statements executed with ``db.execute()`` are now committed automatically, unless a transaction is already open in which case they join it. Previously they opened an implicit transaction that stayed open until something committed it - writes appeared to work when read on the same connection but were silently rolled back when the connection closed. Code that relied on rolling back uncommitted ``db.execute()`` writes should use the new ``db.begin()`` method to open an explicit transaction first. The transaction model is documented in full at :ref:`python_api_transactions`. -- ``db.query()`` now executes its SQL as soon as it is called, rather than waiting until the returned generator is first iterated. Rows are still fetched lazily during iteration. SQL errors are now raised at the call site, statements such as ``INSERT ... RETURNING`` take effect without needing to iterate over their results, and passing a statement that returns no rows - previously a silent no-op - now raises a ``ValueError`` recommending ``db.execute()`` instead. +- ``db.query()`` now executes its SQL as soon as it is called, rather than waiting until the returned generator is first iterated. Rows are still fetched lazily during iteration. SQL errors are now raised at the call site, statements such as ``INSERT ... RETURNING`` are executed and committed immediately without needing to iterate over their results, and passing a statement that returns no rows - previously a silent no-op - now raises a ``ValueError`` recommending ``db.execute()`` instead. A statement rejected this way is rolled back before the error is raised, so it has no effect on the database. - Python API validation errors now raise ``ValueError`` instead of ``AssertionError``. Previously invalid arguments - such as ``create_table()`` with no columns, ``transform()`` on a table that does not exist, or passing both ``ignore=True`` and ``replace=True`` - were rejected using bare ``assert`` statements, which are silently skipped when Python runs with the ``-O`` flag. Code that caught ``AssertionError`` for these cases should catch ``ValueError`` instead. - ``table.upsert()`` and ``table.upsert_all()`` now raise ``PrimaryKeyRequired`` if a record is missing a value for any primary key column, or has a value of ``None`` for one. Previously such records - which can never match an existing row - were quietly inserted as brand new rows, or triggered a confusing ``KeyError`` after the insert had already taken place. - ``db.enable_wal()`` and ``db.disable_wal()`` now raise a ``sqlite_utils.db.TransactionError`` if called while a transaction is open. Previously they would silently commit the open transaction as a side effect of changing the journal mode, breaking the rollback guarantee of ``db.atomic()`` and of user-managed transactions. diff --git a/docs/python-api.rst b/docs/python-api.rst index e9eb371..3f7e12c 100644 --- a/docs/python-api.rst +++ b/docs/python-api.rst @@ -229,9 +229,9 @@ The ``db.query(sql)`` function executes a SQL query and returns an iterator over # {'name': 'Cleo'} # {'name': 'Pancakes'} -The SQL query is executed as soon as ``db.query()`` is called. The resulting rows are fetched lazily as you iterate, so large result sets are not loaded into memory all at once. Because execution is immediate, an error in your SQL will raise an exception straight away, and a statement such as ``INSERT ... RETURNING`` will take effect even if you do not iterate over its results. +The SQL query is executed as soon as ``db.query()`` is called. The resulting rows are fetched lazily as you iterate, so large result sets are not loaded into memory all at once. Because execution is immediate, an error in your SQL will raise an exception straight away, and a statement such as ``INSERT ... RETURNING`` will take effect - and be committed, unless a transaction is open - even if you do not iterate over its results. -``db.query()`` can only be used with SQL that returns rows. Passing a statement that returns no rows - an ``INSERT`` or ``UPDATE`` without a ``RETURNING`` clause, for example - will raise a ``ValueError``. Use :ref:`db.execute() ` for those statements instead. +``db.query()`` can only be used with SQL that returns rows. Passing a statement that returns no rows - an ``INSERT`` or ``UPDATE`` without a ``RETURNING`` clause, for example - will raise a ``ValueError``. The rejected statement is rolled back, so it has no effect on the database. Use :ref:`db.execute() ` for those statements instead. .. _python_api_execute: @@ -356,7 +356,7 @@ If a transaction is open - because the call happens inside a ``db.atomic()`` blo db.execute("insert into news (headline) values (?)", ["Cat unimpressed"]) # Both rows committed together -One corner case: a row-returning write such as ``INSERT ... RETURNING`` executed through ``db.execute()`` cannot be auto-committed, because its rows have not been read yet - call ``db.commit()`` after fetching them, or use :ref:`db.query() ` for those statements, which commits automatically once you have iterated over the results. +One corner case: a row-returning write such as ``INSERT ... RETURNING`` executed through ``db.execute()`` cannot be auto-committed, because its rows have not been read yet - call ``db.commit()`` after fetching them, or use :ref:`db.query() ` for those statements, which executes the write and commits it immediately. .. _python_api_transactions_manual: diff --git a/docs/upgrading.rst b/docs/upgrading.rst index 3109a7c..1974ad6 100644 --- a/docs/upgrading.rst +++ b/docs/upgrading.rst @@ -58,7 +58,7 @@ Python API changes **db.query() executes immediately.** ``db.query(sql)`` previously returned a generator that did not execute the SQL until you started iterating over it. The SQL now runs as soon as the method is called - rows are still fetched lazily. Two consequences: - Errors in your SQL now raise at the ``db.query()`` call site rather than on first iteration. -- Passing a statement that returns no rows - such as an ``INSERT`` or ``UPDATE`` without a ``RETURNING`` clause - previously did nothing at all, silently. It now raises a ``ValueError``. Use ``db.execute()`` for statements that do not return rows. +- Passing a statement that returns no rows - such as an ``INSERT`` or ``UPDATE`` without a ``RETURNING`` clause - previously did nothing at all, silently. It now raises a ``ValueError``, and the statement is rolled back so it has no effect on the database. Use ``db.execute()`` for statements that do not return rows. **Upserts use INSERT ... ON CONFLICT.** Upsert operations now use SQLite's ``INSERT ... ON CONFLICT SET`` syntax rather than the previous ``INSERT OR IGNORE`` followed by ``UPDATE``. If your code depends on the old behavior, pass ``use_old_upsert=True`` to the ``Database()`` constructor - see :ref:`python_api_old_upsert`. diff --git a/sqlite_utils/db.py b/sqlite_utils/db.py index ef367e1..8e6a194 100644 --- a/sqlite_utils/db.py +++ b/sqlite_utils/db.py @@ -652,33 +652,68 @@ class Database: Execute ``sql`` and return an iterable of dictionaries representing each row. The SQL is executed as soon as this method is called - the resulting rows - are then fetched lazily as the returned iterable is iterated over. + are then fetched lazily as the returned iterable is iterated over. A + row-returning write such as ``INSERT ... RETURNING`` takes effect + immediately, even if the results are never iterated. :param sql: SQL query to execute :param params: Parameters to use in that query - an iterable for ``where id = ?`` parameters, or a dictionary for ``where id = :id`` :raises ValueError: if the SQL statement does not return rows - use - :meth:`execute` for those statements instead + :meth:`execute` for those statements instead. The rejected statement + is rolled back, so it has no effect on the database """ - was_in_transaction = self.conn.in_transaction - cursor = self.execute(sql, params or tuple()) - if cursor.description is None: - raise ValueError( - "query() can only be used with SQL that returns rows - " - "use execute() for other statements" - ) - keys = [d[0] for d in cursor.description] - - def rows() -> Generator[dict, None, None]: - for row in cursor: - yield dict(zip(keys, row)) - if not was_in_transaction and self.conn.in_transaction: - # A row-returning write such as INSERT ... RETURNING opened - # an implicit transaction - commit it now that its rows have - # been consumed - self.conn.execute("COMMIT") - - return rows() + message = ( + "query() can only be used with SQL that returns rows - " + "use execute() for other statements" + ) + prefix = sql.lstrip().upper() + if prefix.startswith( + _TRANSACTION_CONTROL_PREFIXES + ("VACUUM", "ATTACH", "DETACH") + ): + # None of these return rows - reject them without executing anything + raise ValueError(message) + if self._tracer: + self._tracer(sql, params) + args: tuple = (params,) if params is not None else () + if prefix.startswith("PRAGMA"): + # Some PRAGMA statements refuse to run inside a transaction, so + # execute these without the savepoint guard used below. PRAGMAs + # never open an implicit transaction, so there is nothing to + # undo if this one turns out not to return rows + cursor = self.conn.execute(sql, *args) + if cursor.description is None: + raise ValueError(message) + keys = [d[0] for d in cursor.description] + return (dict(zip(keys, row)) for row in cursor) + # Execute inside a savepoint, so a statement that turns out not to + # return rows can be rolled back before the ValueError is raised + self.conn.execute('SAVEPOINT "sqlite_utils_query"') + released = False + try: + cursor = self.conn.execute(sql, *args) + if cursor.description is None: + raise ValueError(message) + keys = [d[0] for d in cursor.description] + try: + self.conn.execute('RELEASE "sqlite_utils_query"') + released = True + except sqlite3.OperationalError: + # The savepoint cannot be released while a write statement is + # still executing - this is INSERT ... RETURNING or similar, + # with unfetched rows. Fetch them so the write completes, then + # release again - committing the write immediately, unless an + # outer transaction is open + fetched = cursor.fetchall() + self.conn.execute('RELEASE "sqlite_utils_query"') + released = True + return (dict(zip(keys, row)) for row in fetched) + return (dict(zip(keys, row)) for row in cursor) + finally: + if not released: + # An error occurred - undo anything the statement changed + self.conn.execute('ROLLBACK TO "sqlite_utils_query"') + self.conn.execute('RELEASE "sqlite_utils_query"') def execute( self, sql: str, parameters: Optional[Union[Sequence, Dict[str, Any]]] = None diff --git a/tests/test_query.py b/tests/test_query.py index 91bf1ac..cf8d9f5 100644 --- a/tests/test_query.py +++ b/tests/test_query.py @@ -23,6 +23,56 @@ def test_query_rejects_statements_that_return_no_rows(fresh_db): with pytest.raises(ValueError) as ex: fresh_db.query("update dogs set name = 'Cleopaws'") assert "execute()" in str(ex.value) + # The rejected update was rolled back, and no transaction is left open + assert not fresh_db.conn.in_transaction + assert [row["name"] for row in fresh_db["dogs"].rows] == ["Cleo"] + + +def test_query_rejected_ddl_is_rolled_back(fresh_db): + with pytest.raises(ValueError): + fresh_db.query("create table dogs (id integer primary key)") + assert not fresh_db.conn.in_transaction + assert fresh_db.table_names() == [] + + +def test_query_rejected_write_inside_transaction_is_rolled_back(fresh_db): + fresh_db["dogs"].insert({"name": "Cleo"}) + fresh_db.begin() + fresh_db.execute("insert into dogs (name) values ('Pancakes')") + with pytest.raises(ValueError): + fresh_db.query("update dogs set name = 'Cleopaws'") + # The transaction is still open and the earlier insert is intact + assert fresh_db.conn.in_transaction + fresh_db.commit() + assert [row["name"] for row in fresh_db["dogs"].rows] == ["Cleo", "Pancakes"] + + +@pytest.mark.parametrize( + "sql", ["begin", "commit", "rollback", "vacuum", "detach database foo"] +) +def test_query_rejects_transaction_control_and_vacuum(fresh_db, sql): + with pytest.raises(ValueError) as ex: + fresh_db.query(sql) + assert "execute()" in str(ex.value) + assert not fresh_db.conn.in_transaction + + +def test_query_error_leaves_no_transaction_open(fresh_db): + with pytest.raises(sqlite3.OperationalError): + fresh_db.query("select * from missing_table") + assert not fresh_db.conn.in_transaction + + +def test_query_pragma(tmpdir): + from sqlite_utils import Database + + db = Database(str(tmpdir / "test.db")) + # A row-returning PRAGMA works, including one that cannot run in a transaction + assert list(db.query("pragma journal_mode = wal")) == [{"journal_mode": "wal"}] + # A PRAGMA that returns no rows raises ValueError + with pytest.raises(ValueError): + db.query("pragma user_version = 5") + db.close() @pytest.mark.skipif( @@ -38,6 +88,66 @@ def test_query_insert_returning(fresh_db): assert fresh_db["dogs"].count == 2 +@pytest.mark.skipif( + sqlite3.sqlite_version_info < (3, 35, 0), + reason="RETURNING requires SQLite 3.35.0 or higher", +) +def test_query_insert_returning_commits_without_iteration(tmpdir): + from sqlite_utils import Database + + path = str(tmpdir / "test.db") + db = Database(path) + db["dogs"].insert({"name": "Cleo"}) + # Never iterate over the results + db.query("insert into dogs (name) values ('Pancakes') returning name") + assert not db.conn.in_transaction + # A completely separate connection sees the new row straight away + other = sqlite3.connect(path) + assert other.execute("select count(*) from dogs").fetchone()[0] == 2 + other.close() + db.close() + + +@pytest.mark.skipif( + sqlite3.sqlite_version_info < (3, 35, 0), + reason="RETURNING requires SQLite 3.35.0 or higher", +) +def test_query_insert_returning_partial_iteration_still_commits(tmpdir): + from sqlite_utils import Database + + path = str(tmpdir / "test.db") + db = Database(path) + db["dogs"].insert({"name": "Cleo"}) + row = next( + db.query( + "insert into dogs (name) values ('Pancakes'), ('Marnie') returning name" + ) + ) + assert row == {"name": "Pancakes"} + assert not db.conn.in_transaction + other = sqlite3.connect(path) + assert other.execute("select count(*) from dogs").fetchone()[0] == 3 + other.close() + db.close() + + +@pytest.mark.skipif( + sqlite3.sqlite_version_info < (3, 35, 0), + reason="RETURNING requires SQLite 3.35.0 or higher", +) +def test_query_insert_returning_respects_explicit_transaction(fresh_db): + fresh_db["dogs"].insert({"name": "Cleo"}) + fresh_db.begin() + rows = list( + fresh_db.query("insert into dogs (name) values ('Pancakes') returning name") + ) + assert rows == [{"name": "Pancakes"}] + # Still inside the explicit transaction - not committed + assert fresh_db.conn.in_transaction + fresh_db.rollback() + assert [row["name"] for row in fresh_db["dogs"].rows] == ["Cleo"] + + def test_execute_returning_dicts(fresh_db): # Like db.query() but returns a list, included for backwards compatibility # see https://github.com/simonw/sqlite-utils/issues/290 From 0566a9f128e19475f5cb423043820fb4119051cc Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Sat, 4 Jul 2026 17:40:48 -0700 Subject: [PATCH 02/86] Improve db.query() error handling and transaction safety db.query() rejected statements roll back, RETURNING commits immediately. Fixes two transaction bugs in db.query(): - A non-row-returning statement such as an UPDATE was executed and auto-committed by execute() before the ValueError was raised, so the write took effect despite being rejected. query() now runs the statement inside a savepoint and rolls it back before raising, so a rejected statement has no effect on the database - in every connection mode, including autocommit=True. - INSERT ... RETURNING only committed once the returned generator was fully exhausted, so calling query() without iterating - or partially iterating with next() - left the transaction open and the write could be rolled back on close. The write is now completed and committed at call time, as the documentation already promised. Plain SELECTs are still fetched lazily. Transaction control statements, VACUUM, ATTACH and DETACH never return rows, so query() now rejects them without executing them. PRAGMA statements skip the savepoint guard because some of them - such as pragma journal_mode=wal - refuse to run inside a transaction. Claude-Session: https://claude.ai/code/session_012U3iRfJoTZ5vd22cBSF2nJ --- docs/changelog.rst | 2 +- docs/python-api.rst | 6 +-- docs/upgrading.rst | 2 +- sqlite_utils/db.py | 77 ++++++++++++++++++++++--------- tests/test_query.py | 110 ++++++++++++++++++++++++++++++++++++++++++++ 5 files changed, 171 insertions(+), 26 deletions(-) diff --git a/docs/changelog.rst b/docs/changelog.rst index fbc58d8..f110e2e 100644 --- a/docs/changelog.rst +++ b/docs/changelog.rst @@ -12,7 +12,7 @@ Unreleased Breaking changes: - Write statements executed with ``db.execute()`` are now committed automatically, unless a transaction is already open in which case they join it. Previously they opened an implicit transaction that stayed open until something committed it - writes appeared to work when read on the same connection but were silently rolled back when the connection closed. Code that relied on rolling back uncommitted ``db.execute()`` writes should use the new ``db.begin()`` method to open an explicit transaction first. The transaction model is documented in full at :ref:`python_api_transactions`. -- ``db.query()`` now executes its SQL as soon as it is called, rather than waiting until the returned generator is first iterated. Rows are still fetched lazily during iteration. SQL errors are now raised at the call site, statements such as ``INSERT ... RETURNING`` take effect without needing to iterate over their results, and passing a statement that returns no rows - previously a silent no-op - now raises a ``ValueError`` recommending ``db.execute()`` instead. +- ``db.query()`` now executes its SQL as soon as it is called, rather than waiting until the returned generator is first iterated. Rows are still fetched lazily during iteration. SQL errors are now raised at the call site, statements such as ``INSERT ... RETURNING`` are executed and committed immediately without needing to iterate over their results, and passing a statement that returns no rows - previously a silent no-op - now raises a ``ValueError`` recommending ``db.execute()`` instead. A statement rejected this way is rolled back before the error is raised, so it has no effect on the database. - Python API validation errors now raise ``ValueError`` instead of ``AssertionError``. Previously invalid arguments - such as ``create_table()`` with no columns, ``transform()`` on a table that does not exist, or passing both ``ignore=True`` and ``replace=True`` - were rejected using bare ``assert`` statements, which are silently skipped when Python runs with the ``-O`` flag. Code that caught ``AssertionError`` for these cases should catch ``ValueError`` instead. - ``table.upsert()`` and ``table.upsert_all()`` now raise ``PrimaryKeyRequired`` if a record is missing a value for any primary key column, or has a value of ``None`` for one. Previously such records - which can never match an existing row - were quietly inserted as brand new rows, or triggered a confusing ``KeyError`` after the insert had already taken place. - ``db.enable_wal()`` and ``db.disable_wal()`` now raise a ``sqlite_utils.db.TransactionError`` if called while a transaction is open. Previously they would silently commit the open transaction as a side effect of changing the journal mode, breaking the rollback guarantee of ``db.atomic()`` and of user-managed transactions. diff --git a/docs/python-api.rst b/docs/python-api.rst index e9eb371..3f7e12c 100644 --- a/docs/python-api.rst +++ b/docs/python-api.rst @@ -229,9 +229,9 @@ The ``db.query(sql)`` function executes a SQL query and returns an iterator over # {'name': 'Cleo'} # {'name': 'Pancakes'} -The SQL query is executed as soon as ``db.query()`` is called. The resulting rows are fetched lazily as you iterate, so large result sets are not loaded into memory all at once. Because execution is immediate, an error in your SQL will raise an exception straight away, and a statement such as ``INSERT ... RETURNING`` will take effect even if you do not iterate over its results. +The SQL query is executed as soon as ``db.query()`` is called. The resulting rows are fetched lazily as you iterate, so large result sets are not loaded into memory all at once. Because execution is immediate, an error in your SQL will raise an exception straight away, and a statement such as ``INSERT ... RETURNING`` will take effect - and be committed, unless a transaction is open - even if you do not iterate over its results. -``db.query()`` can only be used with SQL that returns rows. Passing a statement that returns no rows - an ``INSERT`` or ``UPDATE`` without a ``RETURNING`` clause, for example - will raise a ``ValueError``. Use :ref:`db.execute() ` for those statements instead. +``db.query()`` can only be used with SQL that returns rows. Passing a statement that returns no rows - an ``INSERT`` or ``UPDATE`` without a ``RETURNING`` clause, for example - will raise a ``ValueError``. The rejected statement is rolled back, so it has no effect on the database. Use :ref:`db.execute() ` for those statements instead. .. _python_api_execute: @@ -356,7 +356,7 @@ If a transaction is open - because the call happens inside a ``db.atomic()`` blo db.execute("insert into news (headline) values (?)", ["Cat unimpressed"]) # Both rows committed together -One corner case: a row-returning write such as ``INSERT ... RETURNING`` executed through ``db.execute()`` cannot be auto-committed, because its rows have not been read yet - call ``db.commit()`` after fetching them, or use :ref:`db.query() ` for those statements, which commits automatically once you have iterated over the results. +One corner case: a row-returning write such as ``INSERT ... RETURNING`` executed through ``db.execute()`` cannot be auto-committed, because its rows have not been read yet - call ``db.commit()`` after fetching them, or use :ref:`db.query() ` for those statements, which executes the write and commits it immediately. .. _python_api_transactions_manual: diff --git a/docs/upgrading.rst b/docs/upgrading.rst index 3109a7c..1974ad6 100644 --- a/docs/upgrading.rst +++ b/docs/upgrading.rst @@ -58,7 +58,7 @@ Python API changes **db.query() executes immediately.** ``db.query(sql)`` previously returned a generator that did not execute the SQL until you started iterating over it. The SQL now runs as soon as the method is called - rows are still fetched lazily. Two consequences: - Errors in your SQL now raise at the ``db.query()`` call site rather than on first iteration. -- Passing a statement that returns no rows - such as an ``INSERT`` or ``UPDATE`` without a ``RETURNING`` clause - previously did nothing at all, silently. It now raises a ``ValueError``. Use ``db.execute()`` for statements that do not return rows. +- Passing a statement that returns no rows - such as an ``INSERT`` or ``UPDATE`` without a ``RETURNING`` clause - previously did nothing at all, silently. It now raises a ``ValueError``, and the statement is rolled back so it has no effect on the database. Use ``db.execute()`` for statements that do not return rows. **Upserts use INSERT ... ON CONFLICT.** Upsert operations now use SQLite's ``INSERT ... ON CONFLICT SET`` syntax rather than the previous ``INSERT OR IGNORE`` followed by ``UPDATE``. If your code depends on the old behavior, pass ``use_old_upsert=True`` to the ``Database()`` constructor - see :ref:`python_api_old_upsert`. diff --git a/sqlite_utils/db.py b/sqlite_utils/db.py index ef367e1..8e6a194 100644 --- a/sqlite_utils/db.py +++ b/sqlite_utils/db.py @@ -652,33 +652,68 @@ class Database: Execute ``sql`` and return an iterable of dictionaries representing each row. The SQL is executed as soon as this method is called - the resulting rows - are then fetched lazily as the returned iterable is iterated over. + are then fetched lazily as the returned iterable is iterated over. A + row-returning write such as ``INSERT ... RETURNING`` takes effect + immediately, even if the results are never iterated. :param sql: SQL query to execute :param params: Parameters to use in that query - an iterable for ``where id = ?`` parameters, or a dictionary for ``where id = :id`` :raises ValueError: if the SQL statement does not return rows - use - :meth:`execute` for those statements instead + :meth:`execute` for those statements instead. The rejected statement + is rolled back, so it has no effect on the database """ - was_in_transaction = self.conn.in_transaction - cursor = self.execute(sql, params or tuple()) - if cursor.description is None: - raise ValueError( - "query() can only be used with SQL that returns rows - " - "use execute() for other statements" - ) - keys = [d[0] for d in cursor.description] - - def rows() -> Generator[dict, None, None]: - for row in cursor: - yield dict(zip(keys, row)) - if not was_in_transaction and self.conn.in_transaction: - # A row-returning write such as INSERT ... RETURNING opened - # an implicit transaction - commit it now that its rows have - # been consumed - self.conn.execute("COMMIT") - - return rows() + message = ( + "query() can only be used with SQL that returns rows - " + "use execute() for other statements" + ) + prefix = sql.lstrip().upper() + if prefix.startswith( + _TRANSACTION_CONTROL_PREFIXES + ("VACUUM", "ATTACH", "DETACH") + ): + # None of these return rows - reject them without executing anything + raise ValueError(message) + if self._tracer: + self._tracer(sql, params) + args: tuple = (params,) if params is not None else () + if prefix.startswith("PRAGMA"): + # Some PRAGMA statements refuse to run inside a transaction, so + # execute these without the savepoint guard used below. PRAGMAs + # never open an implicit transaction, so there is nothing to + # undo if this one turns out not to return rows + cursor = self.conn.execute(sql, *args) + if cursor.description is None: + raise ValueError(message) + keys = [d[0] for d in cursor.description] + return (dict(zip(keys, row)) for row in cursor) + # Execute inside a savepoint, so a statement that turns out not to + # return rows can be rolled back before the ValueError is raised + self.conn.execute('SAVEPOINT "sqlite_utils_query"') + released = False + try: + cursor = self.conn.execute(sql, *args) + if cursor.description is None: + raise ValueError(message) + keys = [d[0] for d in cursor.description] + try: + self.conn.execute('RELEASE "sqlite_utils_query"') + released = True + except sqlite3.OperationalError: + # The savepoint cannot be released while a write statement is + # still executing - this is INSERT ... RETURNING or similar, + # with unfetched rows. Fetch them so the write completes, then + # release again - committing the write immediately, unless an + # outer transaction is open + fetched = cursor.fetchall() + self.conn.execute('RELEASE "sqlite_utils_query"') + released = True + return (dict(zip(keys, row)) for row in fetched) + return (dict(zip(keys, row)) for row in cursor) + finally: + if not released: + # An error occurred - undo anything the statement changed + self.conn.execute('ROLLBACK TO "sqlite_utils_query"') + self.conn.execute('RELEASE "sqlite_utils_query"') def execute( self, sql: str, parameters: Optional[Union[Sequence, Dict[str, Any]]] = None diff --git a/tests/test_query.py b/tests/test_query.py index 91bf1ac..cf8d9f5 100644 --- a/tests/test_query.py +++ b/tests/test_query.py @@ -23,6 +23,56 @@ def test_query_rejects_statements_that_return_no_rows(fresh_db): with pytest.raises(ValueError) as ex: fresh_db.query("update dogs set name = 'Cleopaws'") assert "execute()" in str(ex.value) + # The rejected update was rolled back, and no transaction is left open + assert not fresh_db.conn.in_transaction + assert [row["name"] for row in fresh_db["dogs"].rows] == ["Cleo"] + + +def test_query_rejected_ddl_is_rolled_back(fresh_db): + with pytest.raises(ValueError): + fresh_db.query("create table dogs (id integer primary key)") + assert not fresh_db.conn.in_transaction + assert fresh_db.table_names() == [] + + +def test_query_rejected_write_inside_transaction_is_rolled_back(fresh_db): + fresh_db["dogs"].insert({"name": "Cleo"}) + fresh_db.begin() + fresh_db.execute("insert into dogs (name) values ('Pancakes')") + with pytest.raises(ValueError): + fresh_db.query("update dogs set name = 'Cleopaws'") + # The transaction is still open and the earlier insert is intact + assert fresh_db.conn.in_transaction + fresh_db.commit() + assert [row["name"] for row in fresh_db["dogs"].rows] == ["Cleo", "Pancakes"] + + +@pytest.mark.parametrize( + "sql", ["begin", "commit", "rollback", "vacuum", "detach database foo"] +) +def test_query_rejects_transaction_control_and_vacuum(fresh_db, sql): + with pytest.raises(ValueError) as ex: + fresh_db.query(sql) + assert "execute()" in str(ex.value) + assert not fresh_db.conn.in_transaction + + +def test_query_error_leaves_no_transaction_open(fresh_db): + with pytest.raises(sqlite3.OperationalError): + fresh_db.query("select * from missing_table") + assert not fresh_db.conn.in_transaction + + +def test_query_pragma(tmpdir): + from sqlite_utils import Database + + db = Database(str(tmpdir / "test.db")) + # A row-returning PRAGMA works, including one that cannot run in a transaction + assert list(db.query("pragma journal_mode = wal")) == [{"journal_mode": "wal"}] + # A PRAGMA that returns no rows raises ValueError + with pytest.raises(ValueError): + db.query("pragma user_version = 5") + db.close() @pytest.mark.skipif( @@ -38,6 +88,66 @@ def test_query_insert_returning(fresh_db): assert fresh_db["dogs"].count == 2 +@pytest.mark.skipif( + sqlite3.sqlite_version_info < (3, 35, 0), + reason="RETURNING requires SQLite 3.35.0 or higher", +) +def test_query_insert_returning_commits_without_iteration(tmpdir): + from sqlite_utils import Database + + path = str(tmpdir / "test.db") + db = Database(path) + db["dogs"].insert({"name": "Cleo"}) + # Never iterate over the results + db.query("insert into dogs (name) values ('Pancakes') returning name") + assert not db.conn.in_transaction + # A completely separate connection sees the new row straight away + other = sqlite3.connect(path) + assert other.execute("select count(*) from dogs").fetchone()[0] == 2 + other.close() + db.close() + + +@pytest.mark.skipif( + sqlite3.sqlite_version_info < (3, 35, 0), + reason="RETURNING requires SQLite 3.35.0 or higher", +) +def test_query_insert_returning_partial_iteration_still_commits(tmpdir): + from sqlite_utils import Database + + path = str(tmpdir / "test.db") + db = Database(path) + db["dogs"].insert({"name": "Cleo"}) + row = next( + db.query( + "insert into dogs (name) values ('Pancakes'), ('Marnie') returning name" + ) + ) + assert row == {"name": "Pancakes"} + assert not db.conn.in_transaction + other = sqlite3.connect(path) + assert other.execute("select count(*) from dogs").fetchone()[0] == 3 + other.close() + db.close() + + +@pytest.mark.skipif( + sqlite3.sqlite_version_info < (3, 35, 0), + reason="RETURNING requires SQLite 3.35.0 or higher", +) +def test_query_insert_returning_respects_explicit_transaction(fresh_db): + fresh_db["dogs"].insert({"name": "Cleo"}) + fresh_db.begin() + rows = list( + fresh_db.query("insert into dogs (name) values ('Pancakes') returning name") + ) + assert rows == [{"name": "Pancakes"}] + # Still inside the explicit transaction - not committed + assert fresh_db.conn.in_transaction + fresh_db.rollback() + assert [row["name"] for row in fresh_db["dogs"].rows] == ["Cleo"] + + def test_execute_returning_dicts(fresh_db): # Like db.query() but returns a list, included for backwards compatibility # see https://github.com/simonw/sqlite-utils/issues/290 From a8e3a643f2a7ad1933eb900713f18f1650e45c1b Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Sat, 4 Jul 2026 17:45:42 -0700 Subject: [PATCH 03/86] Release 4.0rc2 Refs #759, #629, #761, #767, #768 --- docs/changelog.rst | 6 +++--- pyproject.toml | 2 +- 2 files changed, 4 insertions(+), 4 deletions(-) diff --git a/docs/changelog.rst b/docs/changelog.rst index f110e2e..2e1f69c 100644 --- a/docs/changelog.rst +++ b/docs/changelog.rst @@ -4,10 +4,10 @@ Changelog =========== -.. _unreleased: +.. _v4_0rc2: -Unreleased ----------- +4.0rc2 (2026-07-04) +------------------- Breaking changes: diff --git a/pyproject.toml b/pyproject.toml index 4ad1bd9..9d1f041 100644 --- a/pyproject.toml +++ b/pyproject.toml @@ -1,6 +1,6 @@ [project] name = "sqlite-utils" -version = "4.0rc1" +version = "4.0rc2" description = "CLI tool and Python library for manipulating SQLite databases" readme = { file = "README.md", content-type = "text/markdown" } authors = [ From 87cf1c5a007ff724ed803111898d9658d53e320a Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Sun, 5 Jul 2026 11:49:59 -0700 Subject: [PATCH 04/86] Replace SQL prefix matching with a first-token scanner query() and execute() classified statements using sql.lstrip().upper().startswith(...), which a leading SQL comment defeated: db.query('/* c */ COMMIT') inside db.atomic() committed the caller's transaction and masked the ValueError with a 'no such savepoint' error, comment-prefixed PRAGMAs ran inside the savepoint guard where journal mode changes are refused, and a comment-prefixed BEGIN passed to db.execute() was instantly auto-committed. The new _first_keyword() helper skips leading whitespace and -- or /* */ comments - the only things SQLite's tokenizer allows before the first token - and returns that token for exact comparison, so keyword detection now matches what SQLite itself will see. Co-Authored-By: Claude Fable 5 --- sqlite_utils/db.py | 51 +++++++++++++++++++++++++++++------ tests/test_atomic.py | 11 ++++++++ tests/test_query.py | 63 +++++++++++++++++++++++++++++++++++++++++++- 3 files changed, 116 insertions(+), 9 deletions(-) diff --git a/sqlite_utils/db.py b/sqlite_utils/db.py index 8e6a194..f796a63 100644 --- a/sqlite_utils/db.py +++ b/sqlite_utils/db.py @@ -324,14 +324,51 @@ CREATE TABLE IF NOT EXISTS "{}"( """.strip() -_TRANSACTION_CONTROL_PREFIXES = ( +_TRANSACTION_CONTROL_KEYWORDS = { "BEGIN", "COMMIT", "END", "ROLLBACK", "SAVEPOINT", "RELEASE", -) +} + +# Statements that never return rows and cannot run inside (or would break +# out of) the savepoint guard used by query() +_QUERY_REJECTED_KEYWORDS = _TRANSACTION_CONTROL_KEYWORDS | { + "VACUUM", + "ATTACH", + "DETACH", +} + + +def _first_keyword(sql: str) -> str: + """ + Return the first keyword of a SQL statement, uppercased, skipping any + leading whitespace and ``--`` or ``/* ... */`` comments - the only + things SQLite's tokenizer allows before the first token. Returns an + empty string if there is no leading keyword. + """ + i, n = 0, len(sql) + while i < n: + if sql[i].isspace(): + i += 1 + elif sql.startswith("--", i): + newline = sql.find("\n", i) + if newline == -1: + return "" + i = newline + 1 + elif sql.startswith("/*", i): + end = sql.find("*/", i + 2) + if end == -1: + return "" + i = end + 2 + else: + break + j = i + while j < n and (sql[j].isalpha() or sql[j] == "_"): + j += 1 + return sql[i:j].upper() class Database: @@ -667,16 +704,14 @@ class Database: "query() can only be used with SQL that returns rows - " "use execute() for other statements" ) - prefix = sql.lstrip().upper() - if prefix.startswith( - _TRANSACTION_CONTROL_PREFIXES + ("VACUUM", "ATTACH", "DETACH") - ): + keyword = _first_keyword(sql) + if keyword in _QUERY_REJECTED_KEYWORDS: # None of these return rows - reject them without executing anything raise ValueError(message) if self._tracer: self._tracer(sql, params) args: tuple = (params,) if params is not None else () - if prefix.startswith("PRAGMA"): + if keyword == "PRAGMA": # Some PRAGMA statements refuse to run inside a transaction, so # execute these without the savepoint guard used below. PRAGMAs # never open an implicit transaction, so there is nothing to @@ -741,7 +776,7 @@ class Database: not was_in_transaction and self.conn.in_transaction and cursor.description is None - and not sql.lstrip().upper().startswith(_TRANSACTION_CONTROL_PREFIXES) + and _first_keyword(sql) not in _TRANSACTION_CONTROL_KEYWORDS ): # The statement opened an implicit transaction - commit it, so # that execute() behaves consistently with the rest of the diff --git a/tests/test_atomic.py b/tests/test_atomic.py index a8e3563..f75c4ee 100644 --- a/tests/test_atomic.py +++ b/tests/test_atomic.py @@ -247,6 +247,17 @@ def test_execute_write_respects_explicit_transaction(fresh_db): assert [r["id"] for r in fresh_db["t"].rows] == [1] +def test_execute_comment_prefixed_begin_leaves_transaction_open(fresh_db): + # A BEGIN hidden behind a leading comment must not be auto-committed + # out from under the caller + fresh_db["t"].insert({"id": 1}, pk="id") + fresh_db.execute("-- start a transaction\nbegin") + assert fresh_db.conn.in_transaction + fresh_db.execute("insert into t (id) values (2)") + fresh_db.rollback() + assert [r["id"] for r in fresh_db["t"].rows] == [1] + + def test_query_returning_commits_after_iteration(tmpdir): if sqlite3.sqlite_version_info < (3, 35, 0): import pytest as _pytest diff --git a/tests/test_query.py b/tests/test_query.py index cf8d9f5..1d5c1f6 100644 --- a/tests/test_query.py +++ b/tests/test_query.py @@ -48,7 +48,18 @@ def test_query_rejected_write_inside_transaction_is_rolled_back(fresh_db): @pytest.mark.parametrize( - "sql", ["begin", "commit", "rollback", "vacuum", "detach database foo"] + "sql", + [ + "begin", + "commit", + "rollback", + "vacuum", + "detach database foo", + "/* comment */ commit", + "-- comment\nbegin", + "/* multi\nline */ -- and another\n vacuum", + "\t /* a */ /* b */ savepoint s1", + ], ) def test_query_rejects_transaction_control_and_vacuum(fresh_db, sql): with pytest.raises(ValueError) as ex: @@ -57,6 +68,21 @@ def test_query_rejects_transaction_control_and_vacuum(fresh_db, sql): assert not fresh_db.conn.in_transaction +def test_query_comment_prefixed_commit_does_not_commit_transaction(fresh_db): + # A COMMIT hidden behind a leading comment must not slip past the + # keyword check - previously it committed the caller's open + # transaction before the ValueError was raised + fresh_db["dogs"].insert({"name": "Cleo"}) + fresh_db.begin() + fresh_db.execute("insert into dogs (name) values ('Pancakes')") + with pytest.raises(ValueError): + fresh_db.query("/* comment */ COMMIT") + # The explicit transaction is still open and can still be rolled back + assert fresh_db.conn.in_transaction + fresh_db.rollback() + assert [row["name"] for row in fresh_db["dogs"].rows] == ["Cleo"] + + def test_query_error_leaves_no_transaction_open(fresh_db): with pytest.raises(sqlite3.OperationalError): fresh_db.query("select * from missing_table") @@ -75,6 +101,41 @@ def test_query_pragma(tmpdir): db.close() +def test_query_comment_prefixed_pragma(tmpdir): + from sqlite_utils import Database + + db = Database(str(tmpdir / "test.db")) + # A leading comment must not stop a PRAGMA being recognized as one - + # previously it was executed inside the savepoint guard, where + # journal mode changes are refused + assert list(db.query("-- set WAL mode\npragma journal_mode = wal")) == [ + {"journal_mode": "wal"} + ] + db.close() + + +@pytest.mark.parametrize( + "sql,expected", + [ + ("select 1", "SELECT"), + (" \t\n select 1", "SELECT"), + ("-- comment\nbegin", "BEGIN"), + ("/* one */ /* two */ pragma user_version", "PRAGMA"), + ("/* multi\nline */vacuum", "VACUUM"), + ("insert into t values (1)", "INSERT"), + ("-- only a comment", ""), + ("/* unterminated", ""), + ("", ""), + (" ", ""), + ("123", ""), + ], +) +def test_first_keyword(sql, expected): + from sqlite_utils.db import _first_keyword + + assert _first_keyword(sql) == expected + + @pytest.mark.skipif( sqlite3.sqlite_version_info < (3, 35, 0), reason="RETURNING requires SQLite 3.35.0 or higher", From 61619498fa730b4bdaa6f4891465840ad177f2f2 Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Sun, 5 Jul 2026 11:50:10 -0700 Subject: [PATCH 05/86] Give db.query() SQL restrictions top billing in upgrading docs Rejecting statements that return no rows is the change most likely to require code edits when upgrading to 4.0, so it now leads the Python API changes section - before db.table() - with the full list of rejected statement types and a before/after example. Co-Authored-By: Claude Fable 5 --- docs/upgrading.rst | 17 ++++++++++++----- 1 file changed, 12 insertions(+), 5 deletions(-) diff --git a/docs/upgrading.rst b/docs/upgrading.rst index 1974ad6..f743a3e 100644 --- a/docs/upgrading.rst +++ b/docs/upgrading.rst @@ -46,6 +46,18 @@ Two related things have been removed: Python API changes ------------------ +**db.query() now rejects SQL that does not return rows.** This is likely the most common change you will need to make to existing code. ``db.query()`` used to accept any SQL statement - passing one that returns no rows, such as an ``INSERT`` or ``UPDATE`` without a ``RETURNING`` clause or a ``CREATE TABLE``, did nothing at all, silently. Those statements now raise a ``ValueError``, and are rolled back so they have no effect on the database. Transaction control statements (``BEGIN``, ``COMMIT``, ``END``, ``ROLLBACK``, ``SAVEPOINT``, ``RELEASE``) plus ``VACUUM``, ``ATTACH`` and ``DETACH`` are also rejected with a ``ValueError``, without being executed at all. Use ``db.execute()`` for statements that do not return rows: + +.. code-block:: python + + # 3.x accepted this but silently did nothing: + db.query("update dogs set name = 'Cleopaws'") + + # In 4.0 use execute() for SQL that does not return rows: + db.execute("update dogs set name = 'Cleopaws'") + +**db.query() executes immediately.** ``db.query(sql)`` previously returned a generator that did not execute the SQL until you started iterating over it. The SQL now runs as soon as the method is called - rows are still fetched lazily, but errors in your SQL raise at the ``db.query()`` call site rather than on first iteration, and a write with a ``RETURNING`` clause takes effect even if you never iterate over its results. + **db.table() no longer returns views.** ``db.table(name)`` now raises a ``sqlite_utils.db.NoTable`` exception if ``name`` is a SQL view. Use the new ``db.view(name)`` method for views: .. code-block:: python @@ -55,11 +67,6 @@ Python API changes ``db["name"]`` still returns either a ``Table`` or a ``View`` depending on what exists in the database. -**db.query() executes immediately.** ``db.query(sql)`` previously returned a generator that did not execute the SQL until you started iterating over it. The SQL now runs as soon as the method is called - rows are still fetched lazily. Two consequences: - -- Errors in your SQL now raise at the ``db.query()`` call site rather than on first iteration. -- Passing a statement that returns no rows - such as an ``INSERT`` or ``UPDATE`` without a ``RETURNING`` clause - previously did nothing at all, silently. It now raises a ``ValueError``, and the statement is rolled back so it has no effect on the database. Use ``db.execute()`` for statements that do not return rows. - **Upserts use INSERT ... ON CONFLICT.** Upsert operations now use SQLite's ``INSERT ... ON CONFLICT SET`` syntax rather than the previous ``INSERT OR IGNORE`` followed by ``UPDATE``. If your code depends on the old behavior, pass ``use_old_upsert=True`` to the ``Database()`` constructor - see :ref:`python_api_old_upsert`. **Upsert records must include their primary keys.** ``table.upsert()`` and ``table.upsert_all()`` now raise ``sqlite_utils.db.PrimaryKeyRequired`` if a record is missing a value for any primary key column (or has ``None`` for one). Previously such records were quietly inserted as new rows. Relatedly, ``pk=`` is now optional when the table already exists with a primary key - it is detected automatically. From 623331b3f4ee78e1ed2618ccfc2e18e0bd77cd36 Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Sun, 5 Jul 2026 12:16:11 -0700 Subject: [PATCH 06/86] Fix for test failure against sqlean --- sqlite_utils/db.py | 12 ++++++++---- tests/test_query.py | 9 +++++++++ 2 files changed, 17 insertions(+), 4 deletions(-) diff --git a/sqlite_utils/db.py b/sqlite_utils/db.py index f796a63..a8f8e17 100644 --- a/sqlite_utils/db.py +++ b/sqlite_utils/db.py @@ -713,10 +713,14 @@ class Database: args: tuple = (params,) if params is not None else () if keyword == "PRAGMA": # Some PRAGMA statements refuse to run inside a transaction, so - # execute these without the savepoint guard used below. PRAGMAs - # never open an implicit transaction, so there is nothing to - # undo if this one turns out not to return rows - cursor = self.conn.execute(sql, *args) + # execute these without the savepoint guard used below. Some + # adapters open an implicit transaction before comment-prefixed + # PRAGMAs, so temporarily use driver autocommit when it is safe. + if self.conn.in_transaction: + cursor = self.conn.execute(sql, *args) + else: + with self.ensure_autocommit_off(): + cursor = self.conn.execute(sql, *args) if cursor.description is None: raise ValueError(message) keys = [d[0] for d in cursor.description] diff --git a/tests/test_query.py b/tests/test_query.py index 1d5c1f6..b9822e1 100644 --- a/tests/test_query.py +++ b/tests/test_query.py @@ -114,6 +114,15 @@ def test_query_comment_prefixed_pragma(tmpdir): db.close() +def test_query_comment_prefixed_pragma_inside_transaction(fresh_db): + fresh_db.begin() + assert list(fresh_db.query("-- check version\npragma user_version")) == [ + {"user_version": 0} + ] + assert fresh_db.conn.in_transaction + fresh_db.rollback() + + @pytest.mark.parametrize( "sql,expected", [ From f10459cffbdaaada4a61b6b1b1b4bdb59041e88d Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Sun, 5 Jul 2026 13:59:25 -0700 Subject: [PATCH 07/86] table.foreign_keys now handles compound foreign keys, refs #594 This is a breaking change, since the return type is now a dataclass and not a namedtuple. --- docs/changelog.rst | 9 +++ docs/python-api.rst | 32 +++++++-- docs/upgrading.rst | 16 +++++ sqlite_utils/db.py | 129 ++++++++++++++++++++++++++++--------- tests/test_foreign_keys.py | 82 +++++++++++++++++++++++ 5 files changed, 232 insertions(+), 36 deletions(-) create mode 100644 tests/test_foreign_keys.py diff --git a/docs/changelog.rst b/docs/changelog.rst index 2e1f69c..01d59a0 100644 --- a/docs/changelog.rst +++ b/docs/changelog.rst @@ -4,6 +4,15 @@ Changelog =========== +.. _v_unreleased: + +Unreleased +---------- + +Breaking changes: + +- ``table.foreign_keys`` now returns ``ForeignKey`` objects that are dataclasses rather than ``namedtuple`` instances, so they can no longer be unpacked or indexed as ``(table, column, other_table, other_column)`` tuples - access their fields by name instead. Compound (multi-column) foreign keys are now represented as a single ``ForeignKey`` with ``is_compound=True`` and populated ``columns``/``other_columns`` lists, where ``column`` and ``other_column`` are ``None``. Previously they were returned as one ``ForeignKey`` per column, misleadingly suggesting several independent foreign keys. See :ref:`upgrading_3_to_4` for details. (:issue:`594`) + .. _v4_0rc2: 4.0rc2 (2026-07-04) diff --git a/docs/python-api.rst b/docs/python-api.rst index 3f7e12c..d1f466f 100644 --- a/docs/python-api.rst +++ b/docs/python-api.rst @@ -2204,17 +2204,35 @@ Almost all SQLite tables have a ``rowid`` column, but a table with no explicitly .foreign_keys ------------- -The ``.foreign_keys`` property returns any foreign key relationships for the table, as a list of ``ForeignKey(table, column, other_table, other_column)`` named tuples. It is not available on views. +The ``.foreign_keys`` property returns any foreign key relationships for the table, as a list of ``ForeignKey`` objects. It is not available on views. + +Each ``ForeignKey`` has the following attributes: + +``table`` + The table the foreign key is defined on. +``column`` + The column on this table, or ``None`` for a compound foreign key. +``other_table`` + The table being referenced. +``other_column`` + The referenced column, or ``None`` for a compound foreign key. +``columns`` + A list of the columns on this table, always populated (a single-item list for single-column foreign keys). +``other_columns`` + A list of the referenced columns. +``is_compound`` + ``True`` if this is a compound (multi-column) foreign key. + +``ForeignKey`` was a ``namedtuple`` prior to sqlite-utils 4.0. It is now a dataclass and can no longer be unpacked or indexed as a tuple - access its fields by name instead. See :ref:`upgrading_3_to_4` for details. :: >>> db.table("Street_Tree_List").foreign_keys - [ForeignKey(table='Street_Tree_List', column='qLegalStatus', other_table='qLegalStatus', other_column='id'), - ForeignKey(table='Street_Tree_List', column='qCareAssistant', other_table='qCareAssistant', other_column='id'), - ForeignKey(table='Street_Tree_List', column='qSiteInfo', other_table='qSiteInfo', other_column='id'), - ForeignKey(table='Street_Tree_List', column='qSpecies', other_table='qSpecies', other_column='id'), - ForeignKey(table='Street_Tree_List', column='qCaretaker', other_table='qCaretaker', other_column='id'), - ForeignKey(table='Street_Tree_List', column='PlantType', other_table='PlantType', other_column='id')] + [ForeignKey(table='Street_Tree_List', column='qLegalStatus', other_table='qLegalStatus', other_column='id', columns=['qLegalStatus'], other_columns=['id'], is_compound=False), + ForeignKey(table='Street_Tree_List', column='qCareAssistant', other_table='qCareAssistant', other_column='id', columns=['qCareAssistant'], other_columns=['id'], is_compound=False), + ...] + +Compound foreign keys - defined with ``FOREIGN KEY (col_a, col_b) REFERENCES other(col_a, col_b)`` - are returned as a single ``ForeignKey`` with ``is_compound=True``, ``column`` and ``other_column`` set to ``None``, and the participating columns available in the ``columns`` and ``other_columns`` lists. .. _python_api_introspection_schema: diff --git a/docs/upgrading.rst b/docs/upgrading.rst index f743a3e..eb997f6 100644 --- a/docs/upgrading.rst +++ b/docs/upgrading.rst @@ -79,6 +79,22 @@ Python API changes **View.enable_fts() has been removed.** The ``View`` class previously had an ``enable_fts()`` method that existed only to raise ``NotImplementedError`` - full-text search is not supported for views. Calling it now raises ``AttributeError`` like any other missing method. +**ForeignKey is now a dataclass, not a namedtuple.** The ``ForeignKey`` objects returned by ``table.foreign_keys`` gained three new fields - ``columns``, ``other_columns`` and ``is_compound`` - so that compound (multi-column) foreign keys can be represented as a single object. To make room for those fields cleanly ``ForeignKey`` is now a dataclass rather than a ``namedtuple``, so it can no longer be unpacked or indexed as a tuple. Access its fields by name instead: + +.. code-block:: python + + # 3.x - tuple unpacking, no longer works: + for table, column, other_table, other_column in db["courses"].foreign_keys: + ... + + # 4.0 - access fields by name: + for fk in db["courses"].foreign_keys: + fk.table, fk.column, fk.other_table, fk.other_column + +Attempting the old unpacking or ``fk[0]`` indexing now raises ``TypeError``, so any code using those patterns will fail loudly rather than silently misbehave. + +Compound foreign keys - previously returned as one ``ForeignKey`` per column, misleadingly suggesting several independent single-column keys - are now returned as a single ``ForeignKey`` with ``is_compound=True``. For these the scalar ``column`` and ``other_column`` fields are ``None``; use the ``columns`` and ``other_columns`` lists instead. Single-column foreign keys are unaffected apart from the class change: ``column``/``other_column`` behave as before and ``columns``/``other_columns`` are single-item lists. + **Validation errors raise ValueError.** Invalid arguments to Python API methods - for example ``create_table()`` with no columns, or ``ignore=True`` together with ``replace=True`` - now raise ``ValueError``. They previously raised ``AssertionError`` from bare ``assert`` statements, which were silently skipped under ``python -O``. **Transaction behavior is now well-defined.** 4.0 introduces the :ref:`db.atomic() ` context manager and uses it consistently for every write operation - the full model is described in :ref:`python_api_transactions`. Changes you may notice: diff --git a/sqlite_utils/db.py b/sqlite_utils/db.py index a8f8e17..b1cdf83 100644 --- a/sqlite_utils/db.py +++ b/sqlite_utils/db.py @@ -11,6 +11,7 @@ from .utils import ( ) import binascii from collections import namedtuple +from dataclasses import dataclass, field from collections.abc import Mapping import contextlib import datetime @@ -161,9 +162,43 @@ Summary information about a column, see :ref:`python_api_analyze_column`. The ``N`` least common values as a list of ``(value, count)`` tuples, or ``None`` if the table is entirely distinct or if the number of distinct values is less than N (since they will already have been returned in ``most_common``) """ -ForeignKey = namedtuple( - "ForeignKey", ("table", "column", "other_table", "other_column") -) + + +@dataclass(order=True) +class ForeignKey: + """ + A foreign key defined on a table. + + For single-column foreign keys ``column`` and ``other_column`` hold the + column names, and ``columns``/``other_columns`` are single-item lists. + + For compound (multi-column) foreign keys ``column`` and ``other_column`` + are ``None`` - use ``columns`` and ``other_columns`` instead, and check + ``is_compound``. + + Prior to sqlite-utils 4.0 this was a ``namedtuple`` and could be unpacked + or indexed as ``(table, column, other_table, other_column)``. It is now a + dataclass - access its fields by name instead. + """ + + table: str + column: Optional[str] + other_table: str + other_column: Optional[str] + columns: List[str] = field(default_factory=list) + other_columns: List[str] = field(default_factory=list) + is_compound: bool = False + + def __post_init__(self): + # Populate columns/other_columns for single-column foreign keys + if not self.columns: + self.columns = [self.column] if self.column is not None else [] + if not self.other_columns: + self.other_columns = ( + [self.other_column] if self.other_column is not None else [] + ) + + Index = namedtuple("Index", ("seq", "name", "unique", "origin", "partial", "columns")) XIndex = namedtuple("XIndex", ("name", "columns")) XIndexColumn = namedtuple( @@ -1124,7 +1159,7 @@ class Database: if not all(isinstance(fk, (tuple, list)) for fk in foreign_keys): raise ValueError("foreign_keys= should be a list of tuples") fks = [] - for tuple_or_list in foreign_keys: + for tuple_or_list in cast(Iterable[Sequence[str]], foreign_keys): if len(tuple_or_list) == 4: if tuple_or_list[0] != name: raise ValueError( @@ -1234,7 +1269,7 @@ class Database: pk = hash_id # Soundness check foreign_keys point to existing tables for fk in foreign_keys: - if fk.other_table == name and columns.get(fk.other_column): + if fk.other_table == name and columns.get(cast(str, fk.other_column)): continue if fk.other_column != "rowid" and not any( c for c in self[fk.other_table].columns if c.name == fk.other_column @@ -1269,7 +1304,7 @@ class Database: foreign_keys_by_column[column_name].other_table ), quote_identifier( - foreign_keys_by_column[column_name].other_column + cast(str, foreign_keys_by_column[column_name].other_column) ), ) ) @@ -1499,7 +1534,9 @@ class Database: """ # foreign_keys is a list of explicit 4-tuples if not all( - len(fk) == 4 and isinstance(fk, (list, tuple)) for fk in foreign_keys + isinstance(fk, ForeignKey) + or (isinstance(fk, (list, tuple)) and len(fk) == 4) + for fk in foreign_keys ): raise ValueError( "foreign_keys must be a list of 4-tuples, " @@ -1509,7 +1546,16 @@ class Database: foreign_keys_to_create = [] # Verify that all tables and columns exist - for table, column, other_table, other_column in foreign_keys: + for fk in foreign_keys: + if isinstance(fk, ForeignKey): + table, column, other_table, other_column = ( + fk.table, + fk.column, + fk.other_table, + fk.other_column, + ) + else: + table, column, other_table, other_column = fk if not self.table(table).exists(): raise AlterError("No such table: {}".format(table)) table_obj = self.table(table) @@ -1555,8 +1601,11 @@ class Database: i.columns[0] for i in table.indexes if len(i.columns) == 1 } for fk in table.foreign_keys: - if fk.column not in existing_indexes: - table.create_index([fk.column], find_unique_name=True) + # Compound foreign keys expose their columns via fk.columns; + # single-column keys yield a one-item list + for column in fk.columns: + if column not in existing_indexes: + table.create_index([column], find_unique_name=True) def vacuum(self) -> None: "Run a SQLite ``VACUUM`` against the database." @@ -1907,21 +1956,40 @@ class Table(Queryable): @property def foreign_keys(self) -> List["ForeignKey"]: - "List of foreign keys defined on this table." - fks = [] + """ + List of foreign keys defined on this table. + + Compound (multi-column) foreign keys are returned as a single + ``ForeignKey`` with ``is_compound=True`` and populated + ``columns``/``other_columns`` lists. + """ + # PRAGMA foreign_key_list returns one row per column, grouped by "id" + # with "seq" giving the column order within a compound foreign key. + by_id: Dict[int, list] = {} for row in self.db.execute( "PRAGMA foreign_key_list({})".format(quote_identifier(self.name)) ).fetchall(): if row is not None: id, seq, table_name, from_, to_, on_update, on_delete, match = row - fks.append( - ForeignKey( - table=self.name, - column=from_, - other_table=table_name, - other_column=to_, - ) + by_id.setdefault(id, []).append((seq, table_name, from_, to_)) + fks = [] + for id in sorted(by_id): + rows = sorted(by_id[id]) # order columns by seq + other_table = rows[0][1] + columns = [row[2] for row in rows] + other_columns = [row[3] for row in rows] + is_compound = len(rows) > 1 + fks.append( + ForeignKey( + table=self.name, + column=None if is_compound else columns[0], + other_table=other_table, + other_column=None if is_compound else other_columns[0], + columns=columns, + other_columns=other_columns, + is_compound=is_compound, ) + ) return fks @property @@ -2254,17 +2322,20 @@ class Table(Queryable): else: # Construct foreign_keys from current, plus add_foreign_keys, minus drop_foreign_keys create_table_foreign_keys = [] - for table, column, other_table, other_column in self.foreign_keys: - # Copy over old foreign keys, unless we are dropping them - if (drop_foreign_keys is None) or (column not in drop_foreign_keys): - create_table_foreign_keys.append( - ForeignKey( - table, - rename.get(column) or column, - other_table, - other_column, + for fk in self.foreign_keys: + # Expand compound foreign keys into per-column references; + # for single-column keys this iterates exactly once + for column, other_column in zip(fk.columns, fk.other_columns): + # Copy over old foreign keys, unless we are dropping them + if (drop_foreign_keys is None) or (column not in drop_foreign_keys): + create_table_foreign_keys.append( + ForeignKey( + fk.table, + rename.get(column) or column, + fk.other_table, + other_column, + ) ) - ) # Add new foreign keys if add_foreign_keys is not None: for fk in self.db.resolve_foreign_keys(self.name, add_foreign_keys): diff --git a/tests/test_foreign_keys.py b/tests/test_foreign_keys.py new file mode 100644 index 0000000..c785b59 --- /dev/null +++ b/tests/test_foreign_keys.py @@ -0,0 +1,82 @@ +"""Tests for reading compound (multi-column) foreign keys - issue #594.""" + +import pytest +from sqlite_utils import Database + +COMPOUND_SCHEMA = """ +CREATE TABLE departments ( + campus_name TEXT NOT NULL, + dept_code TEXT NOT NULL, + dept_name TEXT, + PRIMARY KEY (campus_name, dept_code) +); +CREATE TABLE courses ( + course_code TEXT PRIMARY KEY, + course_name TEXT, + campus_name TEXT NOT NULL, + dept_code TEXT NOT NULL, + FOREIGN KEY (campus_name, dept_code) + REFERENCES departments(campus_name, dept_code) +); +""" + + +@pytest.fixture +def compound_db(): + db = Database(memory=True) + db.executescript(COMPOUND_SCHEMA) + return db + + +def test_compound_foreign_key(compound_db): + fks = compound_db["courses"].foreign_keys + assert len(fks) == 1 + fk = fks[0] + assert fk.is_compound is True + assert fk.table == "courses" + assert fk.other_table == "departments" + assert fk.columns == ["campus_name", "dept_code"] + assert fk.other_columns == ["campus_name", "dept_code"] + # Scalar column/other_column can't sensibly hold a compound key + assert fk.column is None + assert fk.other_column is None + + +def test_single_foreign_key_gets_columns_fields(fresh_db): + fresh_db["authors"].insert({"id": 1, "name": "Sally"}, pk="id") + fresh_db["books"].insert({"title": "Hedgehogs", "author_id": 1}) + fresh_db["books"].add_foreign_key("author_id", "authors", "id") + fk = fresh_db["books"].foreign_keys[0] + assert fk.is_compound is False + assert fk.column == "author_id" + assert fk.other_column == "id" + assert fk.columns == ["author_id"] + assert fk.other_columns == ["id"] + + +def test_foreign_key_no_longer_unpacks_as_tuple(fresh_db): + # Clean break in 4.0: ForeignKey is a dataclass, not a namedtuple, so the + # old tuple unpacking and indexing patterns now fail hard. + fresh_db["authors"].insert({"id": 1, "name": "Sally"}, pk="id") + fresh_db["books"].insert({"title": "Hedgehogs", "author_id": 1}) + fresh_db["books"].add_foreign_key("author_id", "authors", "id") + fk = fresh_db["books"].foreign_keys[0] + with pytest.raises(TypeError): + table, column, other_table, other_column = fk + with pytest.raises(TypeError): + fk[0] + + +def test_foreign_keys_are_sortable(fresh_db): + fresh_db["authors"].insert({"id": 1, "name": "Sally"}, pk="id") + fresh_db["categories"].insert({"id": 1, "name": "Wildlife"}, pk="id") + fresh_db["books"].insert({"title": "Hedgehogs", "author_id": 1, "category_id": 1}) + fresh_db.add_foreign_keys( + [ + ("books", "author_id", "authors", "id"), + ("books", "category_id", "categories", "id"), + ] + ) + fks = sorted(fresh_db["books"].foreign_keys) + assert fks[0].column == "author_id" + assert fks[1].column == "category_id" From d5bf51df35a527aae19417c97eb3949ffb6df5bc Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Sun, 5 Jul 2026 14:26:56 -0700 Subject: [PATCH 08/86] Fix TypeError when sorting mixed compound/single foreign keys sorted() on a foreign_keys list containing both compound (column=None) and single-column ForeignKeys raised TypeError because dataclass ordering compared None against str. column/other_column are now excluded from comparison - ordering and equality use the always populated columns/other_columns lists instead. Co-Authored-By: Claude Fable 5 --- sqlite_utils/db.py | 6 ++++-- tests/test_foreign_keys.py | 28 ++++++++++++++++++++++++++++ 2 files changed, 32 insertions(+), 2 deletions(-) diff --git a/sqlite_utils/db.py b/sqlite_utils/db.py index b1cdf83..67623fe 100644 --- a/sqlite_utils/db.py +++ b/sqlite_utils/db.py @@ -182,9 +182,11 @@ class ForeignKey: """ table: str - column: Optional[str] + # column/other_column are None for compound keys, which would break + # ordering against str values - comparison uses columns/other_columns + column: Optional[str] = field(compare=False) other_table: str - other_column: Optional[str] + other_column: Optional[str] = field(compare=False) columns: List[str] = field(default_factory=list) other_columns: List[str] = field(default_factory=list) is_compound: bool = False diff --git a/tests/test_foreign_keys.py b/tests/test_foreign_keys.py index c785b59..4b371a9 100644 --- a/tests/test_foreign_keys.py +++ b/tests/test_foreign_keys.py @@ -80,3 +80,31 @@ def test_foreign_keys_are_sortable(fresh_db): fks = sorted(fresh_db["books"].foreign_keys) assert fks[0].column == "author_id" assert fks[1].column == "category_id" + + +def test_mixed_compound_and_single_foreign_keys_are_sortable(): + # compound FKs have column=None, which must not break sorting + # against single-column FKs (None < str raises TypeError) + db = Database(memory=True) + db.executescript(""" + CREATE TABLE departments ( + campus_name TEXT NOT NULL, + dept_code TEXT NOT NULL, + PRIMARY KEY (campus_name, dept_code) + ); + CREATE TABLE accreditations (id INTEGER PRIMARY KEY); + CREATE TABLE courses ( + course_code TEXT PRIMARY KEY, + campus_name TEXT NOT NULL, + dept_code TEXT NOT NULL, + accreditation_id INTEGER REFERENCES accreditations(id), + FOREIGN KEY (campus_name, dept_code) + REFERENCES departments(campus_name, dept_code) + ); + """) + fks = db["courses"].foreign_keys + assert len(fks) == 2 + assert {fk.is_compound for fk in fks} == {True, False} + fks_sorted = sorted(fks) + assert fks_sorted[0].other_table == "accreditations" + assert fks_sorted[1].other_table == "departments" From 577f3011e5bbc5a043b7e4f9d10b06da76e21bf1 Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Sun, 5 Jul 2026 14:30:41 -0700 Subject: [PATCH 09/86] Create tables with compound foreign keys, refs #594 create_table() and friends now accept compound foreign keys, specified as lists of column names in the existing tuple forms: foreign_keys=[ (["campus_name", "dept_code"], "departments"), (["campus_name", "dept_code"], "departments", ["campus_name", "dept_code"]), ] The two-item form guesses the compound primary key of the other table. Compound keys are rendered as table-level FOREIGN KEY constraints, while single-column keys keep their inline REFERENCES clauses. Co-Authored-By: Claude Fable 5 --- docs/python-api.rst | 36 ++++++++++++ sqlite_utils/db.py | 110 +++++++++++++++++++++++++++++-------- tests/test_foreign_keys.py | 96 +++++++++++++++++++++++++++++++- 3 files changed, 218 insertions(+), 24 deletions(-) diff --git a/docs/python-api.rst b/docs/python-api.rst index d1f466f..5e55256 100644 --- a/docs/python-api.rst +++ b/docs/python-api.rst @@ -817,6 +817,42 @@ You can leave off the third item in the tuple to have the referenced column auto ("author_id", "authors") ]) +.. _python_api_compound_foreign_keys: + +Compound foreign keys +~~~~~~~~~~~~~~~~~~~~~ + +To create a compound (multi-column) foreign key, use lists of column names in place of the single column names: + +.. code-block:: python + + db.table("courses").create({ + "course_code": str, + "campus_name": str, + "dept_code": str, + }, pk="course_code", foreign_keys=[ + (["campus_name", "dept_code"], "departments", ["campus_name", "dept_code"]) + ]) + +This creates a table-level constraint: + +.. code-block:: sql + + CREATE TABLE "courses" ( + "course_code" TEXT PRIMARY KEY, + "campus_name" TEXT, + "dept_code" TEXT, + FOREIGN KEY ("campus_name", "dept_code") REFERENCES "departments"("campus_name", "dept_code") + ) + +As with single columns, you can leave off the list of other columns to reference the compound primary key of the other table: + +.. code-block:: python + + foreign_keys=[ + (["campus_name", "dept_code"], "departments") + ] + .. _python_api_table_configuration: Table configuration options diff --git a/sqlite_utils/db.py b/sqlite_utils/db.py index 67623fe..12914d1 100644 --- a/sqlite_utils/db.py +++ b/sqlite_utils/db.py @@ -219,6 +219,10 @@ ForeignKeyIndicator = Union[ Tuple[str, str], Tuple[str, str, str], Tuple[str, str, str, str], + # Compound foreign keys use lists of columns: + Tuple[List[str], str], + Tuple[List[str], str, List[str]], + Tuple[str, List[str], str, List[str]], ] ForeignKeysType = Union[Iterable[ForeignKeyIndicator], List[ForeignKeyIndicator]] @@ -1142,9 +1146,12 @@ class Database: :param name: Name of table that foreign keys are being defined for :param foreign_keys: List of foreign keys, each of which can be a - string, a ForeignKey() named tuple, a tuple of (column, other_table), + string, a ForeignKey() object, a tuple of (column, other_table), or a tuple of (column, other_table, other_column), or a tuple of - (table, column, other_table, other_column) + (table, column, other_table, other_column). For compound foreign + keys the column elements can be lists of column names, e.g. + (["campus_name", "dept_code"], "departments") or + (["campus_name", "dept_code"], "departments", ["campus_name", "dept_code"]) """ table = self.table(name) if all(isinstance(fk, ForeignKey) for fk in foreign_keys): @@ -1161,7 +1168,7 @@ class Database: if not all(isinstance(fk, (tuple, list)) for fk in foreign_keys): raise ValueError("foreign_keys= should be a list of tuples") fks = [] - for tuple_or_list in cast(Iterable[Sequence[str]], foreign_keys): + for tuple_or_list in cast(Iterable[Sequence[Any]], foreign_keys): if len(tuple_or_list) == 4: if tuple_or_list[0] != name: raise ValueError( @@ -1169,28 +1176,61 @@ class Database: tuple_or_list, name ) ) - if len(tuple_or_list) not in (2, 3, 4): + tuple_or_list = tuple_or_list[1:] + if len(tuple_or_list) not in (2, 3): raise ValueError( "foreign_keys= should be a list of tuple pairs or triples" ) - if len(tuple_or_list) in (3, 4): - if len(tuple_or_list) == 4: - tuple_or_list = cast(Tuple[str, str, str], tuple_or_list[1:]) + column_or_columns = tuple_or_list[0] + other_table = tuple_or_list[1] + if isinstance(column_or_columns, (list, tuple)): + # Compound foreign key + columns = list(column_or_columns) + if len(tuple_or_list) == 3: + other_columns = tuple_or_list[2] + if not isinstance(other_columns, (list, tuple)): + raise ValueError( + "Compound foreign key {} should reference a list " + "of other columns".format(tuple(tuple_or_list)) + ) + other_columns = list(other_columns) else: - tuple_or_list = cast(Tuple[str, str, str], tuple_or_list) - fks.append( - ForeignKey( - name, tuple_or_list[0], tuple_or_list[1], tuple_or_list[2] + # Guess the compound primary key of the other table + other_columns = self.table(other_table).pks + if len(columns) != len(other_columns): + raise ValueError( + "Compound foreign key {} should have the same number " + "of columns on both sides".format(tuple(tuple_or_list)) ) + if len(columns) == 1: + # Single-column key passed as a one-item list + fks.append( + ForeignKey(name, columns[0], other_table, other_columns[0]) + ) + else: + fks.append( + ForeignKey( + name, + None, + other_table, + None, + columns=columns, + other_columns=other_columns, + is_compound=True, + ) + ) + elif len(tuple_or_list) == 3: + fks.append( + ForeignKey(name, column_or_columns, other_table, tuple_or_list[2]) ) else: # Guess the primary key fks.append( ForeignKey( name, - tuple_or_list[0], - tuple_or_list[1], - table.guess_foreign_column(tuple_or_list[1]), + column_or_columns, + other_table, + table.guess_foreign_column(other_table), ) ) return fks @@ -1229,7 +1269,11 @@ class Database: if hash_id_columns and (hash_id is None): hash_id = "id" foreign_keys = self.resolve_foreign_keys(name, foreign_keys or []) - foreign_keys_by_column = {fk.column: fk for fk in foreign_keys} + # Compound foreign keys are rendered as table-level constraints; + # single-column ones as inline REFERENCES on their column + foreign_keys_by_column = { + fk.column: fk for fk in foreign_keys if not fk.is_compound + } # any extracts will be treated as integer columns with a foreign key extracts = resolve_extracts(extracts) for extract_column, extract_table in extracts.items(): @@ -1271,14 +1315,15 @@ class Database: pk = hash_id # Soundness check foreign_keys point to existing tables for fk in foreign_keys: - if fk.other_table == name and columns.get(cast(str, fk.other_column)): - continue - if fk.other_column != "rowid" and not any( - c for c in self[fk.other_table].columns if c.name == fk.other_column - ): - raise AlterError( - "No such column: {}.{}".format(fk.other_table, fk.other_column) - ) + for other_column in fk.other_columns: + if fk.other_table == name and columns.get(other_column): + continue + if other_column != "rowid" and not any( + c for c in self[fk.other_table].columns if c.name == other_column + ): + raise AlterError( + "No such column: {}.{}".format(fk.other_table, other_column) + ) column_defs = [] # ensure pk is a tuple @@ -1329,6 +1374,25 @@ class Database: extra_pk = ",\n PRIMARY KEY ({pks})".format( pks=", ".join([quote_identifier(p) for p in pk]) ) + # Compound foreign keys become table-level FOREIGN KEY constraints + column_names = [c[0] for c in column_items] + for fk in foreign_keys: + if not fk.is_compound: + continue + missing = [c for c in fk.columns if c not in column_names] + if missing: + raise AlterError( + "No such column: {}".format(", ".join(sorted(missing))) + ) + column_defs.append( + " FOREIGN KEY ({columns}) REFERENCES {other_table}({other_columns})".format( + columns=", ".join(quote_identifier(c) for c in fk.columns), + other_table=quote_identifier(fk.other_table), + other_columns=", ".join( + quote_identifier(c) for c in fk.other_columns + ), + ) + ) columns_sql = ",\n".join(column_defs) sql = """CREATE TABLE {if_not_exists}{table} ( {columns_sql}{extra_pk} diff --git a/tests/test_foreign_keys.py b/tests/test_foreign_keys.py index 4b371a9..457f307 100644 --- a/tests/test_foreign_keys.py +++ b/tests/test_foreign_keys.py @@ -1,7 +1,8 @@ -"""Tests for reading compound (multi-column) foreign keys - issue #594.""" +"""Tests for compound (multi-column) foreign keys - issue #594.""" import pytest from sqlite_utils import Database +from sqlite_utils.db import AlterError, ForeignKey COMPOUND_SCHEMA = """ CREATE TABLE departments ( @@ -108,3 +109,96 @@ def test_mixed_compound_and_single_foreign_keys_are_sortable(): fks_sorted = sorted(fks) assert fks_sorted[0].other_table == "accreditations" assert fks_sorted[1].other_table == "departments" + + +@pytest.fixture +def departments_db(): + db = Database(memory=True) + db.create_table( + "departments", + {"campus_name": str, "dept_code": str, "dept_name": str}, + pk=("campus_name", "dept_code"), + ) + return db + + +EXPECTED_COURSES_SCHEMA = ( + 'CREATE TABLE "courses" (\n' + ' "course_code" TEXT PRIMARY KEY,\n' + ' "campus_name" TEXT,\n' + ' "dept_code" TEXT,\n' + ' FOREIGN KEY ("campus_name", "dept_code") ' + 'REFERENCES "departments"("campus_name", "dept_code")\n' + ")" +) + + +@pytest.mark.parametrize( + "foreign_keys", + ( + [ + ForeignKey( + table="courses", + column=None, + other_table="departments", + other_column=None, + columns=["campus_name", "dept_code"], + other_columns=["campus_name", "dept_code"], + is_compound=True, + ) + ], + [(["campus_name", "dept_code"], "departments", ["campus_name", "dept_code"])], + # Two-item form guesses the other table's primary key: + [(["campus_name", "dept_code"], "departments")], + ), +) +def test_create_table_with_compound_foreign_key(departments_db, foreign_keys): + departments_db.create_table( + "courses", + {"course_code": str, "campus_name": str, "dept_code": str}, + pk="course_code", + foreign_keys=foreign_keys, + ) + assert departments_db["courses"].schema == EXPECTED_COURSES_SCHEMA + fks = departments_db["courses"].foreign_keys + assert len(fks) == 1 + fk = fks[0] + assert fk.is_compound is True + assert fk.columns == ["campus_name", "dept_code"] + assert fk.other_table == "departments" + assert fk.other_columns == ["campus_name", "dept_code"] + + +def test_create_table_compound_foreign_key_enforced(departments_db): + departments_db.execute("PRAGMA foreign_keys = ON") + departments_db.create_table( + "courses", + {"course_code": str, "campus_name": str, "dept_code": str}, + pk="course_code", + foreign_keys=[(["campus_name", "dept_code"], "departments")], + ) + departments_db["departments"].insert( + {"campus_name": "Berkeley", "dept_code": "CS", "dept_name": "Computer Science"} + ) + departments_db["courses"].insert( + {"course_code": "CS101", "campus_name": "Berkeley", "dept_code": "CS"} + ) + import sqlite3 + + with pytest.raises(sqlite3.IntegrityError): + departments_db.execute( + "insert into courses (course_code, campus_name, dept_code) " + "values ('X1', 'Nowhere', 'NOPE')" + ) + + +def test_create_table_compound_foreign_key_missing_other_column(departments_db): + with pytest.raises(AlterError): + departments_db.create_table( + "courses", + {"course_code": str, "campus_name": str, "dept_code": str}, + pk="course_code", + foreign_keys=[ + (["campus_name", "dept_code"], "departments", ["campus_name", "nope"]) + ], + ) From 7d43fd50e186c359cd0c61a6655f15f0d156da87 Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Sun, 5 Jul 2026 14:33:15 -0700 Subject: [PATCH 10/86] transform() now round-trips compound foreign keys, refs #594 Compound foreign keys are passed through table.transform() intact instead of being degraded to per-column single foreign keys: - rename= applies to each member column of a compound key - drop= of any member column drops the whole constraint, matching the existing single-column behavior - drop_foreign_keys= accepts a bare column name (drops any foreign key that column participates in) or a tuple of columns (drops the compound key with exactly those columns) Co-Authored-By: Claude Fable 5 --- docs/python-api.rst | 10 ++++++ sqlite_utils/db.py | 63 ++++++++++++++++++++++++-------------- tests/test_foreign_keys.py | 48 +++++++++++++++++++++++++++++ 3 files changed, 98 insertions(+), 23 deletions(-) diff --git a/docs/python-api.rst b/docs/python-api.rst index 5e55256..465df79 100644 --- a/docs/python-api.rst +++ b/docs/python-api.rst @@ -1793,6 +1793,16 @@ This example drops two foreign keys - the one from ``places.country`` to ``count drop_foreign_keys=("country", "continent") ) +A bare column name drops any foreign key that column participates in, including compound foreign keys. To target a compound foreign key precisely, pass a tuple of its columns: + +.. code-block:: python + + db.table("courses").transform( + drop_foreign_keys=[("campus_name", "dept_code")] + ) + +Renaming a column with ``rename=`` updates any foreign keys that use it, and dropping a column with ``drop=`` also drops any foreign keys it participates in - for a compound foreign key this removes the whole constraint. + .. _python_api_transform_sql: Custom transformations with .transform_sql() diff --git a/sqlite_utils/db.py b/sqlite_utils/db.py index 12914d1..c37f750 100644 --- a/sqlite_utils/db.py +++ b/sqlite_utils/db.py @@ -2275,7 +2275,9 @@ class Table(Queryable): :param pk: New primary key for the table :param not_null: Columns to set as ``NOT NULL`` :param defaults: Default values for columns - :param drop_foreign_keys: Names of columns that should have their foreign key constraints removed + :param drop_foreign_keys: Foreign key constraints to remove - a column name + drops any foreign key that column participates in, a tuple of column names + drops the compound foreign key with exactly those columns :param add_foreign_keys: List of foreign keys to add to the table :param foreign_keys: List of foreign keys to set for the table, replacing any existing foreign keys :param column_order: List of strings specifying a full or partial column order @@ -2360,7 +2362,9 @@ class Table(Queryable): :param pk: New primary key for the table :param not_null: Columns to set as ``NOT NULL`` :param defaults: Default values for columns - :param drop_foreign_keys: Names of columns that should have their foreign key constraints removed + :param drop_foreign_keys: Foreign key constraints to remove - a column name + drops any foreign key that column participates in, a tuple of column names + drops the compound foreign key with exactly those columns :param add_foreign_keys: List of foreign keys to add to the table :param foreign_keys: List of foreign keys to set for the table, replacing any existing foreign keys :param column_order: List of strings specifying a full or partial column order @@ -2387,32 +2391,45 @@ class Table(Queryable): create_table_foreign_keys.extend(foreign_keys) else: # Construct foreign_keys from current, plus add_foreign_keys, minus drop_foreign_keys + + def fk_should_be_dropped(fk: ForeignKey) -> bool: + if drop_foreign_keys is not None: + for spec in drop_foreign_keys: + if isinstance(spec, str): + # A column name matches any foreign key it participates in + if spec in fk.columns: + return True + elif list(spec) == fk.columns: + # A tuple/list must match a compound key's columns exactly + return True + # Dropping any of a foreign key's columns drops the whole key + return any(column in drop for column in fk.columns) + + def fk_with_renamed_columns(fk: ForeignKey) -> ForeignKey: + columns = [rename.get(column) or column for column in fk.columns] + if fk.is_compound: + return ForeignKey( + self.name, + None, + fk.other_table, + None, + columns=columns, + other_columns=fk.other_columns, + is_compound=True, + ) + return ForeignKey( + self.name, columns[0], fk.other_table, fk.other_columns[0] + ) + create_table_foreign_keys = [] + # Copy over old foreign keys, unless we are dropping them for fk in self.foreign_keys: - # Expand compound foreign keys into per-column references; - # for single-column keys this iterates exactly once - for column, other_column in zip(fk.columns, fk.other_columns): - # Copy over old foreign keys, unless we are dropping them - if (drop_foreign_keys is None) or (column not in drop_foreign_keys): - create_table_foreign_keys.append( - ForeignKey( - fk.table, - rename.get(column) or column, - fk.other_table, - other_column, - ) - ) + if not fk_should_be_dropped(fk): + create_table_foreign_keys.append(fk_with_renamed_columns(fk)) # Add new foreign keys if add_foreign_keys is not None: for fk in self.db.resolve_foreign_keys(self.name, add_foreign_keys): - create_table_foreign_keys.append( - ForeignKey( - self.name, - rename.get(fk.column) or fk.column, - fk.other_table, - fk.other_column, - ) - ) + create_table_foreign_keys.append(fk_with_renamed_columns(fk)) new_table_name = "{}_new_{}".format( self.name, tmp_suffix or os.urandom(6).hex() diff --git a/tests/test_foreign_keys.py b/tests/test_foreign_keys.py index 457f307..d6a7071 100644 --- a/tests/test_foreign_keys.py +++ b/tests/test_foreign_keys.py @@ -202,3 +202,51 @@ def test_create_table_compound_foreign_key_missing_other_column(departments_db): (["campus_name", "dept_code"], "departments", ["campus_name", "nope"]) ], ) + + +def test_transform_preserves_compound_foreign_key(compound_db): + compound_db["courses"].transform(rename={"course_name": "title"}) + fks = compound_db["courses"].foreign_keys + assert len(fks) == 1 + fk = fks[0] + assert fk.is_compound is True + assert fk.columns == ["campus_name", "dept_code"] + assert fk.other_table == "departments" + assert fk.other_columns == ["campus_name", "dept_code"] + + +def test_transform_rename_member_column_updates_compound_foreign_key(compound_db): + compound_db["courses"].transform(rename={"campus_name": "campus"}) + fks = compound_db["courses"].foreign_keys + assert len(fks) == 1 + fk = fks[0] + assert fk.is_compound is True + assert fk.columns == ["campus", "dept_code"] + # Referenced columns in the other table are unchanged + assert fk.other_columns == ["campus_name", "dept_code"] + + +def test_transform_drop_member_column_drops_compound_foreign_key(compound_db): + # Matches single-column behavior: dropping the column silently + # drops the foreign key that used it + compound_db["courses"].transform(drop={"dept_code"}) + assert compound_db["courses"].foreign_keys == [] + assert "FOREIGN KEY" not in compound_db["courses"].schema + + +@pytest.mark.parametrize( + "drop_foreign_keys", + ( + # A bare column name matches any foreign key it participates in: + ["campus_name"], + # A tuple must match the full compound key: + [("campus_name", "dept_code")], + ), +) +def test_transform_drop_compound_foreign_key(compound_db, drop_foreign_keys): + compound_db["courses"].transform(drop_foreign_keys=drop_foreign_keys) + assert compound_db["courses"].foreign_keys == [] + # The columns themselves survive + assert {"campus_name", "dept_code"} <= set( + compound_db["courses"].columns_dict.keys() + ) From b75edf4b30dd500747f6f9f5f1dcb52c03a5b442 Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Sun, 5 Jul 2026 14:37:32 -0700 Subject: [PATCH 11/86] add_foreign_key() accepts compound keys, composite FK indexes, refs #594 table.add_foreign_key() and db.add_foreign_keys() now accept lists of column names to create compound foreign keys: table.add_foreign_key( ["campus_name", "dept_code"], "departments" ) Omitting the other columns uses the compound primary key of the other table. Duplicate detection compares the full column lists. db.index_foreign_keys() now creates a single composite index across the columns of a compound foreign key, rather than an index per column. Co-Authored-By: Claude Fable 5 --- docs/python-api.rst | 12 ++++ sqlite_utils/db.py | 138 ++++++++++++++++++++++++------------- tests/test_foreign_keys.py | 74 ++++++++++++++++++++ 3 files changed, 177 insertions(+), 47 deletions(-) diff --git a/docs/python-api.rst b/docs/python-api.rst index 465df79..882f952 100644 --- a/docs/python-api.rst +++ b/docs/python-api.rst @@ -1562,6 +1562,16 @@ To ignore the case where the key already exists, use ``ignore=True``: db.table("books").add_foreign_key("author_id", "authors", "id", ignore=True) +To add a compound foreign key, pass lists of columns: + +.. code-block:: python + + db.table("courses").add_foreign_key( + ["campus_name", "dept_code"], "departments", ["campus_name", "dept_code"] + ) + +As with single columns, omitting the other columns will use the compound primary key of the other table. ``other_table`` must always be specified for a compound foreign key. + .. _python_api_add_foreign_keys: Adding multiple foreign key constraints at once @@ -1591,6 +1601,8 @@ If you want to ensure that every foreign key column in your database has a corre db.index_foreign_keys() +Compound foreign keys get a single composite index across their columns. + .. _python_api_drop: Dropping a table or view diff --git a/sqlite_utils/db.py b/sqlite_utils/db.py index c37f750..3c78e3b 100644 --- a/sqlite_utils/db.py +++ b/sqlite_utils/db.py @@ -1590,13 +1590,14 @@ class Database: return candidates def add_foreign_keys( - self, foreign_keys: Iterable[Tuple[str, str, str, str]] + self, foreign_keys: Iterable[Union[ForeignKey, Tuple[str, Any, str, Any]]] ) -> None: """ See :ref:`python_api_add_foreign_keys`. :param foreign_keys: A list of ``(table, column, other_table, other_column)`` - tuples + tuples - for compound foreign keys, ``column`` and ``other_column`` can + be lists of column names """ # foreign_keys is a list of explicit 4-tuples if not all( @@ -1609,44 +1610,64 @@ class Database: "(table, column, other_table, other_column)" ) - foreign_keys_to_create = [] + foreign_keys_to_create: List[Tuple[str, Any, str, Any]] = [] # Verify that all tables and columns exist for fk in foreign_keys: if isinstance(fk, ForeignKey): - table, column, other_table, other_column = ( + table, columns, other_table, other_columns = ( fk.table, - fk.column, + fk.columns, fk.other_table, - fk.other_column, + fk.other_columns, ) else: - table, column, other_table, other_column = fk + table, column_or_columns, other_table, other_column_or_columns = fk + # Compound foreign keys use lists of columns + columns = ( + [column_or_columns] + if isinstance(column_or_columns, str) + else list(column_or_columns) + ) + other_columns = ( + [other_column_or_columns] + if isinstance(other_column_or_columns, str) + else list(other_column_or_columns) + ) if not self.table(table).exists(): raise AlterError("No such table: {}".format(table)) table_obj = self.table(table) - if column not in table_obj.columns_dict: - raise AlterError("No such column: {} in {}".format(column, table)) + for column in columns: + if column not in table_obj.columns_dict: + raise AlterError("No such column: {} in {}".format(column, table)) if not self[other_table].exists(): raise AlterError("No such other_table: {}".format(other_table)) - if ( - other_column != "rowid" - and other_column not in self[other_table].columns_dict - ): - raise AlterError( - "No such other_column: {} in {}".format(other_column, other_table) - ) + for other_column in other_columns: + if ( + other_column != "rowid" + and other_column not in self[other_table].columns_dict + ): + raise AlterError( + "No such other_column: {} in {}".format( + other_column, other_table + ) + ) # We will silently skip foreign keys that exist already if not any( fk for fk in table_obj.foreign_keys - if fk.column == column + if fk.columns == columns and fk.other_table == other_table - and fk.other_column == other_column + and fk.other_columns == other_columns ): - foreign_keys_to_create.append( - (table, column, other_table, other_column) - ) + if len(columns) == 1: + foreign_keys_to_create.append( + (table, columns[0], other_table, other_columns[0]) + ) + else: + foreign_keys_to_create.append( + (table, columns, other_table, other_columns) + ) # Group them by table by_table: Dict[str, List] = {} @@ -1663,15 +1684,12 @@ class Database: "Create indexes for every foreign key column on every table in the database." for table_name in self.table_names(): table = self.table(table_name) - existing_indexes = { - i.columns[0] for i in table.indexes if len(i.columns) == 1 - } + existing_indexes = {tuple(i.columns) for i in table.indexes} for fk in table.foreign_keys: - # Compound foreign keys expose their columns via fk.columns; - # single-column keys yield a one-item list - for column in fk.columns: - if column not in existing_indexes: - table.create_index([column], find_unique_name=True) + # A compound foreign key gets a single composite index + if tuple(fk.columns) not in existing_indexes: + table.create_index(fk.columns, find_unique_name=True) + existing_indexes.add(tuple(fk.columns)) def vacuum(self) -> None: "Run a SQLite ``VACUUM`` against the database." @@ -2862,52 +2880,78 @@ class Table(Queryable): def add_foreign_key( self, - column: str, + column: Union[str, List[str]], other_table: Optional[str] = None, - other_column: Optional[str] = None, + other_column: Optional[Union[str, List[str]]] = None, ignore: bool = False, ): """ Alter the schema to mark the specified column as a foreign key to another table. - :param column: The column to mark as a foreign key. + :param column: The column to mark as a foreign key - use a list of columns + for a compound foreign key. :param other_table: The table it refers to - if omitted, will be guessed based on the column name. :param other_column: The column on the other table it - if omitted, will be guessed. + Use a list of columns for a compound foreign key. :param ignore: Set this to ``True`` to ignore an existing foreign key - otherwise a ``AlterError`` will be raised. """ - # Ensure column exists - if column not in self.columns_dict: - raise AlterError("No such column: {}".format(column)) + columns = [column] if isinstance(column, str) else list(column) + # Ensure columns exist + for col in columns: + if col not in self.columns_dict: + raise AlterError("No such column: {}".format(col)) # If other_table is not specified, attempt to guess it from the column if other_table is None: - other_table = self.guess_foreign_table(column) + if len(columns) > 1: + raise ValueError( + "other_table must be specified for a compound foreign key" + ) + other_table = self.guess_foreign_table(columns[0]) # If other_column is not specified, detect the primary key on other_table if other_column is None: - other_column = self.guess_foreign_column(other_table) + if len(columns) > 1: + other_columns = self.db.table(other_table).pks + else: + other_columns = [self.guess_foreign_column(other_table)] + elif isinstance(other_column, str): + other_columns = [other_column] + else: + other_columns = list(other_column) + if len(columns) != len(other_columns): + raise ValueError( + "Compound foreign key must have the same number of columns " + "on both sides" + ) - # Soundness check that the other column exists - if ( - not [c for c in self.db[other_table].columns if c.name == other_column] - and other_column != "rowid" - ): - raise AlterError("No such column: {}.{}".format(other_table, other_column)) + # Soundness check that the other columns exist + for other_col in other_columns: + if ( + not [c for c in self.db[other_table].columns if c.name == other_col] + and other_col != "rowid" + ): + raise AlterError("No such column: {}.{}".format(other_table, other_col)) # Check we do not already have an existing foreign key if any( fk for fk in self.foreign_keys - if fk.column == column + if fk.columns == columns and fk.other_table == other_table - and fk.other_column == other_column + and fk.other_columns == other_columns ): if ignore: return self else: raise AlterError( "Foreign key already exists for {} => {}.{}".format( - column, other_table, other_column + ", ".join(columns), other_table, ", ".join(other_columns) ) ) - self.db.add_foreign_keys([(self.name, column, other_table, other_column)]) + if len(columns) == 1: + self.db.add_foreign_keys( + [(self.name, columns[0], other_table, other_columns[0])] + ) + else: + self.db.add_foreign_keys([(self.name, columns, other_table, other_columns)]) return self def enable_counts(self) -> None: diff --git a/tests/test_foreign_keys.py b/tests/test_foreign_keys.py index d6a7071..1fabce7 100644 --- a/tests/test_foreign_keys.py +++ b/tests/test_foreign_keys.py @@ -250,3 +250,77 @@ def test_transform_drop_compound_foreign_key(compound_db, drop_foreign_keys): assert {"campus_name", "dept_code"} <= set( compound_db["courses"].columns_dict.keys() ) + + +@pytest.fixture +def courses_db(departments_db): + departments_db.create_table( + "courses", + {"course_code": str, "campus_name": str, "dept_code": str}, + pk="course_code", + ) + return departments_db + + +def test_add_compound_foreign_key(courses_db): + t = courses_db["courses"].add_foreign_key( + ["campus_name", "dept_code"], "departments", ["campus_name", "dept_code"] + ) + # Returns self + assert t.name == "courses" + fks = courses_db["courses"].foreign_keys + assert len(fks) == 1 + fk = fks[0] + assert fk.is_compound is True + assert fk.columns == ["campus_name", "dept_code"] + assert fk.other_table == "departments" + assert fk.other_columns == ["campus_name", "dept_code"] + + +def test_add_compound_foreign_key_guesses_other_columns(courses_db): + courses_db["courses"].add_foreign_key(["campus_name", "dept_code"], "departments") + fk = courses_db["courses"].foreign_keys[0] + assert fk.other_columns == ["campus_name", "dept_code"] + + +def test_add_compound_foreign_key_error_if_already_exists(courses_db): + courses_db["courses"].add_foreign_key(["campus_name", "dept_code"], "departments") + with pytest.raises(AlterError) as ex: + courses_db["courses"].add_foreign_key( + ["campus_name", "dept_code"], "departments" + ) + assert "already exists" in ex.value.args[0] + # ignore=True should not raise + courses_db["courses"].add_foreign_key( + ["campus_name", "dept_code"], "departments", ignore=True + ) + + +def test_add_compound_foreign_key_error_if_column_missing(courses_db): + with pytest.raises(AlterError): + courses_db["courses"].add_foreign_key(["campus_name", "nope"], "departments") + + +def test_db_add_foreign_keys_compound(courses_db): + courses_db.add_foreign_keys( + [ + ( + "courses", + ["campus_name", "dept_code"], + "departments", + ["campus_name", "dept_code"], + ) + ] + ) + fk = courses_db["courses"].foreign_keys[0] + assert fk.is_compound is True + assert fk.columns == ["campus_name", "dept_code"] + + +def test_index_foreign_keys_compound_creates_composite_index(compound_db): + compound_db.index_foreign_keys() + index_columns = [i.columns for i in compound_db["courses"].indexes] + assert ["campus_name", "dept_code"] in index_columns + # No separate single-column indexes for the members + assert ["campus_name"] not in index_columns + assert ["dept_code"] not in index_columns From be27a96484b69fcc062b54c384e8713a9d500d9a Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Sun, 5 Jul 2026 14:40:57 -0700 Subject: [PATCH 12/86] Capture and preserve foreign key ON DELETE/ON UPDATE actions ForeignKey gains on_delete and on_update fields (default "NO ACTION"), populated from PRAGMA foreign_key_list. create_table_sql() renders the corresponding clauses for both inline and compound table-level foreign keys, which means table.transform() now preserves actions such as ON DELETE CASCADE - previously they were silently stripped whenever a table was transformed. Co-Authored-By: Claude Fable 5 --- docs/python-api.rst | 23 +++++++++++ sqlite_utils/db.py | 42 ++++++++++++++----- tests/test_foreign_keys.py | 84 ++++++++++++++++++++++++++++++++++++++ 3 files changed, 139 insertions(+), 10 deletions(-) diff --git a/docs/python-api.rst b/docs/python-api.rst index 882f952..f44a10b 100644 --- a/docs/python-api.rst +++ b/docs/python-api.rst @@ -853,6 +853,25 @@ As with single columns, you can leave off the list of other columns to reference (["campus_name", "dept_code"], "departments") ] +To specify ``ON DELETE`` or ``ON UPDATE`` actions, pass ``ForeignKey`` objects instead: + +.. code-block:: python + + from sqlite_utils.db import ForeignKey + + db.table("books").create({ + "id": int, + "author_id": int, + }, pk="id", foreign_keys=[ + ForeignKey( + table="books", column="author_id", + other_table="authors", other_column="id", + on_delete="CASCADE", + ) + ]) + +Foreign key actions are preserved by :ref:`table.transform() ` - prior to sqlite-utils 4.0 they were silently dropped when a table was transformed. + .. _python_api_table_configuration: Table configuration options @@ -2280,6 +2299,10 @@ Each ``ForeignKey`` has the following attributes: A list of the referenced columns. ``is_compound`` ``True`` if this is a compound (multi-column) foreign key. +``on_delete`` + The ``ON DELETE`` action, e.g. ``"CASCADE"`` - ``"NO ACTION"`` if not set. +``on_update`` + The ``ON UPDATE`` action - ``"NO ACTION"`` if not set. ``ForeignKey`` was a ``namedtuple`` prior to sqlite-utils 4.0. It is now a dataclass and can no longer be unpacked or indexed as a tuple - access its fields by name instead. See :ref:`upgrading_3_to_4` for details. diff --git a/sqlite_utils/db.py b/sqlite_utils/db.py index 3c78e3b..1910b84 100644 --- a/sqlite_utils/db.py +++ b/sqlite_utils/db.py @@ -190,6 +190,8 @@ class ForeignKey: columns: List[str] = field(default_factory=list) other_columns: List[str] = field(default_factory=list) is_compound: bool = False + on_delete: str = "NO ACTION" + on_update: str = "NO ACTION" def __post_init__(self): # Populate columns/other_columns for single-column foreign keys @@ -201,6 +203,16 @@ class ForeignKey: ) +def _fk_actions_sql(fk: ForeignKey) -> str: + "ON UPDATE/ON DELETE clauses for a foreign key, or an empty string." + actions = "" + if fk.on_update and fk.on_update != "NO ACTION": + actions += " ON UPDATE {}".format(fk.on_update) + if fk.on_delete and fk.on_delete != "NO ACTION": + actions += " ON DELETE {}".format(fk.on_delete) + return actions + + Index = namedtuple("Index", ("seq", "name", "unique", "origin", "partial", "columns")) XIndex = namedtuple("XIndex", ("name", "columns")) XIndexColumn = namedtuple( @@ -1345,14 +1357,12 @@ class Database: "DEFAULT {}".format(self.quote_default_value(defaults[column_name])) ) if column_name in foreign_keys_by_column: + fk = foreign_keys_by_column[column_name] column_extras.append( - "REFERENCES {}({})".format( - quote_identifier( - foreign_keys_by_column[column_name].other_table - ), - quote_identifier( - cast(str, foreign_keys_by_column[column_name].other_column) - ), + "REFERENCES {}({}){}".format( + quote_identifier(fk.other_table), + quote_identifier(cast(str, fk.other_column)), + _fk_actions_sql(fk), ) ) column_type_str = COLUMN_TYPE_MAPPING[column_type] @@ -1385,12 +1395,13 @@ class Database: "No such column: {}".format(", ".join(sorted(missing))) ) column_defs.append( - " FOREIGN KEY ({columns}) REFERENCES {other_table}({other_columns})".format( + " FOREIGN KEY ({columns}) REFERENCES {other_table}({other_columns}){actions}".format( columns=", ".join(quote_identifier(c) for c in fk.columns), other_table=quote_identifier(fk.other_table), other_columns=", ".join( quote_identifier(c) for c in fk.other_columns ), + actions=_fk_actions_sql(fk), ) ) columns_sql = ",\n".join(column_defs) @@ -2055,7 +2066,9 @@ class Table(Queryable): ).fetchall(): if row is not None: id, seq, table_name, from_, to_, on_update, on_delete, match = row - by_id.setdefault(id, []).append((seq, table_name, from_, to_)) + by_id.setdefault(id, []).append( + (seq, table_name, from_, to_, on_update, on_delete) + ) fks = [] for id in sorted(by_id): rows = sorted(by_id[id]) # order columns by seq @@ -2072,6 +2085,8 @@ class Table(Queryable): columns=columns, other_columns=other_columns, is_compound=is_compound, + on_update=rows[0][4], + on_delete=rows[0][5], ) ) return fks @@ -2434,9 +2449,16 @@ class Table(Queryable): columns=columns, other_columns=fk.other_columns, is_compound=True, + on_delete=fk.on_delete, + on_update=fk.on_update, ) return ForeignKey( - self.name, columns[0], fk.other_table, fk.other_columns[0] + self.name, + columns[0], + fk.other_table, + fk.other_columns[0], + on_delete=fk.on_delete, + on_update=fk.on_update, ) create_table_foreign_keys = [] diff --git a/tests/test_foreign_keys.py b/tests/test_foreign_keys.py index 1fabce7..afe333f 100644 --- a/tests/test_foreign_keys.py +++ b/tests/test_foreign_keys.py @@ -324,3 +324,87 @@ def test_index_foreign_keys_compound_creates_composite_index(compound_db): # No separate single-column indexes for the members assert ["campus_name"] not in index_columns assert ["dept_code"] not in index_columns + + +def test_foreign_key_captures_on_delete_and_on_update(): + db = Database(memory=True) + db.executescript(""" + CREATE TABLE authors (id INTEGER PRIMARY KEY); + CREATE TABLE books ( + id INTEGER PRIMARY KEY, + author_id INTEGER REFERENCES authors(id) + ON DELETE CASCADE ON UPDATE RESTRICT + ); + """) + fk = db["books"].foreign_keys[0] + assert fk.on_delete == "CASCADE" + assert fk.on_update == "RESTRICT" + + +def test_foreign_key_on_delete_defaults_to_no_action(fresh_db): + fresh_db["authors"].insert({"id": 1}, pk="id") + fresh_db["books"].insert({"id": 1, "author_id": 1}, pk="id") + fresh_db["books"].add_foreign_key("author_id", "authors", "id") + fk = fresh_db["books"].foreign_keys[0] + assert fk.on_delete == "NO ACTION" + assert fk.on_update == "NO ACTION" + + +def test_create_table_foreign_key_with_on_delete(fresh_db): + fresh_db["authors"].insert({"id": 1}, pk="id") + fresh_db.create_table( + "books", + {"id": int, "author_id": int}, + pk="id", + foreign_keys=[ + ForeignKey( + table="books", + column="author_id", + other_table="authors", + other_column="id", + on_delete="CASCADE", + ) + ], + ) + assert "ON DELETE CASCADE" in fresh_db["books"].schema + assert fresh_db["books"].foreign_keys[0].on_delete == "CASCADE" + + +def test_transform_preserves_on_delete_cascade(): + db = Database(memory=True) + db.executescript(""" + CREATE TABLE authors (id INTEGER PRIMARY KEY); + CREATE TABLE books ( + id INTEGER PRIMARY KEY, + title TEXT, + author_id INTEGER REFERENCES authors(id) ON DELETE CASCADE + ); + """) + db["books"].transform(rename={"title": "book_title"}) + fk = db["books"].foreign_keys[0] + assert fk.on_delete == "CASCADE" + assert fk.on_update == "NO ACTION" + assert "ON DELETE CASCADE" in db["books"].schema + + +def test_transform_preserves_compound_foreign_key_on_delete(): + db = Database(memory=True) + db.executescript(""" + CREATE TABLE departments ( + campus_name TEXT NOT NULL, + dept_code TEXT NOT NULL, + PRIMARY KEY (campus_name, dept_code) + ); + CREATE TABLE courses ( + course_code TEXT PRIMARY KEY, + campus_name TEXT NOT NULL, + dept_code TEXT NOT NULL, + FOREIGN KEY (campus_name, dept_code) + REFERENCES departments(campus_name, dept_code) ON DELETE CASCADE + ); + """) + db["courses"].transform(rename={"course_code": "code"}) + fk = db["courses"].foreign_keys[0] + assert fk.is_compound is True + assert fk.on_delete == "CASCADE" + assert "ON DELETE CASCADE" in db["courses"].schema From 8443d7f3ba8a6762a1894afcadb342820130323e Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Sun, 5 Jul 2026 14:42:30 -0700 Subject: [PATCH 13/86] Resolve implicit primary key references in table.foreign_keys For foreign keys declared as "REFERENCES other_table" with no explicit columns, PRAGMA foreign_key_list returns None for the referenced column. The foreign_keys property now resolves these to the other table's primary key columns, so other_column=None unambiguously indicates a compound foreign key. Co-Authored-By: Claude Fable 5 --- sqlite_utils/db.py | 6 ++++++ tests/test_foreign_keys.py | 37 +++++++++++++++++++++++++++++++++++++ 2 files changed, 43 insertions(+) diff --git a/sqlite_utils/db.py b/sqlite_utils/db.py index 1910b84..dbb7445 100644 --- a/sqlite_utils/db.py +++ b/sqlite_utils/db.py @@ -2075,6 +2075,12 @@ class Table(Queryable): other_table = rows[0][1] columns = [row[2] for row in rows] other_columns = [row[3] for row in rows] + if all(c is None for c in other_columns): + # "REFERENCES other_table" with no columns - the pragma + # returns None, meaning the other table's primary key + other_table_pks = self.db.table(other_table).pks + if len(other_table_pks) == len(columns): + other_columns = other_table_pks is_compound = len(rows) > 1 fks.append( ForeignKey( diff --git a/tests/test_foreign_keys.py b/tests/test_foreign_keys.py index afe333f..dc4ea68 100644 --- a/tests/test_foreign_keys.py +++ b/tests/test_foreign_keys.py @@ -408,3 +408,40 @@ def test_transform_preserves_compound_foreign_key_on_delete(): assert fk.is_compound is True assert fk.on_delete == "CASCADE" assert "ON DELETE CASCADE" in db["courses"].schema + + +def test_implicit_primary_key_reference_is_resolved(): + # REFERENCES authors (no column) has "to" of None in the pragma - + # it should be resolved to the primary key of the other table + db = Database(memory=True) + db.executescript(""" + CREATE TABLE authors (author_id INTEGER PRIMARY KEY); + CREATE TABLE books ( + id INTEGER PRIMARY KEY, + author_id INTEGER REFERENCES authors + ); + """) + fk = db["books"].foreign_keys[0] + assert fk.is_compound is False + assert fk.other_column == "author_id" + assert fk.other_columns == ["author_id"] + + +def test_implicit_compound_primary_key_reference_is_resolved(): + db = Database(memory=True) + db.executescript(""" + CREATE TABLE departments ( + campus_name TEXT NOT NULL, + dept_code TEXT NOT NULL, + PRIMARY KEY (campus_name, dept_code) + ); + CREATE TABLE courses ( + course_code TEXT PRIMARY KEY, + campus_name TEXT NOT NULL, + dept_code TEXT NOT NULL, + FOREIGN KEY (campus_name, dept_code) REFERENCES departments + ); + """) + fk = db["courses"].foreign_keys[0] + assert fk.is_compound is True + assert fk.other_columns == ["campus_name", "dept_code"] From 42c1dd0d5feeebf295bc3c64a9a4b71544edcabf Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Sun, 5 Jul 2026 14:45:08 -0700 Subject: [PATCH 14/86] Documentation for compound foreign key support, refs #594 - Changelog entries for all the compound foreign key work - Upgrading guide notes the transform() behavior changes - ForeignKey added to the API reference - Updated .foreign_keys introspection example output Co-Authored-By: Claude Fable 5 --- docs/changelog.rst | 13 +++++++++++++ docs/python-api.rst | 4 ++-- docs/reference.rst | 7 +++++++ docs/upgrading.rst | 4 +++- sqlite_utils/db.py | 3 +++ 5 files changed, 28 insertions(+), 3 deletions(-) diff --git a/docs/changelog.rst b/docs/changelog.rst index 01d59a0..b182d88 100644 --- a/docs/changelog.rst +++ b/docs/changelog.rst @@ -13,6 +13,19 @@ Breaking changes: - ``table.foreign_keys`` now returns ``ForeignKey`` objects that are dataclasses rather than ``namedtuple`` instances, so they can no longer be unpacked or indexed as ``(table, column, other_table, other_column)`` tuples - access their fields by name instead. Compound (multi-column) foreign keys are now represented as a single ``ForeignKey`` with ``is_compound=True`` and populated ``columns``/``other_columns`` lists, where ``column`` and ``other_column`` are ``None``. Previously they were returned as one ``ForeignKey`` per column, misleadingly suggesting several independent foreign keys. See :ref:`upgrading_3_to_4` for details. (:issue:`594`) +Compound foreign key support, everywhere: + +- Tables can now be created with compound foreign keys, by passing lists of column names in ``foreign_keys=``: ``foreign_keys=[(["campus_name", "dept_code"], "departments")]``. The referenced columns default to the compound primary key of the other table. Compound keys are rendered as table-level ``FOREIGN KEY`` constraints in the generated schema. See :ref:`python_api_compound_foreign_keys`. +- ``table.transform()`` now preserves compound foreign keys, applying any column renames to them. Dropping a column that is part of a compound foreign key drops the whole constraint, matching the existing single-column behavior. ``drop_foreign_keys=`` accepts a bare column name - dropping any foreign key that column participates in - or a tuple of columns to target a compound key precisely. +- ``table.add_foreign_key()`` and ``db.add_foreign_keys()`` accept lists of column names to add a compound foreign key to an existing table. +- ``db.index_foreign_keys()`` creates a single composite index for a compound foreign key. + +Other foreign key improvements: + +- ``ForeignKey`` now exposes ``on_delete`` and ``on_update`` fields reflecting the foreign key's ``ON DELETE``/``ON UPDATE`` actions, and ``table.transform()`` preserves those actions. Previously a transform silently stripped clauses such as ``ON DELETE CASCADE`` from the table schema. +- Foreign keys declared as ``REFERENCES other_table`` with no explicit column are now resolved to the other table's primary key by ``table.foreign_keys``, instead of reporting ``other_column=None``. +- Fixed a ``TypeError`` when sorting ``ForeignKey`` objects where some were compound. + .. _v4_0rc2: 4.0rc2 (2026-07-04) diff --git a/docs/python-api.rst b/docs/python-api.rst index f44a10b..9c31841 100644 --- a/docs/python-api.rst +++ b/docs/python-api.rst @@ -2309,8 +2309,8 @@ Each ``ForeignKey`` has the following attributes: :: >>> db.table("Street_Tree_List").foreign_keys - [ForeignKey(table='Street_Tree_List', column='qLegalStatus', other_table='qLegalStatus', other_column='id', columns=['qLegalStatus'], other_columns=['id'], is_compound=False), - ForeignKey(table='Street_Tree_List', column='qCareAssistant', other_table='qCareAssistant', other_column='id', columns=['qCareAssistant'], other_columns=['id'], is_compound=False), + [ForeignKey(table='Street_Tree_List', column='qLegalStatus', other_table='qLegalStatus', other_column='id', columns=['qLegalStatus'], other_columns=['id'], is_compound=False, on_delete='NO ACTION', on_update='NO ACTION'), + ForeignKey(table='Street_Tree_List', column='qCareAssistant', other_table='qCareAssistant', other_column='id', columns=['qCareAssistant'], other_columns=['id'], is_compound=False, on_delete='NO ACTION', on_update='NO ACTION'), ...] Compound foreign keys - defined with ``FOREIGN KEY (col_a, col_b) REFERENCES other(col_a, col_b)`` - are returned as a single ``ForeignKey`` with ``is_compound=True``, ``column`` and ``other_column`` set to ``None``, and the participating columns available in the ``columns`` and ``other_columns`` lists. diff --git a/docs/reference.rst b/docs/reference.rst index 5b5fd25..a9fdf29 100644 --- a/docs/reference.rst +++ b/docs/reference.rst @@ -70,6 +70,13 @@ sqlite_utils.db.ColumnDetails .. autoclass:: sqlite_utils.db.ColumnDetails +.. _reference_db_other_foreign_key: + +sqlite_utils.db.ForeignKey +-------------------------- + +.. autoclass:: sqlite_utils.db.ForeignKey + sqlite_utils.utils ================== diff --git a/docs/upgrading.rst b/docs/upgrading.rst index eb997f6..67a6b5e 100644 --- a/docs/upgrading.rst +++ b/docs/upgrading.rst @@ -79,7 +79,7 @@ Python API changes **View.enable_fts() has been removed.** The ``View`` class previously had an ``enable_fts()`` method that existed only to raise ``NotImplementedError`` - full-text search is not supported for views. Calling it now raises ``AttributeError`` like any other missing method. -**ForeignKey is now a dataclass, not a namedtuple.** The ``ForeignKey`` objects returned by ``table.foreign_keys`` gained three new fields - ``columns``, ``other_columns`` and ``is_compound`` - so that compound (multi-column) foreign keys can be represented as a single object. To make room for those fields cleanly ``ForeignKey`` is now a dataclass rather than a ``namedtuple``, so it can no longer be unpacked or indexed as a tuple. Access its fields by name instead: +**ForeignKey is now a dataclass, not a namedtuple.** The ``ForeignKey`` objects returned by ``table.foreign_keys`` gained new fields - ``columns``, ``other_columns``, ``is_compound``, ``on_delete`` and ``on_update`` - so that compound (multi-column) foreign keys and foreign key actions can be represented. To make room for those fields cleanly ``ForeignKey`` is now a dataclass rather than a ``namedtuple``, so it can no longer be unpacked or indexed as a tuple. Access its fields by name instead: .. code-block:: python @@ -95,6 +95,8 @@ Attempting the old unpacking or ``fk[0]`` indexing now raises ``TypeError``, so Compound foreign keys - previously returned as one ``ForeignKey`` per column, misleadingly suggesting several independent single-column keys - are now returned as a single ``ForeignKey`` with ``is_compound=True``. For these the scalar ``column`` and ``other_column`` fields are ``None``; use the ``columns`` and ``other_columns`` lists instead. Single-column foreign keys are unaffected apart from the class change: ``column``/``other_column`` behave as before and ``columns``/``other_columns`` are single-item lists. +Two related behavior changes to ``table.transform()``: compound foreign keys now survive a transform (previously they were split into separate single-column keys), and ``ON DELETE``/``ON UPDATE`` actions such as ``ON DELETE CASCADE`` are now preserved (previously they were silently stripped from the schema). + **Validation errors raise ValueError.** Invalid arguments to Python API methods - for example ``create_table()`` with no columns, or ``ignore=True`` together with ``replace=True`` - now raise ``ValueError``. They previously raised ``AssertionError`` from bare ``assert`` statements, which were silently skipped under ``python -O``. **Transaction behavior is now well-defined.** 4.0 introduces the :ref:`db.atomic() ` context manager and uses it consistently for every write operation - the full model is described in :ref:`python_api_transactions`. Changes you may notice: diff --git a/sqlite_utils/db.py b/sqlite_utils/db.py index dbb7445..009bafc 100644 --- a/sqlite_utils/db.py +++ b/sqlite_utils/db.py @@ -176,6 +176,9 @@ class ForeignKey: are ``None`` - use ``columns`` and ``other_columns`` instead, and check ``is_compound``. + ``on_delete`` and ``on_update`` hold the foreign key actions, e.g. + ``"CASCADE"`` - ``"NO ACTION"`` if not set. + Prior to sqlite-utils 4.0 this was a ``namedtuple`` and could be unpacked or indexed as ``(table, column, other_table, other_column)``. It is now a dataclass - access its fields by name instead. From c16edb2dc4a96490b69cdce2a5e8287ed3133924 Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Sun, 5 Jul 2026 15:03:23 -0700 Subject: [PATCH 15/86] Better type signatures for compound foreign key columns New ForeignKeyColumns type alias - a column name or a list/tuple of column names - used by add_foreign_key(), add_foreign_keys() (via the ForeignKeyTuple alias, replacing its Any slots) and the ForeignKeyIndicator union, which it also simplifies. Tuples now type-check anywhere lists are accepted. ForeignKey.__post_init__ normalizes tuple columns/other_columns to lists so direct construction with tuples compares equal to introspected foreign keys. Refs https://github.com/simonw/sqlite-utils/pull/770/changes#r3525703477 Co-Authored-By: Claude Fable 5 --- sqlite_utils/db.py | 33 ++++++++++++++++++++------------- tests/test_foreign_keys.py | 16 ++++++++++++++++ 2 files changed, 36 insertions(+), 13 deletions(-) diff --git a/sqlite_utils/db.py b/sqlite_utils/db.py index 009bafc..4e67a3c 100644 --- a/sqlite_utils/db.py +++ b/sqlite_utils/db.py @@ -197,10 +197,15 @@ class ForeignKey: on_update: str = "NO ACTION" def __post_init__(self): - # Populate columns/other_columns for single-column foreign keys - if not self.columns: + # Populate columns/other_columns for single-column foreign keys, + # normalizing any tuples to lists + if self.columns: + self.columns = list(self.columns) + else: self.columns = [self.column] if self.column is not None else [] - if not self.other_columns: + if self.other_columns: + self.other_columns = list(self.other_columns) + else: self.other_columns = ( [self.other_column] if self.other_column is not None else [] ) @@ -228,16 +233,18 @@ class TransformError(Exception): pass +# A single column name, or a list/tuple of columns for a compound foreign key +ForeignKeyColumns = Union[str, List[str], Tuple[str, ...]] + +# (table, column(s), other_table, other_column(s)) +ForeignKeyTuple = Tuple[str, ForeignKeyColumns, str, ForeignKeyColumns] + ForeignKeyIndicator = Union[ str, ForeignKey, - Tuple[str, str], - Tuple[str, str, str], - Tuple[str, str, str, str], - # Compound foreign keys use lists of columns: - Tuple[List[str], str], - Tuple[List[str], str, List[str]], - Tuple[str, List[str], str, List[str]], + Tuple[ForeignKeyColumns, str], + Tuple[ForeignKeyColumns, str, ForeignKeyColumns], + ForeignKeyTuple, ] ForeignKeysType = Union[Iterable[ForeignKeyIndicator], List[ForeignKeyIndicator]] @@ -1604,7 +1611,7 @@ class Database: return candidates def add_foreign_keys( - self, foreign_keys: Iterable[Union[ForeignKey, Tuple[str, Any, str, Any]]] + self, foreign_keys: Iterable[Union[ForeignKey, ForeignKeyTuple]] ) -> None: """ See :ref:`python_api_add_foreign_keys`. @@ -2911,9 +2918,9 @@ class Table(Queryable): def add_foreign_key( self, - column: Union[str, List[str]], + column: ForeignKeyColumns, other_table: Optional[str] = None, - other_column: Optional[Union[str, List[str]]] = None, + other_column: Optional[ForeignKeyColumns] = None, ignore: bool = False, ): """ diff --git a/tests/test_foreign_keys.py b/tests/test_foreign_keys.py index dc4ea68..54039b3 100644 --- a/tests/test_foreign_keys.py +++ b/tests/test_foreign_keys.py @@ -445,3 +445,19 @@ def test_implicit_compound_primary_key_reference_is_resolved(): fk = db["courses"].foreign_keys[0] assert fk.is_compound is True assert fk.other_columns == ["campus_name", "dept_code"] + + +def test_foreign_key_normalizes_tuple_columns_to_lists(): + # Compound columns passed as tuples are normalized to lists, so they + # compare equal to introspected ForeignKeys + fk = ForeignKey( + table="courses", + column=None, + other_table="departments", + other_column=None, + columns=("campus_name", "dept_code"), + other_columns=("campus_name", "dept_code"), + is_compound=True, + ) + assert fk.columns == ["campus_name", "dept_code"] + assert fk.other_columns == ["campus_name", "dept_code"] From d100264e9c11f73c2f86328846a33e500eeb2118 Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Sun, 5 Jul 2026 15:10:21 -0700 Subject: [PATCH 16/86] Normalize compound foreign key columns to tuples, promote tuples in docs ForeignKey.columns and .other_columns are now tuples rather than lists, both when introspected and when constructed directly (lists are normalized in __post_init__). Tuples are now the documented form for specifying compound foreign keys everywhere - in create_table() foreign_keys=, add_foreign_key() and drop_foreign_keys=: foreign_keys=[(("campus_name", "dept_code"), "departments")] Lists continue to work as input but are no longer documented. Co-Authored-By: Claude Fable 5 --- docs/changelog.rst | 8 ++-- docs/python-api.rst | 22 +++++------ docs/upgrading.rst | 2 +- sqlite_utils/db.py | 77 +++++++++++++++++++------------------- tests/test_foreign_keys.py | 73 +++++++++++++++++++----------------- 5 files changed, 92 insertions(+), 90 deletions(-) diff --git a/docs/changelog.rst b/docs/changelog.rst index b182d88..cacd5a0 100644 --- a/docs/changelog.rst +++ b/docs/changelog.rst @@ -11,13 +11,13 @@ Unreleased Breaking changes: -- ``table.foreign_keys`` now returns ``ForeignKey`` objects that are dataclasses rather than ``namedtuple`` instances, so they can no longer be unpacked or indexed as ``(table, column, other_table, other_column)`` tuples - access their fields by name instead. Compound (multi-column) foreign keys are now represented as a single ``ForeignKey`` with ``is_compound=True`` and populated ``columns``/``other_columns`` lists, where ``column`` and ``other_column`` are ``None``. Previously they were returned as one ``ForeignKey`` per column, misleadingly suggesting several independent foreign keys. See :ref:`upgrading_3_to_4` for details. (:issue:`594`) +- ``table.foreign_keys`` now returns ``ForeignKey`` objects that are dataclasses rather than ``namedtuple`` instances, so they can no longer be unpacked or indexed as ``(table, column, other_table, other_column)`` tuples - access their fields by name instead. Compound (multi-column) foreign keys are now represented as a single ``ForeignKey`` with ``is_compound=True`` and populated ``columns``/``other_columns`` tuples, where ``column`` and ``other_column`` are ``None``. Previously they were returned as one ``ForeignKey`` per column, misleadingly suggesting several independent foreign keys. See :ref:`upgrading_3_to_4` for details. (:issue:`594`) -Compound foreign key support, everywhere: +Compound foreign key support: -- Tables can now be created with compound foreign keys, by passing lists of column names in ``foreign_keys=``: ``foreign_keys=[(["campus_name", "dept_code"], "departments")]``. The referenced columns default to the compound primary key of the other table. Compound keys are rendered as table-level ``FOREIGN KEY`` constraints in the generated schema. See :ref:`python_api_compound_foreign_keys`. +- Tables can now be created with compound foreign keys, by passing tuples of column names in ``foreign_keys=``: ``foreign_keys=[(("campus_name", "dept_code"), "departments")]``. The referenced columns default to the compound primary key of the other table. Compound keys are rendered as table-level ``FOREIGN KEY`` constraints in the generated schema. See :ref:`python_api_compound_foreign_keys`. - ``table.transform()`` now preserves compound foreign keys, applying any column renames to them. Dropping a column that is part of a compound foreign key drops the whole constraint, matching the existing single-column behavior. ``drop_foreign_keys=`` accepts a bare column name - dropping any foreign key that column participates in - or a tuple of columns to target a compound key precisely. -- ``table.add_foreign_key()`` and ``db.add_foreign_keys()`` accept lists of column names to add a compound foreign key to an existing table. +- ``table.add_foreign_key()`` and ``db.add_foreign_keys()`` accept tuples of column names to add a compound foreign key to an existing table. - ``db.index_foreign_keys()`` creates a single composite index for a compound foreign key. Other foreign key improvements: diff --git a/docs/python-api.rst b/docs/python-api.rst index 9c31841..9ff89eb 100644 --- a/docs/python-api.rst +++ b/docs/python-api.rst @@ -822,7 +822,7 @@ You can leave off the third item in the tuple to have the referenced column auto Compound foreign keys ~~~~~~~~~~~~~~~~~~~~~ -To create a compound (multi-column) foreign key, use lists of column names in place of the single column names: +To create a compound (multi-column) foreign key, use tuples of column names in place of the single column names: .. code-block:: python @@ -831,7 +831,7 @@ To create a compound (multi-column) foreign key, use lists of column names in pl "campus_name": str, "dept_code": str, }, pk="course_code", foreign_keys=[ - (["campus_name", "dept_code"], "departments", ["campus_name", "dept_code"]) + (("campus_name", "dept_code"), "departments", ("campus_name", "dept_code")) ]) This creates a table-level constraint: @@ -845,12 +845,12 @@ This creates a table-level constraint: FOREIGN KEY ("campus_name", "dept_code") REFERENCES "departments"("campus_name", "dept_code") ) -As with single columns, you can leave off the list of other columns to reference the compound primary key of the other table: +As with single columns, you can leave off the tuple of other columns to reference the compound primary key of the other table: .. code-block:: python foreign_keys=[ - (["campus_name", "dept_code"], "departments") + (("campus_name", "dept_code"), "departments") ] To specify ``ON DELETE`` or ``ON UPDATE`` actions, pass ``ForeignKey`` objects instead: @@ -1581,12 +1581,12 @@ To ignore the case where the key already exists, use ``ignore=True``: db.table("books").add_foreign_key("author_id", "authors", "id", ignore=True) -To add a compound foreign key, pass lists of columns: +To add a compound foreign key, pass tuples of columns: .. code-block:: python db.table("courses").add_foreign_key( - ["campus_name", "dept_code"], "departments", ["campus_name", "dept_code"] + ("campus_name", "dept_code"), "departments", ("campus_name", "dept_code") ) As with single columns, omitting the other columns will use the compound primary key of the other table. ``other_table`` must always be specified for a compound foreign key. @@ -2294,9 +2294,9 @@ Each ``ForeignKey`` has the following attributes: ``other_column`` The referenced column, or ``None`` for a compound foreign key. ``columns`` - A list of the columns on this table, always populated (a single-item list for single-column foreign keys). + A tuple of the columns on this table, always populated (a one-item tuple for single-column foreign keys). ``other_columns`` - A list of the referenced columns. + A tuple of the referenced columns. ``is_compound`` ``True`` if this is a compound (multi-column) foreign key. ``on_delete`` @@ -2309,11 +2309,11 @@ Each ``ForeignKey`` has the following attributes: :: >>> db.table("Street_Tree_List").foreign_keys - [ForeignKey(table='Street_Tree_List', column='qLegalStatus', other_table='qLegalStatus', other_column='id', columns=['qLegalStatus'], other_columns=['id'], is_compound=False, on_delete='NO ACTION', on_update='NO ACTION'), - ForeignKey(table='Street_Tree_List', column='qCareAssistant', other_table='qCareAssistant', other_column='id', columns=['qCareAssistant'], other_columns=['id'], is_compound=False, on_delete='NO ACTION', on_update='NO ACTION'), + [ForeignKey(table='Street_Tree_List', column='qLegalStatus', other_table='qLegalStatus', other_column='id', columns=('qLegalStatus',), other_columns=('id',), is_compound=False, on_delete='NO ACTION', on_update='NO ACTION'), + ForeignKey(table='Street_Tree_List', column='qCareAssistant', other_table='qCareAssistant', other_column='id', columns=('qCareAssistant',), other_columns=('id',), is_compound=False, on_delete='NO ACTION', on_update='NO ACTION'), ...] -Compound foreign keys - defined with ``FOREIGN KEY (col_a, col_b) REFERENCES other(col_a, col_b)`` - are returned as a single ``ForeignKey`` with ``is_compound=True``, ``column`` and ``other_column`` set to ``None``, and the participating columns available in the ``columns`` and ``other_columns`` lists. +Compound foreign keys - defined with ``FOREIGN KEY (col_a, col_b) REFERENCES other(col_a, col_b)`` - are returned as a single ``ForeignKey`` with ``is_compound=True``, ``column`` and ``other_column`` set to ``None``, and the participating columns available in the ``columns`` and ``other_columns`` tuples. .. _python_api_introspection_schema: diff --git a/docs/upgrading.rst b/docs/upgrading.rst index 67a6b5e..e7fd273 100644 --- a/docs/upgrading.rst +++ b/docs/upgrading.rst @@ -93,7 +93,7 @@ Python API changes Attempting the old unpacking or ``fk[0]`` indexing now raises ``TypeError``, so any code using those patterns will fail loudly rather than silently misbehave. -Compound foreign keys - previously returned as one ``ForeignKey`` per column, misleadingly suggesting several independent single-column keys - are now returned as a single ``ForeignKey`` with ``is_compound=True``. For these the scalar ``column`` and ``other_column`` fields are ``None``; use the ``columns`` and ``other_columns`` lists instead. Single-column foreign keys are unaffected apart from the class change: ``column``/``other_column`` behave as before and ``columns``/``other_columns`` are single-item lists. +Compound foreign keys - previously returned as one ``ForeignKey`` per column, misleadingly suggesting several independent single-column keys - are now returned as a single ``ForeignKey`` with ``is_compound=True``. For these the scalar ``column`` and ``other_column`` fields are ``None``; use the ``columns`` and ``other_columns`` tuples instead. Single-column foreign keys are unaffected apart from the class change: ``column``/``other_column`` behave as before and ``columns``/``other_columns`` are one-item tuples. Two related behavior changes to ``table.transform()``: compound foreign keys now survive a transform (previously they were split into separate single-column keys), and ``ON DELETE``/``ON UPDATE`` actions such as ``ON DELETE CASCADE`` are now preserved (previously they were silently stripped from the schema). diff --git a/sqlite_utils/db.py b/sqlite_utils/db.py index 4e67a3c..a739486 100644 --- a/sqlite_utils/db.py +++ b/sqlite_utils/db.py @@ -170,7 +170,7 @@ class ForeignKey: A foreign key defined on a table. For single-column foreign keys ``column`` and ``other_column`` hold the - column names, and ``columns``/``other_columns`` are single-item lists. + column names, and ``columns``/``other_columns`` are one-item tuples. For compound (multi-column) foreign keys ``column`` and ``other_column`` are ``None`` - use ``columns`` and ``other_columns`` instead, and check @@ -190,24 +190,24 @@ class ForeignKey: column: Optional[str] = field(compare=False) other_table: str other_column: Optional[str] = field(compare=False) - columns: List[str] = field(default_factory=list) - other_columns: List[str] = field(default_factory=list) + columns: Tuple[str, ...] = () + other_columns: Tuple[str, ...] = () is_compound: bool = False on_delete: str = "NO ACTION" on_update: str = "NO ACTION" def __post_init__(self): # Populate columns/other_columns for single-column foreign keys, - # normalizing any tuples to lists + # normalizing any lists to tuples if self.columns: - self.columns = list(self.columns) + self.columns = tuple(self.columns) else: - self.columns = [self.column] if self.column is not None else [] + self.columns = (self.column,) if self.column is not None else () if self.other_columns: - self.other_columns = list(self.other_columns) + self.other_columns = tuple(self.other_columns) else: self.other_columns = ( - [self.other_column] if self.other_column is not None else [] + (self.other_column,) if self.other_column is not None else () ) @@ -233,8 +233,8 @@ class TransformError(Exception): pass -# A single column name, or a list/tuple of columns for a compound foreign key -ForeignKeyColumns = Union[str, List[str], Tuple[str, ...]] +# A single column name, or a tuple of columns for a compound foreign key +ForeignKeyColumns = Union[str, Tuple[str, ...], List[str]] # (table, column(s), other_table, other_column(s)) ForeignKeyTuple = Tuple[str, ForeignKeyColumns, str, ForeignKeyColumns] @@ -1171,9 +1171,9 @@ class Database: string, a ForeignKey() object, a tuple of (column, other_table), or a tuple of (column, other_table, other_column), or a tuple of (table, column, other_table, other_column). For compound foreign - keys the column elements can be lists of column names, e.g. - (["campus_name", "dept_code"], "departments") or - (["campus_name", "dept_code"], "departments", ["campus_name", "dept_code"]) + keys the column elements can be tuples of column names, e.g. + (("campus_name", "dept_code"), "departments") or + (("campus_name", "dept_code"), "departments", ("campus_name", "dept_code")) """ table = self.table(name) if all(isinstance(fk, ForeignKey) for fk in foreign_keys): @@ -1207,18 +1207,17 @@ class Database: other_table = tuple_or_list[1] if isinstance(column_or_columns, (list, tuple)): # Compound foreign key - columns = list(column_or_columns) + columns = tuple(column_or_columns) if len(tuple_or_list) == 3: - other_columns = tuple_or_list[2] - if not isinstance(other_columns, (list, tuple)): + if not isinstance(tuple_or_list[2], (list, tuple)): raise ValueError( - "Compound foreign key {} should reference a list " + "Compound foreign key {} should reference a tuple " "of other columns".format(tuple(tuple_or_list)) ) - other_columns = list(other_columns) + other_columns = tuple(tuple_or_list[2]) else: # Guess the compound primary key of the other table - other_columns = self.table(other_table).pks + other_columns = tuple(self.table(other_table).pks) if len(columns) != len(other_columns): raise ValueError( "Compound foreign key {} should have the same number " @@ -1618,7 +1617,7 @@ class Database: :param foreign_keys: A list of ``(table, column, other_table, other_column)`` tuples - for compound foreign keys, ``column`` and ``other_column`` can - be lists of column names + be tuples of column names """ # foreign_keys is a list of explicit 4-tuples if not all( @@ -1644,16 +1643,16 @@ class Database: ) else: table, column_or_columns, other_table, other_column_or_columns = fk - # Compound foreign keys use lists of columns + # Compound foreign keys use tuples of columns columns = ( - [column_or_columns] + (column_or_columns,) if isinstance(column_or_columns, str) - else list(column_or_columns) + else tuple(column_or_columns) ) other_columns = ( - [other_column_or_columns] + (other_column_or_columns,) if isinstance(other_column_or_columns, str) - else list(other_column_or_columns) + else tuple(other_column_or_columns) ) if not self.table(table).exists(): raise AlterError("No such table: {}".format(table)) @@ -1708,9 +1707,9 @@ class Database: existing_indexes = {tuple(i.columns) for i in table.indexes} for fk in table.foreign_keys: # A compound foreign key gets a single composite index - if tuple(fk.columns) not in existing_indexes: + if fk.columns not in existing_indexes: table.create_index(fk.columns, find_unique_name=True) - existing_indexes.add(tuple(fk.columns)) + existing_indexes.add(fk.columns) def vacuum(self) -> None: "Run a SQLite ``VACUUM`` against the database." @@ -2083,12 +2082,12 @@ class Table(Queryable): for id in sorted(by_id): rows = sorted(by_id[id]) # order columns by seq other_table = rows[0][1] - columns = [row[2] for row in rows] - other_columns = [row[3] for row in rows] + columns = tuple(row[2] for row in rows) + other_columns = tuple(row[3] for row in rows) if all(c is None for c in other_columns): # "REFERENCES other_table" with no columns - the pragma # returns None, meaning the other table's primary key - other_table_pks = self.db.table(other_table).pks + other_table_pks = tuple(self.db.table(other_table).pks) if len(other_table_pks) == len(columns): other_columns = other_table_pks is_compound = len(rows) > 1 @@ -2448,14 +2447,14 @@ class Table(Queryable): # A column name matches any foreign key it participates in if spec in fk.columns: return True - elif list(spec) == fk.columns: + elif tuple(spec) == fk.columns: # A tuple/list must match a compound key's columns exactly return True # Dropping any of a foreign key's columns drops the whole key return any(column in drop for column in fk.columns) def fk_with_renamed_columns(fk: ForeignKey) -> ForeignKey: - columns = [rename.get(column) or column for column in fk.columns] + columns = tuple(rename.get(column) or column for column in fk.columns) if fk.is_compound: return ForeignKey( self.name, @@ -2926,14 +2925,14 @@ class Table(Queryable): """ Alter the schema to mark the specified column as a foreign key to another table. - :param column: The column to mark as a foreign key - use a list of columns + :param column: The column to mark as a foreign key - use a tuple of columns for a compound foreign key. :param other_table: The table it refers to - if omitted, will be guessed based on the column name. :param other_column: The column on the other table it - if omitted, will be guessed. - Use a list of columns for a compound foreign key. + Use a tuple of columns for a compound foreign key. :param ignore: Set this to ``True`` to ignore an existing foreign key - otherwise a ``AlterError`` will be raised. """ - columns = [column] if isinstance(column, str) else list(column) + columns = (column,) if isinstance(column, str) else tuple(column) # Ensure columns exist for col in columns: if col not in self.columns_dict: @@ -2948,13 +2947,13 @@ class Table(Queryable): # If other_column is not specified, detect the primary key on other_table if other_column is None: if len(columns) > 1: - other_columns = self.db.table(other_table).pks + other_columns = tuple(self.db.table(other_table).pks) else: - other_columns = [self.guess_foreign_column(other_table)] + other_columns = (self.guess_foreign_column(other_table),) elif isinstance(other_column, str): - other_columns = [other_column] + other_columns = (other_column,) else: - other_columns = list(other_column) + other_columns = tuple(other_column) if len(columns) != len(other_columns): raise ValueError( "Compound foreign key must have the same number of columns " diff --git a/tests/test_foreign_keys.py b/tests/test_foreign_keys.py index 54039b3..d1bc498 100644 --- a/tests/test_foreign_keys.py +++ b/tests/test_foreign_keys.py @@ -36,8 +36,8 @@ def test_compound_foreign_key(compound_db): assert fk.is_compound is True assert fk.table == "courses" assert fk.other_table == "departments" - assert fk.columns == ["campus_name", "dept_code"] - assert fk.other_columns == ["campus_name", "dept_code"] + assert fk.columns == ("campus_name", "dept_code") + assert fk.other_columns == ("campus_name", "dept_code") # Scalar column/other_column can't sensibly hold a compound key assert fk.column is None assert fk.other_column is None @@ -51,8 +51,8 @@ def test_single_foreign_key_gets_columns_fields(fresh_db): assert fk.is_compound is False assert fk.column == "author_id" assert fk.other_column == "id" - assert fk.columns == ["author_id"] - assert fk.other_columns == ["id"] + assert fk.columns == ("author_id",) + assert fk.other_columns == ("id",) def test_foreign_key_no_longer_unpacks_as_tuple(fresh_db): @@ -142,14 +142,16 @@ EXPECTED_COURSES_SCHEMA = ( column=None, other_table="departments", other_column=None, - columns=["campus_name", "dept_code"], - other_columns=["campus_name", "dept_code"], + columns=("campus_name", "dept_code"), + other_columns=("campus_name", "dept_code"), is_compound=True, ) ], - [(["campus_name", "dept_code"], "departments", ["campus_name", "dept_code"])], + [(("campus_name", "dept_code"), "departments", ("campus_name", "dept_code"))], # Two-item form guesses the other table's primary key: - [(["campus_name", "dept_code"], "departments")], + [(("campus_name", "dept_code"), "departments")], + # Lists work too, though tuples are the documented form: + [(["campus_name", "dept_code"], "departments", ["campus_name", "dept_code"])], ), ) def test_create_table_with_compound_foreign_key(departments_db, foreign_keys): @@ -164,9 +166,9 @@ def test_create_table_with_compound_foreign_key(departments_db, foreign_keys): assert len(fks) == 1 fk = fks[0] assert fk.is_compound is True - assert fk.columns == ["campus_name", "dept_code"] + assert fk.columns == ("campus_name", "dept_code") assert fk.other_table == "departments" - assert fk.other_columns == ["campus_name", "dept_code"] + assert fk.other_columns == ("campus_name", "dept_code") def test_create_table_compound_foreign_key_enforced(departments_db): @@ -175,7 +177,7 @@ def test_create_table_compound_foreign_key_enforced(departments_db): "courses", {"course_code": str, "campus_name": str, "dept_code": str}, pk="course_code", - foreign_keys=[(["campus_name", "dept_code"], "departments")], + foreign_keys=[(("campus_name", "dept_code"), "departments")], ) departments_db["departments"].insert( {"campus_name": "Berkeley", "dept_code": "CS", "dept_name": "Computer Science"} @@ -199,7 +201,7 @@ def test_create_table_compound_foreign_key_missing_other_column(departments_db): {"course_code": str, "campus_name": str, "dept_code": str}, pk="course_code", foreign_keys=[ - (["campus_name", "dept_code"], "departments", ["campus_name", "nope"]) + (("campus_name", "dept_code"), "departments", ("campus_name", "nope")) ], ) @@ -210,9 +212,9 @@ def test_transform_preserves_compound_foreign_key(compound_db): assert len(fks) == 1 fk = fks[0] assert fk.is_compound is True - assert fk.columns == ["campus_name", "dept_code"] + assert fk.columns == ("campus_name", "dept_code") assert fk.other_table == "departments" - assert fk.other_columns == ["campus_name", "dept_code"] + assert fk.other_columns == ("campus_name", "dept_code") def test_transform_rename_member_column_updates_compound_foreign_key(compound_db): @@ -221,9 +223,9 @@ def test_transform_rename_member_column_updates_compound_foreign_key(compound_db assert len(fks) == 1 fk = fks[0] assert fk.is_compound is True - assert fk.columns == ["campus", "dept_code"] + assert fk.columns == ("campus", "dept_code") # Referenced columns in the other table are unchanged - assert fk.other_columns == ["campus_name", "dept_code"] + assert fk.other_columns == ("campus_name", "dept_code") def test_transform_drop_member_column_drops_compound_foreign_key(compound_db): @@ -264,7 +266,7 @@ def courses_db(departments_db): def test_add_compound_foreign_key(courses_db): t = courses_db["courses"].add_foreign_key( - ["campus_name", "dept_code"], "departments", ["campus_name", "dept_code"] + ("campus_name", "dept_code"), "departments", ("campus_name", "dept_code") ) # Returns self assert t.name == "courses" @@ -272,33 +274,34 @@ def test_add_compound_foreign_key(courses_db): assert len(fks) == 1 fk = fks[0] assert fk.is_compound is True - assert fk.columns == ["campus_name", "dept_code"] + assert fk.columns == ("campus_name", "dept_code") assert fk.other_table == "departments" - assert fk.other_columns == ["campus_name", "dept_code"] + assert fk.other_columns == ("campus_name", "dept_code") def test_add_compound_foreign_key_guesses_other_columns(courses_db): + # Lists work here too, though tuples are the documented form courses_db["courses"].add_foreign_key(["campus_name", "dept_code"], "departments") fk = courses_db["courses"].foreign_keys[0] - assert fk.other_columns == ["campus_name", "dept_code"] + assert fk.other_columns == ("campus_name", "dept_code") def test_add_compound_foreign_key_error_if_already_exists(courses_db): - courses_db["courses"].add_foreign_key(["campus_name", "dept_code"], "departments") + courses_db["courses"].add_foreign_key(("campus_name", "dept_code"), "departments") with pytest.raises(AlterError) as ex: courses_db["courses"].add_foreign_key( - ["campus_name", "dept_code"], "departments" + ("campus_name", "dept_code"), "departments" ) assert "already exists" in ex.value.args[0] # ignore=True should not raise courses_db["courses"].add_foreign_key( - ["campus_name", "dept_code"], "departments", ignore=True + ("campus_name", "dept_code"), "departments", ignore=True ) def test_add_compound_foreign_key_error_if_column_missing(courses_db): with pytest.raises(AlterError): - courses_db["courses"].add_foreign_key(["campus_name", "nope"], "departments") + courses_db["courses"].add_foreign_key(("campus_name", "nope"), "departments") def test_db_add_foreign_keys_compound(courses_db): @@ -306,15 +309,15 @@ def test_db_add_foreign_keys_compound(courses_db): [ ( "courses", - ["campus_name", "dept_code"], + ("campus_name", "dept_code"), "departments", - ["campus_name", "dept_code"], + ("campus_name", "dept_code"), ) ] ) fk = courses_db["courses"].foreign_keys[0] assert fk.is_compound is True - assert fk.columns == ["campus_name", "dept_code"] + assert fk.columns == ("campus_name", "dept_code") def test_index_foreign_keys_compound_creates_composite_index(compound_db): @@ -424,7 +427,7 @@ def test_implicit_primary_key_reference_is_resolved(): fk = db["books"].foreign_keys[0] assert fk.is_compound is False assert fk.other_column == "author_id" - assert fk.other_columns == ["author_id"] + assert fk.other_columns == ("author_id",) def test_implicit_compound_primary_key_reference_is_resolved(): @@ -444,20 +447,20 @@ def test_implicit_compound_primary_key_reference_is_resolved(): """) fk = db["courses"].foreign_keys[0] assert fk.is_compound is True - assert fk.other_columns == ["campus_name", "dept_code"] + assert fk.other_columns == ("campus_name", "dept_code") -def test_foreign_key_normalizes_tuple_columns_to_lists(): - # Compound columns passed as tuples are normalized to lists, so they +def test_foreign_key_normalizes_list_columns_to_tuples(): + # Compound columns passed as lists are normalized to tuples, so they # compare equal to introspected ForeignKeys fk = ForeignKey( table="courses", column=None, other_table="departments", other_column=None, - columns=("campus_name", "dept_code"), - other_columns=("campus_name", "dept_code"), + columns=["campus_name", "dept_code"], + other_columns=["campus_name", "dept_code"], is_compound=True, ) - assert fk.columns == ["campus_name", "dept_code"] - assert fk.other_columns == ["campus_name", "dept_code"] + assert fk.columns == ("campus_name", "dept_code") + assert fk.other_columns == ("campus_name", "dept_code") From 8dfbfa80b8282c14ed9732eff04ba07dbd5388e5 Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Sun, 5 Jul 2026 15:36:46 -0700 Subject: [PATCH 17/86] Fix ty errors in transform_sql foreign key closures The fk_should_be_dropped and fk_with_renamed_columns closures captured the drop and rename parameters, but type checkers do not narrow captured variables inside nested functions so ty saw their Optional declared types. Bind the already-normalized values to fresh names for the closures to use. Co-Authored-By: Claude Fable 5 --- sqlite_utils/db.py | 11 +++++++++-- 1 file changed, 9 insertions(+), 2 deletions(-) diff --git a/sqlite_utils/db.py b/sqlite_utils/db.py index a739486..1c06b62 100644 --- a/sqlite_utils/db.py +++ b/sqlite_utils/db.py @@ -2439,6 +2439,11 @@ class Table(Queryable): create_table_foreign_keys.extend(foreign_keys) else: # Construct foreign_keys from current, plus add_foreign_keys, minus drop_foreign_keys + # Bind fresh names here - type checkers do not narrow captured + # variables inside nested functions, so the closures would + # otherwise see the Optional declared types of drop and rename + dropped_columns = drop + renamed_columns = rename def fk_should_be_dropped(fk: ForeignKey) -> bool: if drop_foreign_keys is not None: @@ -2451,10 +2456,12 @@ class Table(Queryable): # A tuple/list must match a compound key's columns exactly return True # Dropping any of a foreign key's columns drops the whole key - return any(column in drop for column in fk.columns) + return any(column in dropped_columns for column in fk.columns) def fk_with_renamed_columns(fk: ForeignKey) -> ForeignKey: - columns = tuple(rename.get(column) or column for column in fk.columns) + columns = tuple( + renamed_columns.get(column) or column for column in fk.columns + ) if fk.is_compound: return ForeignKey( self.name, From 2f599fc7c6aeede4e97bd91487e5de5f5e190dca Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Sun, 5 Jul 2026 15:38:05 -0700 Subject: [PATCH 18/86] Run ty in just lint, matching CI Co-Authored-By: Claude Fable 5 --- Justfile | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/Justfile b/Justfile index f4f0e7f..5caa120 100644 --- a/Justfile +++ b/Justfile @@ -8,11 +8,12 @@ @run *options: uv run -- {{options}} -# Run linters: black, flake8, mypy, cog +# Run linters: black, flake8, mypy, ty, cog @lint: just run black . --check uv run flake8 uv run mypy sqlite_utils tests + uv run ty check sqlite_utils uv run cog --check README.md docs/*.rst uv run --group docs codespell docs/*.rst --ignore-words docs/codespell-ignore-words.txt From 5b61530965df0d56a09e71f29cc970b4ea3e43d4 Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Sun, 5 Jul 2026 15:42:41 -0700 Subject: [PATCH 19/86] db.add_foreign_keys() no longer drops ON DELETE/ON UPDATE actions ForeignKey objects passed to db.add_foreign_keys() were flattened to plain (table, column, other_table, other_column) tuples before being handed to transform(), silently discarding their on_delete and on_update actions. The method now carries ForeignKey objects through to transform() intact - tuple inputs are converted to ForeignKey objects up front, which also simplifies the validation loop. Co-Authored-By: Claude Fable 5 --- sqlite_utils/db.py | 42 ++++++++++++++++++++++---------------- tests/test_foreign_keys.py | 35 +++++++++++++++++++++++++++++++ 2 files changed, 59 insertions(+), 18 deletions(-) diff --git a/sqlite_utils/db.py b/sqlite_utils/db.py index 1c06b62..7379859 100644 --- a/sqlite_utils/db.py +++ b/sqlite_utils/db.py @@ -1630,17 +1630,12 @@ class Database: "(table, column, other_table, other_column)" ) - foreign_keys_to_create: List[Tuple[str, Any, str, Any]] = [] + foreign_keys_to_create: List[ForeignKey] = [] # Verify that all tables and columns exist for fk in foreign_keys: if isinstance(fk, ForeignKey): - table, columns, other_table, other_columns = ( - fk.table, - fk.columns, - fk.other_table, - fk.other_columns, - ) + fk_object = fk else: table, column_or_columns, other_table, other_column_or_columns = fk # Compound foreign keys use tuples of columns @@ -1654,6 +1649,24 @@ class Database: if isinstance(other_column_or_columns, str) else tuple(other_column_or_columns) ) + if len(columns) == 1: + fk_object = ForeignKey( + table, columns[0], other_table, other_columns[0] + ) + else: + fk_object = ForeignKey( + table, + None, + other_table, + None, + columns=columns, + other_columns=other_columns, + is_compound=True, + ) + table = fk_object.table + columns = fk_object.columns + other_table = fk_object.other_table + other_columns = fk_object.other_columns if not self.table(table).exists(): raise AlterError("No such table: {}".format(table)) table_obj = self.table(table) @@ -1680,19 +1693,12 @@ class Database: and fk.other_table == other_table and fk.other_columns == other_columns ): - if len(columns) == 1: - foreign_keys_to_create.append( - (table, columns[0], other_table, other_columns[0]) - ) - else: - foreign_keys_to_create.append( - (table, columns, other_table, other_columns) - ) + foreign_keys_to_create.append(fk_object) # Group them by table - by_table: Dict[str, List] = {} - for fk in foreign_keys_to_create: - by_table.setdefault(fk[0], []).append(fk) + by_table: Dict[str, List[ForeignKey]] = {} + for fk_object in foreign_keys_to_create: + by_table.setdefault(fk_object.table, []).append(fk_object) for table, fks in by_table.items(): self.table(table).transform(add_foreign_keys=fks) diff --git a/tests/test_foreign_keys.py b/tests/test_foreign_keys.py index d1bc498..b22638a 100644 --- a/tests/test_foreign_keys.py +++ b/tests/test_foreign_keys.py @@ -464,3 +464,38 @@ def test_foreign_key_normalizes_list_columns_to_tuples(): ) assert fk.columns == ("campus_name", "dept_code") assert fk.other_columns == ("campus_name", "dept_code") + + +def test_add_foreign_keys_preserves_actions(fresh_db): + # https://github.com/simonw/sqlite-utils/issues/594 review finding: + # ForeignKey objects passed to db.add_foreign_keys() were flattened + # to plain tuples, losing on_delete/on_update + fresh_db["authors"].insert({"id": 1}, pk="id") + fresh_db["books"].insert({"id": 1, "author_id": 1}, pk="id") + fresh_db.add_foreign_keys( + [ForeignKey("books", "author_id", "authors", "id", on_delete="CASCADE")] + ) + fk = fresh_db["books"].foreign_keys[0] + assert fk.on_delete == "CASCADE" + assert "ON DELETE CASCADE" in fresh_db["books"].schema + + +def test_add_foreign_keys_preserves_actions_compound(courses_db): + courses_db.add_foreign_keys( + [ + ForeignKey( + table="courses", + column=None, + other_table="departments", + other_column=None, + columns=("campus_name", "dept_code"), + other_columns=("campus_name", "dept_code"), + is_compound=True, + on_delete="CASCADE", + ) + ] + ) + fk = courses_db["courses"].foreign_keys[0] + assert fk.is_compound is True + assert fk.on_delete == "CASCADE" + assert "ON DELETE CASCADE" in courses_db["courses"].schema From 658185d297a307d0b9d43fd83019a5acab032f18 Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Sun, 5 Jul 2026 15:46:16 -0700 Subject: [PATCH 20/86] Fix compound foreign key test failure against sqlean test_create_table_compound_foreign_key_enforced caught the stdlib sqlite3.IntegrityError, but sqlean's IntegrityError is not a subclass of it. Import sqlite3 from sqlite_utils.utils like the other test modules, so the right exception is used whichever backend is installed. Co-Authored-By: Claude Fable 5 --- tests/test_foreign_keys.py | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) diff --git a/tests/test_foreign_keys.py b/tests/test_foreign_keys.py index b22638a..7fcd788 100644 --- a/tests/test_foreign_keys.py +++ b/tests/test_foreign_keys.py @@ -3,6 +3,7 @@ import pytest from sqlite_utils import Database from sqlite_utils.db import AlterError, ForeignKey +from sqlite_utils.utils import sqlite3 COMPOUND_SCHEMA = """ CREATE TABLE departments ( @@ -185,8 +186,6 @@ def test_create_table_compound_foreign_key_enforced(departments_db): departments_db["courses"].insert( {"course_code": "CS101", "campus_name": "Berkeley", "dept_code": "CS"} ) - import sqlite3 - with pytest.raises(sqlite3.IntegrityError): departments_db.execute( "insert into courses (course_code, campus_name, dept_code) " From 0ec01804055d7a09acc6c4068562c8bbb05dfdd7 Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Sun, 5 Jul 2026 15:49:20 -0700 Subject: [PATCH 21/86] add_foreign_key() on_delete= and on_update= parameters, closes #530 table.add_foreign_key() now accepts on_delete= and on_update= to create foreign keys with ON DELETE/ON UPDATE actions: table.add_foreign_key("author_id", "authors", "id", on_delete="CASCADE") Works for compound foreign keys too. Co-Authored-By: Claude Fable 5 --- docs/changelog.rst | 1 + docs/python-api.rst | 10 ++++++++++ sqlite_utils/db.py | 27 ++++++++++++++++++++++++--- tests/test_foreign_keys.py | 26 ++++++++++++++++++++++++++ 4 files changed, 61 insertions(+), 3 deletions(-) diff --git a/docs/changelog.rst b/docs/changelog.rst index cacd5a0..f6a8ea2 100644 --- a/docs/changelog.rst +++ b/docs/changelog.rst @@ -23,6 +23,7 @@ Compound foreign key support: Other foreign key improvements: - ``ForeignKey`` now exposes ``on_delete`` and ``on_update`` fields reflecting the foreign key's ``ON DELETE``/``ON UPDATE`` actions, and ``table.transform()`` preserves those actions. Previously a transform silently stripped clauses such as ``ON DELETE CASCADE`` from the table schema. +- ``table.add_foreign_key()`` accepts new ``on_delete=`` and ``on_update=`` parameters for creating foreign keys with actions, e.g. ``table.add_foreign_key("author_id", "authors", "id", on_delete="CASCADE")``. (:issue:`530`) - Foreign keys declared as ``REFERENCES other_table`` with no explicit column are now resolved to the other table's primary key by ``table.foreign_keys``, instead of reporting ``other_column=None``. - Fixed a ``TypeError`` when sorting ``ForeignKey`` objects where some were compound. diff --git a/docs/python-api.rst b/docs/python-api.rst index 9ff89eb..a33b256 100644 --- a/docs/python-api.rst +++ b/docs/python-api.rst @@ -1591,6 +1591,16 @@ To add a compound foreign key, pass tuples of columns: As with single columns, omitting the other columns will use the compound primary key of the other table. ``other_table`` must always be specified for a compound foreign key. +Use ``on_delete=`` and ``on_update=`` to specify ``ON DELETE`` and ``ON UPDATE`` actions for the foreign key: + +.. code-block:: python + + db.table("books").add_foreign_key( + "author_id", "authors", "id", on_delete="CASCADE" + ) + +This creates a foreign key with an ``ON DELETE CASCADE`` clause, so deleting an author will also delete their books (provided foreign key enforcement is enabled with ``PRAGMA foreign_keys = ON``). Valid actions are ``"SET NULL"``, ``"SET DEFAULT"``, ``"CASCADE"``, ``"RESTRICT"`` and the default ``"NO ACTION"``. + .. _python_api_add_foreign_keys: Adding multiple foreign key constraints at once diff --git a/sqlite_utils/db.py b/sqlite_utils/db.py index 7379859..3f23c7d 100644 --- a/sqlite_utils/db.py +++ b/sqlite_utils/db.py @@ -2934,6 +2934,8 @@ class Table(Queryable): other_table: Optional[str] = None, other_column: Optional[ForeignKeyColumns] = None, ignore: bool = False, + on_delete: str = "NO ACTION", + on_update: str = "NO ACTION", ): """ Alter the schema to mark the specified column as a foreign key to another table. @@ -2944,6 +2946,9 @@ class Table(Queryable): :param other_column: The column on the other table it - if omitted, will be guessed. Use a tuple of columns for a compound foreign key. :param ignore: Set this to ``True`` to ignore an existing foreign key - otherwise a ``AlterError`` will be raised. + :param on_delete: ``ON DELETE`` action for the foreign key, e.g. ``"CASCADE"`` + or ``"SET NULL"``. + :param on_update: ``ON UPDATE`` action for the foreign key. """ columns = (column,) if isinstance(column, str) else tuple(column) # Ensure columns exist @@ -2997,11 +3002,27 @@ class Table(Queryable): ) ) if len(columns) == 1: - self.db.add_foreign_keys( - [(self.name, columns[0], other_table, other_columns[0])] + fk_object = ForeignKey( + self.name, + columns[0], + other_table, + other_columns[0], + on_delete=on_delete, + on_update=on_update, ) else: - self.db.add_foreign_keys([(self.name, columns, other_table, other_columns)]) + fk_object = ForeignKey( + self.name, + None, + other_table, + None, + columns=columns, + other_columns=other_columns, + is_compound=True, + on_delete=on_delete, + on_update=on_update, + ) + self.db.add_foreign_keys([fk_object]) return self def enable_counts(self) -> None: diff --git a/tests/test_foreign_keys.py b/tests/test_foreign_keys.py index 7fcd788..8950189 100644 --- a/tests/test_foreign_keys.py +++ b/tests/test_foreign_keys.py @@ -498,3 +498,29 @@ def test_add_foreign_keys_preserves_actions_compound(courses_db): assert fk.is_compound is True assert fk.on_delete == "CASCADE" assert "ON DELETE CASCADE" in courses_db["courses"].schema + + +def test_add_foreign_key_on_delete_on_update(fresh_db): + fresh_db["authors"].insert({"id": 1}, pk="id") + fresh_db["books"].insert({"id": 1, "author_id": 1}, pk="id") + fresh_db["books"].add_foreign_key( + "author_id", "authors", "id", on_delete="CASCADE", on_update="RESTRICT" + ) + fk = fresh_db["books"].foreign_keys[0] + assert fk.on_delete == "CASCADE" + assert fk.on_update == "RESTRICT" + assert "ON UPDATE RESTRICT ON DELETE CASCADE" in fresh_db["books"].schema + # The cascade should actually fire + fresh_db.execute("PRAGMA foreign_keys = ON") + fresh_db.execute("delete from authors where id = 1") + assert fresh_db["books"].count == 0 + + +def test_add_compound_foreign_key_on_delete(courses_db): + courses_db["courses"].add_foreign_key( + ("campus_name", "dept_code"), "departments", on_delete="SET NULL" + ) + fk = courses_db["courses"].foreign_keys[0] + assert fk.is_compound is True + assert fk.on_delete == "SET NULL" + assert "ON DELETE SET NULL" in courses_db["courses"].schema From afbfd95273c51d6a14174e9e07fb63ed2a34f72a Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Sun, 5 Jul 2026 16:00:52 -0700 Subject: [PATCH 22/86] Remove sqlean.py support (#772) Closes #771, refs #769 --- .github/workflows/test.yml | 3 --- docs/changelog.rst | 1 + docs/installation.rst | 10 +++++----- docs/python-api.rst | 2 +- mypy.ini | 3 --- sqlite_utils/utils.py | 11 +++-------- tests/test_gis.py | 8 -------- 7 files changed, 10 insertions(+), 28 deletions(-) diff --git a/.github/workflows/test.yml b/.github/workflows/test.yml index 7e1e953..d85cbf8 100644 --- a/.github/workflows/test.yml +++ b/.github/workflows/test.yml @@ -31,9 +31,6 @@ jobs: - name: Install SpatiaLite if: matrix.os == 'ubuntu-latest' run: sudo apt-get install libsqlite3-mod-spatialite - - name: On macOS with Python 3.10 test with sqlean.py - if: matrix.os == 'macos-latest' && matrix.python-version == '3.10' - run: pip install sqlean.py sqlite-dump - name: Build extension for --load-extension test if: matrix.os == 'ubuntu-latest' run: |- diff --git a/docs/changelog.rst b/docs/changelog.rst index f6a8ea2..7efeb95 100644 --- a/docs/changelog.rst +++ b/docs/changelog.rst @@ -12,6 +12,7 @@ Unreleased Breaking changes: - ``table.foreign_keys`` now returns ``ForeignKey`` objects that are dataclasses rather than ``namedtuple`` instances, so they can no longer be unpacked or indexed as ``(table, column, other_table, other_column)`` tuples - access their fields by name instead. Compound (multi-column) foreign keys are now represented as a single ``ForeignKey`` with ``is_compound=True`` and populated ``columns``/``other_columns`` tuples, where ``column`` and ``other_column`` are ``None``. Previously they were returned as one ``ForeignKey`` per column, misleadingly suggesting several independent foreign keys. See :ref:`upgrading_3_to_4` for details. (:issue:`594`) +- Removed support for using ``sqlean.py`` as a drop-in replacement for the Python standard library ``sqlite3`` module. ``sqlite-utils`` will now use ``pysqlite3`` if it is installed, otherwise it will use ``sqlite3`` from the standard library. Compound foreign key support: diff --git a/docs/installation.rst b/docs/installation.rst index beb6d4a..1333f5d 100644 --- a/docs/installation.rst +++ b/docs/installation.rst @@ -52,15 +52,15 @@ On some platforms the ability to load additional extensions (via ``conn.load_ext You may also see the error ``sqlite3.OperationalError: table sqlite_master may not be modified`` when trying to alter an existing table. -You can work around these limitations by installing either the `pysqlite3 `__ package or the `sqlean.py `__ package, both of which provide drop-in replacements for the standard library ``sqlite3`` module but with a recent version of SQLite and full support for loading extensions. +You can work around these limitations by installing the `pysqlite3 `__ package, which provides a drop-in replacement for the standard library ``sqlite3`` module but with a recent version of SQLite and full support for loading extensions. -To install ``sqlean.py`` (which has compiled binary wheels available for all major platforms) run the following: +To install ``pysqlite3`` run the following: .. code-block:: bash - sqlite-utils install sqlean.py + sqlite-utils install pysqlite3 -``pysqlite3`` and ``sqlean.py`` do not provide implementations of the ``.iterdump()`` method. To use that method (see :ref:`python_api_itedump`) or the ``sqlite-utils dump`` command you should also install the ``sqlite-dump`` package: +``pysqlite3`` does not provide an implementation of the ``.iterdump()`` method. To use that method (see :ref:`python_api_itedump`) or the ``sqlite-utils dump`` command you should also install the ``sqlite-dump`` package: .. code-block:: bash @@ -87,4 +87,4 @@ For ``zsh``: Add this code to ``~/.zshrc`` or ``~/.bashrc`` to automatically run it when you start a new shell. -See `the Click documentation `__ for more details. \ No newline at end of file +See `the Click documentation `__ for more details. diff --git a/docs/python-api.rst b/docs/python-api.rst index a33b256..6723efb 100644 --- a/docs/python-api.rst +++ b/docs/python-api.rst @@ -2184,7 +2184,7 @@ The ``db.iterdump()`` method returns a sequence of SQL strings representing a co This uses the `sqlite3.Connection.iterdump() `__ method. -If you are using ``pysqlite3`` or ``sqlean.py`` the underlying method may be missing. If you install the `sqlite-dump `__ package then the ``db.iterdump()`` method will use that implementation instead: +If you are using ``pysqlite3`` the underlying method may be missing. If you install the `sqlite-dump `__ package then the ``db.iterdump()`` method will use that implementation instead: .. code-block:: bash diff --git a/mypy.ini b/mypy.ini index de0dc83..2f6a875 100644 --- a/mypy.ini +++ b/mypy.ini @@ -16,9 +16,6 @@ ignore_errors = True [mypy-pysqlite3.*] ignore_missing_imports = True -[mypy-sqlean.*] -ignore_missing_imports = True - [mypy-sqlite_dump.*] ignore_missing_imports = True diff --git a/sqlite_utils/utils.py b/sqlite_utils/utils.py index 0ca98fc..865ee79 100644 --- a/sqlite_utils/utils.py +++ b/sqlite_utils/utils.py @@ -43,15 +43,10 @@ else: dbapi2 = importlib.import_module("pysqlite3.dbapi2") OperationalError = dbapi2.OperationalError except ImportError: - try: - sqlite3 = importlib.import_module("sqlean") - dbapi2 = importlib.import_module("sqlean.dbapi2") - OperationalError = dbapi2.OperationalError - except ImportError: - import sqlite3 # noqa: F401 - from sqlite3 import dbapi2 # noqa: F401 + import sqlite3 # noqa: F401 + from sqlite3 import dbapi2 # noqa: F401 - OperationalError = dbapi2.OperationalError + OperationalError = dbapi2.OperationalError SPATIALITE_PATHS = ( diff --git a/tests/test_gis.py b/tests/test_gis.py index 1b5ed70..e8f4c08 100644 --- a/tests/test_gis.py +++ b/tests/test_gis.py @@ -6,11 +6,6 @@ from sqlite_utils.cli import cli from sqlite_utils.db import Database from sqlite_utils.utils import find_spatialite, sqlite3 -try: - import sqlean # type: ignore[import-not-found] -except ImportError: - sqlean = None - pytestmark = [ pytest.mark.skipif( @@ -20,9 +15,6 @@ pytestmark = [ not hasattr(sqlite3.Connection, "enable_load_extension"), reason="sqlite3.Connection missing enable_load_extension", ), - pytest.mark.skipif( - sqlean is not None, reason="sqlean.py is not compatible with SpatiaLite" - ), ] From a00ed60efc3a242de38f666bc4ce1e431f86dcea Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Sun, 5 Jul 2026 21:17:41 -0700 Subject: [PATCH 23/86] Apply Black to tests/test_gis.py Fixes formatting left behind by afbfd95, which was merged while the Black check on its pull request was still failing. Co-Authored-By: Claude Fable 5 --- tests/test_gis.py | 1 - 1 file changed, 1 deletion(-) diff --git a/tests/test_gis.py b/tests/test_gis.py index e8f4c08..f39554e 100644 --- a/tests/test_gis.py +++ b/tests/test_gis.py @@ -6,7 +6,6 @@ from sqlite_utils.cli import cli from sqlite_utils.db import Database from sqlite_utils.utils import find_spatialite, sqlite3 - pytestmark = [ pytest.mark.skipif( not find_spatialite(), reason="Could not find SpatiaLite extension" From 07b603e562af19f80c3d00eb59b9cf29331127ce Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Sun, 5 Jul 2026 21:20:39 -0700 Subject: [PATCH 24/86] Preserve duplicate column names in query results Queries returning duplicate column names - e.g. joins between tables sharing column names - silently lost values because rows were built with dict(zip(keys, row)), where the last duplicate wins. Later occurrences are now renamed with a numeric suffix: id, id becomes id, id_2 - skipping any suffix that would collide with a real column in the same query. The new utils.dedupe_keys() helper transforms the key list once per query, so the per-row dict construction is unchanged and there is no measurable performance impact. Applied in Database.query() (including the PRAGMA and RETURNING paths), Table.rows_where(), Table.search() and the CLI's JSON output. CSV, TSV and table output keep the original duplicate headers. Closes #624 --- docs/cli.rst | 2 ++ docs/python-api.rst | 11 +++++++++++ sqlite_utils/cli.py | 5 +++++ sqlite_utils/db.py | 9 +++++---- sqlite_utils/utils.py | 31 +++++++++++++++++++++++++++++++ tests/test_cli.py | 20 ++++++++++++++++++++ tests/test_fts.py | 14 ++++++++++++++ tests/test_query.py | 16 ++++++++++++++++ tests/test_rows.py | 7 +++++++ tests/test_utils.py | 17 +++++++++++++++++ 10 files changed, 128 insertions(+), 4 deletions(-) diff --git a/docs/cli.rst b/docs/cli.rst index c9389d8..84a65d9 100644 --- a/docs/cli.rst +++ b/docs/cli.rst @@ -45,6 +45,8 @@ The default format returned for queries is JSON: [{"id": 1, "age": 4, "name": "Cleo"}, {"id": 2, "age": 2, "name": "Pancakes"}] +If the query returns more than one column with the same name, later occurrences are renamed with a numeric suffix - ``select 1 as id, 2 as id`` returns ``[{"id": 1, "id_2": 2}]``. This only applies to JSON output: :ref:`CSV and TSV ` and :ref:`table ` output keep the duplicate column headers unchanged. + .. _cli_query_nl: Newline-delimited JSON diff --git a/docs/python-api.rst b/docs/python-api.rst index 6723efb..7ac3951 100644 --- a/docs/python-api.rst +++ b/docs/python-api.rst @@ -233,6 +233,17 @@ The SQL query is executed as soon as ``db.query()`` is called. The resulting row ``db.query()`` can only be used with SQL that returns rows. Passing a statement that returns no rows - an ``INSERT`` or ``UPDATE`` without a ``RETURNING`` clause, for example - will raise a ``ValueError``. The rejected statement is rolled back, so it has no effect on the database. Use :ref:`db.execute() ` for those statements instead. +If a query returns more than one column with the same name - a join between two tables that share column names, for example - later occurrences are renamed with a numeric suffix, so every value is included in the dictionary: + +.. code-block:: python + + row = next(db.query("select 1 as id, 2 as id, 3 as id")) + print(row) + # Outputs: + # {'id': 1, 'id_2': 2, 'id_3': 3} + +A suffix that would collide with another column in the query is skipped - ``select 1 as id, 2 as id, 3 as id_2`` returns ``{'id': 1, 'id_3': 2, 'id_2': 3}``. The same renaming is applied by ``table.rows_where()`` and ``table.search()``. + .. _python_api_execute: db.execute(sql, params) diff --git a/sqlite_utils/cli.py b/sqlite_utils/cli.py index a30f79a..d5a4b1e 100644 --- a/sqlite_utils/cli.py +++ b/sqlite_utils/cli.py @@ -34,6 +34,7 @@ from .utils import ( OperationalError, _compile_code, chunks, + dedupe_keys, file_progress, find_spatialite, flatten as _flatten, @@ -3493,6 +3494,10 @@ FILE_COLUMNS = { def output_rows(iterator, headers, nl, arrays, json_cols): + # Duplicate column names would collide as dictionary keys, so rename + # later occurrences id, id -> id, id_2 - CSV and table output keep + # the original duplicate headers since they never build dictionaries + headers = dedupe_keys(headers) # We have to iterate two-at-a-time so we can know if we # should output a trailing comma or if we have reached # the last row. diff --git a/sqlite_utils/db.py b/sqlite_utils/db.py index 3f23c7d..d009f35 100644 --- a/sqlite_utils/db.py +++ b/sqlite_utils/db.py @@ -1,5 +1,6 @@ from .utils import ( chunks, + dedupe_keys, hash_record, sqlite3, OperationalError, @@ -786,7 +787,7 @@ class Database: cursor = self.conn.execute(sql, *args) if cursor.description is None: raise ValueError(message) - keys = [d[0] for d in cursor.description] + keys = dedupe_keys(d[0] for d in cursor.description) return (dict(zip(keys, row)) for row in cursor) # Execute inside a savepoint, so a statement that turns out not to # return rows can be rolled back before the ValueError is raised @@ -796,7 +797,7 @@ class Database: cursor = self.conn.execute(sql, *args) if cursor.description is None: raise ValueError(message) - keys = [d[0] for d in cursor.description] + keys = dedupe_keys(d[0] for d in cursor.description) try: self.conn.execute('RELEASE "sqlite_utils_query"') released = True @@ -1865,7 +1866,7 @@ class Queryable: if offset is not None: sql += " offset {}".format(offset) cursor = self.db.execute(sql, where_args or []) - columns = [c[0] for c in cursor.description] + columns = dedupe_keys(c[0] for c in cursor.description) for row in cursor: yield dict(zip(columns, row)) @@ -3398,7 +3399,7 @@ class Table(Queryable): ), args, ) - columns = [c[0] for c in cursor.description] + columns = dedupe_keys(c[0] for c in cursor.description) for row in cursor: yield dict(zip(columns, row)) diff --git a/sqlite_utils/utils.py b/sqlite_utils/utils.py index 865ee79..b39b117 100644 --- a/sqlite_utils/utils.py +++ b/sqlite_utils/utils.py @@ -613,6 +613,37 @@ def hash_record(record: Dict[str, Any], keys: Optional[Iterable[str]] = None) -> ).hexdigest() +def dedupe_keys(keys: Iterable[str]) -> List[str]: + """ + Rename duplicates in a list of column names so every name is unique, + by appending ``_2``, ``_3``... to later occurrences - skipping any + suffix that would collide with another column in the list. + + Used when converting SQL query rows to dictionaries, where duplicate + column names would otherwise silently overwrite each other. + + :param keys: List of column names, possibly containing duplicates + """ + keys = list(keys) + taken = set(keys) + if len(taken) == len(keys): + # No duplicates - the common case + return keys + seen: set = set() + result = [] + for key in keys: + if key in seen: + new_key = key + suffix = 2 + while new_key in seen or new_key in taken: + new_key = "{}_{}".format(key, suffix) + suffix += 1 + key = new_key + seen.add(key) + result.append(key) + return result + + def _flatten(d: Dict[str, Any]) -> Generator[Tuple[str, Any], None, None]: for key, value in d.items(): if isinstance(value, dict): diff --git a/tests/test_cli.py b/tests/test_cli.py index d26e4dd..f19a00c 100644 --- a/tests/test_cli.py +++ b/tests/test_cli.py @@ -746,6 +746,26 @@ def test_query_json_empty(db_path): assert result.output.strip() == "[]" +def test_query_json_duplicate_columns_are_deduped(db_path): + # https://github.com/simonw/sqlite-utils/issues/624 + result = CliRunner().invoke( + cli.cli, + [db_path, "select 1 as id, 2 as id, 'x' as value, 'y' as value"], + ) + assert result.output.strip() == ( + '[{"id": 1, "id_2": 2, "value": "x", "value_2": "y"}]' + ) + + +def test_query_csv_duplicate_columns_are_preserved(db_path): + # CSV output should keep the duplicate headers, not rename them + result = CliRunner().invoke( + cli.cli, + [db_path, "select 1 as id, 2 as id", "--csv"], + ) + assert result.output.replace("\r", "").strip() == "id,id\n1,2" + + def test_query_invalid_function(db_path): result = CliRunner().invoke( cli.cli, [db_path, "select bad()", "--functions", "def invalid_python"] diff --git a/tests/test_fts.py b/tests/test_fts.py index 3f7c5a9..64ec645 100644 --- a/tests/test_fts.py +++ b/tests/test_fts.py @@ -83,6 +83,20 @@ def test_enable_fts_escape_table_names(fresh_db): assert [] == list(table.search("bar")) +def test_search_duplicate_columns_are_deduped(fresh_db): + # https://github.com/simonw/sqlite-utils/issues/624 + table = fresh_db["t"] + table.insert_all(search_records) + table.enable_fts(["text", "country"], fts_version="FTS4") + rows = list(table.search("tanuki", columns=["text", "text"])) + assert rows == [ + { + "text": "tanuki are running tricksters", + "text_2": "tanuki are running tricksters", + } + ] + + def test_search_limit_offset(fresh_db): table = fresh_db["t"] table.insert_all(search_records) diff --git a/tests/test_query.py b/tests/test_query.py index b9822e1..0b9f2ae 100644 --- a/tests/test_query.py +++ b/tests/test_query.py @@ -218,6 +218,22 @@ def test_query_insert_returning_respects_explicit_transaction(fresh_db): assert [row["name"] for row in fresh_db["dogs"].rows] == ["Cleo"] +def test_query_duplicate_column_names_are_deduped(fresh_db): + # https://github.com/simonw/sqlite-utils/issues/624 + fresh_db["one"].insert({"id": 1, "value": "left"}) + fresh_db["two"].insert({"id": 2, "value": "right"}) + rows = list( + fresh_db.query("select one.id, two.id, one.value, two.value from one, two") + ) + assert rows == [{"id": 1, "id_2": 2, "value": "left", "value_2": "right"}] + + +def test_query_deduped_column_avoids_existing_names(fresh_db): + # The renamed duplicate must not overwrite a real column called id_2 + rows = list(fresh_db.query("select 1 as id, 2 as id, 3 as id_2")) + assert rows == [{"id": 1, "id_3": 2, "id_2": 3}] + + def test_execute_returning_dicts(fresh_db): # Like db.query() but returns a list, included for backwards compatibility # see https://github.com/simonw/sqlite-utils/issues/290 diff --git a/tests/test_rows.py b/tests/test_rows.py index a8a4ca0..f050d5a 100644 --- a/tests/test_rows.py +++ b/tests/test_rows.py @@ -104,3 +104,10 @@ def test_pks_and_rows_where_compound_pk(fresh_db): (("number", 1), {"type": "number", "number": 1, "plusone": 2}), (("number", 2), {"type": "number", "number": 2, "plusone": 3}), ] + + +def test_rows_where_duplicate_select_columns_are_deduped(fresh_db): + # https://github.com/simonw/sqlite-utils/issues/624 + fresh_db["t"].insert({"id": 1, "name": "Cleo"}) + rows = list(fresh_db["t"].rows_where(select="id, id, name")) + assert rows == [{"id": 1, "id_2": 1, "name": "Cleo"}] diff --git a/tests/test_utils.py b/tests/test_utils.py index f728bcd..3de5e94 100644 --- a/tests/test_utils.py +++ b/tests/test_utils.py @@ -83,3 +83,20 @@ def test_maximize_csv_field_size_limit(): ) def test_flatten(input, expected): assert utils.flatten(input) == expected + + +@pytest.mark.parametrize( + "input,expected", + ( + ([], []), + (["id", "name"], ["id", "name"]), + (["id", "id"], ["id", "id_2"]), + (["id", "id", "id"], ["id", "id_2", "id_3"]), + # A renamed duplicate must not clobber a real column called id_2 + (["id", "id", "id_2"], ["id", "id_3", "id_2"]), + (["id_2", "id", "id"], ["id_2", "id", "id_3"]), + (["id", "id", "id_2", "id_2"], ["id", "id_3", "id_2", "id_2_2"]), + ), +) +def test_dedupe_keys(input, expected): + assert utils.dedupe_keys(input) == expected From 02281f77ed7bfb56ff74a2f09a00aa03298e3268 Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Sun, 5 Jul 2026 21:42:22 -0700 Subject: [PATCH 25/86] Vendor SQLite version setup action (#775) Vendor https://github.com/asg017/sqlite-versions/tree/71ea0de37ae739c33e447af91ba71dda8fcf22e6 and make it more robust against `sqlite.org` timeouts. Closes #774 --- .../actions/setup-sqlite-version/action.yml | 39 +++++ .../setup-sqlite-version.sh | 144 ++++++++++++++++++ .github/workflows/test-sqlite-support.yml | 6 +- 3 files changed, 186 insertions(+), 3 deletions(-) create mode 100644 .github/actions/setup-sqlite-version/action.yml create mode 100644 .github/actions/setup-sqlite-version/setup-sqlite-version.sh diff --git a/.github/actions/setup-sqlite-version/action.yml b/.github/actions/setup-sqlite-version/action.yml new file mode 100644 index 0000000..fdbc71c --- /dev/null +++ b/.github/actions/setup-sqlite-version/action.yml @@ -0,0 +1,39 @@ +name: "Setup SQLite version" +description: "Build and activate a specific SQLite version from its amalgamation archive" +inputs: + version: + description: "The SQLite version to install" + required: true + cflags: + description: "CFLAGS to use when compiling SQLite" + required: false + default: "" + skip-activate: + description: "Set to true to skip modifying the library path" + required: false + default: "false" + fallback-urls: + description: "Whitespace-separated fallback download URLs to try after sqlite.org" + required: false + default: "" +outputs: + sqlite-location: + description: "Directory containing the compiled SQLite library" + value: ${{ steps.build.outputs.sqlite-location }} +runs: + using: "composite" + steps: + - shell: bash + run: mkdir -p "$RUNNER_TEMP/sqlite-versions/downloads" + - uses: actions/cache@v6 + with: + path: ${{ runner.temp }}/sqlite-versions/downloads + key: setup-sqlite-version-${{ inputs.version }}-amalgamation-v1 + - id: build + shell: bash + run: bash "$GITHUB_ACTION_PATH/setup-sqlite-version.sh" + env: + SQLITE_VERSION: ${{ inputs.version }} + SQLITE_CFLAGS: ${{ inputs.cflags }} + SQLITE_SKIP_ACTIVATE: ${{ inputs.skip-activate }} + SQLITE_EXTRA_FALLBACK_URLS: ${{ inputs.fallback-urls }} diff --git a/.github/actions/setup-sqlite-version/setup-sqlite-version.sh b/.github/actions/setup-sqlite-version/setup-sqlite-version.sh new file mode 100644 index 0000000..0df7290 --- /dev/null +++ b/.github/actions/setup-sqlite-version/setup-sqlite-version.sh @@ -0,0 +1,144 @@ +#!/usr/bin/env bash +set -euo pipefail + +version_spec="${SQLITE_VERSION:?SQLITE_VERSION is required}" +cflags="${SQLITE_CFLAGS:-}" +skip_activate="${SQLITE_SKIP_ACTIVATE:-false}" +extra_fallback_urls="${SQLITE_EXTRA_FALLBACK_URLS:-}" + +case "$version_spec" in + 3.46 | 3.46.0) + sqlite_version="3.46.0" + sqlite_year="2024" + amalgamation_id="3460000" + builtin_fallback_urls="https://static.simonwillison.net/static/2026/sqlite-amalgamation-3460000.zip" + ;; + 3.23.1) + sqlite_version="3.23.1" + sqlite_year="2018" + amalgamation_id="3230100" + builtin_fallback_urls="https://static.simonwillison.net/static/2026/sqlite-amalgamation-3230100.zip" + ;; + *) + echo "::error::Unsupported SQLite version '$version_spec'. Add its release year and amalgamation id to $GITHUB_ACTION_PATH/setup-sqlite-version.sh." + exit 1 + ;; +esac + +case "$(uname -s)" in + Linux) + library_name="libsqlite3.so.0" + library_path_var="LD_LIBRARY_PATH" + ;; + Darwin) + library_name="libsqlite3.dylib" + library_path_var="DYLD_LIBRARY_PATH" + ;; + *) + echo "::error::Unsupported platform $(uname -s)" + exit 1 + ;; +esac + +runner_temp="${RUNNER_TEMP:-}" +if [ -z "$runner_temp" ]; then + runner_temp="$(mktemp -d)" +fi + +filename="sqlite-amalgamation-${amalgamation_id}" +official_url="https://www.sqlite.org/${sqlite_year}/${filename}.zip" +download_dir="${runner_temp}/sqlite-versions/downloads" +source_root="${runner_temp}/sqlite-versions/source" +source_dir="${source_root}/${filename}" +build_dir="${runner_temp}/sqlite-versions/build/${sqlite_version}" +archive_path="${download_dir}/${filename}.zip" + +mkdir -p "$download_dir" "$source_root" "$build_dir" + +download_archive() { + local url + local candidate_path="${archive_path}.tmp" + local urls=("$official_url") + + for url in $builtin_fallback_urls $extra_fallback_urls; do + urls+=("$url") + done + + rm -f "$candidate_path" + for url in "${urls[@]}"; do + echo "Downloading SQLite ${sqlite_version} amalgamation from ${url}" + if curl \ + --fail \ + --location \ + --show-error \ + --retry 5 \ + --retry-delay 2 \ + --retry-max-time 180 \ + --retry-all-errors \ + --connect-timeout 20 \ + --max-time 240 \ + --output "$candidate_path" \ + "$url"; then + mv "$candidate_path" "$archive_path" + return 0 + fi + + echo "::warning::Download failed from ${url}" + rm -f "$candidate_path" + done + + echo "::error::Could not download SQLite ${sqlite_version} amalgamation" + return 1 +} + +if [ ! -f "${source_dir}/sqlite3.c" ]; then + if [ ! -f "$archive_path" ]; then + download_archive + fi + + rm -rf "$source_dir" + unzip -q "$archive_path" -d "$source_root" +fi + +if [ ! -f "${source_dir}/sqlite3.c" ]; then + echo "::error::Expected ${source_dir}/sqlite3.c after extracting ${archive_path}" + exit 1 +fi + +read -r -a cflag_args <<< "$cflags" + +echo "Compiling SQLite ${sqlite_version} to ${build_dir}/${library_name}" +gcc \ + -fPIC \ + -shared \ + "${cflag_args[@]}" \ + "${source_dir}/sqlite3.c" \ + "-I${source_dir}" \ + -o "${build_dir}/${library_name}" + +if [ "$library_name" = "libsqlite3.so.0" ]; then + ln -sf "$library_name" "${build_dir}/libsqlite3.so" +fi + +if [ -n "${GITHUB_OUTPUT:-}" ]; then + echo "sqlite-location=${build_dir}" >> "$GITHUB_OUTPUT" +else + echo "sqlite-location=${build_dir}" +fi + +case "$(printf '%s' "$skip_activate" | tr '[:upper:]' '[:lower:]')" in + true | 1 | yes) + echo "Skipping ${library_path_var} activation" + ;; + *) + existing_value="${!library_path_var:-}" + if [ -n "${GITHUB_ENV:-}" ]; then + if [ -n "$existing_value" ]; then + echo "${library_path_var}=${build_dir}:${existing_value}" >> "$GITHUB_ENV" + else + echo "${library_path_var}=${build_dir}" >> "$GITHUB_ENV" + fi + fi + echo "Added ${build_dir} to ${library_path_var}" + ;; +esac diff --git a/.github/workflows/test-sqlite-support.yml b/.github/workflows/test-sqlite-support.yml index aa2ab9b..f195cba 100644 --- a/.github/workflows/test-sqlite-support.yml +++ b/.github/workflows/test-sqlite-support.yml @@ -18,16 +18,16 @@ jobs: "3.23.1", # 2018-04-10, before UPSERT ] steps: - - uses: actions/checkout@v4 + - uses: actions/checkout@v7 - name: Set up Python ${{ matrix.python-version }} - uses: actions/setup-python@v5 + uses: actions/setup-python@v6 with: python-version: ${{ matrix.python-version }} allow-prereleases: true cache: pip cache-dependency-path: pyproject.toml - name: Set up SQLite ${{ matrix.sqlite-version }} - uses: asg017/sqlite-versions@71ea0de37ae739c33e447af91ba71dda8fcf22e6 + uses: ./.github/actions/setup-sqlite-version with: version: ${{ matrix.sqlite-version }} cflags: "-DSQLITE_ENABLE_DESERIALIZE -DSQLITE_ENABLE_FTS5 -DSQLITE_ENABLE_FTS4 -DSQLITE_ENABLE_FTS3_PARENTHESIS -DSQLITE_ENABLE_RTREE -DSQLITE_ENABLE_JSON1" From 50938ee6f846ddd792921f5a0353c73134dbbeda Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Sun, 5 Jul 2026 21:54:30 -0700 Subject: [PATCH 26/86] Rename ensure_autocommit_off() to ensure_autocommit_on(), closes #705 --- docs/changelog.rst | 1 + docs/upgrading.rst | 2 ++ sqlite_utils/db.py | 20 +++++++++++++------- tests/test_wal.py | 11 +++++++++++ 4 files changed, 27 insertions(+), 7 deletions(-) diff --git a/docs/changelog.rst b/docs/changelog.rst index 7efeb95..0220967 100644 --- a/docs/changelog.rst +++ b/docs/changelog.rst @@ -13,6 +13,7 @@ Breaking changes: - ``table.foreign_keys`` now returns ``ForeignKey`` objects that are dataclasses rather than ``namedtuple`` instances, so they can no longer be unpacked or indexed as ``(table, column, other_table, other_column)`` tuples - access their fields by name instead. Compound (multi-column) foreign keys are now represented as a single ``ForeignKey`` with ``is_compound=True`` and populated ``columns``/``other_columns`` tuples, where ``column`` and ``other_column`` are ``None``. Previously they were returned as one ``ForeignKey`` per column, misleadingly suggesting several independent foreign keys. See :ref:`upgrading_3_to_4` for details. (:issue:`594`) - Removed support for using ``sqlean.py`` as a drop-in replacement for the Python standard library ``sqlite3`` module. ``sqlite-utils`` will now use ``pysqlite3`` if it is installed, otherwise it will use ``sqlite3`` from the standard library. +- The ``db.ensure_autocommit_off()`` context manager has been renamed to ``db.ensure_autocommit_on()``, because the old name described the opposite of what it did. The method temporarily puts the connection into driver-level autocommit mode - by setting ``isolation_level = None`` - so that statements such as ``PRAGMA journal_mode=wal`` can run outside of an implicit transaction. (:issue:`705`) Compound foreign key support: diff --git a/docs/upgrading.rst b/docs/upgrading.rst index e7fd273..09a9e2f 100644 --- a/docs/upgrading.rst +++ b/docs/upgrading.rst @@ -77,6 +77,8 @@ Python API changes **table.convert() no longer skips falsey values.** Matching the CLI change above, ``table.convert()`` now converts every value. The ``skip_false`` parameter has been removed - previously it defaulted to ``True``, skipping empty strings and other falsey values. +**ensure_autocommit_off() is now ensure_autocommit_on().** The ``db.ensure_autocommit_off()`` context manager has been renamed to ``db.ensure_autocommit_on()``. The old name described the opposite of what the method did: it temporarily puts the connection into driver-level autocommit mode (by setting ``isolation_level = None``), so that statements such as ``PRAGMA journal_mode=wal`` can run outside of an implicit transaction. The behavior is unchanged - update any calls to use the new name. + **View.enable_fts() has been removed.** The ``View`` class previously had an ``enable_fts()`` method that existed only to raise ``NotImplementedError`` - full-text search is not supported for views. Calling it now raises ``AttributeError`` like any other missing method. **ForeignKey is now a dataclass, not a namedtuple.** The ``ForeignKey`` objects returned by ``table.foreign_keys`` gained new fields - ``columns``, ``other_columns``, ``is_compound``, ``on_delete`` and ``on_update`` - so that compound (multi-column) foreign keys and foreign key actions can be represented. To make room for those fields cleanly ``ForeignKey`` is now a dataclass rather than a ``namedtuple``, so it can no longer be unpacked or indexed as a tuple. Access its fields by name instead: diff --git a/sqlite_utils/db.py b/sqlite_utils/db.py index d009f35..d984bcf 100644 --- a/sqlite_utils/db.py +++ b/sqlite_utils/db.py @@ -610,16 +610,22 @@ class Database: self.conn.execute("ROLLBACK") @contextlib.contextmanager - def ensure_autocommit_off(self) -> Generator[None, None, None]: + def ensure_autocommit_on(self) -> Generator[None, None, None]: """ - Ensure autocommit is off for this database connection. + Ensure the connection is in driver-level autocommit mode for the + duration of a block of code. + + This temporarily sets ``isolation_level = None`` on the underlying + ``sqlite3`` connection, so the driver does not open implicit + transactions. This is useful for statements such as + ``PRAGMA journal_mode=wal`` which cannot run inside a transaction. Example usage:: - with db.ensure_autocommit_off(): + with db.ensure_autocommit_on(): # do stuff here - This will reset to the previous autocommit state at the end of the block. + The previous ``isolation_level`` is restored at the end of the block. """ old_isolation_level = self.conn.isolation_level try: @@ -783,7 +789,7 @@ class Database: if self.conn.in_transaction: cursor = self.conn.execute(sql, *args) else: - with self.ensure_autocommit_off(): + with self.ensure_autocommit_on(): cursor = self.conn.execute(sql, *args) if cursor.description is None: raise ValueError(message) @@ -1085,7 +1091,7 @@ class Database: """ if self.journal_mode != "wal": self._ensure_no_open_transaction("enable_wal()") - with self.ensure_autocommit_off(): + with self.ensure_autocommit_on(): self.execute("PRAGMA journal_mode=wal;") def disable_wal(self) -> None: @@ -1097,7 +1103,7 @@ class Database: """ if self.journal_mode != "delete": self._ensure_no_open_transaction("disable_wal()") - with self.ensure_autocommit_off(): + with self.ensure_autocommit_on(): self.execute("PRAGMA journal_mode=delete;") def _ensure_no_open_transaction(self, operation: str) -> None: diff --git a/tests/test_wal.py b/tests/test_wal.py index ee7ecf0..c5a9c60 100644 --- a/tests/test_wal.py +++ b/tests/test_wal.py @@ -49,6 +49,17 @@ def test_disable_wal_inside_transaction_raises(db_path_tmpdir): assert [r["id"] for r in db["test"].rows] == [1] +def test_ensure_autocommit_on(db_path_tmpdir): + db, path, tmpdir = db_path_tmpdir + previous_isolation_level = db.conn.isolation_level + assert previous_isolation_level is not None + with db.ensure_autocommit_on(): + # isolation_level of None means driver-level autocommit mode + assert db.conn.isolation_level is None + # Restored afterwards + assert db.conn.isolation_level == previous_isolation_level + + def test_enable_wal_noop_inside_transaction_is_allowed(db_path_tmpdir): # Calling enable_wal() when WAL is already enabled is a no-op, # so it is fine inside a transaction From 77d241959c90d34f3b2122c57a5f134ead61f9f0 Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Sun, 5 Jul 2026 22:11:22 -0700 Subject: [PATCH 27/86] Match column names case-insensitively, closes #760 Column names passed to Python API methods are now resolved against the table schema case-insensitively, mirroring how SQLite itself compares identifiers (ASCII-only case folding). Fixes KeyError populating last_pk from insert()/upsert(), silently ignored transform() rename/drop/types options, duplicate column errors from create_table(transform=True), redundant lookup() indexes, and case-sensitive foreign key validation. Co-Authored-By: Claude Fable 5 --- docs/changelog.rst | 13 ++ sqlite_utils/db.py | 238 +++++++++++++++++++++++++++++------- tests/test_column_casing.py | 233 +++++++++++++++++++++++++++++++++++ 3 files changed, 442 insertions(+), 42 deletions(-) create mode 100644 tests/test_column_casing.py diff --git a/docs/changelog.rst b/docs/changelog.rst index 0220967..0614dcd 100644 --- a/docs/changelog.rst +++ b/docs/changelog.rst @@ -29,6 +29,19 @@ Other foreign key improvements: - Foreign keys declared as ``REFERENCES other_table`` with no explicit column are now resolved to the other table's primary key by ``table.foreign_keys``, instead of reporting ``other_column=None``. - Fixed a ``TypeError`` when sorting ``ForeignKey`` objects where some were compound. +Case-insensitive column matching: + +Column names passed to Python API methods are now matched against the table schema case-insensitively, mirroring how SQLite itself treats identifiers. Previously many methods accepted mixed-case identifiers in the SQL they generated but then failed - or silently did nothing - when performing Python-side comparisons against the schema. (:issue:`760`) Fixes include: + +- ``table.insert()`` and ``table.upsert()`` now populate ``table.last_pk`` correctly when the ``pk=`` argument uses different casing to the table schema or the record keys - previously this raised a ``KeyError`` after the row had already been written. +- Upserts no longer raise or misbehave when the casing of ``pk=`` differs from the casing of the record keys. The primary key columns are correctly excluded from the generated ``DO UPDATE SET`` clause. +- ``table.transform()`` arguments ``types=``, ``rename=``, ``drop=``, ``pk=``, ``not_null=``, ``defaults=``, ``column_order=`` and ``drop_foreign_keys=`` all resolve column names case-insensitively. Previously options like ``rename={"name": "title"}`` against a column called ``Name`` were silently ignored. +- ``db.create_table(..., transform=True)`` now recognizes existing columns that differ only by case, instead of attempting to add them again and failing with ``duplicate column name``. The casing used in the existing schema is preserved. +- ``table.lookup()`` returns the primary key value even if ``pk=`` casing differs from the schema, and recognizes existing unique indexes case-insensitively instead of creating redundant ones. +- ``table.extract()`` and ``table.convert()`` - including ``multi=True`` and ``output=`` - accept column names in any casing. +- Foreign key columns are validated and recorded using the casing of the actual schema columns, in ``foreign_keys=`` when creating tables, ``db.add_foreign_keys()``, ``table.add_foreign_key()`` and ``table.add_column(fk_col=...)``. Duplicate foreign key detection is also case-insensitive. +- ``table.create()`` with ``pk=``, ``not_null=``, ``defaults=`` or ``column_order=`` referencing columns using different casing no longer creates an unwanted extra primary key column or raises a ``ValueError``. + .. _v4_0rc2: 4.0rc2 (2026-07-04) diff --git a/sqlite_utils/db.py b/sqlite_utils/db.py index d984bcf..90d49c6 100644 --- a/sqlite_utils/db.py +++ b/sqlite_utils/db.py @@ -85,6 +85,34 @@ def quote_identifier(identifier: str) -> str: return '"{}"'.format(identifier.replace('"', '""')) +_IDENTIFIER_CASEFOLD = str.maketrans( + "ABCDEFGHIJKLMNOPQRSTUVWXYZ", "abcdefghijklmnopqrstuvwxyz" +) + + +def fold_identifier_case(identifier: str) -> str: + """ + Lowercase an identifier using the same rules SQLite uses - only ASCII + characters are folded, other characters are left unchanged. + """ + return identifier.translate(_IDENTIFIER_CASEFOLD) + + +def resolve_casing(name: str, candidates: Iterable[str]) -> str: + """ + SQLite treats identifiers as case-insensitive. Return the entry in + ``candidates`` that matches ``name`` case-insensitively, preferring an + exact match. If nothing matches, return ``name`` unchanged. + """ + if name in candidates: + return name + folded = fold_identifier_case(name) + for candidate in candidates: + if fold_identifier_case(candidate) == folded: + return candidate + return name + + pd: Any = None try: pd = importlib.import_module("pandas") @@ -1263,6 +1291,48 @@ class Database: ) return fks + def _resolve_foreign_key_casing( + self, fk: ForeignKey, columns: Iterable[str] + ) -> ForeignKey: + """ + Return ``fk`` with its column references resolved to match the casing + of the actual columns. ``columns`` provides the column names of + ``fk.table``, which may be a table that is still being created. + """ + resolved_columns = tuple(resolve_casing(c, columns) for c in fk.columns) + if fk.other_table == fk.table: + other_candidates: Iterable[str] = columns + else: + other_candidates = self[fk.other_table].columns_dict + resolved_other_columns = tuple( + resolve_casing(c, other_candidates) for c in fk.other_columns + ) + if ( + resolved_columns == fk.columns + and resolved_other_columns == fk.other_columns + ): + return fk + if fk.is_compound: + return ForeignKey( + fk.table, + None, + fk.other_table, + None, + columns=resolved_columns, + other_columns=resolved_other_columns, + is_compound=True, + on_delete=fk.on_delete, + on_update=fk.on_update, + ) + return ForeignKey( + fk.table, + resolved_columns[0], + fk.other_table, + resolved_other_columns[0], + on_delete=fk.on_delete, + on_update=fk.on_update, + ) + def create_table_sql( self, name: str, @@ -1296,11 +1366,14 @@ class Database: """ if hash_id_columns and (hash_id is None): hash_id = "id" - foreign_keys = self.resolve_foreign_keys(name, foreign_keys or []) + resolved_fks: List[ForeignKey] = [ + self._resolve_foreign_key_casing(fk, columns) + for fk in self.resolve_foreign_keys(name, foreign_keys or []) + ] # Compound foreign keys are rendered as table-level constraints; # single-column ones as inline REFERENCES on their column foreign_keys_by_column = { - fk.column: fk for fk in foreign_keys if not fk.is_compound + fk.column: fk for fk in resolved_fks if not fk.is_compound } # any extracts will be treated as integer columns with a foreign key extracts = resolve_extracts(extracts) @@ -1315,8 +1388,10 @@ class Database: name, extract_column, extract_table, "id" ) # Soundness check not_null, and defaults if provided - not_null = not_null or set() - defaults = defaults or {} + not_null = {resolve_casing(n, columns) for n in not_null or set()} + defaults = {resolve_casing(n, columns): v for n, v in (defaults or {}).items()} + if column_order is not None: + column_order = [resolve_casing(c, columns) for c in column_order] if not columns: raise ValueError("Tables must have at least one column") if not all(n in columns for n in not_null): @@ -1342,7 +1417,7 @@ class Database: column_items.insert(0, (hash_id, str)) pk = hash_id # Soundness check foreign_keys point to existing tables - for fk in foreign_keys: + for fk in resolved_fks: for other_column in fk.other_columns: if fk.other_table == name and columns.get(other_column): continue @@ -1356,12 +1431,14 @@ class Database: column_defs = [] # ensure pk is a tuple single_pk = None - if isinstance(pk, list) and len(pk) == 1 and isinstance(pk[0], str): + if isinstance(pk, (list, tuple)) and len(pk) == 1 and isinstance(pk[0], str): pk = pk[0] if isinstance(pk, str): - single_pk = pk + single_pk = pk = resolve_casing(pk, [c[0] for c in column_items]) if pk not in [c[0] for c in column_items]: column_items.insert(0, (pk, int)) + elif pk: + pk = [resolve_casing(p, [c[0] for c in column_items]) for p in pk] for column_name, column_type in column_items: column_extras = [] if column_name == single_pk: @@ -1402,7 +1479,7 @@ class Database: ) # Compound foreign keys become table-level FOREIGN KEY constraints column_names = [c[0] for c in column_items] - for fk in foreign_keys: + for fk in resolved_fks: if not fk.is_compound: continue missing = [c for c in fk.columns if c not in column_names] @@ -1483,6 +1560,11 @@ class Database: should_transform = False # First add missing columns and figure out columns to drop existing_columns = table.columns_dict + # Match existing columns case-insensitively, the way SQLite does + columns = { + resolve_casing(col_name, existing_columns): col_type + for col_name, col_type in columns.items() + } missing_columns = dict( (col_name, col_type) for col_name, col_type in columns.items() @@ -1506,18 +1588,28 @@ class Database: current_pks = table.pks desired_pk = None if isinstance(pk, str): - desired_pk = [pk] + desired_pk = [resolve_casing(pk, existing_columns)] elif pk: - desired_pk = list(pk) + desired_pk = [resolve_casing(p, existing_columns) for p in pk] if desired_pk and current_pks != desired_pk: should_transform = True # Any not-null changes? current_not_null = {c.name for c in table.columns if c.notnull} - desired_not_null = set(not_null) if not_null else set() + desired_not_null = ( + {resolve_casing(n, existing_columns) for n in not_null} + if not_null + else set() + ) if current_not_null != desired_not_null: should_transform = True # How about defaults? - if defaults and defaults != table.default_values: + if ( + defaults + and { + resolve_casing(c, existing_columns): v for c, v in defaults.items() + } + != table.default_values + ): should_transform = True # Only run .transform() if there is something to do if should_transform: @@ -1671,12 +1763,15 @@ class Database: is_compound=True, ) table = fk_object.table - columns = fk_object.columns other_table = fk_object.other_table - other_columns = fk_object.other_columns if not self.table(table).exists(): raise AlterError("No such table: {}".format(table)) table_obj = self.table(table) + fk_object = self._resolve_foreign_key_casing( + fk_object, table_obj.columns_dict + ) + columns = fk_object.columns + other_columns = fk_object.other_columns for column in columns: if column not in table_obj.columns_dict: raise AlterError("No such column: {} in {}".format(column, table)) @@ -1693,12 +1788,16 @@ class Database: ) ) # We will silently skip foreign keys that exist already + columns_folded = tuple(fold_identifier_case(c) for c in columns) + other_columns_folded = tuple(fold_identifier_case(c) for c in other_columns) if not any( fk for fk in table_obj.foreign_keys - if fk.columns == columns - and fk.other_table == other_table - and fk.other_columns == other_columns + if tuple(fold_identifier_case(c) for c in fk.columns) == columns_folded + and fold_identifier_case(fk.other_table) + == fold_identifier_case(other_table) + and tuple(fold_identifier_case(c) for c in fk.other_columns) + == other_columns_folded ): foreign_keys_to_create.append(fk_object) @@ -2438,6 +2537,31 @@ class Table(Queryable): rename = rename or {} drop = drop or set() + # Resolve column references against the existing schema, matching + # case-insensitively the way SQLite does + existing_columns = self.columns_dict + types = {resolve_casing(c, existing_columns): t for c, t in types.items()} + rename = {resolve_casing(c, existing_columns): v for c, v in rename.items()} + drop = {resolve_casing(c, existing_columns) for c in drop} + if pk is not DEFAULT and pk is not None: + if isinstance(pk, str): + pk = resolve_casing(pk, existing_columns) + else: + pk = [resolve_casing(p, existing_columns) for p in pk] + if isinstance(not_null, dict): + not_null = { + resolve_casing(c, existing_columns): v + for c, v in cast(Dict[str, Any], not_null).items() + } + elif isinstance(not_null, set): + not_null = {resolve_casing(c, existing_columns) for c in not_null} + if defaults is not None: + defaults = { + resolve_casing(c, existing_columns): v for c, v in defaults.items() + } + if column_order is not None: + column_order = [resolve_casing(c, existing_columns) for c in column_order] + create_table_foreign_keys: List[ForeignKeyIndicator] = [] if foreign_keys is not None: @@ -2452,28 +2576,37 @@ class Table(Queryable): create_table_foreign_keys.extend(foreign_keys) else: # Construct foreign_keys from current, plus add_foreign_keys, minus drop_foreign_keys - # Bind fresh names here - type checkers do not narrow captured - # variables inside nested functions, so the closures would - # otherwise see the Optional declared types of drop and rename - dropped_columns = drop - renamed_columns = rename + # The casing of columns in a foreign key definition can differ + # from the casing of the columns themselves, so these comparisons + # are all case-folded + dropped_columns_folded = {fold_identifier_case(c) for c in drop} + renamed_columns_folded = { + fold_identifier_case(k): v for k, v in rename.items() + } def fk_should_be_dropped(fk: ForeignKey) -> bool: + fk_columns_folded = tuple(fold_identifier_case(c) for c in fk.columns) if drop_foreign_keys is not None: for spec in drop_foreign_keys: if isinstance(spec, str): # A column name matches any foreign key it participates in - if spec in fk.columns: + if fold_identifier_case(spec) in fk_columns_folded: return True - elif tuple(spec) == fk.columns: + elif ( + tuple(fold_identifier_case(s) for s in spec) + == fk_columns_folded + ): # A tuple/list must match a compound key's columns exactly return True # Dropping any of a foreign key's columns drops the whole key - return any(column in dropped_columns for column in fk.columns) + return any( + column in dropped_columns_folded for column in fk_columns_folded + ) def fk_with_renamed_columns(fk: ForeignKey) -> ForeignKey: columns = tuple( - renamed_columns.get(column) or column for column in fk.columns + renamed_columns_folded.get(fold_identifier_case(column)) or column + for column in fk.columns ) if fk.is_compound: return ForeignKey( @@ -2662,6 +2795,8 @@ class Table(Queryable): rename = rename or {} if isinstance(columns, str): columns = [columns] + columns = [resolve_casing(c, self.columns_dict) for c in columns] + rename = {resolve_casing(k, self.columns_dict): v for k, v in rename.items()} if not set(columns).issubset(self.columns_dict.keys()): raise InvalidColumns( "Invalid columns {} for table with columns {}".format( @@ -2853,6 +2988,7 @@ class Table(Queryable): raise AlterError("table '{}' does not exist".format(fk)) # if fk_col specified, must be a valid column if fk_col is not None: + fk_col = resolve_casing(fk_col, self.db[fk].columns_dict) if fk_col not in self.db[fk].columns_dict: raise AlterError("table '{}' has no column {}".format(fk, fk_col)) else: @@ -2958,6 +3094,7 @@ class Table(Queryable): :param on_update: ``ON UPDATE`` action for the foreign key. """ columns = (column,) if isinstance(column, str) else tuple(column) + columns = tuple(resolve_casing(c, self.columns_dict) for c in columns) # Ensure columns exist for col in columns: if col not in self.columns_dict: @@ -2979,6 +3116,9 @@ class Table(Queryable): other_columns = (other_column,) else: other_columns = tuple(other_column) + other_columns = tuple( + resolve_casing(c, self.db[other_table].columns_dict) for c in other_columns + ) if len(columns) != len(other_columns): raise ValueError( "Compound foreign key must have the same number of columns " @@ -2996,9 +3136,12 @@ class Table(Queryable): if any( fk for fk in self.foreign_keys - if fk.columns == columns - and fk.other_table == other_table - and fk.other_columns == other_columns + if tuple(fold_identifier_case(c) for c in fk.columns) + == tuple(fold_identifier_case(c) for c in columns) + and fold_identifier_case(fk.other_table) + == fold_identifier_case(other_table) + and tuple(fold_identifier_case(c) for c in fk.other_columns) + == tuple(fold_identifier_case(c) for c in other_columns) ): if ignore: return self @@ -3547,6 +3690,7 @@ class Table(Queryable): """ if isinstance(columns, str): columns = [columns] + columns = [resolve_casing(c, self.columns_dict) for c in columns] if multi: return self._convert_multi( @@ -3561,6 +3705,7 @@ class Table(Queryable): if output is not None: if len(columns) != 1: raise ValueError("output= can only be used with a single column") + output = resolve_casing(output, self.columns_dict) if output not in self.columns_dict: self.add_column(output, output_type or "text") @@ -3760,6 +3905,8 @@ class Table(Queryable): # Everything from here on is for upsert=True pk_cols = [pk] if isinstance(pk, str) else list(pk) + # The records may use different casing for the pk columns than pk= + pk_cols = [resolve_casing(c, all_columns) for c in pk_cols] # Every record must provide a value for every primary key column - a # NULL primary key never matches ON CONFLICT, so the record would be # inserted as a brand new row instead of upserted @@ -3810,10 +3957,7 @@ class Table(Queryable): # At this point we need compatibility UPSERT for SQLite < 3.24.0 # (INSERT OR IGNORE + second UPDATE stage) queries_and_params = [] - if isinstance(pk, str): - pks = [pk] - else: - pks = pk + pks = pk_cols self.last_pk = None for record_values in values: record = dict(zip(all_columns, record_values)) @@ -4226,9 +4370,9 @@ class Table(Queryable): if hash_id: self.last_pk = row[hash_id] elif isinstance(pk, str): - self.last_pk = row[pk] + self.last_pk = row[resolve_casing(pk, row)] else: - self.last_pk = tuple(row[p] for p in pk) + self.last_pk = tuple(row[resolve_casing(p, row)] for p in pk) else: self.last_pk = self.last_rowid else: @@ -4240,11 +4384,14 @@ class Table(Queryable): # hash_id not supported in list mode for last_pk pass elif isinstance(pk, str): - pk_index = column_names.index(pk) + pk_index = column_names.index(resolve_casing(pk, column_names)) self.last_pk = first_record_list[pk_index] else: self.last_pk = tuple( - first_record_list[column_names.index(p)] for p in pk + first_record_list[ + column_names.index(resolve_casing(p, column_names)) + ] + for p in pk ) else: first_record_dict = cast(Dict[str, Any], first_record) @@ -4252,9 +4399,12 @@ class Table(Queryable): self.last_pk = hash_record(first_record_dict, hash_id_columns) else: self.last_pk = ( - first_record_dict[pk] + first_record_dict[resolve_casing(pk, first_record_dict)] if isinstance(pk, str) - else tuple(first_record_dict[p] for p in pk) + else tuple( + first_record_dict[resolve_casing(p, first_record_dict)] + for p in pk + ) ) if analyze: @@ -4399,8 +4549,12 @@ class Table(Queryable): combined_values.update(extra_values) if self.exists(): self.add_missing_columns([combined_values]) - unique_column_sets = [set(i.columns) for i in self.indexes] - if set(lookup_values.keys()) not in unique_column_sets: + unique_column_sets = [ + {fold_identifier_case(c) for c in i.columns} for i in self.indexes + ] + if { + fold_identifier_case(c) for c in lookup_values + } not in unique_column_sets: self.create_index(lookup_values.keys(), unique=True) wheres = [ "{} = ?".format(quote_identifier(column)) for column in lookup_values @@ -4411,7 +4565,7 @@ class Table(Queryable): ) ) try: - return rows[0][pk] + return rows[0][resolve_casing(pk, rows[0])] except IndexError: return self.insert( combined_values, diff --git a/tests/test_column_casing.py b/tests/test_column_casing.py new file mode 100644 index 0000000..ce11345 --- /dev/null +++ b/tests/test_column_casing.py @@ -0,0 +1,233 @@ +""" +SQLite treats column names as case-insensitive. These tests exercise the +places where sqlite-utils performs Python-side lookups of column names +provided by the caller, which should match the schema case-insensitively. + +https://github.com/simonw/sqlite-utils/issues/760 +""" + +import pytest + +from sqlite_utils import Database +from sqlite_utils.db import ForeignKey + + +def test_insert_populates_last_pk_case_insensitively(fresh_db): + books = fresh_db["books"] + books.create({"Id": int, "Title": str}, pk="Id") + books.insert({"Id": 1, "Title": "One"}, pk="id") + assert books.last_pk == 1 + + +def test_insert_populates_last_pk_compound_pk_case_insensitively(fresh_db): + books = fresh_db["books"] + books.create({"Author": str, "Position": int, "Title": str}) + books.insert( + {"Author": "Sue", "Position": 1, "Title": "One"}, pk=("author", "position") + ) + assert books.last_pk == ("Sue", 1) + + +@pytest.mark.parametrize("use_old_upsert", (False, True)) +def test_upsert_pk_case_differs_from_schema(use_old_upsert): + db = Database(memory=True, use_old_upsert=use_old_upsert) + books = db["books"] + books.create({"Id": int, "Title": str}, pk="Id") + books.insert({"Id": 1, "Title": "One"}) + books.upsert({"id": 1, "title": "Won"}, pk="id") + assert list(books.rows) == [{"Id": 1, "Title": "Won"}] + assert books.last_pk == 1 + + +@pytest.mark.parametrize("use_old_upsert", (False, True)) +def test_upsert_record_key_case_differs_from_pk(use_old_upsert): + # all_columns comes from the record keys, pk= from the caller + db = Database(memory=True, use_old_upsert=use_old_upsert) + books = db["books"] + books.create({"Id": int, "Title": str}, pk="Id") + books.upsert({"ID": 1, "Title": "One"}, pk="id") + assert list(books.rows) == [{"Id": 1, "Title": "One"}] + assert books.last_pk == 1 + + +def test_upsert_inferred_pk_case_differs_from_record_keys(fresh_db): + # pk is inferred from the existing schema as "Id", records use "id" + books = fresh_db["books"] + books.create({"Id": int, "Title": str}, pk="Id") + books.upsert({"id": 1, "title": "One"}) + assert list(books.rows) == [{"Id": 1, "Title": "One"}] + assert books.last_pk == 1 + + +def test_upsert_list_mode_pk_case_insensitive(fresh_db): + books = fresh_db["books"] + books.create({"Id": int, "Title": str}, pk="Id") + books.upsert_all([["id", "title"], [1, "One"]], pk="Id") + assert list(books.rows) == [{"Id": 1, "Title": "One"}] + assert books.last_pk == 1 + + +def test_lookup_pk_case_insensitive(fresh_db): + fresh_db["species"].create({"ID": int, "Name": str}, pk="ID") + fresh_db["species"].insert({"ID": 5, "Name": "Palm"}) + fresh_db["species"].create_index(["Name"], unique=True) + assert fresh_db["species"].lookup({"Name": "Palm"}, pk="id") == 5 + + +def test_lookup_does_not_create_redundant_index(fresh_db): + fresh_db["species"].create({"id": int, "Name": str}, pk="id") + fresh_db["species"].create_index(["Name"], unique=True) + fresh_db["species"].lookup({"name": "Palm"}) + assert len(fresh_db["species"].indexes) == 1 + + +def test_create_table_transform_same_columns_different_case(fresh_db): + fresh_db["t"].create({"Name": str, "Age": int}) + fresh_db["t"].insert({"Name": "Cleo", "Age": 5}) + fresh_db.create_table("t", {"name": str, "age": int}, transform=True) + # Schema casing is preserved - SQLite considers these the same columns + assert fresh_db["t"].columns_dict == {"Name": str, "Age": int} + assert list(fresh_db["t"].rows) == [{"Name": "Cleo", "Age": 5}] + + +def test_create_table_transform_case_insensitive_with_changes(fresh_db): + fresh_db["t"].create({"Name": str, "Age": int}) + fresh_db.create_table("t", {"name": str, "age": str, "size": int}, transform=True) + # age changed type, size added, Name untouched + assert fresh_db["t"].columns_dict == {"Name": str, "Age": str, "size": int} + + +def test_transform_types_case_insensitive(fresh_db): + fresh_db["t"].create({"Name": str, "Age": str}) + fresh_db["t"].transform(types={"age": int}) + assert fresh_db["t"].columns_dict == {"Name": str, "Age": int} + + +def test_transform_rename_case_insensitive(fresh_db): + fresh_db["t"].create({"Name": str}) + fresh_db["t"].transform(rename={"name": "title"}) + assert fresh_db["t"].columns_dict == {"title": str} + + +def test_transform_drop_case_insensitive(fresh_db): + fresh_db["t"].create({"Name": str, "Age": int}) + fresh_db["t"].transform(drop=["name"]) + assert fresh_db["t"].columns_dict == {"Age": int} + + +def test_transform_not_null_and_defaults_case_insensitive(fresh_db): + fresh_db["t"].create({"Name": str, "Age": int}) + fresh_db["t"].transform(not_null={"name"}, defaults={"age": 3}) + columns = {c.name: c for c in fresh_db["t"].columns} + assert columns["Name"].notnull + assert fresh_db["t"].default_values == {"Age": 3} + + +def test_transform_pk_case_insensitive(fresh_db): + fresh_db["t"].create({"Id": int, "Name": str}) + fresh_db["t"].transform(pk="id") + assert fresh_db["t"].pks == ["Id"] + assert fresh_db["t"].columns_dict == {"Id": int, "Name": str} + + +def test_transform_drop_foreign_keys_case_insensitive(fresh_db): + fresh_db["parent"].create({"Id": int}, pk="Id") + fresh_db["child"].create( + {"id": int, "Parent_ID": int}, + pk="id", + foreign_keys=[("Parent_ID", "parent", "Id")], + ) + fresh_db["child"].transform(drop_foreign_keys=["parent_id"]) + assert fresh_db["child"].foreign_keys == [] + + +def test_add_foreign_key_case_insensitive(fresh_db): + fresh_db["parent"].create({"Id": int}, pk="Id") + fresh_db["child"].create({"id": int, "Parent_ID": int}, pk="id") + fresh_db["child"].add_foreign_key("parent_id", "parent", "id") + fks = fresh_db["child"].foreign_keys + assert len(fks) == 1 + # The foreign key should use the schema casing of the columns + assert fks[0].column == "Parent_ID" + assert fks[0].other_column == "Id" + + +def test_add_foreign_keys_case_insensitive(fresh_db): + fresh_db["parent"].create({"Id": int}, pk="Id") + fresh_db["child"].create({"id": int, "Parent_ID": int}, pk="id") + fresh_db.add_foreign_keys([("child", "parent_id", "parent", "id")]) + fks = fresh_db["child"].foreign_keys + assert len(fks) == 1 + assert fks[0].column == "Parent_ID" + assert fks[0].other_column == "Id" + + +def test_add_foreign_key_detects_existing_case_insensitively(fresh_db): + fresh_db["parent"].create({"Id": int}, pk="Id") + fresh_db["child"].create( + {"id": int, "Parent_ID": int}, + pk="id", + foreign_keys=[("Parent_ID", "parent", "Id")], + ) + # ignore=True should treat this as already existing, not add a duplicate + fresh_db["child"].add_foreign_key("parent_id", "parent", "id", ignore=True) + assert len(fresh_db["child"].foreign_keys) == 1 + + +def test_add_column_fk_col_case_insensitive(fresh_db): + fresh_db["parent"].create({"Id": int}, pk="Id") + fresh_db["child"].create({"id": int}, pk="id") + fresh_db["child"].add_column("parent_id", int, fk="parent", fk_col="id") + fks = fresh_db["child"].foreign_keys + assert len(fks) == 1 + assert fks[0].other_column == "Id" + + +def test_extract_case_insensitive(fresh_db): + fresh_db["trees"].insert({"id": 1, "Species": "Palm"}, pk="id") + fresh_db["trees"].extract("species") + assert fresh_db["trees"].columns_dict == {"id": int, "Species_id": int} + assert list(fresh_db["Species"].rows) == [{"id": 1, "Species": "Palm"}] + + +def test_convert_multi_case_insensitive(fresh_db): + fresh_db["t"].insert({"id": 1, "Name": "Cleo"}, pk="id") + fresh_db["t"].convert("name", lambda v: {"upper": v.upper()}, multi=True) + assert list(fresh_db["t"].rows) == [{"id": 1, "Name": "Cleo", "upper": "CLEO"}] + + +def test_convert_output_case_insensitive(fresh_db): + fresh_db["t"].insert({"id": 1, "Name": "Cleo", "Upper": None}, pk="id") + fresh_db["t"].convert("name", lambda v: v.upper(), output="upper") + assert list(fresh_db["t"].rows) == [{"id": 1, "Name": "Cleo", "Upper": "CLEO"}] + + +def test_create_table_sql_pk_case_insensitive(fresh_db): + fresh_db["t"].create({"Id": int, "Name": str}, pk="id") + # Should not have created an extra lowercase "id" column + assert fresh_db["t"].columns_dict == {"Id": int, "Name": str} + assert fresh_db["t"].pks == ["Id"] + + +def test_create_table_not_null_and_defaults_case_insensitive(fresh_db): + fresh_db["t"].create( + {"Name": str, "Age": int}, not_null={"name"}, defaults={"age": 1} + ) + columns = {c.name: c for c in fresh_db["t"].columns} + assert columns["Name"].notnull + assert fresh_db["t"].default_values == {"Age": 1} + + +def test_create_table_foreign_keys_case_insensitive(fresh_db): + fresh_db["parent"].create({"Id": int}, pk="Id") + fresh_db["child"].create( + {"id": int, "Parent_ID": int}, + pk="id", + foreign_keys=[("parent_id", "parent", "id")], + ) + fks = fresh_db["child"].foreign_keys + assert fks == [ + ForeignKey( + table="child", column="Parent_ID", other_table="parent", other_column="Id" + ) + ] From a4acc3958c456298bb017cbdea515fe4d5374ed3 Mon Sep 17 00:00:00 2001 From: Vincent Gao Date: Mon, 6 Jul 2026 07:33:55 +0200 Subject: [PATCH 28/86] Fix transform() corrupting TRUE/FALSE/NULL column defaults into strings * Fix transform() corrupting TRUE/FALSE/NULL column defaults quote_default_value() passed keyword-literal defaults (TRUE, FALSE, NULL) through self.quote(), wrapping them in quotes. So transform() rebuilt a column declared "INTEGER DEFAULT TRUE" as "INTEGER DEFAULT 'TRUE'", and a later default insert stored the text 'TRUE' instead of the integer 1 (likewise 'FALSE' for 0 and 'NULL' for null) - silent value corruption. Return these keyword literals unquoted, as already done for the CURRENT_TIME/DATE/TIMESTAMP literals. PR #764 --- docs/changelog.rst | 4 ++++ sqlite_utils/db.py | 5 +++++ tests/test_default_value.py | 5 +++++ tests/test_transform.py | 34 ++++++++++++++++++++++++++++++++++ 4 files changed, 48 insertions(+) diff --git a/docs/changelog.rst b/docs/changelog.rst index 0614dcd..43a4653 100644 --- a/docs/changelog.rst +++ b/docs/changelog.rst @@ -42,6 +42,10 @@ Column names passed to Python API methods are now matched against the table sche - Foreign key columns are validated and recorded using the casing of the actual schema columns, in ``foreign_keys=`` when creating tables, ``db.add_foreign_keys()``, ``table.add_foreign_key()`` and ``table.add_column(fk_col=...)``. Duplicate foreign key detection is also case-insensitive. - ``table.create()`` with ``pk=``, ``not_null=``, ``defaults=`` or ``column_order=`` referencing columns using different casing no longer creates an unwanted extra primary key column or raises a ``ValueError``. +Everything else: + +- Fixed a bug where ``table.transform()`` could convert ``DEFAULT TRUE``, ``DEFAULT FALSE`` and ``DEFAULT NULL`` column defaults into quoted string defaults when rebuilding a table. Thanks, `Vincent Gao `__. (`#764 `__) + .. _v4_0rc2: 4.0rc2 (2026-07-04) diff --git a/sqlite_utils/db.py b/sqlite_utils/db.py index 90d49c6..0807892 100644 --- a/sqlite_utils/db.py +++ b/sqlite_utils/db.py @@ -983,6 +983,11 @@ class Database: if str(value).upper() in ("CURRENT_TIME", "CURRENT_DATE", "CURRENT_TIMESTAMP"): return value + if isinstance(value, str) and value.upper() in ("TRUE", "FALSE", "NULL"): + # Keyword literals must stay unquoted; quoting them would turn the + # default into a string ('TRUE' instead of 1, 'NULL' instead of null). + return value + if str(value).endswith(")"): # Expr return "({})".format(value) diff --git a/tests/test_default_value.py b/tests/test_default_value.py index c5e4b17..3724d99 100644 --- a/tests/test_default_value.py +++ b/tests/test_default_value.py @@ -21,6 +21,11 @@ EXAMPLES = [ # Strings ("TEXT DEFAULT 'CURRENT_TIMESTAMP'", "'CURRENT_TIMESTAMP'", "'CURRENT_TIMESTAMP'"), ('TEXT DEFAULT "CURRENT_TIMESTAMP"', '"CURRENT_TIMESTAMP"', '"CURRENT_TIMESTAMP"'), + # Boolean and null keyword literals must stay unquoted + ("INTEGER DEFAULT TRUE", "TRUE", "TRUE"), + ("INTEGER DEFAULT FALSE", "FALSE", "FALSE"), + ("INTEGER DEFAULT true", "true", "true"), + ("TEXT DEFAULT NULL", "NULL", "NULL"), ] diff --git a/tests/test_transform.py b/tests/test_transform.py index 5eb501d..71518be 100644 --- a/tests/test_transform.py +++ b/tests/test_transform.py @@ -224,6 +224,40 @@ def test_transform_rename_pk(fresh_db): ) +def test_transform_preserves_keyword_literal_defaults(fresh_db): + # transform() used to requote keyword-literal defaults (DEFAULT TRUE became + # DEFAULT 'TRUE'), so a default insert stored the text 'TRUE' instead of the + # integer 1 -- silent value corruption on every rebuilt table. + fresh_db.execute( + "CREATE TABLE t (" + " id INTEGER PRIMARY KEY," + " is_active INTEGER DEFAULT TRUE," + " flag INTEGER DEFAULT FALSE," + " note TEXT DEFAULT NULL" + ")" + ) + table = fresh_db["t"] + table.insert({"id": 1}) + before = fresh_db.execute("SELECT is_active, flag, note FROM t").fetchone() + assert before == (1, 0, None) + + # Rebuild the table via an unrelated change. + table.transform(rename={"note": "note2"}) + + # The keyword literals stay unquoted in the schema ... + assert "DEFAULT TRUE" in table.schema + assert "DEFAULT FALSE" in table.schema + assert "DEFAULT NULL" in table.schema + assert "'TRUE'" not in table.schema + + # ... and a fresh default insert still yields 1 / 0 / NULL, not strings. + table.insert({"id": 2}) + after = fresh_db.execute( + "SELECT is_active, flag, note2 FROM t WHERE id = 2" + ).fetchone() + assert after == (1, 0, None) + + def test_transform_not_null(fresh_db): dogs = fresh_db["dogs"] dogs.insert({"id": 1, "name": "Cleo", "age": "5"}, pk="id") From 3e8b7403a286e0f9ae4cfb9b8ff2e010c3edb2de Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Sun, 5 Jul 2026 22:35:35 -0700 Subject: [PATCH 29/86] Release 4.0rc3 Refs https://github.com/simonw/sqlite-utils/issues/769#issuecomment-4889420844 Refs #530, #594, #624, #705, #760, #764, #769, #770, #771, #772, #774, #775 --- docs/changelog.rst | 6 +++--- pyproject.toml | 2 +- 2 files changed, 4 insertions(+), 4 deletions(-) diff --git a/docs/changelog.rst b/docs/changelog.rst index 43a4653..4b48a40 100644 --- a/docs/changelog.rst +++ b/docs/changelog.rst @@ -4,10 +4,10 @@ Changelog =========== -.. _v_unreleased: +.. _v4_0rc3: -Unreleased ----------- +4.0rc3 (2026-07-05) +------------------- Breaking changes: diff --git a/pyproject.toml b/pyproject.toml index 9d1f041..bff9390 100644 --- a/pyproject.toml +++ b/pyproject.toml @@ -1,6 +1,6 @@ [project] name = "sqlite-utils" -version = "4.0rc2" +version = "4.0rc3" description = "CLI tool and Python library for manipulating SQLite databases" readme = { file = "README.md", content-type = "text/markdown" } authors = [ From d1f5e06816f2fdaa2f626c8cd6814a209bf51a7b Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Sun, 5 Jul 2026 23:12:54 -0700 Subject: [PATCH 30/86] Add subheadings to 4.0rc3 release notes So I can link to those sections. --- docs/changelog.rst | 15 ++++++++++----- 1 file changed, 10 insertions(+), 5 deletions(-) diff --git a/docs/changelog.rst b/docs/changelog.rst index 4b48a40..6412101 100644 --- a/docs/changelog.rst +++ b/docs/changelog.rst @@ -9,27 +9,31 @@ 4.0rc3 (2026-07-05) ------------------- -Breaking changes: +Breaking changes +~~~~~~~~~~~~~~~~ - ``table.foreign_keys`` now returns ``ForeignKey`` objects that are dataclasses rather than ``namedtuple`` instances, so they can no longer be unpacked or indexed as ``(table, column, other_table, other_column)`` tuples - access their fields by name instead. Compound (multi-column) foreign keys are now represented as a single ``ForeignKey`` with ``is_compound=True`` and populated ``columns``/``other_columns`` tuples, where ``column`` and ``other_column`` are ``None``. Previously they were returned as one ``ForeignKey`` per column, misleadingly suggesting several independent foreign keys. See :ref:`upgrading_3_to_4` for details. (:issue:`594`) - Removed support for using ``sqlean.py`` as a drop-in replacement for the Python standard library ``sqlite3`` module. ``sqlite-utils`` will now use ``pysqlite3`` if it is installed, otherwise it will use ``sqlite3`` from the standard library. - The ``db.ensure_autocommit_off()`` context manager has been renamed to ``db.ensure_autocommit_on()``, because the old name described the opposite of what it did. The method temporarily puts the connection into driver-level autocommit mode - by setting ``isolation_level = None`` - so that statements such as ``PRAGMA journal_mode=wal`` can run outside of an implicit transaction. (:issue:`705`) -Compound foreign key support: +Compound foreign key support +~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - Tables can now be created with compound foreign keys, by passing tuples of column names in ``foreign_keys=``: ``foreign_keys=[(("campus_name", "dept_code"), "departments")]``. The referenced columns default to the compound primary key of the other table. Compound keys are rendered as table-level ``FOREIGN KEY`` constraints in the generated schema. See :ref:`python_api_compound_foreign_keys`. - ``table.transform()`` now preserves compound foreign keys, applying any column renames to them. Dropping a column that is part of a compound foreign key drops the whole constraint, matching the existing single-column behavior. ``drop_foreign_keys=`` accepts a bare column name - dropping any foreign key that column participates in - or a tuple of columns to target a compound key precisely. - ``table.add_foreign_key()`` and ``db.add_foreign_keys()`` accept tuples of column names to add a compound foreign key to an existing table. - ``db.index_foreign_keys()`` creates a single composite index for a compound foreign key. -Other foreign key improvements: +Other foreign key improvements +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - ``ForeignKey`` now exposes ``on_delete`` and ``on_update`` fields reflecting the foreign key's ``ON DELETE``/``ON UPDATE`` actions, and ``table.transform()`` preserves those actions. Previously a transform silently stripped clauses such as ``ON DELETE CASCADE`` from the table schema. - ``table.add_foreign_key()`` accepts new ``on_delete=`` and ``on_update=`` parameters for creating foreign keys with actions, e.g. ``table.add_foreign_key("author_id", "authors", "id", on_delete="CASCADE")``. (:issue:`530`) - Foreign keys declared as ``REFERENCES other_table`` with no explicit column are now resolved to the other table's primary key by ``table.foreign_keys``, instead of reporting ``other_column=None``. - Fixed a ``TypeError`` when sorting ``ForeignKey`` objects where some were compound. -Case-insensitive column matching: +Case-insensitive column matching +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Column names passed to Python API methods are now matched against the table schema case-insensitively, mirroring how SQLite itself treats identifiers. Previously many methods accepted mixed-case identifiers in the SQL they generated but then failed - or silently did nothing - when performing Python-side comparisons against the schema. (:issue:`760`) Fixes include: @@ -42,7 +46,8 @@ Column names passed to Python API methods are now matched against the table sche - Foreign key columns are validated and recorded using the casing of the actual schema columns, in ``foreign_keys=`` when creating tables, ``db.add_foreign_keys()``, ``table.add_foreign_key()`` and ``table.add_column(fk_col=...)``. Duplicate foreign key detection is also case-insensitive. - ``table.create()`` with ``pk=``, ``not_null=``, ``defaults=`` or ``column_order=`` referencing columns using different casing no longer creates an unwanted extra primary key column or raises a ``ValueError``. -Everything else: +Everything else +~~~~~~~~~~~~~~~ - Fixed a bug where ``table.transform()`` could convert ``DEFAULT TRUE``, ``DEFAULT FALSE`` and ``DEFAULT NULL`` column defaults into quoted string defaults when rebuilding a table. Thanks, `Vincent Gao `__. (`#764 `__) From af3894a096ed44433ca4d40bf7bf701a71f4f097 Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Sun, 5 Jul 2026 23:19:00 -0700 Subject: [PATCH 31/86] Changelog headings for 4.0rc3 --- docs/changelog.rst | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/docs/changelog.rst b/docs/changelog.rst index 6412101..085d9d5 100644 --- a/docs/changelog.rst +++ b/docs/changelog.rst @@ -12,14 +12,14 @@ Breaking changes ~~~~~~~~~~~~~~~~ -- ``table.foreign_keys`` now returns ``ForeignKey`` objects that are dataclasses rather than ``namedtuple`` instances, so they can no longer be unpacked or indexed as ``(table, column, other_table, other_column)`` tuples - access their fields by name instead. Compound (multi-column) foreign keys are now represented as a single ``ForeignKey`` with ``is_compound=True`` and populated ``columns``/``other_columns`` tuples, where ``column`` and ``other_column`` are ``None``. Previously they were returned as one ``ForeignKey`` per column, misleadingly suggesting several independent foreign keys. See :ref:`upgrading_3_to_4` for details. (:issue:`594`) +- :ref:`table.foreign_keys ` now returns ``ForeignKey`` objects that are dataclasses rather than ``namedtuple`` instances, so they can no longer be unpacked or indexed as ``(table, column, other_table, other_column)`` tuples - access their fields by name instead. Compound (multi-column) foreign keys are now represented as a single ``ForeignKey`` with ``is_compound=True`` and populated ``columns``/``other_columns`` tuples, where ``column`` and ``other_column`` are ``None``. Previously they were returned as one ``ForeignKey`` per column, misleadingly suggesting several independent foreign keys. See :ref:`upgrading_3_to_4` for details. (:issue:`594`) - Removed support for using ``sqlean.py`` as a drop-in replacement for the Python standard library ``sqlite3`` module. ``sqlite-utils`` will now use ``pysqlite3`` if it is installed, otherwise it will use ``sqlite3`` from the standard library. - The ``db.ensure_autocommit_off()`` context manager has been renamed to ``db.ensure_autocommit_on()``, because the old name described the opposite of what it did. The method temporarily puts the connection into driver-level autocommit mode - by setting ``isolation_level = None`` - so that statements such as ``PRAGMA journal_mode=wal`` can run outside of an implicit transaction. (:issue:`705`) Compound foreign key support ~~~~~~~~~~~~~~~~~~~~~~~~~~~~ -- Tables can now be created with compound foreign keys, by passing tuples of column names in ``foreign_keys=``: ``foreign_keys=[(("campus_name", "dept_code"), "departments")]``. The referenced columns default to the compound primary key of the other table. Compound keys are rendered as table-level ``FOREIGN KEY`` constraints in the generated schema. See :ref:`python_api_compound_foreign_keys`. +- Tables can now be created with :ref:`compound foreign keys `, by passing tuples of column names in ``foreign_keys=``: ``foreign_keys=[(("campus_name", "dept_code"), "departments")]``. The referenced columns default to the compound primary key of the other table. Compound keys are rendered as table-level ``FOREIGN KEY`` constraints in the generated schema. - ``table.transform()`` now preserves compound foreign keys, applying any column renames to them. Dropping a column that is part of a compound foreign key drops the whole constraint, matching the existing single-column behavior. ``drop_foreign_keys=`` accepts a bare column name - dropping any foreign key that column participates in - or a tuple of columns to target a compound key precisely. - ``table.add_foreign_key()`` and ``db.add_foreign_keys()`` accept tuples of column names to add a compound foreign key to an existing table. - ``db.index_foreign_keys()`` creates a single composite index for a compound foreign key. From 5f81752cf5e2a9a3f6ef5bff481887582b71b03f Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Sun, 5 Jul 2026 23:19:40 -0700 Subject: [PATCH 32/86] Fix for frustrating error in 'just docs' locally fatal: refusing to fetch into branch 'refs/heads/main' checked out at '/Users/simon/Dropbox/dev/sqlite-utils' I'm not going to pretend to understand this fix, it works. --- docs/conf.py | 58 ++++++++++++++++++++++++++++++++++++++++++++-------- 1 file changed, 49 insertions(+), 9 deletions(-) diff --git a/docs/conf.py b/docs/conf.py index 04a2301..4f29b39 100644 --- a/docs/conf.py +++ b/docs/conf.py @@ -1,8 +1,10 @@ #!/usr/bin/env python3 # -*- coding: utf-8 -*- -from subprocess import Popen, PIPE -from beanbag_docutils.sphinx.ext.github import github_linkcode_resolve +import inspect +from pathlib import Path +from subprocess import Popen, PIPE, check_output +import sys # This file is execfile()d with the current directory set to its # containing dir. @@ -45,14 +47,52 @@ extlinks = { } +def _linkcode_git_ref(): + try: + return check_output(["git", "rev-parse", "HEAD"]).decode("utf8").strip() + except Exception: + return "main" + + def linkcode_resolve(domain, info): - return github_linkcode_resolve( - domain=domain, - info=info, - allowed_module_names=["sqlite_utils"], - github_org_id="simonw", - github_repo_id="sqlite-utils", - branch="main", + if domain != "py": + return None + + module_name = info.get("module") + if not module_name or module_name.split(".")[0] != "sqlite_utils": + return None + + module = sys.modules.get(module_name) + if module is None: + return None + + obj = module + for part in info.get("fullname", "").split("."): + obj = getattr(obj, part, None) + if obj is None: + return None + + if isinstance(obj, property): + obj = obj.fget + + try: + obj = inspect.unwrap(obj) + source_file = inspect.getsourcefile(obj) + _, line_number = inspect.getsourcelines(obj) + except Exception: + return None + + if source_file is None: + return None + + try: + filename = Path(source_file).resolve().relative_to(Path(__file__).parent.parent) + except ValueError: + return None + + return ( + "https://github.com/simonw/sqlite-utils/blob/" + f"{_linkcode_git_ref()}/{filename}#L{line_number}" ) From d516e585433792c2bc2e99f2bc6eaa3227c462ae Mon Sep 17 00:00:00 2001 From: Johnson K C Date: Mon, 6 Jul 2026 09:54:17 -0700 Subject: [PATCH 33/86] Honor --no-headers for --fmt and --table output (#566) (#751) Co-authored-by: Claude Opus 4.8 (1M context) --- docs/cli-reference.rst | 16 +++++++-------- sqlite_utils/cli.py | 18 ++++++++++++++--- tests/test_cli.py | 44 ++++++++++++++++++++++++++++++++++++++++++ 3 files changed, 67 insertions(+), 11 deletions(-) diff --git a/docs/cli-reference.rst b/docs/cli-reference.rst index cbac48b..1eeaabf 100644 --- a/docs/cli-reference.rst +++ b/docs/cli-reference.rst @@ -116,7 +116,7 @@ See :ref:`cli_query`. --arrays Output rows as arrays instead of objects --csv Output CSV --tsv Output TSV - --no-headers Omit CSV headers + --no-headers Omit headers from CSV/TSV and table/--fmt output -t, --table Output as a formatted table --fmt TEXT Table format - one of asciidoc, colon_grid, double_grid, double_outline, fancy_grid, @@ -185,7 +185,7 @@ See :ref:`cli_memory`. --arrays Output rows as arrays instead of objects --csv Output CSV --tsv Output TSV - --no-headers Omit CSV headers + --no-headers Omit headers from CSV/TSV and table/--fmt output -t, --table Output as a formatted table --fmt TEXT Table format - one of asciidoc, colon_grid, double_grid, double_outline, fancy_grid, @@ -422,7 +422,7 @@ See :ref:`cli_search`. --arrays Output rows as arrays instead of objects --csv Output CSV --tsv Output TSV - --no-headers Omit CSV headers + --no-headers Omit headers from CSV/TSV and table/--fmt output -t, --table Output as a formatted table --fmt TEXT Table format - one of asciidoc, colon_grid, double_grid, double_outline, fancy_grid, fancy_outline, @@ -688,7 +688,7 @@ See :ref:`cli_tables`. --arrays Output rows as arrays instead of objects --csv Output CSV --tsv Output TSV - --no-headers Omit CSV headers + --no-headers Omit headers from CSV/TSV and table/--fmt output -t, --table Output as a formatted table --fmt TEXT Table format - one of asciidoc, colon_grid, double_grid, double_outline, fancy_grid, fancy_outline, @@ -730,7 +730,7 @@ See :ref:`cli_views`. --arrays Output rows as arrays instead of objects --csv Output CSV --tsv Output TSV - --no-headers Omit CSV headers + --no-headers Omit headers from CSV/TSV and table/--fmt output -t, --table Output as a formatted table --fmt TEXT Table format - one of asciidoc, colon_grid, double_grid, double_outline, fancy_grid, fancy_outline, @@ -777,7 +777,7 @@ See :ref:`cli_rows`. --arrays Output rows as arrays instead of objects --csv Output CSV --tsv Output TSV - --no-headers Omit CSV headers + --no-headers Omit headers from CSV/TSV and table/--fmt output -t, --table Output as a formatted table --fmt TEXT Table format - one of asciidoc, colon_grid, double_grid, double_outline, fancy_grid, @@ -817,7 +817,7 @@ See :ref:`cli_triggers`. --arrays Output rows as arrays instead of objects --csv Output CSV --tsv Output TSV - --no-headers Omit CSV headers + --no-headers Omit headers from CSV/TSV and table/--fmt output -t, --table Output as a formatted table --fmt TEXT Table format - one of asciidoc, colon_grid, double_grid, double_outline, fancy_grid, fancy_outline, @@ -857,7 +857,7 @@ See :ref:`cli_indexes`. --arrays Output rows as arrays instead of objects --csv Output CSV --tsv Output TSV - --no-headers Omit CSV headers + --no-headers Omit headers from CSV/TSV and table/--fmt output -t, --table Output as a formatted table --fmt TEXT Table format - one of asciidoc, colon_grid, double_grid, double_outline, fancy_grid, fancy_outline, diff --git a/sqlite_utils/cli.py b/sqlite_utils/cli.py index d5a4b1e..8918439 100644 --- a/sqlite_utils/cli.py +++ b/sqlite_utils/cli.py @@ -111,7 +111,11 @@ def output_options(fn): ), click.option("--csv", is_flag=True, help="Output CSV"), click.option("--tsv", is_flag=True, help="Output TSV"), - click.option("--no-headers", is_flag=True, help="Omit CSV headers"), + click.option( + "--no-headers", + is_flag=True, + help="Omit headers from CSV/TSV and table/--fmt output", + ), click.option( "-t", "--table", is_flag=True, help="Output as a formatted table" ), @@ -240,7 +244,13 @@ def tables( yield row if table or fmt: - print(tabulate.tabulate(_iter(), headers=headers, tablefmt=fmt or "simple")) + print( + tabulate.tabulate( + _iter(), + headers=() if no_headers else headers, + tablefmt=fmt or "simple", + ) + ) elif csv or tsv: writer = csv_std.writer(sys.stdout, dialect="excel-tab" if tsv else "excel") if not no_headers: @@ -2145,7 +2155,9 @@ def _execute_query( elif fmt or table: print( tabulate.tabulate( - list(cursor), headers=headers, tablefmt=fmt or "simple" + list(cursor), + headers=() if no_headers else headers, + tablefmt=fmt or "simple", ) ) elif csv or tsv: diff --git a/tests/test_cli.py b/tests/test_cli.py index f19a00c..d11c42f 100644 --- a/tests/test_cli.py +++ b/tests/test_cli.py @@ -195,6 +195,50 @@ def test_output_table(db_path, options, expected): assert expected == result.output.strip() +@pytest.mark.parametrize( + "fmt_option", [["--fmt", "simple"], ["-t"], ["--fmt", "github"]] +) +def test_output_table_no_headers(db_path, fmt_option): + # --no-headers should omit the header row from --fmt/--table output too, not + # just from --csv/--tsv (#566). Previously the flag was silently ignored for + # tabulate formats and the column names were always printed. + db = Database(db_path) + with db.conn: + db["dogs"].insert_all( + [ + {"id": 1, "name": "Cleo", "age": 4}, + {"id": 2, "name": "Pancakes", "age": 2}, + ] + ) + sql = "select id, name, age from dogs order by id" + + with_headers = CliRunner().invoke(cli.cli, ["query", db_path, sql] + fmt_option) + without_headers = CliRunner().invoke( + cli.cli, ["query", db_path, sql] + fmt_option + ["--no-headers"] + ) + assert with_headers.exit_code == 0 + assert without_headers.exit_code == 0 + + # The column names appear when headers are shown, and must not appear at all + # once --no-headers is passed. + assert "name" in with_headers.output + for header in ("id", "name", "age"): + assert ( + header not in without_headers.output + ), f"header {header!r} leaked into --no-headers output" + # The data is still all present. + for value in ("Cleo", "Pancakes", "1", "2", "4"): + assert value in without_headers.output + + # The rows command shares the same code path. + rows_no_headers = CliRunner().invoke( + cli.cli, ["rows", db_path, "dogs"] + fmt_option + ["--no-headers"] + ) + assert rows_no_headers.exit_code == 0 + assert "name" not in rows_no_headers.output + assert "Cleo" in rows_no_headers.output + + def test_create_index(db_path): db = Database(db_path) assert [] == db["Gosh"].indexes From 815b6a7d3dcd262b9ef415482d197a2141b60427 Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Mon, 6 Jul 2026 11:10:07 -0700 Subject: [PATCH 34/86] JSON output no longer escapes non-ASCII characters, new --ascii option (#777) Closes #625 Co-Authored-By: Claude Fable 5 Claude-Session: https://claude.ai/code/session_01JaHan1NhaTRAxJ9LQtSLf9 --- docs/changelog.rst | 7 +++++++ docs/cli-reference.rst | 11 +++++++++++ docs/cli.rst | 27 +++++++++++++++++++++++++++ sqlite_utils/cli.py | 35 +++++++++++++++++++++++++++++------ tests/test_cli.py | 28 ++++++++++++++++++++++++++++ tests/test_cli_convert.py | 19 +++++++++++++++++++ 6 files changed, 121 insertions(+), 6 deletions(-) diff --git a/docs/changelog.rst b/docs/changelog.rst index 085d9d5..556e497 100644 --- a/docs/changelog.rst +++ b/docs/changelog.rst @@ -4,6 +4,13 @@ Changelog =========== +.. _unreleased: + +Unreleased +---------- + +- JSON output from the command-line tool no longer escapes non-ASCII characters, so ``sqlite-utils data.db "select '日本語' as text"`` now outputs ``[{"text": "日本語"}]``. This matches how values were already stored by ``insert`` and how CSV/TSV output already behaved. A new ``--ascii`` option restores the previous behavior of escaping non-ASCII characters, for output destinations that cannot handle UTF-8 - see :ref:`cli_query_json_ascii`. The option is available on the ``query``, ``rows``, ``search``, ``tables``, ``views``, ``triggers``, ``indexes`` and ``memory`` commands. The ``convert --multi --dry-run`` preview and ``plugins`` output also no longer escape non-ASCII characters. (:issue:`625`) + .. _v4_0rc3: 4.0rc3 (2026-07-05) diff --git a/docs/cli-reference.rst b/docs/cli-reference.rst index 1eeaabf..49eba52 100644 --- a/docs/cli-reference.rst +++ b/docs/cli-reference.rst @@ -129,6 +129,8 @@ See :ref:`cli_query`. simple_outline, textile, tsv, unsafehtml, youtrack --json-cols Detect JSON cols and output them as JSON, not escaped strings + --ascii Escape non-ASCII characters in JSON output as + \uXXXX -r, --raw Raw output, first column of first row --raw-lines Raw output, first column of each row -p, --param ... Named :parameters for SQL query @@ -198,6 +200,8 @@ See :ref:`cli_memory`. simple_outline, textile, tsv, unsafehtml, youtrack --json-cols Detect JSON cols and output them as JSON, not escaped strings + --ascii Escape non-ASCII characters in JSON output as + \uXXXX -r, --raw Raw output, first column of first row --raw-lines Raw output, first column of each row -p, --param ... Named :parameters for SQL query @@ -435,6 +439,7 @@ See :ref:`cli_search`. youtrack --json-cols Detect JSON cols and output them as JSON, not escaped strings + --ascii Escape non-ASCII characters in JSON output as \uXXXX --load-extension TEXT Path to SQLite extension, with optional :entrypoint -h, --help Show this message and exit. @@ -701,6 +706,7 @@ See :ref:`cli_tables`. youtrack --json-cols Detect JSON cols and output them as JSON, not escaped strings + --ascii Escape non-ASCII characters in JSON output as \uXXXX --columns Include list of columns for each table --schema Include schema for each table --load-extension TEXT Path to SQLite extension, with optional :entrypoint @@ -743,6 +749,7 @@ See :ref:`cli_views`. youtrack --json-cols Detect JSON cols and output them as JSON, not escaped strings + --ascii Escape non-ASCII characters in JSON output as \uXXXX --columns Include list of columns for each view --schema Include schema for each view --load-extension TEXT Path to SQLite extension, with optional :entrypoint @@ -790,6 +797,8 @@ See :ref:`cli_rows`. simple_outline, textile, tsv, unsafehtml, youtrack --json-cols Detect JSON cols and output them as JSON, not escaped strings + --ascii Escape non-ASCII characters in JSON output as + \uXXXX --load-extension TEXT Path to SQLite extension, with optional :entrypoint -h, --help Show this message and exit. @@ -830,6 +839,7 @@ See :ref:`cli_triggers`. youtrack --json-cols Detect JSON cols and output them as JSON, not escaped strings + --ascii Escape non-ASCII characters in JSON output as \uXXXX --load-extension TEXT Path to SQLite extension, with optional :entrypoint -h, --help Show this message and exit. @@ -870,6 +880,7 @@ See :ref:`cli_indexes`. youtrack --json-cols Detect JSON cols and output them as JSON, not escaped strings + --ascii Escape non-ASCII characters in JSON output as \uXXXX --load-extension TEXT Path to SQLite extension, with optional :entrypoint -h, --help Show this message and exit. diff --git a/docs/cli.rst b/docs/cli.rst index 84a65d9..957ee64 100644 --- a/docs/cli.rst +++ b/docs/cli.rst @@ -111,6 +111,33 @@ If you want to pretty-print the output further, you can pipe it through ``python } ] +.. _cli_query_json_ascii: + +Unicode characters in JSON +~~~~~~~~~~~~~~~~~~~~~~~~~~ + +JSON output includes unicode characters directly, without escaping them: + +.. code-block:: bash + + sqlite-utils dogs.db "select '日本語' as text" + +.. code-block:: output + + [{"text": "日本語"}] + +Use ``--ascii`` to escape non-ASCII characters as ``\uXXXX`` sequences instead: + +.. code-block:: bash + + sqlite-utils dogs.db "select '日本語' as text" --ascii + +.. code-block:: output + + [{"text": "\u65e5\u672c\u8a9e"}] + +The ``--ascii`` option can help on systems that cannot display or process UTF-8, such as Windows consoles using a legacy code page. On Windows, setting the ``PYTHONUTF8=1`` environment variable is an alternative fix for ``UnicodeEncodeError`` crashes when redirecting output to a file. + .. _cli_query_binary_json: Binary data in JSON diff --git a/sqlite_utils/cli.py b/sqlite_utils/cli.py index 8918439..8ee2091 100644 --- a/sqlite_utils/cli.py +++ b/sqlite_utils/cli.py @@ -131,6 +131,13 @@ def output_options(fn): is_flag=True, default=False, ), + click.option( + "--ascii", + "ascii_", + help="Escape non-ASCII characters in JSON output as \\uXXXX", + is_flag=True, + default=False, + ), ) ): fn = decorator(fn) @@ -199,6 +206,7 @@ def tables( table, fmt, json_cols, + ascii_, columns, schema, load_extension, @@ -258,7 +266,7 @@ def tables( for row in _iter(): writer.writerow(row) else: - for line in output_rows(_iter(), headers, nl, arrays, json_cols): + for line in output_rows(_iter(), headers, nl, arrays, json_cols, ascii_): click.echo(line) @@ -296,6 +304,7 @@ def views( table, fmt, json_cols, + ascii_, columns, schema, load_extension, @@ -321,6 +330,7 @@ def views( table=table, fmt=fmt, json_cols=json_cols, + ascii_=ascii_, columns=columns, schema=schema, load_extension=load_extension, @@ -1857,6 +1867,7 @@ def query( table, fmt, json_cols, + ascii_, raw, raw_lines, param, @@ -1895,6 +1906,7 @@ def query( nl, arrays, json_cols, + ascii_, ) @@ -1969,6 +1981,7 @@ def memory( table, fmt, json_cols, + ascii_, raw, raw_lines, param, @@ -2108,6 +2121,7 @@ def memory( nl, arrays, json_cols, + ascii_, ) @@ -2125,6 +2139,7 @@ def _execute_query( nl, arrays, json_cols, + ascii_, ): with db.conn: try: @@ -2167,7 +2182,7 @@ def _execute_query( for row in cursor: writer.writerow(row) else: - for line in output_rows(cursor, headers, nl, arrays, json_cols): + for line in output_rows(cursor, headers, nl, arrays, json_cols, ascii_): click.echo(line) @@ -2211,6 +2226,7 @@ def search( table, fmt, json_cols, + ascii_, load_extension, ): """Execute a full-text search against this table @@ -2257,6 +2273,7 @@ def search( table=table, fmt=fmt, json_cols=json_cols, + ascii_=ascii_, param=[("query", q)], load_extension=load_extension, ) @@ -2317,6 +2334,7 @@ def rows( table, fmt, json_cols, + ascii_, load_extension, ): """Output all rows in the specified table @@ -2351,6 +2369,7 @@ def rows( fmt=fmt, param=param, json_cols=json_cols, + ascii_=ascii_, load_extension=load_extension, ) @@ -2377,6 +2396,7 @@ def triggers( table, fmt, json_cols, + ascii_, load_extension, ): """Show triggers configured in this database @@ -2406,6 +2426,7 @@ def triggers( table=table, fmt=fmt, json_cols=json_cols, + ascii_=ascii_, load_extension=load_extension, ) @@ -2434,6 +2455,7 @@ def indexes( table, fmt, json_cols, + ascii_, load_extension, ): """Show indexes for the whole database or specific tables @@ -2475,6 +2497,7 @@ def indexes( table=table, fmt=fmt, json_cols=json_cols, + ascii_=ascii_, load_extension=load_extension, ) @@ -3112,7 +3135,7 @@ def convert( if multi: def preview(v): - return json.dumps(fn(v), default=repr) if v else v + return json.dumps(fn(v), default=repr, ensure_ascii=False) if v else v else: @@ -3458,7 +3481,7 @@ def migrate(db_path, migrations, stop_before, list_, verbose): @cli.command(name="plugins") def plugins_list(): "List installed plugins" - click.echo(json.dumps(get_plugins(), indent=2)) + click.echo(json.dumps(get_plugins(), indent=2, ensure_ascii=False)) ensure_plugins_loaded() @@ -3505,7 +3528,7 @@ FILE_COLUMNS = { } -def output_rows(iterator, headers, nl, arrays, json_cols): +def output_rows(iterator, headers, nl, arrays, json_cols, ascii_=False): # Duplicate column names would collide as dictionary keys, so rename # later occurrences id, id -> id, id_2 - CSV and table output keep # the original duplicate headers since they never build dictionaries @@ -3526,7 +3549,7 @@ def output_rows(iterator, headers, nl, arrays, json_cols): data = dict(zip(headers, data)) line = "{firstchar}{serialized}{maybecomma}{lastchar}".format( firstchar=("[" if first else " ") if not nl else "", - serialized=json.dumps(data, default=json_binary), + serialized=json.dumps(data, default=json_binary, ensure_ascii=ascii_), maybecomma="," if (not nl and not is_last) else "", lastchar="]" if (is_last and not nl) else "", ) diff --git a/tests/test_cli.py b/tests/test_cli.py index d11c42f..9e17969 100644 --- a/tests/test_cli.py +++ b/tests/test_cli.py @@ -1047,6 +1047,34 @@ def test_query_json_with_json_cols(db_path): assert expected == result_rows.output.strip() +def test_query_json_unicode_not_escaped_by_default(db_path): + db = Database(db_path) + with db.conn: + db["text"].insert({"id": 1, "text": "Japanese 日本語"}, pk="id") + result = CliRunner().invoke(cli.cli, [db_path, "select id, text from text"]) + assert result.exit_code == 0 + assert result.output.strip() == '[{"id": 1, "text": "Japanese 日本語"}]' + # Same for --nl + result = CliRunner().invoke(cli.cli, [db_path, "select id, text from text", "--nl"]) + assert result.exit_code == 0 + assert result.output.strip() == '{"id": 1, "text": "Japanese 日本語"}' + + +@pytest.mark.parametrize("command", ["query", "rows"]) +def test_query_json_ascii_option(db_path, command): + db = Database(db_path) + with db.conn: + db["text"].insert({"id": 1, "text": "Japanese 日本語"}, pk="id") + if command == "query": + args = [db_path, "select id, text from text", "--ascii"] + else: + args = ["rows", db_path, "text", "--ascii"] + result = CliRunner().invoke(cli.cli, args) + assert result.exit_code == 0 + expected = '[{"id": 1, "text": "Japanese ' + "\\u65e5\\u672c\\u8a9e" + '"}]' + assert result.output.strip() == expected + + @pytest.mark.parametrize( "content,is_binary", [(b"\x00\x0fbinary", True), ("this is text", False), (1, False), (1.5, False)], diff --git a/tests/test_cli_convert.py b/tests/test_cli_convert.py index 443e72c..6c3f5c5 100644 --- a/tests/test_cli_convert.py +++ b/tests/test_cli_convert.py @@ -215,6 +215,25 @@ def test_convert_multi_dryrun(test_db_and_path): ) +def test_convert_multi_dryrun_unicode_not_escaped(test_db_and_path): + db_path = test_db_and_path[1] + result = CliRunner().invoke( + cli.cli, + [ + "convert", + db_path, + "example", + "dt", + "{'text': 'Japanese 日本語'}", + "--dry-run", + "--multi", + ], + ) + assert result.exit_code == 0 + # Preview should match what jsonify_if_needed() would actually store + assert '{"text": "Japanese 日本語"}' in result.output + + @pytest.mark.parametrize("drop", (True, False)) def test_convert_output_column(test_db_and_path, drop): db, db_path = test_db_and_path From 6225eba5c8daf97a5e459e9fcce2e46bcd022f50 Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Mon, 6 Jul 2026 20:13:52 -0700 Subject: [PATCH 35/86] Raise InvalidColumns on insert(pk="invalid") Closes #732 --- docs/changelog.rst | 1 + sqlite_utils/db.py | 18 ++++++++++++++++++ tests/test_create.py | 16 ++++++++++++++++ 3 files changed, 35 insertions(+) diff --git a/docs/changelog.rst b/docs/changelog.rst index 556e497..a68a9b6 100644 --- a/docs/changelog.rst +++ b/docs/changelog.rst @@ -10,6 +10,7 @@ Unreleased ---------- - JSON output from the command-line tool no longer escapes non-ASCII characters, so ``sqlite-utils data.db "select '日本語' as text"`` now outputs ``[{"text": "日本語"}]``. This matches how values were already stored by ``insert`` and how CSV/TSV output already behaved. A new ``--ascii`` option restores the previous behavior of escaping non-ASCII characters, for output destinations that cannot handle UTF-8 - see :ref:`cli_query_json_ascii`. The option is available on the ``query``, ``rows``, ``search``, ``tables``, ``views``, ``triggers``, ``indexes`` and ``memory`` commands. The ``convert --multi --dry-run`` preview and ``plugins`` output also no longer escape non-ASCII characters. (:issue:`625`) +- ``table.insert_all(..., pk=...)`` now raises ``InvalidColumns`` if ``pk=`` names columns that do not exist in an existing table. Previously this behaved inconsistently, with single-row inserts raising a ``KeyError`` while other row counts succeeded. (:issue:`732`) .. _v4_0rc3: diff --git a/sqlite_utils/db.py b/sqlite_utils/db.py index 0807892..68aee3e 100644 --- a/sqlite_utils/db.py +++ b/sqlite_utils/db.py @@ -4229,6 +4229,24 @@ class Table(Queryable): if hash_id: pk = hash_id + if pk and not hash_id and self.exists(): + pk_cols = [pk] if isinstance(pk, str) else list(pk) + existing_columns = self.columns_dict + missing_pk_cols = [ + col + for col in pk_cols + if resolve_casing(col, existing_columns) not in existing_columns + ] + if missing_pk_cols: + raise InvalidColumns( + "Invalid primary key column{} {} for table {} with columns {}".format( + "s" if len(missing_pk_cols) > 1 else "", + missing_pk_cols, + self.name, + list(existing_columns), + ) + ) + if ignore and replace: raise ValueError("Use either ignore=True or replace=True, not both") all_columns = [] diff --git a/tests/test_create.py b/tests/test_create.py index decefcf..7fbd7d5 100644 --- a/tests/test_create.py +++ b/tests/test_create.py @@ -3,6 +3,7 @@ from sqlite_utils.db import ( Database, DescIndex, AlterError, + InvalidColumns, NoObviousTable, OperationalError, ForeignKey, @@ -925,6 +926,21 @@ def test_insert_thousands_adds_extra_columns_after_first_100_with_alter(fresh_db assert rows == [{"i": 101, "word": None, "extra": "Should trigger ALTER"}] +@pytest.mark.parametrize("num_rows", (0, 1, 2, 3, 10)) +def test_insert_all_pk_not_in_records_raises(fresh_db, num_rows): + # https://github.com/simonw/sqlite-utils/issues/732 + fresh_db.conn.execute("CREATE TABLE t (a TEXT, b INT, PRIMARY KEY (a, b))") + rows = [{"a": "x{}".format(i), "b": i} for i in range(num_rows)] + + with pytest.raises(InvalidColumns) as ex: + fresh_db["t"].insert_all(rows, pk="not_a_column", alter=True) + + assert ex.value.args == ( + "Invalid primary key column ['not_a_column'] for table t with columns ['a', 'b']", + ) + assert fresh_db["t"].count == 0 + + def test_insert_ignore(fresh_db): fresh_db["test"].insert({"id": 1, "bar": 2}, pk="id") # Should raise an error if we try this again From 221774f25a53c76874b28af1c2dd170c76bbd9f1 Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Mon, 6 Jul 2026 20:22:08 -0700 Subject: [PATCH 36/86] Fix for table.insert(..., pk=..., ignore=True), closes #554 --- docs/changelog.rst | 1 + sqlite_utils/db.py | 56 ++++++++++++++++++++++++++++++++++++-------- tests/test_create.py | 18 ++++++++++++++ 3 files changed, 65 insertions(+), 10 deletions(-) diff --git a/docs/changelog.rst b/docs/changelog.rst index a68a9b6..2b2e5cf 100644 --- a/docs/changelog.rst +++ b/docs/changelog.rst @@ -11,6 +11,7 @@ Unreleased - JSON output from the command-line tool no longer escapes non-ASCII characters, so ``sqlite-utils data.db "select '日本語' as text"`` now outputs ``[{"text": "日本語"}]``. This matches how values were already stored by ``insert`` and how CSV/TSV output already behaved. A new ``--ascii`` option restores the previous behavior of escaping non-ASCII characters, for output destinations that cannot handle UTF-8 - see :ref:`cli_query_json_ascii`. The option is available on the ``query``, ``rows``, ``search``, ``tables``, ``views``, ``triggers``, ``indexes`` and ``memory`` commands. The ``convert --multi --dry-run`` preview and ``plugins`` output also no longer escape non-ASCII characters. (:issue:`625`) - ``table.insert_all(..., pk=...)`` now raises ``InvalidColumns`` if ``pk=`` names columns that do not exist in an existing table. Previously this behaved inconsistently, with single-row inserts raising a ``KeyError`` while other row counts succeeded. (:issue:`732`) +- Fixed an ``IndexError`` from ``table.insert(..., pk=..., ignore=True)`` when an ignored insert followed writes to another table on the same connection. ``last_pk`` is now populated from the explicit primary key value instead of looking up a stale ``lastrowid``. (:issue:`554`) .. _v4_0rc3: diff --git a/sqlite_utils/db.py b/sqlite_utils/db.py index 68aee3e..916d2f0 100644 --- a/sqlite_utils/db.py +++ b/sqlite_utils/db.py @@ -4386,18 +4386,54 @@ class Table(Queryable): if num_records_processed == 1: # For an insert we need to use result.lastrowid if not upsert and result is not None: - self.last_rowid = result.lastrowid - if (hash_id or pk) and self.last_rowid: - # Set self.last_pk to the pk(s) for that rowid - row = list(self.rows_where("rowid = ?", [self.last_rowid]))[0] - if hash_id: - self.last_pk = row[hash_id] - elif isinstance(pk, str): - self.last_pk = row[resolve_casing(pk, row)] + ignored_insert = ignore and result.rowcount == 0 + if ignored_insert: + if list_mode: + first_record_list = cast(Sequence[Any], first_record) + if hash_id: + pass + elif isinstance(pk, str): + pk_index = column_names.index( + resolve_casing(pk, column_names) + ) + self.last_pk = first_record_list[pk_index] + elif pk: + self.last_pk = tuple( + first_record_list[ + column_names.index(resolve_casing(p, column_names)) + ] + for p in pk + ) else: - self.last_pk = tuple(row[resolve_casing(p, row)] for p in pk) + first_record_dict = cast(Dict[str, Any], first_record) + if hash_id: + self.last_pk = hash_record( + first_record_dict, hash_id_columns + ) + elif isinstance(pk, str): + self.last_pk = first_record_dict[ + resolve_casing(pk, first_record_dict) + ] + elif pk: + self.last_pk = tuple( + first_record_dict[resolve_casing(p, first_record_dict)] + for p in pk + ) else: - self.last_pk = self.last_rowid + self.last_rowid = result.lastrowid + if (hash_id or pk) and self.last_rowid: + # Set self.last_pk to the pk(s) for that rowid + row = list(self.rows_where("rowid = ?", [self.last_rowid]))[0] + if hash_id: + self.last_pk = row[hash_id] + elif isinstance(pk, str): + self.last_pk = row[resolve_casing(pk, row)] + else: + self.last_pk = tuple( + row[resolve_casing(p, row)] for p in pk + ) + else: + self.last_pk = self.last_rowid else: # For an upsert use first_record from earlier if list_mode: diff --git a/tests/test_create.py b/tests/test_create.py index 7fbd7d5..2788920 100644 --- a/tests/test_create.py +++ b/tests/test_create.py @@ -953,6 +953,24 @@ def test_insert_ignore(fresh_db): assert rows == [{"id": 1, "bar": 2}] +def test_insert_ignore_with_pk_after_other_table_insert(fresh_db): + # https://github.com/simonw/sqlite-utils/issues/554 + user = {"id": "abc", "name": "david"} + + fresh_db["users"].insert(user, pk="id") + fresh_db["comments"].insert_all( + [ + {"id": "def", "text": "ok"}, + {"id": "ghi", "text": "great"}, + ], + ) + + table = fresh_db["users"].insert(user, pk="id", ignore=True) + + assert table.last_pk == "abc" + assert list(fresh_db["users"].rows) == [user] + + def test_insert_hash_id(fresh_db): dogs = fresh_db["dogs"] id = dogs.insert({"name": "Cleo", "twitter": "cleopaws"}, hash_id="id").last_pk From b8aa1368571b09f765af37783579224f74812bf3 Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Mon, 6 Jul 2026 20:37:02 -0700 Subject: [PATCH 37/86] Remove beanbag-docutils - we stopped needing that in 5f81752 --- pyproject.toml | 1 - 1 file changed, 1 deletion(-) diff --git a/pyproject.toml b/pyproject.toml index bff9390..01f1735 100644 --- a/pyproject.toml +++ b/pyproject.toml @@ -53,7 +53,6 @@ dev = [ "tabulate>=0.10.0", ] docs = [ - "beanbag-docutils>=2.0", "codespell", "furo", "pygments-csv-lexer", From 2616dec7957a366da3260cf87eb79c78bae39d14 Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Mon, 6 Jul 2026 20:39:29 -0700 Subject: [PATCH 38/86] .extract() and .lookup() no longer extract null values, closes #186 - table.extract() and the sqlite-utils extract command now skip rows where every extracted column is null: the new foreign key column is left null and no all-null record is added to the lookup table. Rows with at least one non-null extracted column are extracted as before. - The extracts= option to insert() and friends keeps None values as null instead of creating a lookup record for them - previously each insert batch added a duplicate null row to the lookup table. - table.lookup() compares values using IS rather than = so lookup values containing None match existing rows correctly, instead of inserting a duplicate row on every call. Co-Authored-By: Claude Fable 5 --- docs/changelog.rst | 3 ++ docs/cli.rst | 2 ++ docs/python-api.rst | 6 ++++ docs/upgrading.rst | 2 ++ sqlite_utils/db.py | 18 +++++++--- tests/test_extract.py | 80 ++++++++++++++++++++++++++++++++++++++++-- tests/test_extracts.py | 48 +++++++++++++++++++++++++ tests/test_lookup.py | 24 +++++++++++++ 8 files changed, 176 insertions(+), 7 deletions(-) diff --git a/docs/changelog.rst b/docs/changelog.rst index 2b2e5cf..e773b11 100644 --- a/docs/changelog.rst +++ b/docs/changelog.rst @@ -9,6 +9,9 @@ Unreleased ---------- +- **Breaking change**: ``table.extract()`` - and the ``sqlite-utils extract`` command - no longer extract rows where every extracted column is ``null``. Those rows now keep a ``null`` value in the new foreign key column instead of pointing at an all-``null`` record in the lookup table. When extracting multiple columns, rows are still extracted if at least one of the columns has a value. (:issue:`186`) +- The ``extracts=`` option to ``table.insert()`` and friends no longer creates a lookup table record for ``None`` values - the column value stays ``null``. Previously every batch of inserted rows containing a ``None`` value would add a duplicate ``null`` record to the lookup table. +- Fixed a bug where ``table.lookup()`` inserted a duplicate row on every call if any of the lookup values were ``None``. Lookup values are now compared using ``IS`` so that ``None`` values match existing rows correctly. - JSON output from the command-line tool no longer escapes non-ASCII characters, so ``sqlite-utils data.db "select '日本語' as text"`` now outputs ``[{"text": "日本語"}]``. This matches how values were already stored by ``insert`` and how CSV/TSV output already behaved. A new ``--ascii`` option restores the previous behavior of escaping non-ASCII characters, for output destinations that cannot handle UTF-8 - see :ref:`cli_query_json_ascii`. The option is available on the ``query``, ``rows``, ``search``, ``tables``, ``views``, ``triggers``, ``indexes`` and ``memory`` commands. The ``convert --multi --dry-run`` preview and ``plugins`` output also no longer escape non-ASCII characters. (:issue:`625`) - ``table.insert_all(..., pk=...)`` now raises ``InvalidColumns`` if ``pk=`` names columns that do not exist in an existing table. Previously this behaved inconsistently, with single-row inserts raising a ``KeyError`` while other row counts succeeded. (:issue:`732`) - Fixed an ``IndexError`` from ``table.insert(..., pk=..., ignore=True)`` when an ignored insert followed writes to another table on the same connection. ``last_pk`` is now populated from the explicit primary key value instead of looking up a stale ``lastrowid``. (:issue:`554`) diff --git a/docs/cli.rst b/docs/cli.rst index 957ee64..1f95bbd 100644 --- a/docs/cli.rst +++ b/docs/cli.rst @@ -2286,6 +2286,8 @@ The ``sqlite-utils extract`` command can be used to extract specified columns in Take a look at the Python API documentation for :ref:`python_api_extract` for a detailed description of how this works, including examples of table schemas before and after running an extraction operation. +Rows where every extracted column is ``null`` are not extracted - those rows get a ``null`` value in their new foreign key column and no record is created for them in the lookup table. + The command takes a database, table and one or more columns that should be extracted. To extract the ``species`` column from the ``trees`` table you would run: .. code-block:: bash diff --git a/docs/python-api.rst b/docs/python-api.rst index 7ac3951..0e61cd4 100644 --- a/docs/python-api.rst +++ b/docs/python-api.rst @@ -1279,6 +1279,8 @@ To create a species record with a note on when it was first seen, you can use th The first time this is called the record will be created for ``name="Palm"``. Any subsequent calls with that name will ignore the second argument, even if it includes different values. +``None`` values are matched correctly: calling ``.lookup()`` a second time with the same values will return the primary key of the existing row even if some of those values are ``None``. + ``.lookup()`` also accepts keyword arguments, which are passed through to the :ref:`insert() method ` and can be used to influence the shape of the created table. Supported parameters are: - ``pk`` - which defaults to ``id`` @@ -1324,6 +1326,8 @@ To extract the ``species`` column out to a separate ``Species`` table, you can d "species": "Common Juniper" }, extracts={"species": "Species"}) +``None`` values are not extracted: no record is created for them in the lookup table and the column value stays ``null``. + .. _python_api_m2m: Working with many-to-many relationships @@ -2022,6 +2026,8 @@ This produces a lookup table like so: "latin" TEXT ) +Rows where every extracted column is ``null`` are not extracted: no record is created for them in the lookup table and their foreign key column is left as ``null``. When extracting multiple columns, rows where at least one of the extracted columns has a value will be extracted as usual. + .. _python_api_hash: Setting an ID based on the hash of the row contents diff --git a/docs/upgrading.rst b/docs/upgrading.rst index 09a9e2f..88e987d 100644 --- a/docs/upgrading.rst +++ b/docs/upgrading.rst @@ -77,6 +77,8 @@ Python API changes **table.convert() no longer skips falsey values.** Matching the CLI change above, ``table.convert()`` now converts every value. The ``skip_false`` parameter has been removed - previously it defaulted to ``True``, skipping empty strings and other falsey values. +**Null values are no longer extracted into lookup tables.** ``table.extract()`` and the ``sqlite-utils extract`` command leave rows alone if every extracted column is ``null`` - the new foreign key column is left as ``null`` instead of pointing at an all-``null`` record in the lookup table. The ``extracts=`` insert option similarly keeps ``None`` values as ``null``. Relatedly, ``table.lookup()`` now compares values using ``IS`` so that looking up a value containing ``None`` returns the existing matching row - previously it inserted a duplicate row on every call. + **ensure_autocommit_off() is now ensure_autocommit_on().** The ``db.ensure_autocommit_off()`` context manager has been renamed to ``db.ensure_autocommit_on()``. The old name described the opposite of what the method did: it temporarily puts the connection into driver-level autocommit mode (by setting ``isolation_level = None``), so that statements such as ``PRAGMA journal_mode=wal`` can run outside of an implicit transaction. The behavior is unchanged - update any calls to use the new name. **View.enable_fts() has been removed.** The ``View`` class previously had an ``enable_fts()`` method that existed only to raise ``NotImplementedError`` - full-text search is not supported for views. Calling it now raises ``AttributeError`` like any other missing method. diff --git a/sqlite_utils/db.py b/sqlite_utils/db.py index 916d2f0..0f2924c 100644 --- a/sqlite_utils/db.py +++ b/sqlite_utils/db.py @@ -2841,14 +2841,20 @@ class Table(Queryable): ) lookup_columns = [(rename.get(col) or col) for col in columns] lookup_table.create_index(lookup_columns, unique=True, if_not_exists=True) + # Rows where every extracted column is null are left alone - they + # get a null foreign key and no lookup table record, see #186 + all_columns_are_null = " AND ".join( + "{} IS NULL".format(quote_identifier(c)) for c in columns + ) self.db.execute( - "INSERT OR IGNORE INTO {} ({lookup_columns}) SELECT DISTINCT {table_cols} FROM {}".format( + "INSERT OR IGNORE INTO {} ({lookup_columns}) SELECT DISTINCT {table_cols} FROM {} WHERE NOT ({all_null})".format( quote_identifier(table), quote_identifier(self.name), lookup_columns=", ".join( quote_identifier(c) for c in lookup_columns ), table_cols=", ".join(quote_identifier(c) for c in columns), + all_null=all_columns_are_null, ) ) @@ -2857,7 +2863,7 @@ class Table(Queryable): # And populate it self.db.execute( - "UPDATE {} SET {} = (SELECT id FROM {} WHERE {where})".format( + "UPDATE {} SET {} = (SELECT id FROM {} WHERE {where}) WHERE NOT ({all_null})".format( quote_identifier(self.name), quote_identifier(magic_lookup_column), quote_identifier(table), @@ -2870,6 +2876,7 @@ class Table(Queryable): ) for column in columns ), + all_null=all_columns_are_null, ) ) # Figure out the right column order @@ -3858,7 +3865,7 @@ class Table(Queryable): # Only process extracts if there are any if has_extracts: for i, key in enumerate(all_columns): - if key in extracts: + if key in extracts and record_values[i] is not None: record_values[i] = self.db.table(extracts[key]).lookup( {"value": record_values[i]} ) @@ -3878,7 +3885,7 @@ class Table(Queryable): ), ) ) - if key in extracts: + if key in extracts and value is not None: extract_table = extracts[key] value = self.db.table(extract_table).lookup({"value": value}) record_values.append(value) @@ -4615,8 +4622,9 @@ class Table(Queryable): fold_identifier_case(c) for c in lookup_values } not in unique_column_sets: self.create_index(lookup_values.keys(), unique=True) + # IS rather than = so that null values are matched correctly wheres = [ - "{} = ?".format(quote_identifier(column)) for column in lookup_values + "{} IS ?".format(quote_identifier(column)) for column in lookup_values ] rows = list( self.rows_where( diff --git a/tests/test_extract.py b/tests/test_extract.py index d24c597..1c0fa01 100644 --- a/tests/test_extract.py +++ b/tests/test_extract.py @@ -189,9 +189,85 @@ def test_extract_works_with_null_values(fresh_db): ) assert list(fresh_db["listens"].rows) == [ {"id": 1, "track_title": "foo", "album_id": 1}, - {"id": 2, "track_title": "baz", "album_id": 2}, + {"id": 2, "track_title": "baz", "album_id": None}, ] assert list(fresh_db["albums"].rows) == [ {"id": 1, "album_title": "bar"}, - {"id": 2, "album_title": None}, + ] + + +def test_extract_null_values_single_column(fresh_db): + # https://github.com/simonw/sqlite-utils/issues/186 + fresh_db["species"].insert({"id": 1, "species": "Wolf"}, pk="id") + fresh_db["individuals"].insert_all( + [ + {"id": 10, "name": "Terriana", "species": "Fox"}, + {"id": 11, "name": "Spenidorm", "species": None}, + {"id": 12, "name": "Grantheim", "species": "Wolf"}, + {"id": 13, "name": "Turnutopia", "species": None}, + {"id": 14, "name": "Wargal", "species": "Wolf"}, + ], + pk="id", + ) + fresh_db["individuals"].extract("species") + # No null row should have been added to species + assert list(fresh_db["species"].rows) == [ + {"id": 1, "species": "Wolf"}, + {"id": 2, "species": "Fox"}, + ] + assert list(fresh_db["individuals"].rows) == [ + {"id": 10, "name": "Terriana", "species_id": 2}, + {"id": 11, "name": "Spenidorm", "species_id": None}, + {"id": 12, "name": "Grantheim", "species_id": 1}, + {"id": 13, "name": "Turnutopia", "species_id": None}, + {"id": 14, "name": "Wargal", "species_id": 1}, + ] + + +def test_extract_null_values_multiple_columns(fresh_db): + # A row should be extracted if at least one column is not null - + # only rows where ALL extracted columns are null are left alone + fresh_db["circulation"].insert_all( + [ + {"id": 1, "title": "title one", "creator": "creator one", "year": 2018}, + {"id": 2, "title": "title two", "creator": None, "year": 2019}, + {"id": 3, "title": None, "creator": None, "year": 2020}, + {"id": 4, "title": None, "creator": None, "year": 2021}, + ], + pk="id", + ) + fresh_db["circulation"].extract( + ["title", "creator"], table="books", fk_column="book_id" + ) + assert list(fresh_db["books"].rows) == [ + {"id": 1, "title": "title one", "creator": "creator one"}, + {"id": 2, "title": "title two", "creator": None}, + ] + assert list(fresh_db["circulation"].rows) == [ + {"id": 1, "book_id": 1, "year": 2018}, + {"id": 2, "book_id": 2, "year": 2019}, + {"id": 3, "book_id": None, "year": 2020}, + {"id": 4, "book_id": None, "year": 2021}, + ] + + +def test_extract_null_values_existing_lookup_table_with_null_row(fresh_db): + # Even if the lookup table already contains an all-null row, rows where + # every extracted column is null should keep a null foreign key + fresh_db["species"].insert({"id": 1, "species": None}, pk="id") + fresh_db["individuals"].insert_all( + [ + {"id": 10, "name": "Terriana", "species": "Fox"}, + {"id": 11, "name": "Spenidorm", "species": None}, + ], + pk="id", + ) + fresh_db["individuals"].extract("species") + assert list(fresh_db["species"].rows) == [ + {"id": 1, "species": None}, + {"id": 2, "species": "Fox"}, + ] + assert list(fresh_db["individuals"].rows) == [ + {"id": 10, "name": "Terriana", "species_id": 2}, + {"id": 11, "name": "Spenidorm", "species_id": None}, ] diff --git a/tests/test_extracts.py b/tests/test_extracts.py index eb4f37e..7add79a 100644 --- a/tests/test_extracts.py +++ b/tests/test_extracts.py @@ -67,3 +67,51 @@ def test_extracts(fresh_db, kwargs, expected_table, use_table_factory): {"id": 2, "species_id": 1}, {"id": 3, "species_id": 2}, ] == list(fresh_db["Trees"].rows) + + +def test_extracts_null_values(fresh_db): + # https://github.com/simonw/sqlite-utils/issues/186 + # Null values should stay null, not be extracted into the lookup table + fresh_db["Trees"].insert_all( + [ + {"id": 1, "species_id": "Oak"}, + {"id": 2, "species_id": None}, + {"id": 3, "species_id": "Palm"}, + {"id": 4, "species_id": None}, + ], + extracts={"species_id": "Species"}, + ) + assert list(fresh_db["Species"].rows) == [ + {"id": 1, "value": "Oak"}, + {"id": 2, "value": "Palm"}, + ] + assert list(fresh_db["Trees"].rows) == [ + {"id": 1, "species_id": 1}, + {"id": 2, "species_id": None}, + {"id": 3, "species_id": 2}, + {"id": 4, "species_id": None}, + ] + + +def test_extracts_null_values_list_mode(fresh_db): + # Same as test_extracts_null_values but for list-based records + fresh_db["Trees"].insert_all( + [ + ["id", "species_id"], + [1, "Oak"], + [2, None], + [3, "Palm"], + [4, None], + ], + extracts={"species_id": "Species"}, + ) + assert list(fresh_db["Species"].rows) == [ + {"id": 1, "value": "Oak"}, + {"id": 2, "value": "Palm"}, + ] + assert list(fresh_db["Trees"].rows) == [ + {"id": 1, "species_id": 1}, + {"id": 2, "species_id": None}, + {"id": 3, "species_id": 2}, + {"id": 4, "species_id": None}, + ] diff --git a/tests/test_lookup.py b/tests/test_lookup.py index a36b464..da4f18b 100644 --- a/tests/test_lookup.py +++ b/tests/test_lookup.py @@ -157,3 +157,27 @@ def test_lookup_with_extra_insert_parameters(fresh_db): def test_lookup_new_table_strict(fresh_db, strict): fresh_db["species"].lookup({"name": "Palm"}, strict=strict) assert fresh_db["species"].strict == strict or not fresh_db.supports_strict + + +def test_lookup_null_value_idempotent(fresh_db): + # https://github.com/simonw/sqlite-utils/issues/186 + # Repeated lookups of a null value should return the same row, + # not insert a duplicate row each time + species = fresh_db["species"] + first_id = species.lookup({"name": None}) + second_id = species.lookup({"name": None}) + assert first_id == second_id + assert list(species.rows) == [{"id": first_id, "name": None}] + + +def test_lookup_compound_key_with_null_idempotent(fresh_db): + species = fresh_db["species"] + palm_id = species.lookup({"name": "Palm", "type": None}) + oak_id = species.lookup({"name": "Oak", "type": "Tree"}) + assert palm_id == species.lookup({"name": "Palm", "type": None}) + assert oak_id == species.lookup({"name": "Oak", "type": "Tree"}) + assert palm_id != oak_id + assert list(species.rows) == [ + {"id": palm_id, "name": "Palm", "type": None}, + {"id": oak_id, "name": "Oak", "type": "Tree"}, + ] From f2fbcf60d8bbc95ca996fd078b42fc60831c4f6c Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Mon, 6 Jul 2026 21:03:15 -0700 Subject: [PATCH 39/86] --no-headers fix in changelog, refs #566 --- docs/changelog.rst | 1 + 1 file changed, 1 insertion(+) diff --git a/docs/changelog.rst b/docs/changelog.rst index e773b11..14e26a1 100644 --- a/docs/changelog.rst +++ b/docs/changelog.rst @@ -13,6 +13,7 @@ Unreleased - The ``extracts=`` option to ``table.insert()`` and friends no longer creates a lookup table record for ``None`` values - the column value stays ``null``. Previously every batch of inserted rows containing a ``None`` value would add a duplicate ``null`` record to the lookup table. - Fixed a bug where ``table.lookup()`` inserted a duplicate row on every call if any of the lookup values were ``None``. Lookup values are now compared using ``IS`` so that ``None`` values match existing rows correctly. - JSON output from the command-line tool no longer escapes non-ASCII characters, so ``sqlite-utils data.db "select '日本語' as text"`` now outputs ``[{"text": "日本語"}]``. This matches how values were already stored by ``insert`` and how CSV/TSV output already behaved. A new ``--ascii`` option restores the previous behavior of escaping non-ASCII characters, for output destinations that cannot handle UTF-8 - see :ref:`cli_query_json_ascii`. The option is available on the ``query``, ``rows``, ``search``, ``tables``, ``views``, ``triggers``, ``indexes`` and ``memory`` commands. The ``convert --multi --dry-run`` preview and ``plugins`` output also no longer escape non-ASCII characters. (:issue:`625`) +- ``--no-headers`` now omits the header row from ``--fmt`` and ``--table`` output, not just CSV and TSV output. (:issue:`566`) - ``table.insert_all(..., pk=...)`` now raises ``InvalidColumns`` if ``pk=`` names columns that do not exist in an existing table. Previously this behaved inconsistently, with single-row inserts raising a ``KeyError`` while other row counts succeeded. (:issue:`732`) - Fixed an ``IndexError`` from ``table.insert(..., pk=..., ignore=True)`` when an ignored insert followed writes to another table on the same connection. ``last_pk`` is now populated from the explicit primary key value instead of looking up a stale ``lastrowid``. (:issue:`554`) From 7d861181687e88194931c52ebfd57eb31f34c8d8 Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Mon, 6 Jul 2026 21:19:20 -0700 Subject: [PATCH 40/86] Fix failed db.execute() write leaves a phantom transaction open Refs https://github.com/simonw/sqlite-utils/issues/769#issuecomment-4900034150 --- docs/changelog.rst | 1 + sqlite_utils/db.py | 16 ++++++++++++---- tests/test_atomic.py | 41 +++++++++++++++++++++++++++++++++++++++++ 3 files changed, 54 insertions(+), 4 deletions(-) diff --git a/docs/changelog.rst b/docs/changelog.rst index 14e26a1..3bdab36 100644 --- a/docs/changelog.rst +++ b/docs/changelog.rst @@ -16,6 +16,7 @@ Unreleased - ``--no-headers`` now omits the header row from ``--fmt`` and ``--table`` output, not just CSV and TSV output. (:issue:`566`) - ``table.insert_all(..., pk=...)`` now raises ``InvalidColumns`` if ``pk=`` names columns that do not exist in an existing table. Previously this behaved inconsistently, with single-row inserts raising a ``KeyError`` while other row counts succeeded. (:issue:`732`) - Fixed an ``IndexError`` from ``table.insert(..., pk=..., ignore=True)`` when an ignored insert followed writes to another table on the same connection. ``last_pk`` is now populated from the explicit primary key value instead of looking up a stale ``lastrowid``. (:issue:`554`) +- Fixed a bug where a failed write statement executed with ``db.execute()`` left the driver's implicit transaction open. Every subsequent write then joined that phantom transaction, which nothing committed, so their work was silently rolled back when the connection was closed. The implicit transaction opened by a failed statement is now rolled back before the exception is raised. A failed write inside a transaction opened with ``db.begin()`` or ``db.atomic()`` leaves that transaction open and untouched, as before. .. _v4_0rc3: diff --git a/sqlite_utils/db.py b/sqlite_utils/db.py index 0f2924c..534f145 100644 --- a/sqlite_utils/db.py +++ b/sqlite_utils/db.py @@ -870,10 +870,18 @@ class Database: if self._tracer: self._tracer(sql, parameters) was_in_transaction = self.conn.in_transaction - if parameters is not None: - cursor = self.conn.execute(sql, parameters) - else: - cursor = self.conn.execute(sql) + try: + if parameters is not None: + cursor = self.conn.execute(sql, parameters) + else: + cursor = self.conn.execute(sql) + except Exception: + if not was_in_transaction and self.conn.in_transaction: + # The failed statement opened an implicit transaction that + # nothing would ever commit - roll it back, otherwise it + # would capture every subsequent write + self.conn.execute("ROLLBACK") + raise if ( not was_in_transaction and self.conn.in_transaction diff --git a/tests/test_atomic.py b/tests/test_atomic.py index f75c4ee..1a4b4ae 100644 --- a/tests/test_atomic.py +++ b/tests/test_atomic.py @@ -258,6 +258,47 @@ def test_execute_comment_prefixed_begin_leaves_transaction_open(fresh_db): assert [r["id"] for r in fresh_db["t"].rows] == [1] +def test_execute_failed_write_rolls_back_implicit_transaction(tmpdir): + # A failed write must not leave the driver's implicit transaction open - + # that would silently disable auto-commit for every subsequent write + path = str(tmpdir / "test.db") + db = Database(path) + db["t"].insert({"id": 1}, pk="id") + with pytest.raises(sqlite3.IntegrityError): + db.execute("insert into t (id) values (1)") + assert not db.conn.in_transaction + # Subsequent writes commit as normal and survive closing the connection + db["other"].insert({"id": 2}) + db.close() + db2 = Database(path) + assert db2["other"].exists() + db2.close() + + +def test_execute_failed_write_preserves_explicit_transaction(fresh_db): + # A failed write inside an explicit transaction must not roll back + # the caller's earlier work - only the caller decides that + fresh_db["t"].insert({"id": 1}, pk="id") + fresh_db.begin() + fresh_db.execute("insert into t (id) values (2)") + with pytest.raises(sqlite3.IntegrityError): + fresh_db.execute("insert into t (id) values (1)") + assert fresh_db.conn.in_transaction + fresh_db.commit() + assert [r["id"] for r in fresh_db["t"].rows] == [1, 2] + + +def test_execute_failed_write_inside_atomic_preserves_block(fresh_db): + # A caught failure inside an atomic() block must leave the block's + # transaction open so its other work still commits + fresh_db["t"].insert({"id": 1}, pk="id") + with fresh_db.atomic(): + fresh_db.execute("insert into t (id) values (2)") + with pytest.raises(sqlite3.IntegrityError): + fresh_db.execute("insert into t (id) values (1)") + assert [r["id"] for r in fresh_db["t"].rows] == [1, 2] + + def test_query_returning_commits_after_iteration(tmpdir): if sqlite3.sqlite_version_info < (3, 35, 0): import pytest as _pytest From adc10df98102c76c86c77108615792ba238a0ae3 Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Mon, 6 Jul 2026 21:22:29 -0700 Subject: [PATCH 41/86] Fix for db.query("; COMMIT") bypasses the first-token scanner Refs https://github.com/simonw/sqlite-utils/issues/769#issuecomment-4900034150 --- docs/changelog.rst | 1 + sqlite_utils/db.py | 11 ++++++----- tests/test_atomic.py | 13 +++++++++++++ tests/test_query.py | 27 +++++++++++++++++++++++++++ 4 files changed, 47 insertions(+), 5 deletions(-) diff --git a/docs/changelog.rst b/docs/changelog.rst index 3bdab36..30ad989 100644 --- a/docs/changelog.rst +++ b/docs/changelog.rst @@ -17,6 +17,7 @@ Unreleased - ``table.insert_all(..., pk=...)`` now raises ``InvalidColumns`` if ``pk=`` names columns that do not exist in an existing table. Previously this behaved inconsistently, with single-row inserts raising a ``KeyError`` while other row counts succeeded. (:issue:`732`) - Fixed an ``IndexError`` from ``table.insert(..., pk=..., ignore=True)`` when an ignored insert followed writes to another table on the same connection. ``last_pk`` is now populated from the explicit primary key value instead of looking up a stale ``lastrowid``. (:issue:`554`) - Fixed a bug where a failed write statement executed with ``db.execute()`` left the driver's implicit transaction open. Every subsequent write then joined that phantom transaction, which nothing committed, so their work was silently rolled back when the connection was closed. The implicit transaction opened by a failed statement is now rolled back before the exception is raised. A failed write inside a transaction opened with ``db.begin()`` or ``db.atomic()`` leaves that transaction open and untouched, as before. +- Fixed a bug where transaction-control statements prefixed with an empty statement - ``db.query("; COMMIT")`` - or a UTF-8 byte order mark slipped past the check that rejects them, committing the caller's open transaction before raising a confusing ``OperationalError``. The keyword scanner used by ``db.query()`` and ``db.execute()`` now skips leading ``;`` and byte order marks, matching what the ``sqlite3`` driver tolerates before the first token, so these statements are rejected with a ``ValueError`` without being executed. The same fix means ``db.execute("; BEGIN")`` no longer auto-commits the transaction it just opened. .. _v4_0rc3: diff --git a/sqlite_utils/db.py b/sqlite_utils/db.py index 534f145..bd023c5 100644 --- a/sqlite_utils/db.py +++ b/sqlite_utils/db.py @@ -436,14 +436,15 @@ _QUERY_REJECTED_KEYWORDS = _TRANSACTION_CONTROL_KEYWORDS | { def _first_keyword(sql: str) -> str: """ - Return the first keyword of a SQL statement, uppercased, skipping any - leading whitespace and ``--`` or ``/* ... */`` comments - the only - things SQLite's tokenizer allows before the first token. Returns an - empty string if there is no leading keyword. + Return the first keyword of a SQL statement, uppercased, skipping + everything the sqlite3 driver tolerates before the first real token: + whitespace, ``--`` or ``/* ... */`` comments, empty statements + (bare ``;``) and a UTF-8 byte order mark. Returns an empty string if + there is no leading keyword. """ i, n = 0, len(sql) while i < n: - if sql[i].isspace(): + if sql[i].isspace() or sql[i] in (";", "\ufeff"): i += 1 elif sql.startswith("--", i): newline = sql.find("\n", i) diff --git a/tests/test_atomic.py b/tests/test_atomic.py index 1a4b4ae..0d25b84 100644 --- a/tests/test_atomic.py +++ b/tests/test_atomic.py @@ -258,6 +258,19 @@ def test_execute_comment_prefixed_begin_leaves_transaction_open(fresh_db): assert [r["id"] for r in fresh_db["t"].rows] == [1] +@pytest.mark.parametrize("begin_sql", ["; begin", "\ufeffbegin"]) +def test_execute_prefixed_begin_leaves_transaction_open(fresh_db, begin_sql): + # sqlite3 tolerates empty statements and a UTF-8 BOM before the first + # real token, so a BEGIN behind either must not be auto-committed + # out from under the caller + fresh_db["t"].insert({"id": 1}, pk="id") + fresh_db.execute(begin_sql) + assert fresh_db.conn.in_transaction + fresh_db.execute("insert into t (id) values (2)") + fresh_db.rollback() + assert [r["id"] for r in fresh_db["t"].rows] == [1] + + def test_execute_failed_write_rolls_back_implicit_transaction(tmpdir): # A failed write must not leave the driver's implicit transaction open - # that would silently disable auto-commit for every subsequent write diff --git a/tests/test_query.py b/tests/test_query.py index 0b9f2ae..c2f6731 100644 --- a/tests/test_query.py +++ b/tests/test_query.py @@ -59,6 +59,10 @@ def test_query_rejected_write_inside_transaction_is_rolled_back(fresh_db): "-- comment\nbegin", "/* multi\nline */ -- and another\n vacuum", "\t /* a */ /* b */ savepoint s1", + "; commit", + ";;\n ; rollback", + "; /* comment */ vacuum", + "\ufeffbegin", ], ) def test_query_rejects_transaction_control_and_vacuum(fresh_db, sql): @@ -83,6 +87,23 @@ def test_query_comment_prefixed_commit_does_not_commit_transaction(fresh_db): assert [row["name"] for row in fresh_db["dogs"].rows] == ["Cleo"] +@pytest.mark.parametrize("sql", ["; COMMIT", "\ufeffCOMMIT"]) +def test_query_prefixed_commit_does_not_commit_transaction(fresh_db, sql): + # sqlite3 tolerates empty statements and a UTF-8 BOM before the first + # real token, so the keyword scanner must skip them too - previously + # '; COMMIT' slipped past the check and committed the caller's open + # transaction before raising OperationalError + fresh_db["dogs"].insert({"name": "Cleo"}) + fresh_db.begin() + fresh_db.execute("insert into dogs (name) values ('Pancakes')") + with pytest.raises(ValueError): + fresh_db.query(sql) + # The explicit transaction is still open and can still be rolled back + assert fresh_db.conn.in_transaction + fresh_db.rollback() + assert [row["name"] for row in fresh_db["dogs"].rows] == ["Cleo"] + + def test_query_error_leaves_no_transaction_open(fresh_db): with pytest.raises(sqlite3.OperationalError): fresh_db.query("select * from missing_table") @@ -137,6 +158,12 @@ def test_query_comment_prefixed_pragma_inside_transaction(fresh_db): ("", ""), (" ", ""), ("123", ""), + ("; commit", "COMMIT"), + (";;\n ; rollback", "ROLLBACK"), + ("; -- comment\n begin", "BEGIN"), + ("\ufeffcommit", "COMMIT"), + ("\ufeff ; select 1", "SELECT"), + (";", ""), ], ) def test_first_keyword(sql, expected): From 66934918c689238b19a74b2f09794002cc985094 Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Mon, 6 Jul 2026 21:30:47 -0700 Subject: [PATCH 42/86] Document that rejected write PRAGMAs in db.query() still take effect db.query() promises that a statement rejected with ValueError is rolled back and has no effect. PRAGMA statements are the exception: some of them refuse to run inside a transaction, so they execute outside the savepoint guard - a row-less PRAGMA such as "PRAGMA user_version = 5" therefore takes effect despite the ValueError. Documenting this as a known limitation rather than fixing it, since a fix would need a hardcoded list of row-returning PRAGMAs. Refs https://github.com/simonw/sqlite-utils/issues/769#issuecomment-4900034150 Co-Authored-By: Claude Fable 5 --- docs/changelog.rst | 1 + docs/python-api.rst | 2 ++ sqlite_utils/db.py | 5 ++++- tests/test_query.py | 12 ++++++++++++ 4 files changed, 19 insertions(+), 1 deletion(-) diff --git a/docs/changelog.rst b/docs/changelog.rst index 30ad989..027765c 100644 --- a/docs/changelog.rst +++ b/docs/changelog.rst @@ -18,6 +18,7 @@ Unreleased - Fixed an ``IndexError`` from ``table.insert(..., pk=..., ignore=True)`` when an ignored insert followed writes to another table on the same connection. ``last_pk`` is now populated from the explicit primary key value instead of looking up a stale ``lastrowid``. (:issue:`554`) - Fixed a bug where a failed write statement executed with ``db.execute()`` left the driver's implicit transaction open. Every subsequent write then joined that phantom transaction, which nothing committed, so their work was silently rolled back when the connection was closed. The implicit transaction opened by a failed statement is now rolled back before the exception is raised. A failed write inside a transaction opened with ``db.begin()`` or ``db.atomic()`` leaves that transaction open and untouched, as before. - Fixed a bug where transaction-control statements prefixed with an empty statement - ``db.query("; COMMIT")`` - or a UTF-8 byte order mark slipped past the check that rejects them, committing the caller's open transaction before raising a confusing ``OperationalError``. The keyword scanner used by ``db.query()`` and ``db.execute()`` now skips leading ``;`` and byte order marks, matching what the ``sqlite3`` driver tolerates before the first token, so these statements are rejected with a ``ValueError`` without being executed. The same fix means ``db.execute("; BEGIN")`` no longer auto-commits the transaction it just opened. +- Documented a limitation of ``db.query()``: a ``PRAGMA`` statement that returns no rows raises a ``ValueError`` but still takes effect, because PRAGMA statements run outside the savepoint guard used to roll back other rejected statements. Use ``db.execute()`` for row-less PRAGMA statements. .. _v4_0rc3: diff --git a/docs/python-api.rst b/docs/python-api.rst index 0e61cd4..516e2aa 100644 --- a/docs/python-api.rst +++ b/docs/python-api.rst @@ -233,6 +233,8 @@ The SQL query is executed as soon as ``db.query()`` is called. The resulting row ``db.query()`` can only be used with SQL that returns rows. Passing a statement that returns no rows - an ``INSERT`` or ``UPDATE`` without a ``RETURNING`` clause, for example - will raise a ``ValueError``. The rejected statement is rolled back, so it has no effect on the database. Use :ref:`db.execute() ` for those statements instead. +There is one exception to the rolled-back guarantee: a ``PRAGMA`` statement that returns no rows, such as ``PRAGMA user_version = 5``, still raises a ``ValueError`` but will already have taken effect. Some PRAGMA statements refuse to run inside a transaction, so PRAGMAs are executed outside the savepoint that is used to roll back other rejected statements. Use ``db.execute()`` for PRAGMA statements that do not return rows. + If a query returns more than one column with the same name - a join between two tables that share column names, for example - later occurrences are renamed with a numeric suffix, so every value is included in the dictionary: .. code-block:: python diff --git a/sqlite_utils/db.py b/sqlite_utils/db.py index bd023c5..f6153cd 100644 --- a/sqlite_utils/db.py +++ b/sqlite_utils/db.py @@ -797,7 +797,10 @@ class Database: parameters, or a dictionary for ``where id = :id`` :raises ValueError: if the SQL statement does not return rows - use :meth:`execute` for those statements instead. The rejected statement - is rolled back, so it has no effect on the database + is rolled back, so it has no effect on the database. One exception: + a row-less ``PRAGMA`` statement takes effect despite the + ``ValueError``, because PRAGMAs run outside the savepoint guard - + some of them refuse to run inside a transaction """ message = ( "query() can only be used with SQL that returns rows - " diff --git a/tests/test_query.py b/tests/test_query.py index c2f6731..f4aa336 100644 --- a/tests/test_query.py +++ b/tests/test_query.py @@ -122,6 +122,18 @@ def test_query_pragma(tmpdir): db.close() +def test_query_rejected_pragma_still_takes_effect(fresh_db): + # Documented limitation: PRAGMAs run outside the savepoint guard, + # because some of them refuse to run inside a transaction - so a + # row-less PRAGMA takes effect even though it raises ValueError. + # If this test starts failing because the pragma was rolled back, + # the limitation has been fixed - update the docs in python-api.rst + # and the query() docstring to remove the carve-out + with pytest.raises(ValueError): + fresh_db.query("pragma user_version = 5") + assert fresh_db.execute("pragma user_version").fetchone()[0] == 5 + + def test_query_comment_prefixed_pragma(tmpdir): from sqlite_utils import Database From 8e015d024c34f2a5e5059e0f10fcdf41410613fe Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Mon, 6 Jul 2026 21:34:02 -0700 Subject: [PATCH 43/86] Compound primary keys now resolve in PRIMARY KEY declaration order PRAGMA table_info sets is_pk to the 1-based position of each column within the PRIMARY KEY, which can differ from table column order. table.pks previously returned table column order, so an implicit compound FOREIGN KEY ... REFERENCES other was introspected with its referenced columns inverted, and transform() baked that inverted order into the rewritten schema - failing with IntegrityError on valid data, or silently reversing the constraint with foreign keys off. table.pks, compound foreign key guessing (create, add_foreign_key) and transform() now all use declaration order, and transform() no longer reorders a compound PRIMARY KEY (b, a) into table column order. Refs https://github.com/simonw/sqlite-utils/issues/769#issuecomment-4900034150 Co-Authored-By: Claude Fable 5 --- docs/changelog.rst | 1 + sqlite_utils/db.py | 31 +++++++++++++++------- tests/test_foreign_keys.py | 54 ++++++++++++++++++++++++++++++++++++++ tests/test_introspect.py | 15 +++++++++++ 4 files changed, 91 insertions(+), 10 deletions(-) diff --git a/docs/changelog.rst b/docs/changelog.rst index 027765c..e7c582e 100644 --- a/docs/changelog.rst +++ b/docs/changelog.rst @@ -19,6 +19,7 @@ Unreleased - Fixed a bug where a failed write statement executed with ``db.execute()`` left the driver's implicit transaction open. Every subsequent write then joined that phantom transaction, which nothing committed, so their work was silently rolled back when the connection was closed. The implicit transaction opened by a failed statement is now rolled back before the exception is raised. A failed write inside a transaction opened with ``db.begin()`` or ``db.atomic()`` leaves that transaction open and untouched, as before. - Fixed a bug where transaction-control statements prefixed with an empty statement - ``db.query("; COMMIT")`` - or a UTF-8 byte order mark slipped past the check that rejects them, committing the caller's open transaction before raising a confusing ``OperationalError``. The keyword scanner used by ``db.query()`` and ``db.execute()`` now skips leading ``;`` and byte order marks, matching what the ``sqlite3`` driver tolerates before the first token, so these statements are rejected with a ``ValueError`` without being executed. The same fix means ``db.execute("; BEGIN")`` no longer auto-commits the transaction it just opened. - Documented a limitation of ``db.query()``: a ``PRAGMA`` statement that returns no rows raises a ``ValueError`` but still takes effect, because PRAGMA statements run outside the savepoint guard used to roll back other rejected statements. Use ``db.execute()`` for row-less PRAGMA statements. +- Fixed a bug where compound primary key columns were returned in table column order instead of ``PRIMARY KEY`` declaration order. For a table declared as ``CREATE TABLE other (b TEXT, a TEXT, PRIMARY KEY (a, b))`` an implicit ``FOREIGN KEY (x, y) REFERENCES other`` was introspected as referencing ``(b, a)`` when SQLite resolves it as ``(a, b)`` - running ``transform()`` on such a table then rewrote the schema with the inverted column order, silently reversing the meaning of the constraint and causing foreign key errors on valid data. ``table.pks``, compound foreign key guessing and ``transform()`` now all use the primary key declaration order, and ``transform()`` no longer reorders a compound ``PRIMARY KEY (b, a)`` into table column order. .. _v4_0rc3: diff --git a/sqlite_utils/db.py b/sqlite_utils/db.py index f6153cd..1e10b11 100644 --- a/sqlite_utils/db.py +++ b/sqlite_utils/db.py @@ -2012,10 +2012,9 @@ class Queryable: :param offset: Integer for SQL offset """ column_names = [column.name for column in self.columns] - pks = [column.name for column in self.columns if column.is_pk] - if not pks: + pks = self.pks + if self.use_rowid: column_names.insert(0, "rowid") - pks = ["rowid"] select = ",".join(quote_identifier(column_name) for column_name in column_names) for row in self.rows_where( select=select, @@ -2148,8 +2147,18 @@ class Table(Queryable): @property def pks(self) -> List[str]: - "Primary key columns for this table." - names = [column.name for column in self.columns if column.is_pk] + """ + Primary key columns for this table, in PRIMARY KEY declaration order - + ``PRAGMA table_info`` sets ``is_pk`` to the 1-based position of each + column within the primary key, which can differ from the order of the + columns in the table. SQLite uses the declaration order to resolve + implicit foreign key references, so this order matters. + """ + pk_columns = sorted( + (column for column in self.columns if column.is_pk), + key=lambda column: column.is_pk, + ) + names = [column.name for column in pk_columns] if not names: names = ["rowid"] return names @@ -2673,7 +2682,8 @@ class Table(Queryable): if pk is DEFAULT: pks_renamed = tuple( - rename.get(p.name) or p.name for p in self.columns if p.is_pk + rename.get(pk_name) or pk_name + for pk_name in (self.pks if not self.use_rowid else []) ) if len(pks_renamed) == 1: pk = pks_renamed[0] @@ -3017,7 +3027,10 @@ class Table(Queryable): raise AlterError("table '{}' has no column {}".format(fk, fk_col)) else: # automatically set fk_col to first primary_key of fk table - pks = [c for c in self.db[fk].columns if c.is_pk] + pks = sorted( + (c for c in self.db[fk].columns if c.is_pk), + key=lambda c: c.is_pk, + ) if pks: fk_col = pks[0].name fk_col_type = pks[0].type @@ -3770,9 +3783,7 @@ class Table(Queryable): # First we execute the function pk_to_values = {} new_column_types: Dict[str, Set[type]] = {} - pks = [column.name for column in self.columns if column.is_pk] - if not pks: - pks = ["rowid"] + pks = self.pks with progressbar( length=self.count, silent=not show_progress, label="1: Evaluating" diff --git a/tests/test_foreign_keys.py b/tests/test_foreign_keys.py index 8950189..1672069 100644 --- a/tests/test_foreign_keys.py +++ b/tests/test_foreign_keys.py @@ -524,3 +524,57 @@ def test_add_compound_foreign_key_on_delete(courses_db): assert fk.is_compound is True assert fk.on_delete == "SET NULL" assert "ON DELETE SET NULL" in courses_db["courses"].schema + + +def test_implicit_compound_foreign_key_resolves_pk_declaration_order(fresh_db): + # The other table's PRIMARY KEY declares its columns in a different + # order to the table's column order. SQLite resolves the implicit + # "REFERENCES other" using PRIMARY KEY declaration order, so the + # introspected other_columns must too + fresh_db.execute("create table other (b text, a text, primary key (a, b))") + fresh_db.execute( + "create table child (x text, y text, foreign key (x, y) references other)" + ) + fk = fresh_db["child"].foreign_keys[0] + assert fk.other_columns == ("a", "b") + + +def test_transform_implicit_compound_foreign_key_stays_valid(fresh_db): + # transform() rewrites the implicit FK with explicit columns - they + # must be in PRIMARY KEY declaration order or valid data fails the + # foreign key check with an IntegrityError + fresh_db.execute("create table other (b text, a text, primary key (a, b))") + fresh_db.execute( + "create table child (x text, y text, foreign key (x, y) references other)" + ) + fresh_db.execute("PRAGMA foreign_keys = ON") + fresh_db["other"].insert({"a": "A", "b": "B"}) + fresh_db["child"].insert({"x": "A", "y": "B"}) + fresh_db["child"].transform(types={"x": str}) + assert fresh_db["child"].foreign_keys[0].other_columns == ("a", "b") + # The constraint still points the right way around + fresh_db["child"].insert({"x": "A", "y": "B"}) + with pytest.raises(sqlite3.IntegrityError): + fresh_db["child"].insert({"x": "B", "y": "A"}) + + +def test_create_compound_foreign_key_guesses_pk_declaration_order(fresh_db): + fresh_db.execute("create table other (b text, a text, primary key (a, b))") + fresh_db["other"].insert({"a": "A", "b": "B"}) + fresh_db["child"].create( + {"id": int, "x": str, "y": str}, + pk="id", + foreign_keys=[(("x", "y"), "other")], + ) + assert fresh_db["child"].foreign_keys[0].other_columns == ("a", "b") + fresh_db.execute("PRAGMA foreign_keys = ON") + fresh_db["child"].insert({"id": 1, "x": "A", "y": "B"}) + with pytest.raises(sqlite3.IntegrityError): + fresh_db["child"].insert({"id": 2, "x": "B", "y": "A"}) + + +def test_add_compound_foreign_key_guesses_pk_declaration_order(fresh_db): + fresh_db.execute("create table other (b text, a text, primary key (a, b))") + fresh_db["child"].insert({"id": 1, "x": "A", "y": "B"}, pk="id") + fresh_db["child"].add_foreign_key(("x", "y"), "other") + assert fresh_db["child"].foreign_keys[0].other_columns == ("a", "b") diff --git a/tests/test_introspect.py b/tests/test_introspect.py index 4ff3f77..8b6765d 100644 --- a/tests/test_introspect.py +++ b/tests/test_introspect.py @@ -321,3 +321,18 @@ def test_table_default_values(fresh_db, value): ) default_values = fresh_db["default_values"].default_values assert default_values == {"value": value} + + +def test_pks_use_primary_key_declaration_order(fresh_db): + # PRIMARY KEY (a, b) declared against columns stored in order (b, a) - + # pks must follow the declaration order, which is what SQLite uses to + # resolve implicit foreign key references and compound pk lookups + fresh_db.execute("create table t (b text, a text, primary key (a, b))") + assert fresh_db["t"].pks == ["a", "b"] + + +def test_transform_preserves_compound_pk_declaration_order(fresh_db): + fresh_db.execute("create table t (a text, b text, c text, primary key (b, a))") + fresh_db["t"].transform(drop={"c"}) + assert fresh_db["t"].pks == ["b", "a"] + assert 'PRIMARY KEY ("b", "a")' in fresh_db["t"].schema From 404e935b6348c57e507fba36eb06daf462257af8 Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Mon, 6 Jul 2026 21:35:45 -0700 Subject: [PATCH 44/86] ForeignKey is now a frozen dataclass, restoring hashability The namedtuple-to-dataclass change made ForeignKey unhashable, breaking set(table.foreign_keys) and dict-key usage that worked in 3.x. frozen=True restores immutability and hashability. Equality and hashing cover all compared fields including on_delete/on_update - two foreign keys differing only in their actions are different constraints. Refs https://github.com/simonw/sqlite-utils/issues/769#issuecomment-4900034150 Co-Authored-By: Claude Fable 5 --- docs/changelog.rst | 1 + docs/upgrading.rst | 2 +- sqlite_utils/db.py | 24 +++++++++++++++++------- tests/test_foreign_keys.py | 32 ++++++++++++++++++++++++++++++++ 4 files changed, 51 insertions(+), 8 deletions(-) diff --git a/docs/changelog.rst b/docs/changelog.rst index e7c582e..b1c6108 100644 --- a/docs/changelog.rst +++ b/docs/changelog.rst @@ -19,6 +19,7 @@ Unreleased - Fixed a bug where a failed write statement executed with ``db.execute()`` left the driver's implicit transaction open. Every subsequent write then joined that phantom transaction, which nothing committed, so their work was silently rolled back when the connection was closed. The implicit transaction opened by a failed statement is now rolled back before the exception is raised. A failed write inside a transaction opened with ``db.begin()`` or ``db.atomic()`` leaves that transaction open and untouched, as before. - Fixed a bug where transaction-control statements prefixed with an empty statement - ``db.query("; COMMIT")`` - or a UTF-8 byte order mark slipped past the check that rejects them, committing the caller's open transaction before raising a confusing ``OperationalError``. The keyword scanner used by ``db.query()`` and ``db.execute()`` now skips leading ``;`` and byte order marks, matching what the ``sqlite3`` driver tolerates before the first token, so these statements are rejected with a ``ValueError`` without being executed. The same fix means ``db.execute("; BEGIN")`` no longer auto-commits the transaction it just opened. - Documented a limitation of ``db.query()``: a ``PRAGMA`` statement that returns no rows raises a ``ValueError`` but still takes effect, because PRAGMA statements run outside the savepoint guard used to roll back other rejected statements. Use ``db.execute()`` for row-less PRAGMA statements. +- ``ForeignKey`` objects are hashable again. The 4.0 change from ``namedtuple`` to dataclass accidentally made them unhashable, breaking patterns like ``set(table.foreign_keys)`` that worked in 3.x. ``ForeignKey`` is now a frozen dataclass - immutable and hashable, like the namedtuple was. - Fixed a bug where compound primary key columns were returned in table column order instead of ``PRIMARY KEY`` declaration order. For a table declared as ``CREATE TABLE other (b TEXT, a TEXT, PRIMARY KEY (a, b))`` an implicit ``FOREIGN KEY (x, y) REFERENCES other`` was introspected as referencing ``(b, a)`` when SQLite resolves it as ``(a, b)`` - running ``transform()`` on such a table then rewrote the schema with the inverted column order, silently reversing the meaning of the constraint and causing foreign key errors on valid data. ``table.pks``, compound foreign key guessing and ``transform()`` now all use the primary key declaration order, and ``transform()`` no longer reorders a compound ``PRIMARY KEY (b, a)`` into table column order. .. _v4_0rc3: diff --git a/docs/upgrading.rst b/docs/upgrading.rst index 88e987d..92b582a 100644 --- a/docs/upgrading.rst +++ b/docs/upgrading.rst @@ -95,7 +95,7 @@ Python API changes for fk in db["courses"].foreign_keys: fk.table, fk.column, fk.other_table, fk.other_column -Attempting the old unpacking or ``fk[0]`` indexing now raises ``TypeError``, so any code using those patterns will fail loudly rather than silently misbehave. +Attempting the old unpacking or ``fk[0]`` indexing now raises ``TypeError``, so any code using those patterns will fail loudly rather than silently misbehave. Like the old namedtuple, ``ForeignKey`` instances are immutable and hashable - they can be collected into sets and used as dictionary keys. Note that equality now includes the ``on_delete`` and ``on_update`` actions: a ``ForeignKey`` with ``ON DELETE CASCADE`` is not equal to one without. Compound foreign keys - previously returned as one ``ForeignKey`` per column, misleadingly suggesting several independent single-column keys - are now returned as a single ``ForeignKey`` with ``is_compound=True``. For these the scalar ``column`` and ``other_column`` fields are ``None``; use the ``columns`` and ``other_columns`` tuples instead. Single-column foreign keys are unaffected apart from the class change: ``column``/``other_column`` behave as before and ``columns``/``other_columns`` are one-item tuples. diff --git a/sqlite_utils/db.py b/sqlite_utils/db.py index 1e10b11..e94434a 100644 --- a/sqlite_utils/db.py +++ b/sqlite_utils/db.py @@ -193,7 +193,7 @@ Summary information about a column, see :ref:`python_api_analyze_column`. """ -@dataclass(order=True) +@dataclass(order=True, frozen=True) class ForeignKey: """ A foreign key defined on a table. @@ -208,6 +208,11 @@ class ForeignKey: ``on_delete`` and ``on_update`` hold the foreign key actions, e.g. ``"CASCADE"`` - ``"NO ACTION"`` if not set. + Instances are immutable and hashable, so they can be collected into + sets and used as dictionary keys. Equality covers every compared field, + including ``on_delete`` and ``on_update`` - two foreign keys differing + only in their actions are different constraints. + Prior to sqlite-utils 4.0 this was a ``namedtuple`` and could be unpacked or indexed as ``(table, column, other_table, other_column)``. It is now a dataclass - access its fields by name instead. @@ -227,16 +232,21 @@ class ForeignKey: def __post_init__(self): # Populate columns/other_columns for single-column foreign keys, - # normalizing any lists to tuples + # normalizing any lists to tuples. object.__setattr__ because the + # dataclass is frozen if self.columns: - self.columns = tuple(self.columns) + object.__setattr__(self, "columns", tuple(self.columns)) else: - self.columns = (self.column,) if self.column is not None else () + object.__setattr__( + self, "columns", (self.column,) if self.column is not None else () + ) if self.other_columns: - self.other_columns = tuple(self.other_columns) + object.__setattr__(self, "other_columns", tuple(self.other_columns)) else: - self.other_columns = ( - (self.other_column,) if self.other_column is not None else () + object.__setattr__( + self, + "other_columns", + (self.other_column,) if self.other_column is not None else (), ) diff --git a/tests/test_foreign_keys.py b/tests/test_foreign_keys.py index 1672069..49d3352 100644 --- a/tests/test_foreign_keys.py +++ b/tests/test_foreign_keys.py @@ -578,3 +578,35 @@ def test_add_compound_foreign_key_guesses_pk_declaration_order(fresh_db): fresh_db["child"].insert({"id": 1, "x": "A", "y": "B"}, pk="id") fresh_db["child"].add_foreign_key(("x", "y"), "other") assert fresh_db["child"].foreign_keys[0].other_columns == ("a", "b") + + +def test_foreign_keys_are_hashable(fresh_db): + # set() over foreign_keys worked with the 3.x namedtuple and must + # keep working with the dataclass + fresh_db["p"].insert({"id": 1}, pk="id") + fresh_db["c"].insert( + {"id": 1, "pid": 1}, pk="id", foreign_keys=[("pid", "p", "id")] + ) + fks = set(fresh_db["c"].foreign_keys) + assert len(fks) == 1 + assert ForeignKey("c", "pid", "p", "id") in fks + # Usable as dict keys too + assert {fk: True for fk in fks} + + +def test_foreign_key_is_immutable(): + import dataclasses + + fk = ForeignKey("c", "pid", "p", "id") + with pytest.raises(dataclasses.FrozenInstanceError): + fk.table = "other" + + +def test_foreign_key_equality_and_hash_include_actions(): + # Two foreign keys differing only in ON DELETE behavior are different + # constraints - they compare unequal and hash separately + plain = ForeignKey("c", "pid", "p", "id") + cascade = ForeignKey("c", "pid", "p", "id", on_delete="CASCADE") + assert plain != cascade + assert len({plain, cascade}) == 2 + assert plain == ForeignKey("c", "pid", "p", "id") From 29ca9d27e24be818a26fff41450664ef771a03a0 Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Mon, 6 Jul 2026 21:37:56 -0700 Subject: [PATCH 45/86] foreign_keys= accepts mixed ForeignKey objects, tuples and strings again resolve_foreign_keys() used all-or-nothing isinstance checks, so mixing ForeignKey objects with tuples raised "foreign_keys= should be a list of tuples" - a regression from 3.x, where ForeignKey was a namedtuple and passed the tuple check. Each item is now normalized individually, which also allows bare column-name strings in a mixed list. Refs https://github.com/simonw/sqlite-utils/issues/769#issuecomment-4900034150 Co-Authored-By: Claude Fable 5 --- docs/changelog.rst | 1 + sqlite_utils/db.py | 30 ++++++++++++++++-------------- tests/test_foreign_keys.py | 33 +++++++++++++++++++++++++++++++++ 3 files changed, 50 insertions(+), 14 deletions(-) diff --git a/docs/changelog.rst b/docs/changelog.rst index b1c6108..6e6fa30 100644 --- a/docs/changelog.rst +++ b/docs/changelog.rst @@ -19,6 +19,7 @@ Unreleased - Fixed a bug where a failed write statement executed with ``db.execute()`` left the driver's implicit transaction open. Every subsequent write then joined that phantom transaction, which nothing committed, so their work was silently rolled back when the connection was closed. The implicit transaction opened by a failed statement is now rolled back before the exception is raised. A failed write inside a transaction opened with ``db.begin()`` or ``db.atomic()`` leaves that transaction open and untouched, as before. - Fixed a bug where transaction-control statements prefixed with an empty statement - ``db.query("; COMMIT")`` - or a UTF-8 byte order mark slipped past the check that rejects them, committing the caller's open transaction before raising a confusing ``OperationalError``. The keyword scanner used by ``db.query()`` and ``db.execute()`` now skips leading ``;`` and byte order marks, matching what the ``sqlite3`` driver tolerates before the first token, so these statements are rejected with a ``ValueError`` without being executed. The same fix means ``db.execute("; BEGIN")`` no longer auto-commits the transaction it just opened. - Documented a limitation of ``db.query()``: a ``PRAGMA`` statement that returns no rows raises a ``ValueError`` but still takes effect, because PRAGMA statements run outside the savepoint guard used to roll back other rejected statements. Use ``db.execute()`` for row-less PRAGMA statements. +- The ``foreign_keys=`` argument to ``create()`` and ``insert()`` accepts a mixed list of ``ForeignKey`` objects, tuples and column name strings again. In 4.0 pre-releases mixing ``ForeignKey`` objects with tuples raised a ``ValueError`` - a regression from 3.x, where ``ForeignKey`` was a ``namedtuple`` and passed the tuple checks. - ``ForeignKey`` objects are hashable again. The 4.0 change from ``namedtuple`` to dataclass accidentally made them unhashable, breaking patterns like ``set(table.foreign_keys)`` that worked in 3.x. ``ForeignKey`` is now a frozen dataclass - immutable and hashable, like the namedtuple was. - Fixed a bug where compound primary key columns were returned in table column order instead of ``PRIMARY KEY`` declaration order. For a table declared as ``CREATE TABLE other (b TEXT, a TEXT, PRIMARY KEY (a, b))`` an implicit ``FOREIGN KEY (x, y) REFERENCES other`` was introspected as referencing ``(b, a)`` when SQLite resolves it as ``(a, b)`` - running ``transform()`` on such a table then rewrote the schema with the inverted column order, silently reversing the meaning of the constraint and causing foreign key errors on valid data. ``table.pks``, compound foreign key guessing and ``transform()`` now all use the primary key declaration order, and ``transform()`` no longer reorders a compound ``PRIMARY KEY (b, a)`` into table column order. diff --git a/sqlite_utils/db.py b/sqlite_utils/db.py index e94434a..6105a64 100644 --- a/sqlite_utils/db.py +++ b/sqlite_utils/db.py @@ -1238,21 +1238,23 @@ class Database: (("campus_name", "dept_code"), "departments", ("campus_name", "dept_code")) """ table = self.table(name) - if all(isinstance(fk, ForeignKey) for fk in foreign_keys): - return cast(List[ForeignKey], foreign_keys) - if all(isinstance(fk, str) for fk in foreign_keys): - # It's a list of columns - fks = [] - for column in foreign_keys: - column = cast(str, column) - other_table = table.guess_foreign_table(column) - other_column = table.guess_foreign_column(other_table) - fks.append(ForeignKey(name, column, other_table, other_column)) - return fks - if not all(isinstance(fk, (tuple, list)) for fk in foreign_keys): - raise ValueError("foreign_keys= should be a list of tuples") fks = [] - for tuple_or_list in cast(Iterable[Sequence[Any]], foreign_keys): + for fk in foreign_keys: + if isinstance(fk, ForeignKey): + fks.append(fk) + continue + if isinstance(fk, str): + # A bare column name - guess the other table and column + other_table = table.guess_foreign_table(fk) + other_column = table.guess_foreign_column(other_table) + fks.append(ForeignKey(name, fk, other_table, other_column)) + continue + if not isinstance(fk, (tuple, list)): + raise ValueError( + "foreign_keys= should be a list of tuples, " + "ForeignKey objects or column name strings" + ) + tuple_or_list = cast(Sequence[Any], fk) if len(tuple_or_list) == 4: if tuple_or_list[0] != name: raise ValueError( diff --git a/tests/test_foreign_keys.py b/tests/test_foreign_keys.py index 49d3352..d7c20f6 100644 --- a/tests/test_foreign_keys.py +++ b/tests/test_foreign_keys.py @@ -610,3 +610,36 @@ def test_foreign_key_equality_and_hash_include_actions(): assert plain != cascade assert len({plain, cascade}) == 2 assert plain == ForeignKey("c", "pid", "p", "id") + + +def test_create_table_mixed_foreign_keys_list(fresh_db): + # 3.x accepted a mix of ForeignKey objects, tuples and bare column + # strings in foreign_keys= (ForeignKey was a namedtuple, so it passed + # the tuple check) - keep accepting the mix + fresh_db["authors"].insert({"id": 1}, pk="id") + fresh_db["publishers"].insert({"id": 1}, pk="id") + fresh_db["books"].create( + {"id": int, "author_id": int, "publisher_id": int}, + pk="id", + foreign_keys=[ + ForeignKey("books", "author_id", "authors", "id"), + ("publisher_id", "publishers", "id"), + ], + ) + fks = {fk.column: fk.other_table for fk in fresh_db["books"].foreign_keys} + assert fks == {"author_id": "authors", "publisher_id": "publishers"} + + +def test_create_table_mixed_foreign_keys_with_string(fresh_db): + fresh_db["authors"].insert({"id": 1}, pk="id") + fresh_db["publishers"].insert({"id": 1}, pk="id") + fresh_db["books"].create( + {"id": int, "author_id": int, "publisher_id": int}, + pk="id", + foreign_keys=[ + "author_id", # bare column, table and column guessed + ("publisher_id", "publishers", "id"), + ], + ) + fks = {fk.column: fk.other_table for fk in fresh_db["books"].foreign_keys} + assert fks == {"author_id": "authors", "publisher_id": "publishers"} From 1ed95e4ad2676b7f2f0725919bba9c4b051456c4 Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Mon, 6 Jul 2026 21:42:02 -0700 Subject: [PATCH 46/86] Fix pks_and_rows_where() on views The previous compound-pk-ordering commit switched pks_and_rows_where() to Table-only properties, but the method is defined on Queryable and views exposed it too - calling it on a View raised AttributeError. Restored Queryable-safe logic, and stopped double-quoting the synthesized rowid column: SQLite turns a double-quoted identifier that does not resolve into a string literal, so on a view the generated select "rowid" silently produced the string 'rowid' and a confusing KeyError, where 3.x's [rowid] quoting raised OperationalError cleanly. Refs https://github.com/simonw/sqlite-utils/issues/769#issuecomment-4900034150 Co-Authored-By: Claude Fable 5 --- docs/changelog.rst | 1 + sqlite_utils/db.py | 21 ++++++++++++++++----- tests/test_rows.py | 21 +++++++++++++++++++++ 3 files changed, 38 insertions(+), 5 deletions(-) diff --git a/docs/changelog.rst b/docs/changelog.rst index 6e6fa30..5d6c901 100644 --- a/docs/changelog.rst +++ b/docs/changelog.rst @@ -19,6 +19,7 @@ Unreleased - Fixed a bug where a failed write statement executed with ``db.execute()`` left the driver's implicit transaction open. Every subsequent write then joined that phantom transaction, which nothing committed, so their work was silently rolled back when the connection was closed. The implicit transaction opened by a failed statement is now rolled back before the exception is raised. A failed write inside a transaction opened with ``db.begin()`` or ``db.atomic()`` leaves that transaction open and untouched, as before. - Fixed a bug where transaction-control statements prefixed with an empty statement - ``db.query("; COMMIT")`` - or a UTF-8 byte order mark slipped past the check that rejects them, committing the caller's open transaction before raising a confusing ``OperationalError``. The keyword scanner used by ``db.query()`` and ``db.execute()`` now skips leading ``;`` and byte order marks, matching what the ``sqlite3`` driver tolerates before the first token, so these statements are rejected with a ``ValueError`` without being executed. The same fix means ``db.execute("; BEGIN")`` no longer auto-commits the transaction it just opened. - Documented a limitation of ``db.query()``: a ``PRAGMA`` statement that returns no rows raises a ``ValueError`` but still takes effect, because PRAGMA statements run outside the savepoint guard used to roll back other rejected statements. Use ``db.execute()`` for row-less PRAGMA statements. +- Fixed ``pks_and_rows_where()`` raising ``AttributeError`` when called on a view, and no longer double-quotes the synthesized ``rowid`` column in its generated SQL - SQLite turns a double-quoted identifier that does not resolve into a string literal, which on a view produced a confusing ``KeyError`` instead of the ``OperationalError`` raised in 3.x. Compound primary keys returned by this method now follow ``PRIMARY KEY`` declaration order. - The ``foreign_keys=`` argument to ``create()`` and ``insert()`` accepts a mixed list of ``ForeignKey`` objects, tuples and column name strings again. In 4.0 pre-releases mixing ``ForeignKey`` objects with tuples raised a ``ValueError`` - a regression from 3.x, where ``ForeignKey`` was a ``namedtuple`` and passed the tuple checks. - ``ForeignKey`` objects are hashable again. The 4.0 change from ``namedtuple`` to dataclass accidentally made them unhashable, breaking patterns like ``set(table.foreign_keys)`` that worked in 3.x. ``ForeignKey`` is now a frozen dataclass - immutable and hashable, like the namedtuple was. - Fixed a bug where compound primary key columns were returned in table column order instead of ``PRIMARY KEY`` declaration order. For a table declared as ``CREATE TABLE other (b TEXT, a TEXT, PRIMARY KEY (a, b))`` an implicit ``FOREIGN KEY (x, y) REFERENCES other`` was introspected as referencing ``(b, a)`` when SQLite resolves it as ``(a, b)`` - running ``transform()`` on such a table then rewrote the schema with the inverted column order, silently reversing the meaning of the constraint and causing foreign key errors on valid data. ``table.pks``, compound foreign key guessing and ``transform()`` now all use the primary key declaration order, and ``transform()`` no longer reorders a compound ``PRIMARY KEY (b, a)`` into table column order. diff --git a/sqlite_utils/db.py b/sqlite_utils/db.py index 6105a64..4c7ef52 100644 --- a/sqlite_utils/db.py +++ b/sqlite_utils/db.py @@ -2023,11 +2023,22 @@ class Queryable: :param limit: Integer number of rows to limit to :param offset: Integer for SQL offset """ - column_names = [column.name for column in self.columns] - pks = self.pks - if self.use_rowid: - column_names.insert(0, "rowid") - select = ",".join(quote_identifier(column_name) for column_name in column_names) + # This method is defined on Queryable so it serves views too, which + # have no pks property - sort pk columns into declaration order here + pk_columns = sorted( + (column for column in self.columns if column.is_pk), + key=lambda column: column.is_pk, + ) + pks = [column.name for column in pk_columns] + select_parts = [quote_identifier(column.name) for column in self.columns] + if not pks: + # rowid is left unquoted: it is not a real column, and SQLite + # turns a double-quoted identifier that does not resolve into a + # string literal - on a view that would silently select the + # string 'rowid' instead of raising an error + select_parts.insert(0, "rowid") + pks = ["rowid"] + select = ",".join(select_parts) for row in self.rows_where( select=select, where=where, diff --git a/tests/test_rows.py b/tests/test_rows.py index f050d5a..46417ef 100644 --- a/tests/test_rows.py +++ b/tests/test_rows.py @@ -111,3 +111,24 @@ def test_rows_where_duplicate_select_columns_are_deduped(fresh_db): fresh_db["t"].insert({"id": 1, "name": "Cleo"}) rows = list(fresh_db["t"].rows_where(select="id, id, name")) assert rows == [{"id": 1, "id_2": 1, "name": "Cleo"}] + + +def test_pks_and_rows_where_view(fresh_db): + # pks_and_rows_where() lives on Queryable so views expose it, but + # SQLite views have no rowid - it has always failed with an + # OperationalError from the generated SQL. Guard against it failing + # earlier with an AttributeError from View lacking Table properties + from sqlite_utils.utils import sqlite3 + + fresh_db["dogs"].insert({"id": 1, "name": "Cleo"}, pk="id") + fresh_db.create_view("dog_names", "select name from dogs") + with pytest.raises(sqlite3.OperationalError): + list(fresh_db["dog_names"].pks_and_rows_where()) + + +def test_pks_and_rows_where_compound_pk_declaration_order(fresh_db): + # Compound pks are returned in PRIMARY KEY declaration order + fresh_db.execute("create table t (b text, a text, primary key (a, b))") + fresh_db["t"].insert({"a": "A", "b": "B"}) + pks_and_rows = list(fresh_db["t"].pks_and_rows_where()) + assert pks_and_rows == [(("A", "B"), {"b": "B", "a": "A"})] From 884574685fde886863dcadb0d297960c88ba5446 Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Mon, 6 Jul 2026 21:44:56 -0700 Subject: [PATCH 47/86] insert/upsert --csv no longer rewrites column types of existing tables Type detection is the 4.0 default for CSV/TSV data, and the detected-type transform ran even when the target table already existed - inserting a CSV into a table with a TEXT zip column converted the column to INTEGER, corrupting values with leading zeros ('01234' became 1234) with no warning. Detected types now only apply to tables the command created. Refs https://github.com/simonw/sqlite-utils/issues/769#issuecomment-4900034150 Co-Authored-By: Claude Fable 5 --- docs/changelog.rst | 1 + docs/cli.rst | 2 ++ sqlite_utils/cli.py | 12 +++++++++- tests/test_cli_insert.py | 48 ++++++++++++++++++++++++++++++++++++++++ 4 files changed, 62 insertions(+), 1 deletion(-) diff --git a/docs/changelog.rst b/docs/changelog.rst index 5d6c901..80bcfb7 100644 --- a/docs/changelog.rst +++ b/docs/changelog.rst @@ -19,6 +19,7 @@ Unreleased - Fixed a bug where a failed write statement executed with ``db.execute()`` left the driver's implicit transaction open. Every subsequent write then joined that phantom transaction, which nothing committed, so their work was silently rolled back when the connection was closed. The implicit transaction opened by a failed statement is now rolled back before the exception is raised. A failed write inside a transaction opened with ``db.begin()`` or ``db.atomic()`` leaves that transaction open and untouched, as before. - Fixed a bug where transaction-control statements prefixed with an empty statement - ``db.query("; COMMIT")`` - or a UTF-8 byte order mark slipped past the check that rejects them, committing the caller's open transaction before raising a confusing ``OperationalError``. The keyword scanner used by ``db.query()`` and ``db.execute()`` now skips leading ``;`` and byte order marks, matching what the ``sqlite3`` driver tolerates before the first token, so these statements are rejected with a ``ValueError`` without being executed. The same fix means ``db.execute("; BEGIN")`` no longer auto-commits the transaction it just opened. - Documented a limitation of ``db.query()``: a ``PRAGMA`` statement that returns no rows raises a ``ValueError`` but still takes effect, because PRAGMA statements run outside the savepoint guard used to roll back other rejected statements. Use ``db.execute()`` for row-less PRAGMA statements. +- Fixed a bug where inserting CSV or TSV data into an existing table rewrote that table's column types to match the incoming file. Type detection is the default in 4.0, so ``sqlite-utils insert data.db places places.csv --csv`` against a table with a ``TEXT`` zip code column would convert the column to ``INTEGER`` and corrupt values with leading zeros - ``"01234"`` became ``1234``. Detected types are now only applied when the ``insert`` or ``upsert`` command creates the table. - Fixed ``pks_and_rows_where()`` raising ``AttributeError`` when called on a view, and no longer double-quotes the synthesized ``rowid`` column in its generated SQL - SQLite turns a double-quoted identifier that does not resolve into a string literal, which on a view produced a confusing ``KeyError`` instead of the ``OperationalError`` raised in 3.x. Compound primary keys returned by this method now follow ``PRIMARY KEY`` declaration order. - The ``foreign_keys=`` argument to ``create()`` and ``insert()`` accepts a mixed list of ``ForeignKey`` objects, tuples and column name strings again. In 4.0 pre-releases mixing ``ForeignKey`` objects with tuples raised a ``ValueError`` - a regression from 3.x, where ``ForeignKey`` was a ``namedtuple`` and passed the tuple checks. - ``ForeignKey`` objects are hashable again. The 4.0 change from ``namedtuple`` to dataclass accidentally made them unhashable, breaking patterns like ``set(table.foreign_keys)`` that worked in 3.x. ``ForeignKey`` is now a frozen dataclass - immutable and hashable, like the namedtuple was. diff --git a/docs/cli.rst b/docs/cli.rst index 1f95bbd..063e80f 100644 --- a/docs/cli.rst +++ b/docs/cli.rst @@ -1326,6 +1326,8 @@ A progress bar is displayed when inserting data from a file. You can hide the pr By default, column types are automatically detected for CSV or TSV files - resulting in a mix of ``TEXT``, ``INTEGER`` and ``REAL`` columns. To disable type detection and treat all columns as ``TEXT``, use the ``--no-detect-types`` option. +Detected types are only applied when the table is created by the command. Inserting CSV or TSV data into a table that already exists leaves the existing column types unchanged - values are inserted using the table's existing schema. + For example, given a ``creatures.csv`` file containing this: .. code-block:: diff --git a/sqlite_utils/cli.py b/sqlite_utils/cli.py index 8ee2091..0ef7cc4 100644 --- a/sqlite_utils/cli.py +++ b/sqlite_utils/cli.py @@ -1165,6 +1165,9 @@ def insert_upsert_implementation( db.conn.cursor().executemany(bulk_sql, doc_chunk) return + # table_names() rather than db.table(), which raises NoTable for + # views before the error handling below can deal with them + table_existed_before_insert = table in db.table_names() try: db.table(table).insert_all( docs, pk=pk, batch_size=batch_size, alter=alter, **extra_kwargs @@ -1194,7 +1197,14 @@ def insert_upsert_implementation( ) else: raise - if tracker is not None and db.table(table).exists(): + # Apply detected types only to a table this command created - + # transforming a pre-existing table would rewrite its column types + # and corrupt values such as TEXT zip codes with leading zeros + if ( + tracker is not None + and not table_existed_before_insert + and db.table(table).exists() + ): db.table(table).transform(types=tracker.types) # Clean up open file-like objects diff --git a/tests/test_cli_insert.py b/tests/test_cli_insert.py index 2df1e0c..196590e 100644 --- a/tests/test_cli_insert.py +++ b/tests/test_cli_insert.py @@ -628,3 +628,51 @@ def test_insert_into_view_errors(tmpdir): ) assert result.exit_code == 1 assert result.output.strip() == "Error: Table v is actually a view" + + +def test_insert_csv_detect_types_leaves_existing_table_alone(db_path): + # Type detection is the default for CSV/TSV inserts, but it must only + # apply to tables created by this command - transforming a pre-existing + # table would rewrite its column types and corrupt data such as + # TEXT zip codes with leading zeros + db = Database(db_path) + db["places"].insert({"name": "Boston", "zip": "01234"}) + result = CliRunner().invoke( + cli.cli, + ["insert", db_path, "places", "-", "--csv"], + catch_exceptions=False, + input="name,zip\nSF,94107", + ) + assert result.exit_code == 0, result.output + assert db["places"].columns_dict["zip"] is str + assert list(db["places"].rows) == [ + {"name": "Boston", "zip": "01234"}, + {"name": "SF", "zip": "94107"}, + ] + + +def test_insert_csv_detect_types_new_table(db_path): + # A table created by the insert still gets detected types + result = CliRunner().invoke( + cli.cli, + ["insert", db_path, "data", "-", "--csv"], + catch_exceptions=False, + input="name,age,weight\nCleo,5,12.5", + ) + assert result.exit_code == 0, result.output + db = Database(db_path) + assert db["data"].columns_dict == {"name": str, "age": int, "weight": float} + + +def test_upsert_csv_detect_types_leaves_existing_table_alone(db_path): + db = Database(db_path) + db["places"].insert({"id": 1, "name": "Boston", "zip": "01234"}, pk="id") + result = CliRunner().invoke( + cli.cli, + ["upsert", db_path, "places", "-", "--csv", "--pk", "id"], + catch_exceptions=False, + input="id,name,zip\n2,SF,94107", + ) + assert result.exit_code == 0, result.output + assert db["places"].columns_dict["zip"] is str + assert db["places"].get(1)["zip"] == "01234" From 3de8507c6b7d28288d220d5627d4efe4f21b7abc Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Mon, 6 Jul 2026 21:47:59 -0700 Subject: [PATCH 48/86] insert(pk=, alter=True) can add the pk column from records again The InvalidColumns check added for #732 fired before alter=True had a chance to add the missing pk column - a regression from 3.x, where insert({"id": 5, "a": 2}, pk="id", alter=True) against a table without an id column worked. With alter=True the check is now deferred until the first batch of record keys is known: a pk column supplied by the records passes, one found in neither the table nor the records still raises InvalidColumns before anything is inserted. An empty record iterator with alter=True returns without error, matching the 3.x no-op. Refs https://github.com/simonw/sqlite-utils/issues/769#issuecomment-4900034150 Co-Authored-By: Claude Fable 5 --- docs/changelog.rst | 1 + sqlite_utils/db.py | 19 ++++++++++++++++++- tests/test_create.py | 37 +++++++++++++++++++++++++++++++++++++ 3 files changed, 56 insertions(+), 1 deletion(-) diff --git a/docs/changelog.rst b/docs/changelog.rst index 80bcfb7..df41884 100644 --- a/docs/changelog.rst +++ b/docs/changelog.rst @@ -19,6 +19,7 @@ Unreleased - Fixed a bug where a failed write statement executed with ``db.execute()`` left the driver's implicit transaction open. Every subsequent write then joined that phantom transaction, which nothing committed, so their work was silently rolled back when the connection was closed. The implicit transaction opened by a failed statement is now rolled back before the exception is raised. A failed write inside a transaction opened with ``db.begin()`` or ``db.atomic()`` leaves that transaction open and untouched, as before. - Fixed a bug where transaction-control statements prefixed with an empty statement - ``db.query("; COMMIT")`` - or a UTF-8 byte order mark slipped past the check that rejects them, committing the caller's open transaction before raising a confusing ``OperationalError``. The keyword scanner used by ``db.query()`` and ``db.execute()`` now skips leading ``;`` and byte order marks, matching what the ``sqlite3`` driver tolerates before the first token, so these statements are rejected with a ``ValueError`` without being executed. The same fix means ``db.execute("; BEGIN")`` no longer auto-commits the transaction it just opened. - Documented a limitation of ``db.query()``: a ``PRAGMA`` statement that returns no rows raises a ``ValueError`` but still takes effect, because PRAGMA statements run outside the savepoint guard used to roll back other rejected statements. Use ``db.execute()`` for row-less PRAGMA statements. +- Fixed a regression where ``table.insert(..., pk=..., alter=True)`` raised ``InvalidColumns`` if the primary key column did not exist in the table yet. With ``alter=True`` the check now waits until the record keys are known, so a pk column supplied by the records is added by the alter as it was in 3.x. A pk column found in neither the table nor the records still raises ``InvalidColumns``. - Fixed a bug where inserting CSV or TSV data into an existing table rewrote that table's column types to match the incoming file. Type detection is the default in 4.0, so ``sqlite-utils insert data.db places places.csv --csv`` against a table with a ``TEXT`` zip code column would convert the column to ``INTEGER`` and corrupt values with leading zeros - ``"01234"`` became ``1234``. Detected types are now only applied when the ``insert`` or ``upsert`` command creates the table. - Fixed ``pks_and_rows_where()`` raising ``AttributeError`` when called on a view, and no longer double-quotes the synthesized ``rowid`` column in its generated SQL - SQLite turns a double-quoted identifier that does not resolve into a string literal, which on a view produced a confusing ``KeyError`` instead of the ``OperationalError`` raised in 3.x. Compound primary keys returned by this method now follow ``PRIMARY KEY`` declaration order. - The ``foreign_keys=`` argument to ``create()`` and ``insert()`` accepts a mixed list of ``ForeignKey`` objects, tuples and column name strings again. In 4.0 pre-releases mixing ``ForeignKey`` objects with tuples raised a ``ValueError`` - a regression from 3.x, where ``ForeignKey`` was a ``namedtuple`` and passed the tuple checks. diff --git a/sqlite_utils/db.py b/sqlite_utils/db.py index 4c7ef52..faf588f 100644 --- a/sqlite_utils/db.py +++ b/sqlite_utils/db.py @@ -4282,6 +4282,10 @@ class Table(Queryable): if hash_id: pk = hash_id + # pk columns missing from an existing table are an error - unless + # alter=True, where a pk column supplied by the records will be + # added, so validation waits until the record keys are known + deferred_invalid_pk_check = None if pk and not hash_id and self.exists(): pk_cols = [pk] if isinstance(pk, str) else list(pk) existing_columns = self.columns_dict @@ -4291,7 +4295,7 @@ class Table(Queryable): if resolve_casing(col, existing_columns) not in existing_columns ] if missing_pk_cols: - raise InvalidColumns( + invalid_pk_error = InvalidColumns( "Invalid primary key column{} {} for table {} with columns {}".format( "s" if len(missing_pk_cols) > 1 else "", missing_pk_cols, @@ -4299,6 +4303,9 @@ class Table(Queryable): list(existing_columns), ) ) + if not alter: + raise invalid_pk_error + deferred_invalid_pk_check = (missing_pk_cols, invalid_pk_error) if ignore and replace: raise ValueError("Use either ignore=True or replace=True, not both") @@ -4409,6 +4416,16 @@ class Table(Queryable): all_columns = list(sorted(all_columns_set)) if hash_id: all_columns.insert(0, hash_id) + if deferred_invalid_pk_check is not None: + # alter=True - pk columns the table lacks are valid if + # the records supply them, otherwise raise the error + missing_pk_cols, invalid_pk_error = deferred_invalid_pk_check + record_columns = {column: True for column in all_columns} + if any( + resolve_casing(col, record_columns) not in record_columns + for col in missing_pk_cols + ): + raise invalid_pk_error else: if not list_mode: for record in cast(List[Dict[str, Any]], chunk): diff --git a/tests/test_create.py b/tests/test_create.py index 2788920..42fa1c4 100644 --- a/tests/test_create.py +++ b/tests/test_create.py @@ -932,6 +932,23 @@ def test_insert_all_pk_not_in_records_raises(fresh_db, num_rows): fresh_db.conn.execute("CREATE TABLE t (a TEXT, b INT, PRIMARY KEY (a, b))") rows = [{"a": "x{}".format(i), "b": i} for i in range(num_rows)] + with pytest.raises(InvalidColumns) as ex: + fresh_db["t"].insert_all(rows, pk="not_a_column") + + assert ex.value.args == ( + "Invalid primary key column ['not_a_column'] for table t with columns ['a', 'b']", + ) + assert fresh_db["t"].count == 0 + + +@pytest.mark.parametrize("num_rows", (1, 2, 3, 10)) +def test_insert_all_pk_not_in_records_alter_raises(fresh_db, num_rows): + # With alter=True the check is deferred until the record keys are + # known - a pk column that is in neither the table nor the records + # still raises + fresh_db.conn.execute("CREATE TABLE t (a TEXT, b INT, PRIMARY KEY (a, b))") + rows = [{"a": "x{}".format(i), "b": i} for i in range(num_rows)] + with pytest.raises(InvalidColumns) as ex: fresh_db["t"].insert_all(rows, pk="not_a_column", alter=True) @@ -941,6 +958,26 @@ def test_insert_all_pk_not_in_records_raises(fresh_db, num_rows): assert fresh_db["t"].count == 0 +def test_insert_pk_in_records_with_alter_adds_column(fresh_db): + # 3.x allowed insert(pk=..., alter=True) to add the pk column from the + # records - the InvalidColumns check must not fire in that case + fresh_db["t"].insert({"a": 1}) + fresh_db["t"].insert({"id": 5, "a": 2}, pk="id", alter=True) + assert fresh_db["t"].columns_dict.keys() == {"a", "id"} + assert list(fresh_db.query("select * from t order by a")) == [ + {"a": 1, "id": None}, + {"a": 2, "id": 5}, + ] + + +def test_insert_all_invalid_pk_alter_empty_records_is_noop(fresh_db): + # With alter=True the pk check needs record keys, so an empty iterator + # returns without error - matching the 3.x no-op for empty inserts + fresh_db.conn.execute("CREATE TABLE t (a TEXT)") + fresh_db["t"].insert_all([], pk="not_a_column", alter=True) + assert fresh_db["t"].count == 0 + + def test_insert_ignore(fresh_db): fresh_db["test"].insert({"id": 1, "bar": 2}, pk="id") # Should raise an error if we try this again From b3aa3f47b717bba2c166f11e59a35bcd2f8e6b09 Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Mon, 6 Jul 2026 21:50:05 -0700 Subject: [PATCH 49/86] migrate --stop-before an already-applied migration is now an error The CLI validated --stop-before names against both pending and applied migrations, but Migrations.apply() only looked for the stop name among pending ones - naming an applied migration passed validation and then silently applied every migration after it, the exact outcome the option exists to prevent. apply() now raises ValueError before applying anything if a stop_before name matches an applied migration in its set; names not in the set are still ignored, since unqualified CLI values are offered to every set. The migrate command reports it as a clean error. Refs https://github.com/simonw/sqlite-utils/issues/769#issuecomment-4900034150 Co-Authored-By: Claude Fable 5 --- docs/changelog.rst | 1 + docs/migrations.rst | 2 +- sqlite_utils/cli.py | 5 ++++- sqlite_utils/migrations.py | 22 +++++++++++++++++++++- tests/test_cli_migrate.py | 22 ++++++++++++++++++++++ tests/test_migrations.py | 30 ++++++++++++++++++++++++++++++ 6 files changed, 79 insertions(+), 3 deletions(-) diff --git a/docs/changelog.rst b/docs/changelog.rst index df41884..4339e81 100644 --- a/docs/changelog.rst +++ b/docs/changelog.rst @@ -19,6 +19,7 @@ Unreleased - Fixed a bug where a failed write statement executed with ``db.execute()`` left the driver's implicit transaction open. Every subsequent write then joined that phantom transaction, which nothing committed, so their work was silently rolled back when the connection was closed. The implicit transaction opened by a failed statement is now rolled back before the exception is raised. A failed write inside a transaction opened with ``db.begin()`` or ``db.atomic()`` leaves that transaction open and untouched, as before. - Fixed a bug where transaction-control statements prefixed with an empty statement - ``db.query("; COMMIT")`` - or a UTF-8 byte order mark slipped past the check that rejects them, committing the caller's open transaction before raising a confusing ``OperationalError``. The keyword scanner used by ``db.query()`` and ``db.execute()`` now skips leading ``;`` and byte order marks, matching what the ``sqlite3`` driver tolerates before the first token, so these statements are rejected with a ``ValueError`` without being executed. The same fix means ``db.execute("; BEGIN")`` no longer auto-commits the transaction it just opened. - Documented a limitation of ``db.query()``: a ``PRAGMA`` statement that returns no rows raises a ``ValueError`` but still takes effect, because PRAGMA statements run outside the savepoint guard used to roll back other rejected statements. Use ``db.execute()`` for row-less PRAGMA statements. +- ``sqlite-utils migrate --stop-before`` now exits with an error if the named migration has already been applied. Previously the name passed validation but was only checked against pending migrations, so every migration after it was silently applied - the exact outcome ``--stop-before`` exists to prevent. ``Migrations.apply(db, stop_before=...)`` raises ``ValueError`` in the same situation, before applying anything. - Fixed a regression where ``table.insert(..., pk=..., alter=True)`` raised ``InvalidColumns`` if the primary key column did not exist in the table yet. With ``alter=True`` the check now waits until the record keys are known, so a pk column supplied by the records is added by the alter as it was in 3.x. A pk column found in neither the table nor the records still raises ``InvalidColumns``. - Fixed a bug where inserting CSV or TSV data into an existing table rewrote that table's column types to match the incoming file. Type detection is the default in 4.0, so ``sqlite-utils insert data.db places places.csv --csv`` against a table with a ``TEXT`` zip code column would convert the column to ``INTEGER`` and corrupt values with leading zeros - ``"01234"`` became ``1234``. Detected types are now only applied when the ``insert`` or ``upsert`` command creates the table. - Fixed ``pks_and_rows_where()`` raising ``AttributeError`` when called on a view, and no longer double-quotes the synthesized ``rowid`` column in its generated SQL - SQLite turns a double-quoted identifier that does not resolve into a string literal, which on a view produced a confusing ``KeyError`` instead of the ``OperationalError`` raised in 3.x. Compound primary keys returned by this method now follow ``PRIMARY KEY`` declaration order. diff --git a/docs/migrations.rst b/docs/migrations.rst index e685936..cfdbf13 100644 --- a/docs/migrations.rst +++ b/docs/migrations.rst @@ -157,7 +157,7 @@ You can also target a specific migration set using ``migration_set:migration_nam The ``--stop-before`` option can be passed more than once. -If a ``--stop-before`` value does not match any known migration the command exits with an error, rather than silently applying everything. +If a ``--stop-before`` value does not match any known migration the command exits with an error, rather than silently applying everything. Naming a migration that has already been applied is also an error - stopping before it is impossible to honor - and no pending migrations are applied. Verbose output ============== diff --git a/sqlite_utils/cli.py b/sqlite_utils/cli.py index 0ef7cc4..188eff6 100644 --- a/sqlite_utils/cli.py +++ b/sqlite_utils/cli.py @@ -3462,7 +3462,10 @@ def migrate(db_path, migrations, stop_before, list_, verbose): for migration_set in migration_sets: matches = _stop_before_for_migration_set(stop_before, migration_set.name) if isinstance(migration_set, sqlite_utils.Migrations): - migration_set.apply(db, stop_before=matches) + try: + migration_set.apply(db, stop_before=matches) + except ValueError as e: + raise click.ClickException(str(e)) else: # Legacy sqlite-migrate Migrations objects take a single string # for stop_before, not a list diff --git a/sqlite_utils/migrations.py b/sqlite_utils/migrations.py index 36e053d..00d0fa5 100644 --- a/sqlite_utils/migrations.py +++ b/sqlite_utils/migrations.py @@ -96,14 +96,34 @@ class Migrations: changes are rolled back, no record is written and the migration stays pending. Migrations registered with ``transactional=False`` run outside of a transaction. + + :raises ValueError: if a ``stop_before`` name matches a migration in + this set that has already been applied - stopping before it is + impossible to honor, and no pending migrations are applied """ - self.ensure_migrations_table(db) if stop_before is None: stop_before_names = set() elif isinstance(stop_before, str): stop_before_names = {stop_before} else: stop_before_names = set(stop_before) + # A stop_before naming an already-applied migration cannot be + # honored - error rather than applying everything after it. Names + # not in this set at all are ignored, because unqualified CLI + # values are offered to every migration set + already_applied = stop_before_names.intersection( + migration.name for migration in self.applied(db) + ) + if already_applied: + raise ValueError( + "Cannot stop before migration{} {} in set '{}' - already " + "been applied".format( + "s" if len(already_applied) > 1 else "", + ", ".join(sorted(already_applied)), + self.name, + ) + ) + self.ensure_migrations_table(db) for migration in self.pending(db): name = migration.name if name in stop_before_names: diff --git a/tests/test_cli_migrate.py b/tests/test_cli_migrate.py index 1cdf8e7..0f1c7ea 100644 --- a/tests/test_cli_migrate.py +++ b/tests/test_cli_migrate.py @@ -463,3 +463,25 @@ def test_list_does_not_upgrade_legacy_migrations_table(two_migrations): db2 = sqlite_utils.Database(db_path) assert db2["_sqlite_migrations"].pks == ["migration_set", "name"] db2.close() + + +def test_stop_before_applied_migration_errors(two_migrations): + path, _ = two_migrations + db_path = str(path / "test.db") + migrations_path = str(path / "foo" / "migrations.py") + # Apply everything first + first = CliRunner().invoke( + sqlite_utils.cli.cli, + ["migrate", db_path, migrations_path, "--stop-before", "bar"], + ) + assert first.exit_code == 0 + # foo is now applied - stopping before it is an error, and bar + # must not be applied as a side effect + result = CliRunner().invoke( + sqlite_utils.cli.cli, + ["migrate", db_path, migrations_path, "--stop-before", "foo"], + ) + assert result.exit_code != 0 + assert "already been applied" in result.output + db = sqlite_utils.Database(db_path) + assert not db["bar"].exists() diff --git a/tests/test_migrations.py b/tests/test_migrations.py index 5634d79..04185fc 100644 --- a/tests/test_migrations.py +++ b/tests/test_migrations.py @@ -214,3 +214,33 @@ def test_duplicate_migration_name_errors(): pass assert "m001" in str(ex.value) + + +def test_stop_before_applied_migration_errors(migrations): + # Stopping before a migration that has already been applied is + # impossible to honor - previously the stop name was only checked + # against pending migrations, so everything after it was applied + db = sqlite_utils.Database(memory=True) + migrations.apply(db, stop_before="m002") # applies m001 only + with pytest.raises(ValueError) as ex: + migrations.apply(db, stop_before="m001") + assert "m001" in str(ex.value) + assert "already been applied" in str(ex.value) + # Nothing else was applied + assert not db["cats"].exists() + + +def test_stop_before_applied_migration_errors_before_any_apply(migrations): + # The error fires before any pending migration runs, even those that + # come before the already-applied stop target in registration order + db = sqlite_utils.Database(memory=True) + only_second = Migrations("test") + + @only_second() + def m002(db): + db["cats"].create({"name": str}) + + only_second.apply(db) # m002 applied, m001 still pending + with pytest.raises(ValueError): + migrations.apply(db, stop_before="m002") + assert not db["dogs"].exists() From a0387791e511a17eb96e0ed667da222550a00412 Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Mon, 6 Jul 2026 21:51:31 -0700 Subject: [PATCH 50/86] ensure_autocommit_on() raises TransactionError inside a transaction Assigning conn.isolation_level commits any pending transaction as a side effect, so entering the context manager with a transaction open silently committed the caller's work and made a later rollback() a no-op. All internal callers already ensure no transaction is open; the public API now enforces it, matching enable_wal() and disable_wal(). Refs https://github.com/simonw/sqlite-utils/issues/769#issuecomment-4900034150 Co-Authored-By: Claude Fable 5 --- docs/changelog.rst | 1 + sqlite_utils/db.py | 9 +++++++++ tests/test_wal.py | 17 +++++++++++++++++ 3 files changed, 27 insertions(+) diff --git a/docs/changelog.rst b/docs/changelog.rst index 4339e81..87a1ab7 100644 --- a/docs/changelog.rst +++ b/docs/changelog.rst @@ -19,6 +19,7 @@ Unreleased - Fixed a bug where a failed write statement executed with ``db.execute()`` left the driver's implicit transaction open. Every subsequent write then joined that phantom transaction, which nothing committed, so their work was silently rolled back when the connection was closed. The implicit transaction opened by a failed statement is now rolled back before the exception is raised. A failed write inside a transaction opened with ``db.begin()`` or ``db.atomic()`` leaves that transaction open and untouched, as before. - Fixed a bug where transaction-control statements prefixed with an empty statement - ``db.query("; COMMIT")`` - or a UTF-8 byte order mark slipped past the check that rejects them, committing the caller's open transaction before raising a confusing ``OperationalError``. The keyword scanner used by ``db.query()`` and ``db.execute()`` now skips leading ``;`` and byte order marks, matching what the ``sqlite3`` driver tolerates before the first token, so these statements are rejected with a ``ValueError`` without being executed. The same fix means ``db.execute("; BEGIN")`` no longer auto-commits the transaction it just opened. - Documented a limitation of ``db.query()``: a ``PRAGMA`` statement that returns no rows raises a ``ValueError`` but still takes effect, because PRAGMA statements run outside the savepoint guard used to roll back other rejected statements. Use ``db.execute()`` for row-less PRAGMA statements. +- ``db.ensure_autocommit_on()`` now raises ``TransactionError`` if called while a transaction is open. Assigning ``isolation_level`` commits any pending transaction as a side effect, so entering the block silently committed the caller's open transaction and made a later ``rollback()`` a no-op. - ``sqlite-utils migrate --stop-before`` now exits with an error if the named migration has already been applied. Previously the name passed validation but was only checked against pending migrations, so every migration after it was silently applied - the exact outcome ``--stop-before`` exists to prevent. ``Migrations.apply(db, stop_before=...)`` raises ``ValueError`` in the same situation, before applying anything. - Fixed a regression where ``table.insert(..., pk=..., alter=True)`` raised ``InvalidColumns`` if the primary key column did not exist in the table yet. With ``alter=True`` the check now waits until the record keys are known, so a pk column supplied by the records is added by the alter as it was in 3.x. A pk column found in neither the table nor the records still raises ``InvalidColumns``. - Fixed a bug where inserting CSV or TSV data into an existing table rewrote that table's column types to match the incoming file. Type detection is the default in 4.0, so ``sqlite-utils insert data.db places places.csv --csv`` against a table with a ``TEXT`` zip code column would convert the column to ``INTEGER`` and corrupt values with leading zeros - ``"01234"`` became ``1234``. Detected types are now only applied when the ``insert`` or ``upsert`` command creates the table. diff --git a/sqlite_utils/db.py b/sqlite_utils/db.py index faf588f..90e7a83 100644 --- a/sqlite_utils/db.py +++ b/sqlite_utils/db.py @@ -665,7 +665,16 @@ class Database: # do stuff here The previous ``isolation_level`` is restored at the end of the block. + + :raises TransactionError: if a transaction is open - assigning + ``isolation_level`` would commit it as a side effect, silently + breaking the caller's ability to roll back """ + if self.conn.in_transaction: + raise TransactionError( + "ensure_autocommit_on() cannot be used inside a transaction - " + "changing isolation_level would commit the open transaction" + ) old_isolation_level = self.conn.isolation_level try: self.conn.isolation_level = None diff --git a/tests/test_wal.py b/tests/test_wal.py index c5a9c60..2ddcf54 100644 --- a/tests/test_wal.py +++ b/tests/test_wal.py @@ -69,3 +69,20 @@ def test_enable_wal_noop_inside_transaction_is_allowed(db_path_tmpdir): db["test"].insert({"id": 1}, pk="id") db.enable_wal() assert [r["id"] for r in db["test"].rows] == [1] + + +def test_ensure_autocommit_on_inside_transaction_raises(db_path_tmpdir): + # Setting isolation_level commits any pending transaction as a side + # effect, silently breaking the caller's rollback guarantee - so + # entering autocommit mode with a transaction open is an error + db, path, tmpdir = db_path_tmpdir + db["test"].insert({"id": 1}, pk="id") + db.begin() + db.execute("insert into test (id) values (2)") + with pytest.raises(TransactionError): + with db.ensure_autocommit_on(): + pass + # The transaction is still open and can still be rolled back + assert db.conn.in_transaction + db.rollback() + assert [r["id"] for r in db["test"].rows] == [1] From 16bbfb582d6acf33a8c05317b64d93c85e184a5c Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Mon, 6 Jul 2026 21:53:48 -0700 Subject: [PATCH 51/86] add_foreign_keys() errors instead of silently dropping action changes The existing-foreign-key dedup compared only columns and other_table, so requesting an FK that already exists with different ON DELETE/ON UPDATE actions was a silent no-op - and with add_foreign_key() raising "already exists", there was no signal that the requested actions were dropped. An exact match (including actions) is still skipped for idempotency; a mismatch now raises AlterError pointing at table.transform(). Also validates that compound foreign keys passed as 4-tuples have the same number of columns on both sides - extra other-columns were being silently discarded, e.g. ("t", ("a",), "other", ("a", "b")) created a single-column key referencing just "a". Refs https://github.com/simonw/sqlite-utils/issues/769#issuecomment-4900034150 Co-Authored-By: Claude Fable 5 --- docs/changelog.rst | 1 + docs/python-api.rst | 2 ++ sqlite_utils/db.py | 25 ++++++++++++++++++--- tests/test_foreign_keys.py | 46 ++++++++++++++++++++++++++++++++++++++ 4 files changed, 71 insertions(+), 3 deletions(-) diff --git a/docs/changelog.rst b/docs/changelog.rst index 87a1ab7..ff78738 100644 --- a/docs/changelog.rst +++ b/docs/changelog.rst @@ -19,6 +19,7 @@ Unreleased - Fixed a bug where a failed write statement executed with ``db.execute()`` left the driver's implicit transaction open. Every subsequent write then joined that phantom transaction, which nothing committed, so their work was silently rolled back when the connection was closed. The implicit transaction opened by a failed statement is now rolled back before the exception is raised. A failed write inside a transaction opened with ``db.begin()`` or ``db.atomic()`` leaves that transaction open and untouched, as before. - Fixed a bug where transaction-control statements prefixed with an empty statement - ``db.query("; COMMIT")`` - or a UTF-8 byte order mark slipped past the check that rejects them, committing the caller's open transaction before raising a confusing ``OperationalError``. The keyword scanner used by ``db.query()`` and ``db.execute()`` now skips leading ``;`` and byte order marks, matching what the ``sqlite3`` driver tolerates before the first token, so these statements are rejected with a ``ValueError`` without being executed. The same fix means ``db.execute("; BEGIN")`` no longer auto-commits the transaction it just opened. - Documented a limitation of ``db.query()``: a ``PRAGMA`` statement that returns no rows raises a ``ValueError`` but still takes effect, because PRAGMA statements run outside the savepoint guard used to roll back other rejected statements. Use ``db.execute()`` for row-less PRAGMA statements. +- ``db.add_foreign_keys()`` no longer silently ignores requested ``ON DELETE``/``ON UPDATE`` actions when a foreign key with the same columns already exists - it raises ``AlterError`` suggesting ``table.transform()``, since the actions of an existing foreign key cannot be changed in place. Exact duplicates, including actions, are still skipped so repeated calls stay idempotent. The method also now validates that compound foreign keys have the same number of columns on both sides, instead of silently discarding the extra columns. - ``db.ensure_autocommit_on()`` now raises ``TransactionError`` if called while a transaction is open. Assigning ``isolation_level`` commits any pending transaction as a side effect, so entering the block silently committed the caller's open transaction and made a later ``rollback()`` a no-op. - ``sqlite-utils migrate --stop-before`` now exits with an error if the named migration has already been applied. Previously the name passed validation but was only checked against pending migrations, so every migration after it was silently applied - the exact outcome ``--stop-before`` exists to prevent. ``Migrations.apply(db, stop_before=...)`` raises ``ValueError`` in the same situation, before applying anything. - Fixed a regression where ``table.insert(..., pk=..., alter=True)`` raised ``InvalidColumns`` if the primary key column did not exist in the table yet. With ``alter=True`` the check now waits until the record keys are known, so a pk column supplied by the records is added by the alter as it was in 3.x. A pk column found in neither the table nor the records still raises ``InvalidColumns``. diff --git a/docs/python-api.rst b/docs/python-api.rst index 516e2aa..11af1f4 100644 --- a/docs/python-api.rst +++ b/docs/python-api.rst @@ -1636,6 +1636,8 @@ Here's an example adding two foreign keys at once: This method runs the same checks as ``.add_foreign_keys()`` and will raise ``sqlite_utils.db.AlterError`` if those checks fail. +Foreign keys that already exist are silently skipped, so repeated calls are idempotent - but only if they match exactly. Requesting a foreign key that exists with different ``ON DELETE``/``ON UPDATE`` actions raises ``AlterError``: use ``table.transform()`` to change the actions of an existing foreign key. + .. _python_api_index_foreign_keys: Adding indexes for all foreign keys diff --git a/sqlite_utils/db.py b/sqlite_utils/db.py index 90e7a83..eb59c50 100644 --- a/sqlite_utils/db.py +++ b/sqlite_utils/db.py @@ -1786,6 +1786,11 @@ class Database: if isinstance(other_column_or_columns, str) else tuple(other_column_or_columns) ) + if len(columns) != len(other_columns): + raise ValueError( + "Compound foreign key must have the same number of " + "columns on both sides" + ) if len(columns) == 1: fk_object = ForeignKey( table, columns[0], other_table, other_columns[0] @@ -1825,10 +1830,11 @@ class Database: other_column, other_table ) ) - # We will silently skip foreign keys that exist already + # Silently skip foreign keys that exist already - but only if + # they match exactly, including ON DELETE/ON UPDATE actions columns_folded = tuple(fold_identifier_case(c) for c in columns) other_columns_folded = tuple(fold_identifier_case(c) for c in other_columns) - if not any( + existing = [ fk for fk in table_obj.foreign_keys if tuple(fold_identifier_case(c) for c in fk.columns) == columns_folded @@ -1836,8 +1842,21 @@ class Database: == fold_identifier_case(other_table) and tuple(fold_identifier_case(c) for c in fk.other_columns) == other_columns_folded - ): + ] + if not existing: foreign_keys_to_create.append(fk_object) + elif any( + fk.on_delete != fk_object.on_delete + or fk.on_update != fk_object.on_update + for fk in existing + ): + raise AlterError( + "Foreign key already exists for {} => {}.{} but with " + "different ON DELETE/ON UPDATE actions - use " + "table.transform() to change them".format( + ", ".join(columns), other_table, ", ".join(other_columns) + ) + ) # Group them by table by_table: Dict[str, List[ForeignKey]] = {} diff --git a/tests/test_foreign_keys.py b/tests/test_foreign_keys.py index d7c20f6..b37d374 100644 --- a/tests/test_foreign_keys.py +++ b/tests/test_foreign_keys.py @@ -643,3 +643,49 @@ def test_create_table_mixed_foreign_keys_with_string(fresh_db): ) fks = {fk.column: fk.other_table for fk in fresh_db["books"].foreign_keys} assert fks == {"author_id": "authors", "publisher_id": "publishers"} + + +def test_add_foreign_keys_existing_with_different_actions_errors(fresh_db): + # Requesting an existing foreign key with different ON DELETE/ON UPDATE + # actions was silently skipped, dropping the requested change + fresh_db["authors"].insert({"id": 1}, pk="id") + fresh_db["books"].insert( + {"id": 1, "author_id": 1}, + pk="id", + foreign_keys=[("author_id", "authors", "id")], + ) + with pytest.raises(AlterError) as ex: + fresh_db.add_foreign_keys( + [ForeignKey("books", "author_id", "authors", "id", on_delete="CASCADE")] + ) + assert "ON DELETE" in str(ex.value) + assert fresh_db["books"].foreign_keys[0].on_delete == "NO ACTION" + + +def test_add_foreign_keys_identical_existing_is_noop(fresh_db): + # An exact match, including actions, is silently skipped so repeated + # calls stay idempotent + fresh_db["authors"].insert({"id": 1}, pk="id") + fresh_db["books"].insert({"id": 1, "author_id": 1}, pk="id") + fresh_db["books"].add_foreign_key("author_id", "authors", "id", on_delete="CASCADE") + fresh_db.add_foreign_keys( + [ForeignKey("books", "author_id", "authors", "id", on_delete="CASCADE")] + ) + fks = fresh_db["books"].foreign_keys + assert len(fks) == 1 + assert fks[0].on_delete == "CASCADE" + + +def test_add_foreign_keys_compound_column_count_mismatch_errors(fresh_db): + # Previously the extra other-column was silently discarded, creating + # a single-column foreign key to just ("id") + fresh_db["departments"].insert( + {"campus": "north", "code": "cs"}, pk=("campus", "code") + ) + fresh_db["courses"].insert({"id": 1, "campus": "north"}, pk="id") + with pytest.raises(ValueError) as ex: + fresh_db.add_foreign_keys( + [("courses", ("campus",), "departments", ("campus", "code"))] + ) + assert "same number of columns" in str(ex.value) + assert fresh_db["courses"].foreign_keys == [] From 8572d1e39c3c807bc0643249411fb33bd0881eb6 Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Mon, 6 Jul 2026 21:55:21 -0700 Subject: [PATCH 52/86] extract() no longer duplicates NULL-containing rows in shared lookups The lookup-table insert relied on INSERT OR IGNORE and the unique index to dedupe against existing rows, but SQLite unique indexes treat NULLs as distinct - extracting a second table into the same lookup table re-inserted every NULL-containing value, growing orphan rows on each extract. The insert now also has an IS-based NOT EXISTS guard, matching how the foreign keys themselves are resolved. Refs https://github.com/simonw/sqlite-utils/issues/769#issuecomment-4900034150 Co-Authored-By: Claude Fable 5 --- docs/changelog.rst | 1 + sqlite_utils/db.py | 18 +++++++++++++++++- tests/test_extract.py | 31 +++++++++++++++++++++++++++++++ 3 files changed, 49 insertions(+), 1 deletion(-) diff --git a/docs/changelog.rst b/docs/changelog.rst index ff78738..d2b8a67 100644 --- a/docs/changelog.rst +++ b/docs/changelog.rst @@ -19,6 +19,7 @@ Unreleased - Fixed a bug where a failed write statement executed with ``db.execute()`` left the driver's implicit transaction open. Every subsequent write then joined that phantom transaction, which nothing committed, so their work was silently rolled back when the connection was closed. The implicit transaction opened by a failed statement is now rolled back before the exception is raised. A failed write inside a transaction opened with ``db.begin()`` or ``db.atomic()`` leaves that transaction open and untouched, as before. - Fixed a bug where transaction-control statements prefixed with an empty statement - ``db.query("; COMMIT")`` - or a UTF-8 byte order mark slipped past the check that rejects them, committing the caller's open transaction before raising a confusing ``OperationalError``. The keyword scanner used by ``db.query()`` and ``db.execute()`` now skips leading ``;`` and byte order marks, matching what the ``sqlite3`` driver tolerates before the first token, so these statements are rejected with a ``ValueError`` without being executed. The same fix means ``db.execute("; BEGIN")`` no longer auto-commits the transaction it just opened. - Documented a limitation of ``db.query()``: a ``PRAGMA`` statement that returns no rows raises a ``ValueError`` but still takes effect, because PRAGMA statements run outside the savepoint guard used to roll back other rejected statements. Use ``db.execute()`` for row-less PRAGMA statements. +- Fixed a bug where running ``table.extract()`` more than once against the same lookup table inserted duplicate rows for values containing ``null`` - SQLite unique indexes treat ``NULL`` values as distinct, so ``INSERT OR IGNORE`` alone could not dedupe them. Each repeat extract added another copy that nothing referenced. The insert now uses an ``IS``-based ``NOT EXISTS`` guard so ``null``-containing rows match existing lookup rows. - ``db.add_foreign_keys()`` no longer silently ignores requested ``ON DELETE``/``ON UPDATE`` actions when a foreign key with the same columns already exists - it raises ``AlterError`` suggesting ``table.transform()``, since the actions of an existing foreign key cannot be changed in place. Exact duplicates, including actions, are still skipped so repeated calls stay idempotent. The method also now validates that compound foreign keys have the same number of columns on both sides, instead of silently discarding the extra columns. - ``db.ensure_autocommit_on()`` now raises ``TransactionError`` if called while a transaction is open. Assigning ``isolation_level`` commits any pending transaction as a side effect, so entering the block silently committed the caller's open transaction and made a later ``rollback()`` a no-op. - ``sqlite-utils migrate --stop-before`` now exits with an error if the named migration has already been applied. Previously the name passed validation but was only checked against pending migrations, so every migration after it was silently applied - the exact outcome ``--stop-before`` exists to prevent. ``Migrations.apply(db, stop_before=...)`` raises ``ValueError`` in the same situation, before applying anything. diff --git a/sqlite_utils/db.py b/sqlite_utils/db.py index eb59c50..e69f996 100644 --- a/sqlite_utils/db.py +++ b/sqlite_utils/db.py @@ -2919,8 +2919,22 @@ class Table(Queryable): all_columns_are_null = " AND ".join( "{} IS NULL".format(quote_identifier(c)) for c in columns ) + # INSERT OR IGNORE dedupes against the unique index, but unique + # indexes treat NULLs as distinct - the NOT EXISTS guard uses IS + # comparison so NULL-containing rows match existing lookup rows + # instead of being inserted again + already_in_lookup = " AND ".join( + "{lookup}.{lookup_col} IS {source}.{source_col}".format( + lookup=quote_identifier(table), + lookup_col=quote_identifier(rename.get(column) or column), + source=quote_identifier(self.name), + source_col=quote_identifier(column), + ) + for column in columns + ) self.db.execute( - "INSERT OR IGNORE INTO {} ({lookup_columns}) SELECT DISTINCT {table_cols} FROM {} WHERE NOT ({all_null})".format( + "INSERT OR IGNORE INTO {} ({lookup_columns}) SELECT DISTINCT {table_cols} FROM {} " + "WHERE NOT ({all_null}) AND NOT EXISTS (SELECT 1 FROM {lookup} WHERE {already_in_lookup})".format( quote_identifier(table), quote_identifier(self.name), lookup_columns=", ".join( @@ -2928,6 +2942,8 @@ class Table(Queryable): ), table_cols=", ".join(quote_identifier(c) for c in columns), all_null=all_columns_are_null, + lookup=quote_identifier(table), + already_in_lookup=already_in_lookup, ) ) diff --git a/tests/test_extract.py b/tests/test_extract.py index 1c0fa01..c73ee7a 100644 --- a/tests/test_extract.py +++ b/tests/test_extract.py @@ -271,3 +271,34 @@ def test_extract_null_values_existing_lookup_table_with_null_row(fresh_db): {"id": 10, "name": "Terriana", "species_id": 2}, {"id": 11, "name": "Spenidorm", "species_id": None}, ] + + +def test_extract_repeated_into_shared_lookup_with_nulls(fresh_db): + # Unique indexes treat NULLs as distinct, so INSERT OR IGNORE alone + # cannot dedupe NULL-containing rows against the existing lookup + # table - extracting a second table into the same lookup previously + # inserted duplicate rows that nothing pointed to + fresh_db["t1"].insert_all( + [ + {"id": 1, "species": None, "common": "X"}, + {"id": 2, "species": "Oak", "common": "Oak"}, + ], + pk="id", + ) + fresh_db["t2"].insert_all([{"id": 1, "species": None, "common": "X"}], pk="id") + fresh_db["t1"].extract(["species", "common"], table="lk") + fresh_db["t2"].extract(["species", "common"], table="lk") + assert fresh_db["lk"].count == 2 + # Both tables point at the same lookup row + t1_fk = fresh_db.execute("select lk_id from t1 where id = 1").fetchone()[0] + t2_fk = fresh_db.execute("select lk_id from t2 where id = 1").fetchone()[0] + assert t1_fk == t2_fk + + +def test_extract_repeated_into_shared_lookup_no_nulls(fresh_db): + # Non-NULL rows were already deduped by the unique index - keep it so + fresh_db["t1"].insert_all([{"id": 1, "species": "Oak"}], pk="id") + fresh_db["t2"].insert_all([{"id": 1, "species": "Oak"}], pk="id") + fresh_db["t1"].extract(["species"], table="lk") + fresh_db["t2"].extract(["species"], table="lk") + assert fresh_db["lk"].count == 1 From 548a886ca1d2ba6c259676da2c8e00f022bacc2c Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Mon, 6 Jul 2026 21:57:14 -0700 Subject: [PATCH 53/86] Clean CLI errors for InvalidColumns from insert --pk and extract sqlite-utils insert db t - --pk badcol and sqlite-utils extract db t nosuchcol dumped raw InvalidColumns tracebacks - the insert error handling caught NoTable and OperationalError but not the InvalidColumns introduced for #732, and the extract command had no handling at all (including for NoTable when pointed at a view). Both now exit with click-style Error: messages. Refs https://github.com/simonw/sqlite-utils/issues/769#issuecomment-4900034150 Co-Authored-By: Claude Fable 5 --- docs/changelog.rst | 1 + sqlite_utils/cli.py | 8 ++++++-- tests/test_cli.py | 19 +++++++++++++++++++ tests/test_cli_insert.py | 15 +++++++++++++++ 4 files changed, 41 insertions(+), 2 deletions(-) diff --git a/docs/changelog.rst b/docs/changelog.rst index d2b8a67..8c5b3e6 100644 --- a/docs/changelog.rst +++ b/docs/changelog.rst @@ -19,6 +19,7 @@ Unreleased - Fixed a bug where a failed write statement executed with ``db.execute()`` left the driver's implicit transaction open. Every subsequent write then joined that phantom transaction, which nothing committed, so their work was silently rolled back when the connection was closed. The implicit transaction opened by a failed statement is now rolled back before the exception is raised. A failed write inside a transaction opened with ``db.begin()`` or ``db.atomic()`` leaves that transaction open and untouched, as before. - Fixed a bug where transaction-control statements prefixed with an empty statement - ``db.query("; COMMIT")`` - or a UTF-8 byte order mark slipped past the check that rejects them, committing the caller's open transaction before raising a confusing ``OperationalError``. The keyword scanner used by ``db.query()`` and ``db.execute()`` now skips leading ``;`` and byte order marks, matching what the ``sqlite3`` driver tolerates before the first token, so these statements are rejected with a ``ValueError`` without being executed. The same fix means ``db.execute("; BEGIN")`` no longer auto-commits the transaction it just opened. - Documented a limitation of ``db.query()``: a ``PRAGMA`` statement that returns no rows raises a ``ValueError`` but still takes effect, because PRAGMA statements run outside the savepoint guard used to roll back other rejected statements. Use ``db.execute()`` for row-less PRAGMA statements. +- ``sqlite-utils insert ... --pk `` and ``sqlite-utils extract `` now show a clean ``Error:`` message instead of a raw Python traceback. The ``extract`` command also shows a clean error when pointed at a view. - Fixed a bug where running ``table.extract()`` more than once against the same lookup table inserted duplicate rows for values containing ``null`` - SQLite unique indexes treat ``NULL`` values as distinct, so ``INSERT OR IGNORE`` alone could not dedupe them. Each repeat extract added another copy that nothing referenced. The insert now uses an ``IS``-based ``NOT EXISTS`` guard so ``null``-containing rows match existing lookup rows. - ``db.add_foreign_keys()`` no longer silently ignores requested ``ON DELETE``/``ON UPDATE`` actions when a foreign key with the same columns already exists - it raises ``AlterError`` suggesting ``table.transform()``, since the actions of an existing foreign key cannot be changed in place. Exact duplicates, including actions, are still skipped so repeated calls stay idempotent. The method also now validates that compound foreign keys have the same number of columns on both sides, instead of silently discarding the extra columns. - ``db.ensure_autocommit_on()`` now raises ``TransactionError`` if called while a transaction is open. Assigning ``isolation_level`` commits any pending transaction as a side effect, so entering the block silently committed the caller's open transaction and made a later ``rollback()`` a no-op. diff --git a/sqlite_utils/cli.py b/sqlite_utils/cli.py index 188eff6..29dcb9d 100644 --- a/sqlite_utils/cli.py +++ b/sqlite_utils/cli.py @@ -13,6 +13,7 @@ from sqlite_utils.db import ( BadMultiValues, DEFAULT, DescIndex, + InvalidColumns, NoTable, NoView, quote_identifier, @@ -1172,7 +1173,7 @@ def insert_upsert_implementation( db.table(table).insert_all( docs, pk=pk, batch_size=batch_size, alter=alter, **extra_kwargs ) - except NoTable as e: + except (NoTable, InvalidColumns) as e: raise click.ClickException(str(e)) except Exception as e: if ( @@ -2734,7 +2735,10 @@ def extract( fk_column=fk_column, rename=dict(rename), ) - db.table(table).extract(**kwargs) + try: + db.table(table).extract(**kwargs) + except (NoTable, InvalidColumns) as e: + raise click.ClickException(str(e)) @cli.command(name="insert-files") diff --git a/tests/test_cli.py b/tests/test_cli.py index 9e17969..06e14ea 100644 --- a/tests/test_cli.py +++ b/tests/test_cli.py @@ -2758,3 +2758,22 @@ def test_insert_upsert_strict(tmpdir, method, strict): assert result.exit_code == 0 db = Database(db_path) assert db["items"].strict == strict or not db.supports_strict + + +def test_extract_bad_column_clean_error(db_path): + db = Database(db_path) + db["trees"].insert({"id": 1, "species": "Palm"}, pk="id") + result = CliRunner().invoke(cli.cli, ["extract", db_path, "trees", "nope"]) + assert result.exit_code == 1 + assert result.exception is None or isinstance(result.exception, SystemExit) + assert result.output.startswith("Error: Invalid columns") + + +def test_extract_view_clean_error(db_path): + db = Database(db_path) + db["trees"].insert({"id": 1, "species": "Palm"}, pk="id") + db.create_view("v", "select * from trees") + result = CliRunner().invoke(cli.cli, ["extract", db_path, "v", "species"]) + assert result.exit_code == 1 + assert result.exception is None or isinstance(result.exception, SystemExit) + assert result.output.startswith("Error:") diff --git a/tests/test_cli_insert.py b/tests/test_cli_insert.py index 196590e..5af8a2f 100644 --- a/tests/test_cli_insert.py +++ b/tests/test_cli_insert.py @@ -676,3 +676,18 @@ def test_upsert_csv_detect_types_leaves_existing_table_alone(db_path): assert result.exit_code == 0, result.output assert db["places"].columns_dict["zip"] is str assert db["places"].get(1)["zip"] == "01234" + + +def test_insert_invalid_pk_clean_error(db_path): + # An invalid --pk against an existing table should be a clean CLI + # error, not a raw InvalidColumns traceback + db = Database(db_path) + db["t"].insert({"a": 1}) + result = CliRunner().invoke( + cli.cli, + ["insert", db_path, "t", "-", "--pk", "badcol"], + input='{"a": 2}', + ) + assert result.exit_code == 1 + assert result.exception is None or isinstance(result.exception, SystemExit) + assert result.output.startswith("Error: Invalid primary key column") From 93640a7ddedf008036427a3e5bbd033d17cbe9df Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Mon, 6 Jul 2026 21:58:52 -0700 Subject: [PATCH 54/86] migrate --list is read-only with legacy sqlite-migrate classes too The docs promise --list will not create the database file or the _sqlite_migrations table, but legacy sqlite_migrate.Migrations classes create the table (in the legacy schema) from their pending()/applied() methods. The listing now runs inside a transaction that is rolled back, keeping --list read-only regardless of what the migration class does. Refs https://github.com/simonw/sqlite-utils/issues/769#issuecomment-4900034150 Co-Authored-By: Claude Fable 5 --- docs/changelog.rst | 1 + sqlite_utils/cli.py | 9 ++++++++- tests/test_cli_migrate.py | 20 ++++++++++++++++++++ 3 files changed, 29 insertions(+), 1 deletion(-) diff --git a/docs/changelog.rst b/docs/changelog.rst index 8c5b3e6..881423c 100644 --- a/docs/changelog.rst +++ b/docs/changelog.rst @@ -19,6 +19,7 @@ Unreleased - Fixed a bug where a failed write statement executed with ``db.execute()`` left the driver's implicit transaction open. Every subsequent write then joined that phantom transaction, which nothing committed, so their work was silently rolled back when the connection was closed. The implicit transaction opened by a failed statement is now rolled back before the exception is raised. A failed write inside a transaction opened with ``db.begin()`` or ``db.atomic()`` leaves that transaction open and untouched, as before. - Fixed a bug where transaction-control statements prefixed with an empty statement - ``db.query("; COMMIT")`` - or a UTF-8 byte order mark slipped past the check that rejects them, committing the caller's open transaction before raising a confusing ``OperationalError``. The keyword scanner used by ``db.query()`` and ``db.execute()`` now skips leading ``;`` and byte order marks, matching what the ``sqlite3`` driver tolerates before the first token, so these statements are rejected with a ``ValueError`` without being executed. The same fix means ``db.execute("; BEGIN")`` no longer auto-commits the transaction it just opened. - Documented a limitation of ``db.query()``: a ``PRAGMA`` statement that returns no rows raises a ``ValueError`` but still takes effect, because PRAGMA statements run outside the savepoint guard used to roll back other rejected statements. Use ``db.execute()`` for row-less PRAGMA statements. +- ``sqlite-utils migrate --list`` is now read-only even when the migrations file uses the legacy ``sqlite_migrate.Migrations`` class, whose listing methods create the ``_sqlite_migrations`` table as a side effect. The listing now runs inside a transaction that is rolled back. - ``sqlite-utils insert ... --pk `` and ``sqlite-utils extract `` now show a clean ``Error:`` message instead of a raw Python traceback. The ``extract`` command also shows a clean error when pointed at a view. - Fixed a bug where running ``table.extract()`` more than once against the same lookup table inserted duplicate rows for values containing ``null`` - SQLite unique indexes treat ``NULL`` values as distinct, so ``INSERT OR IGNORE`` alone could not dedupe them. Each repeat extract added another copy that nothing referenced. The insert now uses an ``IS``-based ``NOT EXISTS`` guard so ``null``-containing rows match existing lookup rows. - ``db.add_foreign_keys()`` no longer silently ignores requested ``ON DELETE``/``ON UPDATE`` actions when a foreign key with the same columns already exists - it raises ``AlterError`` suggesting ``table.transform()``, since the actions of an existing foreign key cannot be changed in place. Exact duplicates, including actions, are still skipped so repeated calls stay idempotent. The method also now validates that compound foreign keys have the same number of columns on both sides, instead of silently discarding the extra columns. diff --git a/sqlite_utils/cli.py b/sqlite_utils/cli.py index 29dcb9d..fe7dbe4 100644 --- a/sqlite_utils/cli.py +++ b/sqlite_utils/cli.py @@ -3434,7 +3434,14 @@ def migrate(db_path, migrations, stop_before, list_, verbose): # Listing is read-only - don't create the database file db = sqlite_utils.Database(memory=True) _register_db_for_cleanup(db) - _display_migration_list(db, migration_sets) + # Legacy sqlite-migrate classes create the migrations table from + # their pending()/applied() methods - run the listing inside a + # transaction and roll it back so --list stays read-only + db.begin() + try: + _display_migration_list(db, migration_sets) + finally: + db.rollback() return db = sqlite_utils.Database(db_path) diff --git a/tests/test_cli_migrate.py b/tests/test_cli_migrate.py index 0f1c7ea..0f29e36 100644 --- a/tests/test_cli_migrate.py +++ b/tests/test_cli_migrate.py @@ -485,3 +485,23 @@ def test_stop_before_applied_migration_errors(two_migrations): assert "already been applied" in result.output db = sqlite_utils.Database(db_path) assert not db["bar"].exists() + + +def test_list_with_legacy_class_is_read_only(tmpdir): + # Legacy sqlite-migrate classes create the _sqlite_migrations table + # from their pending()/applied() methods - --list must roll that + # back so it stays a read-only operation as documented + path = pathlib.Path(tmpdir) + (path / "migrations.py").write_text(LEGACY_MIGRATIONS, "utf-8") + db_path = str(path / "test.db") + db = sqlite_utils.Database(db_path) + db["existing"].insert({"id": 1}) + db.close() + result = CliRunner().invoke( + sqlite_utils.cli.cli, ["migrate", db_path, str(path), "--list"] + ) + assert result.exit_code == 0, result.output + assert "first" in result.output + db2 = sqlite_utils.Database(db_path) + assert "_sqlite_migrations" not in db2.table_names() + db2.close() From c2a17744095553c7fcc4706cc5da1d20045a299e Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Mon, 6 Jul 2026 22:02:41 -0700 Subject: [PATCH 55/86] Fix two tests that assumed modern SQLite behavior SQLite 3.23.1 rejects a UTF-8 byte order mark before the first token, so the BOM variant of the execute()-prefixed-BEGIN test now skips when the SQLite version does not accept a leading BOM. And versions before 3.36 allowed selecting rowid from a view, returning NULL, rather than raising an error - the pks_and_rows_where() view test now accepts either behavior. Refs https://github.com/simonw/sqlite-utils/issues/769#issuecomment-4900034150 Co-Authored-By: Claude Fable 5 --- tests/test_atomic.py | 53 ++++++++++++++++++++++++++++++++++++++++++++ tests/test_rows.py | 16 ++++++++----- 2 files changed, 64 insertions(+), 5 deletions(-) diff --git a/tests/test_atomic.py b/tests/test_atomic.py index 0d25b84..c3fd02f 100644 --- a/tests/test_atomic.py +++ b/tests/test_atomic.py @@ -258,11 +258,21 @@ def test_execute_comment_prefixed_begin_leaves_transaction_open(fresh_db): assert [r["id"] for r in fresh_db["t"].rows] == [1] +def _sqlite_accepts_bom(): + try: + sqlite3.connect(":memory:").execute("\ufeffselect 1") + return True + except sqlite3.OperationalError: + return False + + @pytest.mark.parametrize("begin_sql", ["; begin", "\ufeffbegin"]) def test_execute_prefixed_begin_leaves_transaction_open(fresh_db, begin_sql): # sqlite3 tolerates empty statements and a UTF-8 BOM before the first # real token, so a BEGIN behind either must not be auto-committed # out from under the caller + if begin_sql.startswith("\ufeff") and not _sqlite_accepts_bom(): + pytest.skip("This SQLite version rejects a leading byte order mark") fresh_db["t"].insert({"id": 1}, pk="id") fresh_db.execute(begin_sql) assert fresh_db.conn.in_transaction @@ -327,3 +337,46 @@ def test_query_returning_commits_after_iteration(tmpdir): assert other.execute("select count(*) from t").fetchone()[0] == 2 other.close() db.close() + + +TRIGGER_SQL = """ +create trigger no_bad before insert on t +when new.v = 'bad' +begin + select raise(rollback, 'trigger says no'); +end +""" + + +def test_atomic_preserves_error_from_transaction_destroying_trigger(fresh_db): + # RAISE(ROLLBACK) rolls back the whole transaction and destroys every + # savepoint - atomic()'s cleanup must not mask the IntegrityError + # with "cannot rollback - no transaction is active" + fresh_db.execute("create table t (id integer primary key, v text)") + fresh_db.execute(TRIGGER_SQL) + with pytest.raises(sqlite3.IntegrityError, match="trigger says no"): + with fresh_db.atomic(): + fresh_db.execute("insert into t (v) values ('bad')") + assert not fresh_db.conn.in_transaction + + +def test_nested_atomic_preserves_error_from_transaction_destroying_trigger( + fresh_db, +): + # The nested savepoint branch previously raised + # "no such savepoint" from ROLLBACK TO SAVEPOINT + fresh_db.execute("create table t (id integer primary key, v text)") + fresh_db.execute(TRIGGER_SQL) + with pytest.raises(sqlite3.IntegrityError, match="trigger says no"): + with fresh_db.atomic(): + with fresh_db.atomic(): + fresh_db.execute("insert into t (v) values ('bad')") + assert not fresh_db.conn.in_transaction + + +def test_atomic_preserves_error_from_insert_or_rollback(fresh_db): + fresh_db["t"].insert({"id": 1}, pk="id") + with pytest.raises(sqlite3.IntegrityError): + with fresh_db.atomic(): + fresh_db.execute("insert or rollback into t (id) values (1)") + assert not fresh_db.conn.in_transaction diff --git a/tests/test_rows.py b/tests/test_rows.py index 46417ef..46d4f53 100644 --- a/tests/test_rows.py +++ b/tests/test_rows.py @@ -115,15 +115,21 @@ def test_rows_where_duplicate_select_columns_are_deduped(fresh_db): def test_pks_and_rows_where_view(fresh_db): # pks_and_rows_where() lives on Queryable so views expose it, but - # SQLite views have no rowid - it has always failed with an - # OperationalError from the generated SQL. Guard against it failing - # earlier with an AttributeError from View lacking Table properties + # SQLite views have no rowid. Modern SQLite (3.36+) raises an + # OperationalError from the generated SQL; older versions returned + # NULL for a view's rowid. Either way it must not fail earlier with + # an AttributeError from View lacking Table-only properties from sqlite_utils.utils import sqlite3 fresh_db["dogs"].insert({"id": 1, "name": "Cleo"}, pk="id") fresh_db.create_view("dog_names", "select name from dogs") - with pytest.raises(sqlite3.OperationalError): - list(fresh_db["dog_names"].pks_and_rows_where()) + try: + result = list(fresh_db["dog_names"].pks_and_rows_where()) + except sqlite3.OperationalError: + pass # SQLite 3.36+: no such column: rowid + else: + # Older SQLite returns NULL rowids for views + assert result == [(None, {"rowid": None, "name": "Cleo"})] def test_pks_and_rows_where_compound_pk_declaration_order(fresh_db): From d9a0fd26e09a7e1c07b7356d0cd25a22695f89c5 Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Mon, 6 Jul 2026 22:04:00 -0700 Subject: [PATCH 56/86] Transaction cleanup no longer masks transaction-destroying errors A RAISE(ROLLBACK) trigger or INSERT OR ROLLBACK conflict rolls back the entire transaction and destroys every savepoint. The cleanup paths in atomic() and query() then raised OperationalError ("no such savepoint" / "cannot rollback - no transaction is active"), masking the original IntegrityError - breaking user code that catches sqlite3.IntegrityError. Cleanup now checks conn.in_transaction first: if the error already destroyed the transaction there is nothing left to undo, and the original exception propagates. Refs https://github.com/simonw/sqlite-utils/issues/769#issuecomment-4900034150 Co-Authored-By: Claude Fable 5 --- docs/changelog.rst | 1 + sqlite_utils/db.py | 22 ++++++++++++++++------ tests/test_query.py | 17 +++++++++++++++++ 3 files changed, 34 insertions(+), 6 deletions(-) diff --git a/docs/changelog.rst b/docs/changelog.rst index 881423c..9228fb6 100644 --- a/docs/changelog.rst +++ b/docs/changelog.rst @@ -19,6 +19,7 @@ Unreleased - Fixed a bug where a failed write statement executed with ``db.execute()`` left the driver's implicit transaction open. Every subsequent write then joined that phantom transaction, which nothing committed, so their work was silently rolled back when the connection was closed. The implicit transaction opened by a failed statement is now rolled back before the exception is raised. A failed write inside a transaction opened with ``db.begin()`` or ``db.atomic()`` leaves that transaction open and untouched, as before. - Fixed a bug where transaction-control statements prefixed with an empty statement - ``db.query("; COMMIT")`` - or a UTF-8 byte order mark slipped past the check that rejects them, committing the caller's open transaction before raising a confusing ``OperationalError``. The keyword scanner used by ``db.query()`` and ``db.execute()`` now skips leading ``;`` and byte order marks, matching what the ``sqlite3`` driver tolerates before the first token, so these statements are rejected with a ``ValueError`` without being executed. The same fix means ``db.execute("; BEGIN")`` no longer auto-commits the transaction it just opened. - Documented a limitation of ``db.query()``: a ``PRAGMA`` statement that returns no rows raises a ``ValueError`` but still takes effect, because PRAGMA statements run outside the savepoint guard used to roll back other rejected statements. Use ``db.execute()`` for row-less PRAGMA statements. +- Fixed exception masking when a statement destroys the enclosing transaction. An error such as a ``RAISE(ROLLBACK)`` trigger or ``INSERT OR ROLLBACK`` conflict rolls back the whole transaction, destroying every savepoint - the cleanup in ``db.atomic()`` and ``db.query()`` then failed with ``OperationalError: no such savepoint`` (or ``cannot rollback - no transaction is active``), hiding the original ``IntegrityError`` from code that tried to catch it. Cleanup now checks whether a transaction is still open first, so the original exception propagates. - ``sqlite-utils migrate --list`` is now read-only even when the migrations file uses the legacy ``sqlite_migrate.Migrations`` class, whose listing methods create the ``_sqlite_migrations`` table as a side effect. The listing now runs inside a transaction that is rolled back. - ``sqlite-utils insert ... --pk `` and ``sqlite-utils extract `` now show a clean ``Error:`` message instead of a raw Python traceback. The ``extract`` command also shows a clean error when pointed at a view. - Fixed a bug where running ``table.extract()`` more than once against the same lookup table inserted duplicate rows for values containing ``null`` - SQLite unique indexes treat ``NULL`` values as distinct, so ``INSERT OR IGNORE`` alone could not dedupe them. Each repeat extract added another copy that nothing referenced. The insert now uses an ``IS``-based ``NOT EXISTS`` guard so ``null``-containing rows match existing lookup rows. diff --git a/sqlite_utils/db.py b/sqlite_utils/db.py index e69f996..bec68fc 100644 --- a/sqlite_utils/db.py +++ b/sqlite_utils/db.py @@ -602,8 +602,13 @@ class Database: try: yield self except BaseException: - self.conn.execute("ROLLBACK TO SAVEPOINT {};".format(savepoint)) - self.conn.execute("RELEASE SAVEPOINT {};".format(savepoint)) + # An error such as a RAISE(ROLLBACK) trigger can destroy + # the whole transaction, savepoints included - cleaning up + # anyway would mask the original exception with + # "no such savepoint" + if self.conn.in_transaction: + self.conn.execute("ROLLBACK TO SAVEPOINT {};".format(savepoint)) + self.conn.execute("RELEASE SAVEPOINT {};".format(savepoint)) raise else: self.conn.execute("RELEASE SAVEPOINT {};".format(savepoint)) @@ -612,13 +617,15 @@ class Database: try: yield self except BaseException: - self.conn.execute("ROLLBACK") + # rollback() is a no-op if the error already destroyed the + # transaction, so the original exception propagates + self.rollback() raise else: try: self.conn.execute("COMMIT") except BaseException: - self.conn.execute("ROLLBACK") + self.rollback() raise def begin(self) -> None: @@ -870,8 +877,11 @@ class Database: return (dict(zip(keys, row)) for row in fetched) return (dict(zip(keys, row)) for row in cursor) finally: - if not released: - # An error occurred - undo anything the statement changed + if not released and self.conn.in_transaction: + # An error occurred - undo anything the statement changed. + # If the error itself destroyed the transaction (such as a + # RAISE(ROLLBACK) trigger) the savepoint is already gone + # and there is nothing left to undo self.conn.execute('ROLLBACK TO "sqlite_utils_query"') self.conn.execute('RELEASE "sqlite_utils_query"') diff --git a/tests/test_query.py b/tests/test_query.py index f4aa336..aac5a29 100644 --- a/tests/test_query.py +++ b/tests/test_query.py @@ -280,3 +280,20 @@ def test_execute_returning_dicts(fresh_db): assert fresh_db.execute_returning_dicts("select * from test") == [ {"id": 1, "bar": 2} ] + + +def test_query_preserves_error_from_transaction_destroying_trigger(fresh_db): + # RAISE(ROLLBACK) destroys the savepoint guard - the original + # IntegrityError must propagate, not "no such savepoint" + fresh_db.execute("create table t (id integer primary key, v text)") + fresh_db.execute(""" + create trigger no_bad before insert on t + when new.v = 'bad' + begin + select raise(rollback, 'trigger says no'); + end + """) + with pytest.raises(sqlite3.IntegrityError, match="trigger says no"): + fresh_db.query("insert into t (id, v) values (1, 'bad') returning id") + assert not fresh_db.conn.in_transaction + assert fresh_db.execute("select count(*) from t").fetchone()[0] == 0 From 25824467846ca75db3bd737cb19b5a8e0d01b1b2 Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Mon, 6 Jul 2026 22:06:51 -0700 Subject: [PATCH 57/86] Skip RETURNING-based trigger test on SQLite older than 3.35 The query() exception-masking test uses INSERT ... RETURNING, which is a syntax error on SQLite 3.23.1 - skip it there like the other RETURNING tests. Refs https://github.com/simonw/sqlite-utils/issues/769#issuecomment-4900034150 Co-Authored-By: Claude Fable 5 --- tests/test_query.py | 4 ++++ 1 file changed, 4 insertions(+) diff --git a/tests/test_query.py b/tests/test_query.py index aac5a29..06847da 100644 --- a/tests/test_query.py +++ b/tests/test_query.py @@ -282,6 +282,10 @@ def test_execute_returning_dicts(fresh_db): ] +@pytest.mark.skipif( + sqlite3.sqlite_version_info < (3, 35, 0), + reason="RETURNING requires SQLite 3.35.0 or higher", +) def test_query_preserves_error_from_transaction_destroying_trigger(fresh_db): # RAISE(ROLLBACK) destroys the savepoint guard - the original # IntegrityError must propagate, not "no such savepoint" From 9a2c58246528109c63afa48af3c0c5b5f943f2b0 Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Mon, 6 Jul 2026 22:26:58 -0700 Subject: [PATCH 58/86] README tweak ready for v4, refs #769 --- README.md | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) diff --git a/README.md b/README.md index e425461..c444c64 100644 --- a/README.md +++ b/README.md @@ -18,9 +18,12 @@ Python CLI utility and library for manipulating SQLite databases. - [Configure SQLite full-text search](https://sqlite-utils.datasette.io/en/stable/cli.html#configuring-full-text-search) against your database tables and run search queries against them, ordered by relevance - Run [transformations against your tables](https://sqlite-utils.datasette.io/en/stable/cli.html#transforming-tables) to make schema changes that SQLite `ALTER TABLE` does not directly support, such as changing the type of a column - [Extract columns](https://sqlite-utils.datasette.io/en/stable/cli.html#extracting-columns-into-a-separate-table) into separate tables to better normalize your existing data +- [Manage database migrations](https://sqlite-utils.datasette.io/en/stable/migrations.html) using Python migration files and the `sqlite-utils migrate` command - [Install plugins](https://sqlite-utils.datasette.io/en/stable/plugins.html) to add custom SQL functions and additional features -Read more on my blog, in this series of posts on [New features in sqlite-utils](https://simonwillison.net/series/sqlite-utils-features/) and other [entries tagged sqliteutils](https://simonwillison.net/tags/sqliteutils/). +Upgrading from sqlite-utils 3.x? See the [4.0 upgrade guide](https://sqlite-utils.datasette.io/en/stable/upgrading.html#upgrading-from-3-x-to-4-0). + +Read more on my blog, in this series of posts on [New features in sqlite-utils](https://simonwillison.net/series/sqlite-utils-features/) and other [entries tagged sqlite-utils](https://simonwillison.net/tags/sqlite-utils/). ## Installation From d34f1bea0b34b7004a56b75b5dd083f89f8cc970 Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Mon, 6 Jul 2026 22:33:50 -0700 Subject: [PATCH 59/86] Release 4.0rc4 Refs #186, #554, #566, #625, #732, #769 --- docs/changelog.rst | 6 +++--- pyproject.toml | 2 +- 2 files changed, 4 insertions(+), 4 deletions(-) diff --git a/docs/changelog.rst b/docs/changelog.rst index 9228fb6..0943dca 100644 --- a/docs/changelog.rst +++ b/docs/changelog.rst @@ -4,10 +4,10 @@ Changelog =========== -.. _unreleased: +.. _v4_0rc4: -Unreleased ----------- +4.0rc4 (2026-07-06) +------------------- - **Breaking change**: ``table.extract()`` - and the ``sqlite-utils extract`` command - no longer extract rows where every extracted column is ``null``. Those rows now keep a ``null`` value in the new foreign key column instead of pointing at an all-``null`` record in the lookup table. When extracting multiple columns, rows are still extracted if at least one of the columns has a value. (:issue:`186`) - The ``extracts=`` option to ``table.insert()`` and friends no longer creates a lookup table record for ``None`` values - the column value stays ``null``. Previously every batch of inserted rows containing a ``None`` value would add a duplicate ``null`` record to the lookup table. diff --git a/pyproject.toml b/pyproject.toml index 01f1735..c1024c2 100644 --- a/pyproject.toml +++ b/pyproject.toml @@ -1,6 +1,6 @@ [project] name = "sqlite-utils" -version = "4.0rc3" +version = "4.0rc4" description = "CLI tool and Python library for manipulating SQLite databases" readme = { file = "README.md", content-type = "text/markdown" } authors = [ From d314d04215f7337d42c847214861ec7ffe0bf757 Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Mon, 6 Jul 2026 22:40:37 -0700 Subject: [PATCH 60/86] Bump GitHub Actions versions --- .github/workflows/publish.yml | 8 ++++---- .github/workflows/test-coverage.yml | 4 ++-- .github/workflows/test.yml | 4 ++-- 3 files changed, 8 insertions(+), 8 deletions(-) diff --git a/.github/workflows/publish.yml b/.github/workflows/publish.yml index 396e2b8..23b23bd 100644 --- a/.github/workflows/publish.yml +++ b/.github/workflows/publish.yml @@ -12,9 +12,9 @@ jobs: python-version: ["3.10", "3.11", "3.12", "3.13", "3.14"] os: [ubuntu-latest, windows-latest, macos-latest] steps: - - uses: actions/checkout@v4 + - uses: actions/checkout@v7 - name: Set up Python ${{ matrix.python-version }} - uses: actions/setup-python@v5 + uses: actions/setup-python@v6 with: python-version: ${{ matrix.python-version }} cache: pip @@ -29,9 +29,9 @@ jobs: runs-on: ubuntu-latest needs: [test] steps: - - uses: actions/checkout@v4 + - uses: actions/checkout@v7 - name: Set up Python - uses: actions/setup-python@v5 + uses: actions/setup-python@v6 with: python-version: '3.14' cache: pip diff --git a/.github/workflows/test-coverage.yml b/.github/workflows/test-coverage.yml index 1f55f4e..7668f1b 100644 --- a/.github/workflows/test-coverage.yml +++ b/.github/workflows/test-coverage.yml @@ -12,9 +12,9 @@ jobs: runs-on: ubuntu-latest steps: - name: Check out repo - uses: actions/checkout@v4 + uses: actions/checkout@v7 - name: Set up Python - uses: actions/setup-python@v5 + uses: actions/setup-python@v6 with: python-version: "3.11" cache: pip diff --git a/.github/workflows/test.yml b/.github/workflows/test.yml index d85cbf8..923de2e 100644 --- a/.github/workflows/test.yml +++ b/.github/workflows/test.yml @@ -14,9 +14,9 @@ jobs: numpy: [0, 1] os: [ubuntu-latest, macos-latest, windows-latest, macos-14] steps: - - uses: actions/checkout@v4 + - uses: actions/checkout@v7 - name: Set up Python ${{ matrix.python-version }} - uses: actions/setup-python@v5 + uses: actions/setup-python@v6 with: python-version: ${{ matrix.python-version }} allow-prereleases: true From 60811e730509667f702bc08f9bf5fc3fe13b7f45 Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Tue, 7 Jul 2026 08:24:24 -0700 Subject: [PATCH 61/86] Fix rowid pk and last_rowid regressions in insert/upsert Closes #781, #783 Refs https://github.com/simonw/sqlite-utils/issues/769#issuecomment-4900618685 * Fix rowid pk and last_rowid regressions in insert/upsert Two behaviour regressions in the 4.0 insert/upsert rewrite broke callers (notably Datasette's write API) that operate on tables without an explicit primary key. Both are fixed here with regression tests. 1. rowid (and its aliases _rowid_/oid) were rejected as a primary key. Table.pks already reports ["rowid"] for a rowid table, but the new pk validation raised InvalidColumns because rowid is not listed among the table's columns, and the insert success path then raised KeyError when looking up the pk value. rowid aliases are now accepted for rowid tables and resolve directly to the rowid. 2. An ignored insert (INSERT OR IGNORE that matched an existing row) no longer populated last_rowid, and only set last_pk when an explicit pk= was passed. It now locates the existing conflicting row by its primary key values and reports that row's rowid and pk, rather than relying on the connection's last inserted rowid. Add a shared ROWID_ALIASES constant for the rowid alias names. Co-Authored-By: Claude Opus 4.8 Claude-Session: https://claude.ai/code/session_01E7af8SxFZqiCerJB6MqKnY --- sqlite_utils/db.py | 97 +++++++++++++++++++++++++++++++------------- tests/test_create.py | 90 ++++++++++++++++++++++++++++++++++++++++ 2 files changed, 159 insertions(+), 28 deletions(-) diff --git a/sqlite_utils/db.py b/sqlite_utils/db.py index bec68fc..3033a36 100644 --- a/sqlite_utils/db.py +++ b/sqlite_utils/db.py @@ -53,6 +53,11 @@ except ImportError: SQLITE_MAX_VARS = 999 +# Names that refer to a rowid table's implicit integer primary key. These are +# valid primary key targets even though they are not listed among a table's +# columns. See https://www.sqlite.org/lang_createtable.html#rowid +ROWID_ALIASES = frozenset({"rowid", "_rowid_", "oid"}) + _quote_fts_re = re.compile(r'\s+|(".*?")') _virtual_table_using_re = re.compile( @@ -4343,10 +4348,14 @@ class Table(Queryable): if pk and not hash_id and self.exists(): pk_cols = [pk] if isinstance(pk, str) else list(pk) existing_columns = self.columns_dict + # rowid and its aliases are valid primary keys for a rowid table + # even though they are not listed among the table's columns + rowid_aliases = ROWID_ALIASES if self.use_rowid else frozenset() missing_pk_cols = [ col for col in pk_cols - if resolve_casing(col, existing_columns) not in existing_columns + if col.lower() not in rowid_aliases + and resolve_casing(col, existing_columns) not in existing_columns ] if missing_pk_cols: invalid_pk_error = InvalidColumns( @@ -4512,40 +4521,72 @@ class Table(Queryable): if not upsert and result is not None: ignored_insert = ignore and result.rowcount == 0 if ignored_insert: + # The row was not inserted because it conflicts with an + # existing row. Point last_pk / last_rowid at that existing + # row when we can identify it from the record's primary key + # values, rather than leaving them stale or unset. if list_mode: - first_record_list = cast(Sequence[Any], first_record) - if hash_id: - pass - elif isinstance(pk, str): - pk_index = column_names.index( - resolve_casing(pk, column_names) - ) - self.last_pk = first_record_list[pk_index] - elif pk: - self.last_pk = tuple( - first_record_list[ - column_names.index(resolve_casing(p, column_names)) - ] - for p in pk - ) + first_record_dict = dict( + zip(column_names, cast(Sequence[Any], first_record)) + ) else: first_record_dict = cast(Dict[str, Any], first_record) - if hash_id: - self.last_pk = hash_record( - first_record_dict, hash_id_columns - ) - elif isinstance(pk, str): - self.last_pk = first_record_dict[ - resolve_casing(pk, first_record_dict) + if hash_id: + self.last_pk = hash_record(first_record_dict, hash_id_columns) + elif isinstance(pk, str): + self.last_pk = first_record_dict[ + resolve_casing(pk, first_record_dict) + ] + elif pk: + self.last_pk = tuple( + first_record_dict[resolve_casing(p, first_record_dict)] + for p in pk + ) + # Locate the existing conflicting row using its primary key + # columns so we can report its rowid (and pk if not already + # known). Falls back to leaving them unset if the conflict + # cannot be resolved to a pk lookup (e.g. a UNIQUE column). + key_cols: Optional[List[str]] = None + if isinstance(pk, str): + key_cols = [pk] + elif pk: + key_cols = list(pk) + elif not hash_id and not self.use_rowid: + key_cols = self.pks + if key_cols: + try: + key_values = [ + first_record_dict[resolve_casing(c, first_record_dict)] + for c in key_cols ] - elif pk: - self.last_pk = tuple( - first_record_dict[resolve_casing(p, first_record_dict)] - for p in pk + except KeyError: + key_values = None + if key_values is not None: + where = " and ".join( + "{} = ?".format(quote_identifier(c)) for c in key_cols ) + existing = self.db.execute( + "select rowid from {} where {} limit 1".format( + quote_identifier(self.name), where + ), + key_values, + ).fetchone() + if existing is not None: + self.last_rowid = existing[0] + # On a primary key conflict the record's pk + # values identify the existing row + if self.last_pk is None: + self.last_pk = ( + key_values[0] + if len(key_cols) == 1 + else tuple(key_values) + ) else: self.last_rowid = result.lastrowid - if (hash_id or pk) and self.last_rowid: + # A rowid-alias pk resolves directly to the rowid, so there + # is no separate pk column to look up + rowid_pk = isinstance(pk, str) and pk.lower() in ROWID_ALIASES + if (hash_id or (pk and not rowid_pk)) and self.last_rowid: # Set self.last_pk to the pk(s) for that rowid row = list(self.rows_where("rowid = ?", [self.last_rowid]))[0] if hash_id: diff --git a/tests/test_create.py b/tests/test_create.py index 42fa1c4..7fab5e6 100644 --- a/tests/test_create.py +++ b/tests/test_create.py @@ -990,6 +990,96 @@ def test_insert_ignore(fresh_db): assert rows == [{"id": 1, "bar": 2}] +def test_insert_ignore_reports_existing_row(fresh_db): + # An ignored insert (row already exists) should point last_rowid and + # last_pk at the existing conflicting row - see the Datasette insert API + fresh_db["docs"].insert({"id": 1, "title": "Exists"}, pk="id") + # Insert a conflicting row with ignore=True and no explicit pk= + table = fresh_db["docs"].insert({"id": 1, "title": "One"}, ignore=True) + assert table.last_rowid == 1 + assert table.last_pk == 1 + assert list(fresh_db["docs"].rows_where("rowid = ?", [table.last_rowid])) == [ + {"id": 1, "title": "Exists"} + ] + + +@pytest.mark.parametrize("rowid_alias", ("rowid", "_rowid_", "oid")) +@pytest.mark.parametrize("method", ("upsert", "insert_replace", "insert_ignore")) +def test_pk_rowid_alias_on_rowid_table(fresh_db, rowid_alias, method): + # rowid and its aliases are valid primary keys for a rowid table even + # though they are not listed among the table's columns - see the Datasette + # upsert API against tables without an explicit primary key + fresh_db["t"].insert({"title": "Hello"}) + assert fresh_db["t"].pks == ["rowid"] + record = {rowid_alias: 1, "title": "Updated"} + if method == "upsert": + table = fresh_db["t"].upsert(record, pk=rowid_alias) + elif method == "insert_replace": + table = fresh_db["t"].insert(record, pk=rowid_alias, replace=True) + else: + table = fresh_db["t"].insert(record, pk=rowid_alias, ignore=True) + assert table.last_pk == 1 + expected_title = "Hello" if method == "insert_ignore" else "Updated" + assert list(fresh_db["t"].rows) == [{"title": expected_title}] + + +def test_insert_ignore_reports_existing_row_compound_pk(fresh_db): + # Compound primary key variant of the ignored-insert lookup + fresh_db["t"].insert_all([{"a": 1, "b": 2, "note": "first"}], pk=("a", "b")) + table = fresh_db["t"].insert( + {"a": 1, "b": 2, "note": "second"}, pk=("a", "b"), ignore=True + ) + assert table.last_pk == (1, 2) + assert list(fresh_db["t"].rows_where("rowid = ?", [table.last_rowid])) == [ + {"a": 1, "b": 2, "note": "first"} + ] + + +def test_insert_ignore_reports_existing_row_list_mode(fresh_db): + # List-based iteration variant of the ignored-insert lookup + fresh_db["t"].insert_all([["id", "title"], [1, "first"]], pk="id") + table = fresh_db["t"].insert_all( + [["id", "title"], [1, "second"]], pk="id", ignore=True + ) + assert table.last_pk == 1 + assert table.last_rowid == 1 + assert list(fresh_db["t"].rows) == [{"id": 1, "title": "first"}] + + +def test_insert_ignore_hash_id_reports_pk(fresh_db): + # With hash_id the pk is the computed hash; the original record has no id + # column to look up so last_rowid is left unset + first = fresh_db["dogs"].insert({"name": "Cleo"}, hash_id="id") + table = fresh_db["dogs"].insert({"name": "Cleo"}, hash_id="id", ignore=True) + assert table.last_pk == first.last_pk + assert table.last_rowid is None + assert fresh_db["dogs"].count == 1 + + +def test_insert_ignore_unresolvable_conflict_leaves_pk_unset(fresh_db): + # When the conflict cannot be resolved to a primary key lookup, last_pk and + # last_rowid are left unset rather than reporting a misleading value + + # rowid table with a UNIQUE column and no primary key: no pk to look up + fresh_db["u"].db.execute("create table u (title text unique)") + fresh_db["u"].insert({"title": "x"}) + table = fresh_db["u"].insert({"title": "x"}, ignore=True) + assert table.last_pk is None + assert table.last_rowid is None + assert fresh_db["u"].count == 1 + + # Conflict on a UNIQUE column other than the primary key: the pk value from + # the record does not match the existing row, so the lookup finds nothing + fresh_db["docs"].db.execute( + "create table docs (id integer primary key, email text unique)" + ) + fresh_db["docs"].insert({"id": 1, "email": "a"}, pk="id") + table = fresh_db["docs"].insert({"id": 2, "email": "a"}, ignore=True) + assert table.last_pk is None + assert table.last_rowid is None + assert fresh_db["docs"].count == 1 + + def test_insert_ignore_with_pk_after_other_table_insert(fresh_db): # https://github.com/simonw/sqlite-utils/issues/554 user = {"id": "abc", "name": "david"} From 8bc9213a8e9c52a2d46ffc14da72fcd5af276840 Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Tue, 7 Jul 2026 08:40:31 -0700 Subject: [PATCH 62/86] Release 4.0 Refs #781, #783, #769 --- docs/changelog.rst | 31 +++++++++++++++++++++++++++++++ pyproject.toml | 2 +- 2 files changed, 32 insertions(+), 1 deletion(-) diff --git a/docs/changelog.rst b/docs/changelog.rst index 0943dca..9623ec2 100644 --- a/docs/changelog.rst +++ b/docs/changelog.rst @@ -4,6 +4,37 @@ Changelog =========== +.. _v4_0: + +4.0 (2026-07-07) +---------------- + +The 4.0 release includes some minor backwards-incompatible fixes (hence the major version number bump) and introduces three major new features: + +- :ref:`Database migrations `, providing a structured mechanism for evolving a project's schema over time. (:issue:`752`) +- :ref:`Nested transaction support ` via ``db.atomic()``, plus numerous improvements to how transactions work across the library. (:issue:`755`) +- Support for :ref:`compound foreign keys `, including creation, transformation and introspection through :ref:`table.foreign_keys `. (:issue:`594`) + +Other notable changes include: + +- Upserts now use SQLite's ``INSERT ... ON CONFLICT ... DO UPDATE SET`` syntax, detect existing table primary keys automatically and reject records that are missing required primary key values. (:issue:`652`) +- ``db.query()`` now executes immediately and rejects statements that do not return rows; use ``db.execute()`` for writes and DDL. +- CSV and TSV imports now detect column types by default, while inserts into existing tables preserve those tables' column types. (:issue:`679`) +- Foreign key handling now preserves ``ON DELETE``/``ON UPDATE`` actions during transforms and resolves referenced primary keys more accurately. (:issue:`530`) +- Column names passed to Python API methods are now matched case-insensitively, mirroring SQLite's own identifier behavior. (:issue:`760`) +- The command-line tool now emits UTF-8 JSON output by default, with ``--ascii`` available to restore escaped output. (:issue:`625`) +- ``table.extract()`` and ``extracts=`` no longer create lookup table records for all-``null`` values. (:issue:`186`) + +See :ref:`upgrading_3_to_4` for details on backwards-incompatible changes. + +The detailed release notes for the features and fixes shipped during the 4.0 pre-release cycle are available in :ref:`4.0a0 `, :ref:`4.0a1 `, :ref:`4.0rc1 `, :ref:`4.0rc2 `, :ref:`4.0rc3 ` and :ref:`4.0rc4 `. + +Bug fixes since 4.0rc4 +~~~~~~~~~~~~~~~~~~~~~~ + +- Fixed 4.0 regressions in ``insert``/``upsert`` against tables that use SQLite's implicit ``rowid`` primary key. Passing ``pk="rowid"``, ``pk="_rowid_"`` or ``pk="oid"`` now works again for rowid tables, and ``last_pk`` is set correctly. (:issue:`781`) +- Fixed ``insert(..., ignore=True)`` and ``insert_all(..., ignore=True)`` so an ignored insert that conflicts with an existing primary key row now reports that existing row in ``last_rowid`` and ``last_pk`` where possible. This also works for compound primary keys and list-mode inserts. (:issue:`783`) + .. _v4_0rc4: 4.0rc4 (2026-07-06) diff --git a/pyproject.toml b/pyproject.toml index c1024c2..a3a42a9 100644 --- a/pyproject.toml +++ b/pyproject.toml @@ -1,6 +1,6 @@ [project] name = "sqlite-utils" -version = "4.0rc4" +version = "4.0" description = "CLI tool and Python library for manipulating SQLite databases" readme = { file = "README.md", content-type = "text/markdown" } authors = [ From 353baf280d3765b9680a49276945a0e96bc9e238 Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Tue, 7 Jul 2026 11:51:53 -0700 Subject: [PATCH 63/86] Corrected imports in migrations docs --- docs/migrations.rst | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) diff --git a/docs/migrations.rst b/docs/migrations.rst index cfdbf13..23aa9d8 100644 --- a/docs/migrations.rst +++ b/docs/migrations.rst @@ -27,7 +27,7 @@ Here is a simple example of a ``migrations.py`` file which creates a table, then .. code-block:: python - from sqlite_utils import Database, Migrations + from sqlite_utils import Migrations migrations = Migrations("creatures") @@ -51,6 +51,8 @@ Once you have a ``Migrations(name)`` collection with one or more migrations regi .. code-block:: python + from sqlite_utils import Database + db = Database("creatures.db") migrations.apply(db) From 619770bf427e47657099f1a4fdaa8777ec4c68d8 Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Tue, 7 Jul 2026 17:52:26 -0700 Subject: [PATCH 64/86] sqlite-utils query - to read SQL from stdin, closes #765 --- docs/changelog.rst | 7 +++++++ docs/cli-reference.rst | 4 ++++ docs/cli.rst | 10 ++++++++++ sqlite_utils/cli.py | 8 ++++++++ tests/test_cli.py | 19 +++++++++++++++++++ 5 files changed, 48 insertions(+) diff --git a/docs/changelog.rst b/docs/changelog.rst index 9623ec2..cf9486a 100644 --- a/docs/changelog.rst +++ b/docs/changelog.rst @@ -4,6 +4,13 @@ Changelog =========== +.. _v_unreleased: + +Unreleased +---------- + +- ``sqlite-utils query`` can now read the SQL query from standard input by passing ``-`` in place of the query, for example ``echo "select * from dogs" | sqlite-utils query dogs.db -``. (:issue:`765`) + .. _v4_0: 4.0 (2026-07-07) diff --git a/docs/cli-reference.rst b/docs/cli-reference.rst index 49eba52..000f88f 100644 --- a/docs/cli-reference.rst +++ b/docs/cli-reference.rst @@ -109,6 +109,10 @@ See :ref:`cli_query`. "select * from chickens where age > :age" \ -p age 1 + Pass "-" as the SQL to read the query from standard input: + + echo "select * from chickens" | sqlite-utils data.db - + Options: --attach ... Additional databases to attach - specify alias and filepath diff --git a/docs/cli.rst b/docs/cli.rst index 063e80f..bbbbf03 100644 --- a/docs/cli.rst +++ b/docs/cli.rst @@ -29,6 +29,16 @@ The ``sqlite-utils query`` command lets you run queries directly against a SQLit .. note:: In Python: :ref:`db.query() ` CLI reference: :ref:`sqlite-utils query ` +Pass ``-`` as the SQL query to read the query from standard input. This is useful for longer queries that would otherwise require careful shell escaping, or for piping in SQL generated by another tool: + +.. code-block:: bash + + echo "select * from dogs" | sqlite-utils query dogs.db - + +.. code-block:: bash + + sqlite-utils query dogs.db - < query.sql + .. _cli_query_json: Returning JSON diff --git a/sqlite_utils/cli.py b/sqlite_utils/cli.py index fe7dbe4..c6496d4 100644 --- a/sqlite_utils/cli.py +++ b/sqlite_utils/cli.py @@ -1893,7 +1893,15 @@ def query( sqlite-utils data.db \\ "select * from chickens where age > :age" \\ -p age 1 + + Pass "-" as the SQL to read the query from standard input: + + \b + echo "select * from chickens" | sqlite-utils data.db - """ + if sql == "-": + # Read SQL from standard input + sql = sys.stdin.read() db = sqlite_utils.Database(path) _register_db_for_cleanup(db) for alias, attach_path in attach: diff --git a/tests/test_cli.py b/tests/test_cli.py index 06e14ea..8196a08 100644 --- a/tests/test_cli.py +++ b/tests/test_cli.py @@ -782,6 +782,25 @@ def test_query_json(db_path, sql, args, expected): assert expected == result.output.strip() +def test_query_sql_from_stdin(db_path): + # https://github.com/simonw/sqlite-utils/issues/765 + db = Database(db_path) + with db.conn: + db["dogs"].insert_all( + [ + {"id": 1, "age": 4, "name": "Cleo"}, + {"id": 2, "age": 2, "name": "Pancakes"}, + ] + ) + result = CliRunner().invoke( + cli.cli, + ["query", db_path, "-"], + input="select name from dogs order by name", + ) + assert result.exit_code == 0, result.output + assert json.loads(result.output) == [{"name": "Cleo"}, {"name": "Pancakes"}] + + def test_query_json_empty(db_path): result = CliRunner().invoke( cli.cli, From fa5d66bf5377b00e976ebb3bf92d806ed8f54270 Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Tue, 7 Jul 2026 17:56:57 -0700 Subject: [PATCH 65/86] Update create-table --help to mention real --- docs/cli-reference.rst | 4 ++-- sqlite_utils/cli.py | 4 ++-- 2 files changed, 4 insertions(+), 4 deletions(-) diff --git a/docs/cli-reference.rst b/docs/cli-reference.rst index 000f88f..fd1199f 100644 --- a/docs/cli-reference.rst +++ b/docs/cli-reference.rst @@ -930,10 +930,10 @@ See :ref:`cli_create_table`. sqlite-utils create-table my.db people \ id integer \ name text \ - height float \ + height real \ photo blob --pk id - Valid column types are text, integer, float and blob. + Valid column types are text, integer, real, float and blob. Options: --pk TEXT Column to use as primary key diff --git a/sqlite_utils/cli.py b/sqlite_utils/cli.py index c6496d4..905963f 100644 --- a/sqlite_utils/cli.py +++ b/sqlite_utils/cli.py @@ -1634,10 +1634,10 @@ def create_table( sqlite-utils create-table my.db people \\ id integer \\ name text \\ - height float \\ + height real \\ photo blob --pk id - Valid column types are text, integer, float and blob. + Valid column types are text, integer, real, float and blob. """ db = sqlite_utils.Database(path) _register_db_for_cleanup(db) From 8ee0b7c65cfa34f550672872bffa06183bea9963 Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Tue, 7 Jul 2026 17:57:29 -0700 Subject: [PATCH 66/86] Fix for rogue quote in enable-fts --help --- docs/cli-reference.rst | 2 +- sqlite_utils/cli.py | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) diff --git a/docs/cli-reference.rst b/docs/cli-reference.rst index fd1199f..e57f4b3 100644 --- a/docs/cli-reference.rst +++ b/docs/cli-reference.rst @@ -1028,7 +1028,7 @@ See :ref:`cli_fts`. Usage: sqlite-utils enable-fts [OPTIONS] PATH TABLE COLUMN... - Enable full-text search for specific table and columns" + Enable full-text search for specific table and columns Example: diff --git a/sqlite_utils/cli.py b/sqlite_utils/cli.py index 905963f..99a8c64 100644 --- a/sqlite_utils/cli.py +++ b/sqlite_utils/cli.py @@ -706,7 +706,7 @@ def create_index( def enable_fts( path, table, column, fts4, fts5, tokenize, create_triggers, replace, load_extension ): - """Enable full-text search for specific table and columns" + """Enable full-text search for specific table and columns Example: From cf3373e7b7fb9342d705d7b5b8d52fe09ac46a34 Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Tue, 7 Jul 2026 17:59:58 -0700 Subject: [PATCH 67/86] Refactor --functions option, improve help --- docs/cli-reference.rst | 11 ++++++----- sqlite_utils/cli.py | 29 ++++++++++++++--------------- 2 files changed, 20 insertions(+), 20 deletions(-) diff --git a/docs/cli-reference.rst b/docs/cli-reference.rst index e57f4b3..c232881 100644 --- a/docs/cli-reference.rst +++ b/docs/cli-reference.rst @@ -138,8 +138,8 @@ See :ref:`cli_query`. -r, --raw Raw output, first column of first row --raw-lines Raw output, first column of each row -p, --param ... Named :parameters for SQL query - --functions TEXT Python code or file path defining custom SQL - functions + --functions TEXT Python code or a file path defining custom SQL + functions; can be used multiple times --load-extension TEXT Path to SQLite extension, with optional :entrypoint -h, --help Show this message and exit. @@ -181,8 +181,8 @@ See :ref:`cli_memory`. sqlite-utils memory animals.csv --schema Options: - --functions TEXT Python code or file path defining custom SQL - functions + --functions TEXT Python code or a file path defining custom SQL + functions; can be used multiple times --attach ... Additional databases to attach - specify alias and filepath --flatten Flatten nested JSON objects, so {"foo": {"bar": @@ -383,7 +383,8 @@ See :ref:`cli_bulk`. Options: --batch-size INTEGER Commit every X records - --functions TEXT Python code or file path defining custom SQL functions + --functions TEXT Python code or a file path defining custom SQL + functions; can be used multiple times --flatten Flatten nested JSON objects, so {"a": {"b": 1}} becomes {"a_b": 1} --nl Expect newline-delimited JSON diff --git a/sqlite_utils/cli.py b/sqlite_utils/cli.py index 99a8c64..edf1634 100644 --- a/sqlite_utils/cli.py +++ b/sqlite_utils/cli.py @@ -153,6 +153,17 @@ def load_extension_option(fn): )(fn) +def functions_option(fn): + return click.option( + "--functions", + help=( + "Python code or a file path defining custom SQL functions; " + "can be used multiple times" + ), + multiple=True, + )(fn) + + @click.group( cls=DefaultGroup, default="query", @@ -1450,11 +1461,7 @@ def upsert( @click.argument("sql") @click.argument("file", type=click.File("rb"), required=True) @click.option("--batch-size", type=int, default=100, help="Commit every X records") -@click.option( - "--functions", - help="Python code or file path defining custom SQL functions", - multiple=True, -) +@functions_option @import_options @load_extension_option def bulk( @@ -1860,11 +1867,7 @@ def drop_view(path, view, ignore, load_extension): type=(str, str), help="Named :parameters for SQL query", ) -@click.option( - "--functions", - help="Python code or file path defining custom SQL functions", - multiple=True, -) +@functions_option @load_extension_option def query( path, @@ -1937,11 +1940,7 @@ def query( nargs=-1, ) @click.argument("sql") -@click.option( - "--functions", - help="Python code or file path defining custom SQL functions", - multiple=True, -) +@functions_option @click.option( "--attach", type=(str, click.Path(file_okay=True, dir_okay=False, allow_dash=False)), From aa300942bfcba0bd788951db0b9070cf291c193a Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Tue, 7 Jul 2026 18:03:25 -0700 Subject: [PATCH 68/86] Update sqlite-utils convert --help, refs #686 --- docs/cli-reference.rst | 7 +++++-- sqlite_utils/cli.py | 9 ++++++++- 2 files changed, 13 insertions(+), 3 deletions(-) diff --git a/docs/cli-reference.rst b/docs/cli-reference.rst index c232881..105b19a 100644 --- a/docs/cli-reference.rst +++ b/docs/cli-reference.rst @@ -621,6 +621,11 @@ See :ref:`cli_convert`. "value" is a variable with the column value to be converted. + CODE can also be a reference to a callable that takes the value, for example: + + sqlite-utils convert my.db mytable date r.parsedate + sqlite-utils convert my.db mytable data json.loads --import json + Use "-" for CODE to read Python code from standard input. The following common operations are available as recipe functions: @@ -634,7 +639,6 @@ See :ref:`cli_convert`. errors: 'Optional[object]' = None) -> 'Optional[str]' Parse a date and convert it to ISO date format: yyyy-mm-dd - - dayfirst=True: treat xx as the day in xx/yy/zz - yearfirst=True: treat xx as the year in xx/yy/zz - errors=r.IGNORE to ignore values that cannot be parsed @@ -644,7 +648,6 @@ See :ref:`cli_convert`. False, errors: 'Optional[object]' = None) -> 'Optional[str]' Parse a datetime and convert it to ISO datetime format: yyyy-mm-ddTHH:MM:SS - - dayfirst=True: treat xx as the day in xx/yy/zz - yearfirst=True: treat xx as the year in xx/yy/zz - errors=r.IGNORE to ignore values that cannot be parsed diff --git a/sqlite_utils/cli.py b/sqlite_utils/cli.py index edf1634..318b208 100644 --- a/sqlite_utils/cli.py +++ b/sqlite_utils/cli.py @@ -3054,6 +3054,12 @@ def _generate_convert_help(): "value" is a variable with the column value to be converted. + CODE can also be a reference to a callable that takes the value, for example: + + \b + sqlite-utils convert my.db mytable date r.parsedate + sqlite-utils convert my.db mytable data json.loads --import json + Use "-" for CODE to read Python code from standard input. The following common operations are available as recipe functions: @@ -3067,8 +3073,9 @@ def _generate_convert_help(): ] for name in recipe_names: fn = getattr(recipes, name) + doc = textwrap.dedent(fn.__doc__.rstrip()).replace("\b\n", "") help += "\n\nr.{}{}\n\n\b{}".format( - name, str(inspect.signature(fn)), textwrap.dedent(fn.__doc__.rstrip()) + name, str(inspect.signature(fn)), doc ) help += "\n\n" help += textwrap.dedent(""" From ebafb84c93bd3666e0e6acd24bfafe070f25cdee Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Tue, 7 Jul 2026 18:09:12 -0700 Subject: [PATCH 69/86] Allow sqlite-utils upsert to infer --pk from existing table --- docs/cli-reference.rst | 3 ++- docs/cli.rst | 2 ++ sqlite_utils/cli.py | 7 +++++-- tests/test_cli.py | 24 +++++++++++++++++++++--- 4 files changed, 30 insertions(+), 6 deletions(-) diff --git a/docs/cli-reference.rst b/docs/cli-reference.rst index 105b19a..5ea9200 100644 --- a/docs/cli-reference.rst +++ b/docs/cli-reference.rst @@ -321,6 +321,8 @@ See :ref:`cli_upsert`. incoming record has a primary key that matches an existing record the existing record will be updated. + If the table already exists and has a primary key, --pk can be omitted. + Example: echo '[ @@ -330,7 +332,6 @@ See :ref:`cli_upsert`. Options: --pk TEXT Columns to use as the primary key, e.g. id - [required] --flatten Flatten nested JSON objects, so {"a": {"b": 1}} becomes {"a_b": 1} --nl Expect newline-delimited JSON diff --git a/docs/cli.rst b/docs/cli.rst index bbbbf03..e373704 100644 --- a/docs/cli.rst +++ b/docs/cli.rst @@ -1593,6 +1593,8 @@ For example: This will update the dog with an ID of 2 to have an age of 4, creating a new record (with a null name) if one does not exist. If a row DOES exist the name will be left as-is. +If the table already exists and has a primary key, you can omit the ``--pk`` option and ``sqlite-utils`` will use that existing primary key. + The command will fail if you reference columns that do not exist on the table. To automatically create missing columns, use the ``--alter`` option. .. note:: diff --git a/sqlite_utils/cli.py b/sqlite_utils/cli.py index 318b208..5e21f92 100644 --- a/sqlite_utils/cli.py +++ b/sqlite_utils/cli.py @@ -16,6 +16,7 @@ from sqlite_utils.db import ( InvalidColumns, NoTable, NoView, + PrimaryKeyRequired, quote_identifier, ) from sqlite_utils.plugins import ensure_plugins_loaded, pm, get_plugins @@ -1184,7 +1185,7 @@ def insert_upsert_implementation( db.table(table).insert_all( docs, pk=pk, batch_size=batch_size, alter=alter, **extra_kwargs ) - except (NoTable, InvalidColumns) as e: + except (NoTable, InvalidColumns, PrimaryKeyRequired) as e: raise click.ClickException(str(e)) except Exception as e: if ( @@ -1372,7 +1373,7 @@ def insert( @cli.command() -@insert_upsert_options(require_pk=True) +@insert_upsert_options() def upsert( path, table, @@ -1408,6 +1409,8 @@ def upsert( an incoming record has a primary key that matches an existing record the existing record will be updated. + If the table already exists and has a primary key, --pk can be omitted. + Example: \b diff --git a/tests/test_cli.py b/tests/test_cli.py index 8196a08..bc5d492 100644 --- a/tests/test_cli.py +++ b/tests/test_cli.py @@ -1235,20 +1235,38 @@ def test_upsert(db_path, tmpdir): ] -def test_upsert_pk_required(db_path, tmpdir): +def test_upsert_pk_inferred_from_existing_table(db_path, tmpdir): json_path = str(tmpdir / "dogs.json") + db = Database(db_path) insert_dogs = [ {"id": 1, "name": "Cleo", "age": 4}, {"id": 2, "name": "Nixie", "age": 4}, ] write_json(json_path, insert_dogs) + result = CliRunner().invoke( + cli.cli, + ["insert", db_path, "dogs", json_path, "--pk", "id"], + catch_exceptions=False, + ) + assert result.exit_code == 0, result.output + + write_json( + json_path, + [ + {"id": 1, "age": 5}, + {"id": 2, "age": 5}, + ], + ) result = CliRunner().invoke( cli.cli, ["upsert", db_path, "dogs", json_path], catch_exceptions=False, ) - assert result.exit_code == 2 - assert "Error: Missing option '--pk'" in result.output + assert result.exit_code == 0, result.output + assert list(db.query("select * from dogs order by id")) == [ + {"id": 1, "name": "Cleo", "age": 5}, + {"id": 2, "name": "Nixie", "age": 5}, + ] def test_upsert_analyze(db_path, tmpdir): From 23a21c1d6bc8aa1c17a49b37181aadffd7abb8be Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Tue, 7 Jul 2026 18:11:34 -0700 Subject: [PATCH 70/86] Ran Black --- sqlite_utils/cli.py | 4 +--- 1 file changed, 1 insertion(+), 3 deletions(-) diff --git a/sqlite_utils/cli.py b/sqlite_utils/cli.py index 5e21f92..b99dcf1 100644 --- a/sqlite_utils/cli.py +++ b/sqlite_utils/cli.py @@ -3077,9 +3077,7 @@ def _generate_convert_help(): for name in recipe_names: fn = getattr(recipes, name) doc = textwrap.dedent(fn.__doc__.rstrip()).replace("\b\n", "") - help += "\n\nr.{}{}\n\n\b{}".format( - name, str(inspect.signature(fn)), doc - ) + help += "\n\nr.{}{}\n\n\b{}".format(name, str(inspect.signature(fn)), doc) help += "\n\n" help += textwrap.dedent(""" You can use these recipes like so: From 569608e40f28893262d727adadddbf3d077112da Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Tue, 7 Jul 2026 18:48:57 -0700 Subject: [PATCH 71/86] sqlite-utils insert --code option, closes #684 --- docs/changelog.rst | 1 + docs/cli-reference.rst | 18 ++- docs/cli.rst | 21 ++++ sqlite_utils/cli.py | 232 +++++++++++++++++++++++++++------------ tests/test_cli_insert.py | 150 +++++++++++++++++++++++++ 5 files changed, 347 insertions(+), 75 deletions(-) diff --git a/docs/changelog.rst b/docs/changelog.rst index cf9486a..a9dd4a3 100644 --- a/docs/changelog.rst +++ b/docs/changelog.rst @@ -10,6 +10,7 @@ Unreleased ---------- - ``sqlite-utils query`` can now read the SQL query from standard input by passing ``-`` in place of the query, for example ``echo "select * from dogs" | sqlite-utils query dogs.db -``. (:issue:`765`) +- ``sqlite-utils insert`` and ``sqlite-utils upsert`` now accept a ``--code`` option for providing a block of Python code (or a path to a ``.py`` file) that defines a ``rows()`` function or ``rows`` iterable of rows to insert, as an alternative to importing from a file. (:issue:`684`) .. _v4_0: diff --git a/docs/cli-reference.rst b/docs/cli-reference.rst index 5ea9200..5798809 100644 --- a/docs/cli-reference.rst +++ b/docs/cli-reference.rst @@ -230,7 +230,7 @@ See :ref:`cli_inserting_data`, :ref:`cli_insert_csv_tsv`, :ref:`cli_insert_unstr :: - Usage: sqlite-utils insert [OPTIONS] PATH TABLE FILE + Usage: sqlite-utils insert [OPTIONS] PATH TABLE [FILE] Insert records from FILE into a table, creating the table if it does not already exist. @@ -272,8 +272,20 @@ See :ref:`cli_inserting_data`, :ref:`cli_insert_csv_tsv`, :ref:`cli_insert_unstr echo 'A bunch of words' | sqlite-utils insert words.db words - \ --text --convert '({"word": w} for w in text.split())' + Instead of a FILE you can use --code to provide a block of Python code that + defines the rows to insert, as either a rows() function that yields + dictionaries or a "rows" iterable. --code can also be a path to a .py file: + + sqlite-utils insert data.db creatures --code ' + def rows(): + yield {"id": 1, "name": "Cleo"} + yield {"id": 2, "name": "Suna"} + ' --pk id + Options: --pk TEXT Columns to use as the primary key, e.g. id + --code TEXT Python code defining a rows() function or iterable + of rows to insert --flatten Flatten nested JSON objects, so {"a": {"b": 1}} becomes {"a_b": 1} --nl Expect newline-delimited JSON @@ -315,7 +327,7 @@ See :ref:`cli_upsert`. :: - Usage: sqlite-utils upsert [OPTIONS] PATH TABLE FILE + Usage: sqlite-utils upsert [OPTIONS] PATH TABLE [FILE] Upsert records based on their primary key. Works like 'insert' but if an incoming record has a primary key that matches an existing record the existing @@ -332,6 +344,8 @@ See :ref:`cli_upsert`. Options: --pk TEXT Columns to use as the primary key, e.g. id + --code TEXT Python code defining a rows() function or iterable + of rows to insert --flatten Flatten nested JSON objects, so {"a": {"b": 1}} becomes {"a_b": 1} --nl Expect newline-delimited JSON diff --git a/docs/cli.rst b/docs/cli.rst index e373704..4ed92ed 100644 --- a/docs/cli.rst +++ b/docs/cli.rst @@ -1561,6 +1561,27 @@ The result looks like this: COMMIT; +.. _cli_insert_code: + +Inserting rows generated by Python code +======================================= + +Instead of providing a ``FILE`` to import, you can use the ``--code`` option to pass a block of Python code that generates the rows to insert. This is the command-line equivalent of calling ``db["creatures"].insert_all(rows())`` from the :ref:`Python API `. + +Your code should define either a ``rows()`` function that returns or yields dictionaries, or a ``rows`` iterable such as a list of dictionaries: + +.. code-block:: bash + + sqlite-utils insert data.db creatures --code ' + def rows(): + yield {"id": 1, "name": "Cleo"} + yield {"id": 2, "name": "Suna"} + ' --pk id + +``--code`` can also be given a path to a Python ``.py`` file. + +The ``--code`` option works with both ``sqlite-utils insert`` and ``sqlite-utils upsert``, and composes with table options such as ``--pk``, ``--replace``, ``--alter``, ``--not-null`` and ``--default``. It cannot be combined with a ``FILE`` argument or with input format options such as ``--csv`` or ``--convert``. + .. _cli_insert_replace: Insert-replacing data diff --git a/sqlite_utils/cli.py b/sqlite_utils/cli.py index b99dcf1..4eefc70 100644 --- a/sqlite_utils/cli.py +++ b/sqlite_utils/cli.py @@ -944,13 +944,19 @@ def insert_upsert_options(*, require_pk=False): required=True, ), click.argument("table"), - click.argument("file", type=click.File("rb", lazy=True), required=True), + click.argument( + "file", type=click.File("rb", lazy=True), required=False + ), click.option( "--pk", help="Columns to use as the primary key, e.g. id", multiple=True, required=require_pk, ), + click.option( + "--code", + help="Python code defining a rows() function or iterable of rows to insert", + ), ) + _import_options + ( @@ -1035,11 +1041,118 @@ def insert_upsert_implementation( bulk_sql=None, functions=None, strict=False, + code=None, ): db = sqlite_utils.Database(path) _register_db_for_cleanup(db) _load_extensions(db, load_extension) _maybe_register_functions(db, functions) + + def _insert_docs(docs, tracker=None): + extra_kwargs = { + "ignore": ignore, + "replace": replace, + "truncate": truncate, + "analyze": analyze, + "strict": strict, + } + if not_null: + extra_kwargs["not_null"] = set(not_null) + if default: + extra_kwargs["defaults"] = dict(default) + if upsert: + extra_kwargs["upsert"] = upsert + + # docs should all be dictionaries + docs = (verify_is_dict(doc) for doc in docs) + + # Apply {"$base64": true, ...} decoding, if needed + docs = (decode_base64_values(doc) for doc in docs) + + # For bulk_sql= we use cursor.executemany() instead + if bulk_sql: + if batch_size: + doc_chunks = chunks(docs, batch_size) + else: + doc_chunks = [docs] + for doc_chunk in doc_chunks: + with db.atomic(): + db.conn.cursor().executemany(bulk_sql, doc_chunk) + return + + # table_names() rather than db.table(), which raises NoTable for + # views before the error handling below can deal with them + table_existed_before_insert = table in db.table_names() + try: + db.table(table).insert_all( + docs, pk=pk, batch_size=batch_size, alter=alter, **extra_kwargs + ) + except (NoTable, InvalidColumns, PrimaryKeyRequired) as e: + raise click.ClickException(str(e)) + except Exception as e: + if ( + isinstance(e, OperationalError) + and e.args + and ( + "has no column named" in e.args[0] or "no such column" in e.args[0] + ) + ): + raise click.ClickException( + "{}\n\nTry using --alter to add additional columns".format( + e.args[0] + ) + ) + # If we can find sql= and parameters= arguments, show those + variables = _find_variables(e.__traceback__, ["sql", "parameters"]) + if "sql" in variables and "parameters" in variables: + raise click.ClickException( + "{}\n\nsql = {}\nparameters = {}".format( + str(e), variables["sql"], variables["parameters"] + ) + ) + else: + raise + # Apply detected types only to a table this command created - + # transforming a pre-existing table would rewrite its column types + # and corrupt values such as TEXT zip codes with leading zeros + if ( + tracker is not None + and not table_existed_before_insert + and db.table(table).exists() + ): + db.table(table).transform(types=tracker.types) + + if code is not None: + if file is not None: + raise click.ClickException("--code cannot be used with a FILE argument") + if any( + [ + flatten, + nl, + csv, + tsv, + empty_null, + lines, + text, + convert, + sniff, + no_headers, + delimiter, + quotechar, + encoding, + ] + ): + raise click.ClickException( + "--code cannot be used with input format options" + ) + _insert_docs(_rows_from_code(code)) + return + + if file is None: + raise click.ClickException( + "Provide either a FILE argument or --code to specify rows to insert" + ) + if (delimiter or quotechar or sniff or no_headers) and not tsv: csv = True if (nl + csv + tsv) >= 2: @@ -1147,78 +1260,7 @@ def insert_upsert_implementation( else: docs = (fn(doc) or doc for doc in docs) - extra_kwargs = { - "ignore": ignore, - "replace": replace, - "truncate": truncate, - "analyze": analyze, - "strict": strict, - } - if not_null: - extra_kwargs["not_null"] = set(not_null) - if default: - extra_kwargs["defaults"] = dict(default) - if upsert: - extra_kwargs["upsert"] = upsert - - # docs should all be dictionaries - docs = (verify_is_dict(doc) for doc in docs) - - # Apply {"$base64": true, ...} decoding, if needed - docs = (decode_base64_values(doc) for doc in docs) - - # For bulk_sql= we use cursor.executemany() instead - if bulk_sql: - if batch_size: - doc_chunks = chunks(docs, batch_size) - else: - doc_chunks = [docs] - for doc_chunk in doc_chunks: - with db.atomic(): - db.conn.cursor().executemany(bulk_sql, doc_chunk) - return - - # table_names() rather than db.table(), which raises NoTable for - # views before the error handling below can deal with them - table_existed_before_insert = table in db.table_names() - try: - db.table(table).insert_all( - docs, pk=pk, batch_size=batch_size, alter=alter, **extra_kwargs - ) - except (NoTable, InvalidColumns, PrimaryKeyRequired) as e: - raise click.ClickException(str(e)) - except Exception as e: - if ( - isinstance(e, OperationalError) - and e.args - and ( - "has no column named" in e.args[0] or "no such column" in e.args[0] - ) - ): - raise click.ClickException( - "{}\n\nTry using --alter to add additional columns".format( - e.args[0] - ) - ) - # If we can find sql= and parameters= arguments, show those - variables = _find_variables(e.__traceback__, ["sql", "parameters"]) - if "sql" in variables and "parameters" in variables: - raise click.ClickException( - "{}\n\nsql = {}\nparameters = {}".format( - str(e), variables["sql"], variables["parameters"] - ) - ) - else: - raise - # Apply detected types only to a table this command created - - # transforming a pre-existing table would rewrite its column types - # and corrupt values such as TEXT zip codes with leading zeros - if ( - tracker is not None - and not table_existed_before_insert - and db.table(table).exists() - ): - db.table(table).transform(types=tracker.types) + _insert_docs(docs, tracker=tracker) # Clean up open file-like objects if sniff_buffer: @@ -1261,6 +1303,7 @@ def insert( table, file, pk, + code, flatten, nl, csv, @@ -1332,6 +1375,17 @@ def insert( \b echo 'A bunch of words' | sqlite-utils insert words.db words - \\ --text --convert '({"word": w} for w in text.split())' + + Instead of a FILE you can use --code to provide a block of Python code + that defines the rows to insert, as either a rows() function that yields + dictionaries or a "rows" iterable. --code can also be a path to a .py file: + + \b + sqlite-utils insert data.db creatures --code ' + def rows(): + yield {"id": 1, "name": "Cleo"} + yield {"id": 2, "name": "Suna"} + ' --pk id """ try: insert_upsert_implementation( @@ -1367,6 +1421,7 @@ def insert( not_null=not_null, default=default, strict=strict, + code=code, ) except UnicodeDecodeError as ex: raise click.ClickException(UNICODE_ERROR.format(ex)) @@ -1379,6 +1434,7 @@ def upsert( table, file, pk, + code, flatten, nl, csv, @@ -1450,6 +1506,7 @@ def upsert( load_extension=load_extension, silent=silent, strict=strict, + code=code, ) except UnicodeDecodeError as ex: raise click.ClickException(UNICODE_ERROR.format(ex)) @@ -3669,3 +3726,32 @@ def _maybe_register_functions(db, functions_list): for functions in functions_list: if isinstance(functions, str) and functions.strip(): _register_functions(db, functions) + + +def _rows_from_code(code): + # code may be a path to a .py file + if "\n" not in code and code.endswith(".py"): + try: + code = pathlib.Path(code).read_text() + except FileNotFoundError: + raise click.ClickException("File not found: {}".format(code)) + namespace = {} + try: + exec(code, namespace) + except SyntaxError as ex: + raise click.ClickException("Error in --code: {}".format(ex)) + rows = namespace.get("rows") + if callable(rows): + rows = rows() + if isinstance(rows, dict): + rows = [rows] + error = click.ClickException( + "--code must define a 'rows' function or iterable of rows to insert" + ) + if rows is None or isinstance(rows, (str, bytes)): + raise error + try: + iter(rows) + except TypeError: + raise error + return rows diff --git a/tests/test_cli_insert.py b/tests/test_cli_insert.py index 5af8a2f..e290f97 100644 --- a/tests/test_cli_insert.py +++ b/tests/test_cli_insert.py @@ -691,3 +691,153 @@ def test_insert_invalid_pk_clean_error(db_path): assert result.exit_code == 1 assert result.exception is None or isinstance(result.exception, SystemExit) assert result.output.startswith("Error: Invalid primary key column") + + +# --code tests, see https://github.com/simonw/sqlite-utils/issues/684 +CODE_ROWS_FUNCTION = """ +def rows(): + yield {"id": 1, "name": "Cleo"} + yield {"id": 2, "name": "Suna"} +""" + +CODE_ROWS_ITERABLE = """ +rows = [ + {"id": 1, "name": "Cleo"}, + {"id": 2, "name": "Suna"}, +] +""" + + +@pytest.mark.parametrize("code", (CODE_ROWS_FUNCTION, CODE_ROWS_ITERABLE)) +def test_insert_code(tmpdir, code): + db_path = str(tmpdir / "dogs.db") + result = CliRunner().invoke( + cli.cli, + ["insert", db_path, "creatures", "--code", code, "--pk", "id"], + ) + assert result.exit_code == 0, result.output + db = Database(db_path) + assert db["creatures"].pks == ["id"] + assert list(db["creatures"].rows) == [ + {"id": 1, "name": "Cleo"}, + {"id": 2, "name": "Suna"}, + ] + + +def test_insert_code_from_file(tmpdir): + db_path = str(tmpdir / "dogs.db") + code_path = str(tmpdir / "gen.py") + with open(code_path, "w") as fp: + fp.write(CODE_ROWS_FUNCTION) + result = CliRunner().invoke( + cli.cli, + ["insert", db_path, "creatures", "--code", code_path], + ) + assert result.exit_code == 0, result.output + assert list(Database(db_path)["creatures"].rows) == [ + {"id": 1, "name": "Cleo"}, + {"id": 2, "name": "Suna"}, + ] + + +def test_upsert_code(tmpdir): + db_path = str(tmpdir / "dogs.db") + db = Database(db_path) + db["creatures"].insert_all( + [{"id": 1, "name": "old"}, {"id": 2, "name": "Suna"}], pk="id" + ) + result = CliRunner().invoke( + cli.cli, + ["upsert", db_path, "creatures", "--code", CODE_ROWS_FUNCTION, "--pk", "id"], + ) + assert result.exit_code == 0, result.output + assert list(db["creatures"].rows) == [ + {"id": 1, "name": "Cleo"}, + {"id": 2, "name": "Suna"}, + ] + + +def test_insert_code_requires_file_or_code(tmpdir): + db_path = str(tmpdir / "dogs.db") + result = CliRunner().invoke(cli.cli, ["insert", db_path, "creatures"]) + assert result.exit_code == 1 + assert "Provide either a FILE argument or --code" in result.output + + +def test_insert_code_mutually_exclusive_with_file(tmpdir): + db_path = str(tmpdir / "dogs.db") + result = CliRunner().invoke( + cli.cli, + ["insert", db_path, "creatures", "-", "--code", CODE_ROWS_FUNCTION], + input="{}", + ) + assert result.exit_code == 1 + assert "--code cannot be used with a FILE argument" in result.output + + +def test_insert_code_rejects_input_format_options(tmpdir): + db_path = str(tmpdir / "dogs.db") + result = CliRunner().invoke( + cli.cli, + ["insert", db_path, "creatures", "--code", CODE_ROWS_FUNCTION, "--csv"], + ) + assert result.exit_code == 1 + assert "--code cannot be used with input format options" in result.output + + +def test_insert_code_missing_rows(tmpdir): + db_path = str(tmpdir / "dogs.db") + result = CliRunner().invoke( + cli.cli, + ["insert", db_path, "creatures", "--code", "x = 1"], + ) + assert result.exit_code == 1 + assert "must define a 'rows' function or iterable" in result.output + + +def test_insert_code_single_dict(tmpdir): + db_path = str(tmpdir / "dogs.db") + result = CliRunner().invoke( + cli.cli, + [ + "insert", + db_path, + "creatures", + "--code", + 'rows = {"id": 1, "name": "Cleo"}', + "--pk", + "id", + ], + ) + assert result.exit_code == 0, result.output + assert list(Database(db_path)["creatures"].rows) == [{"id": 1, "name": "Cleo"}] + + +def test_insert_code_not_iterable(tmpdir): + db_path = str(tmpdir / "dogs.db") + result = CliRunner().invoke( + cli.cli, + ["insert", db_path, "creatures", "--code", "rows = 5"], + ) + assert result.exit_code == 1 + assert "must define a 'rows' function or iterable" in result.output + + +def test_insert_code_syntax_error(tmpdir): + db_path = str(tmpdir / "dogs.db") + result = CliRunner().invoke( + cli.cli, + ["insert", db_path, "creatures", "--code", "def rows(:"], + ) + assert result.exit_code == 1 + assert "Error in --code" in result.output + + +def test_insert_code_file_not_found(tmpdir): + db_path = str(tmpdir / "dogs.db") + result = CliRunner().invoke( + cli.cli, + ["insert", db_path, "creatures", "--code", "missing.py"], + ) + assert result.exit_code == 1 + assert "File not found: missing.py" in result.output From 092f0919c3c254de0801dda60e7bbf8b73818c6c Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Tue, 7 Jul 2026 18:57:27 -0700 Subject: [PATCH 72/86] Changelog tweak refs #684 --- docs/changelog.rst | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/docs/changelog.rst b/docs/changelog.rst index a9dd4a3..5d1816a 100644 --- a/docs/changelog.rst +++ b/docs/changelog.rst @@ -10,7 +10,7 @@ Unreleased ---------- - ``sqlite-utils query`` can now read the SQL query from standard input by passing ``-`` in place of the query, for example ``echo "select * from dogs" | sqlite-utils query dogs.db -``. (:issue:`765`) -- ``sqlite-utils insert`` and ``sqlite-utils upsert`` now accept a ``--code`` option for providing a block of Python code (or a path to a ``.py`` file) that defines a ``rows()`` function or ``rows`` iterable of rows to insert, as an alternative to importing from a file. (:issue:`684`) +- ``sqlite-utils insert`` and ``sqlite-utils upsert`` now accept a ``--code`` option for :ref:`providing a block of Python code ` (or a path to a ``.py`` file) that defines a ``rows()`` function or ``rows`` iterable of rows to insert, as an alternative to importing from a file. (:issue:`684`) .. _v4_0: From d2ac3765ed9f0516bb0cbc2508a5c3907fb6a71a Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Tue, 7 Jul 2026 19:17:29 -0700 Subject: [PATCH 73/86] sqlite-utils insert/upsert --type colunm-name type option, closes #131 --- docs/changelog.rst | 1 + docs/cli-reference.rst | 8 +++++++ docs/cli.rst | 19 ++++++++++++++++ sqlite_utils/cli.py | 28 +++++++++++++++++++++++- tests/test_cli_insert.py | 47 ++++++++++++++++++++++++++++++++++++++++ 5 files changed, 102 insertions(+), 1 deletion(-) diff --git a/docs/changelog.rst b/docs/changelog.rst index 5d1816a..9e03f39 100644 --- a/docs/changelog.rst +++ b/docs/changelog.rst @@ -11,6 +11,7 @@ Unreleased - ``sqlite-utils query`` can now read the SQL query from standard input by passing ``-`` in place of the query, for example ``echo "select * from dogs" | sqlite-utils query dogs.db -``. (:issue:`765`) - ``sqlite-utils insert`` and ``sqlite-utils upsert`` now accept a ``--code`` option for :ref:`providing a block of Python code ` (or a path to a ``.py`` file) that defines a ``rows()`` function or ``rows`` iterable of rows to insert, as an alternative to importing from a file. (:issue:`684`) +- ``sqlite-utils insert`` and ``sqlite-utils upsert`` now accept ``--type column-name type`` to :ref:`override the type automatically chosen when the table is created `. This is useful for CSV or TSV columns such as ZIP codes that look like integers but should be stored as ``TEXT`` to preserve leading zeros. (:issue:`131`) .. _v4_0: diff --git a/docs/cli-reference.rst b/docs/cli-reference.rst index 5798809..39226ac 100644 --- a/docs/cli-reference.rst +++ b/docs/cli-reference.rst @@ -246,6 +246,9 @@ See :ref:`cli_inserting_data`, :ref:`cli_insert_csv_tsv`, :ref:`cli_insert_unstr - Use --lines to write each incoming line to a column called "line" - Use --text to write the entire input to a column called "text" + Use --type column-name type to override the type automatically chosen when the + table is created. + You can also use --convert to pass a fragment of Python code that will be used to convert each input. @@ -306,6 +309,7 @@ See :ref:`cli_inserting_data`, :ref:`cli_insert_csv_tsv`, :ref:`cli_insert_unstr --alter Alter existing table to add any missing columns --not-null TEXT Columns that should be created as NOT NULL --default ... Default value that should be set for a column + --type ... Column types to use when creating the table --no-detect-types Treat all CSV/TSV columns as TEXT --analyze Run ANALYZE at the end of this operation --load-extension TEXT Path to SQLite extension, with optional :entrypoint @@ -335,6 +339,9 @@ See :ref:`cli_upsert`. If the table already exists and has a primary key, --pk can be omitted. + Use --type column-name type to override the type automatically chosen when the + table is created. + Example: echo '[ @@ -366,6 +373,7 @@ See :ref:`cli_upsert`. --alter Alter existing table to add any missing columns --not-null TEXT Columns that should be created as NOT NULL --default ... Default value that should be set for a column + --type ... Column types to use when creating the table --no-detect-types Treat all CSV/TSV columns as TEXT --analyze Run ANALYZE at the end of this operation --load-extension TEXT Path to SQLite extension, with optional :entrypoint diff --git a/docs/cli.rst b/docs/cli.rst index 4ed92ed..731e93b 100644 --- a/docs/cli.rst +++ b/docs/cli.rst @@ -1366,6 +1366,25 @@ Will produce this schema with automatically detected types: "weight" REAL ); +.. _cli_insert_csv_tsv_column_types: + +Overriding column types +----------------------- + +Use ``--type column-name type`` to override the type automatically chosen when the table is created. This option can be used more than once, and works with both ``insert`` and ``upsert``: + +.. code-block:: bash + + sqlite-utils insert places.db places places.csv --csv \ + --type zipcode text \ + --type score real + +This is useful for values such as ZIP codes, which may look like integers but should be stored as ``TEXT`` to preserve leading zeros. + +The column type should be one of ``TEXT``, ``INTEGER``, ``FLOAT``, ``REAL`` or ``BLOB``. Column types are matched case-insensitively. + +As with detected column types, ``--type`` only affects tables created by the command. If the table already exists, its existing column types are left unchanged. + To disable type detection and treat all columns as TEXT, use ``--no-detect-types``: .. code-block:: bash diff --git a/sqlite_utils/cli.py b/sqlite_utils/cli.py index 4eefc70..cf39ff8 100644 --- a/sqlite_utils/cli.py +++ b/sqlite_utils/cli.py @@ -980,6 +980,16 @@ def insert_upsert_options(*, require_pk=False): type=(str, str), help="Default value that should be set for a column", ), + click.option( + "--type", + "types", + type=( + str, + click.Choice(list(VALID_COLUMN_TYPES), case_sensitive=False), + ), + multiple=True, + help="Column types to use when creating the table", + ), click.option( "--no-detect-types", is_flag=True, @@ -1034,6 +1044,7 @@ def insert_upsert_implementation( truncate=False, not_null=None, default=None, + types=None, no_detect_types=False, analyze=False, load_extension=None, @@ -1047,6 +1058,7 @@ def insert_upsert_implementation( _register_db_for_cleanup(db) _load_extensions(db, load_extension) _maybe_register_functions(db, functions) + column_type_overrides = {column: ctype.upper() for column, ctype in (types or [])} def _insert_docs(docs, tracker=None): extra_kwargs = { @@ -1060,6 +1072,8 @@ def insert_upsert_implementation( extra_kwargs["not_null"] = set(not_null) if default: extra_kwargs["defaults"] = dict(default) + if column_type_overrides: + extra_kwargs["columns"] = column_type_overrides if upsert: extra_kwargs["upsert"] = upsert @@ -1120,7 +1134,9 @@ def insert_upsert_implementation( and not table_existed_before_insert and db.table(table).exists() ): - db.table(table).transform(types=tracker.types) + detected_types = tracker.types + detected_types.update(column_type_overrides) + db.table(table).transform(types=detected_types) if code is not None: if file is not None: @@ -1330,6 +1346,7 @@ def insert( truncate, not_null, default, + types, strict, ): """ @@ -1348,6 +1365,9 @@ def insert( - Use --lines to write each incoming line to a column called "line" - Use --text to write the entire input to a column called "text" + Use --type column-name type to override the type automatically chosen + when the table is created. + You can also use --convert to pass a fragment of Python code that will be used to convert each input. @@ -1420,6 +1440,7 @@ def insert( silent=silent, not_null=not_null, default=default, + types=types, strict=strict, code=code, ) @@ -1454,6 +1475,7 @@ def upsert( alter, not_null, default, + types, no_detect_types, analyze, load_extension, @@ -1467,6 +1489,9 @@ def upsert( If the table already exists and has a primary key, --pk can be omitted. + Use --type column-name type to override the type automatically chosen + when the table is created. + Example: \b @@ -1501,6 +1526,7 @@ def upsert( upsert=True, not_null=not_null, default=default, + types=types, no_detect_types=no_detect_types, analyze=analyze, load_extension=load_extension, diff --git a/tests/test_cli_insert.py b/tests/test_cli_insert.py index e290f97..df6f80c 100644 --- a/tests/test_cli_insert.py +++ b/tests/test_cli_insert.py @@ -664,6 +664,53 @@ def test_insert_csv_detect_types_new_table(db_path): assert db["data"].columns_dict == {"name": str, "age": int, "weight": float} +@pytest.mark.parametrize( + "command,extra_args,input_text,expected_row", + ( + ( + "insert", + [], + "zipcode,score\n01234,9.5\n", + {"zipcode": "01234", "score": 9.5}, + ), + ( + "upsert", + ["--pk", "id"], + "id,zipcode,score\n1,01234,9.5\n", + {"id": 1, "zipcode": "01234", "score": 9.5}, + ), + ), +) +def test_insert_upsert_csv_type_overrides_detected_types( + db_path, command, extra_args, input_text, expected_row +): + result = CliRunner().invoke( + cli.cli, + [ + command, + db_path, + "places", + "-", + "--csv", + ] + + extra_args + + [ + "--type", + "zipcode", + "text", + ], + catch_exceptions=False, + input=input_text, + ) + assert result.exit_code == 0, result.output + db = Database(db_path) + expected_columns = {"zipcode": str, "score": float} + if command == "upsert": + expected_columns = {"id": int, **expected_columns} + assert db["places"].columns_dict == expected_columns + assert list(db["places"].rows) == [expected_row] + + def test_upsert_csv_detect_types_leaves_existing_table_alone(db_path): db = Database(db_path) db["places"].insert({"id": 1, "name": "Boston", "zip": "01234"}, pk="id") From d302835d57bcf53c36c0dc67356ec292f55a5931 Mon Sep 17 00:00:00 2001 From: Sai Asish Y Date: Wed, 13 May 2026 11:55:13 -0700 Subject: [PATCH 74/86] feat(db): document and test Database.memory and Database.memory_name Refs #590, closes #734 --- docs/python-api.rst | 8 ++++++++ tests/test_constructor.py | 31 +++++++++++++++++++++++++++++++ 2 files changed, 39 insertions(+) diff --git a/docs/python-api.rst b/docs/python-api.rst index 11af1f4..f8e0787 100644 --- a/docs/python-api.rst +++ b/docs/python-api.rst @@ -109,6 +109,14 @@ You can also create a named in-memory database. Unlike regular memory databases db = Database(memory_name="my_shared_database") +After creating a ``Database`` you can use ``db.memory`` and ``db.memory_name`` to tell whether it is backed by an in-memory database and to read the shared cache name. ``db.memory`` is ``True`` for any in-memory database and ``db.memory_name`` holds the name passed to ``memory_name=``, or ``None`` otherwise. + +.. code-block:: python + + db = Database(memory_name="shared") + db.memory # True + db.memory_name # "shared" + Connections use ``PRAGMA recursive_triggers=on`` by default. If you don't want to use `recursive triggers `__ you can turn them off using: .. code-block:: python diff --git a/tests/test_constructor.py b/tests/test_constructor.py index 7714f26..7428dcf 100644 --- a/tests/test_constructor.py +++ b/tests/test_constructor.py @@ -87,3 +87,34 @@ def test_legacy_transaction_control_connection_is_accepted(tmpdir): db["t"].insert({"id": 1}, pk="id") assert [r["id"] for r in db["t"].rows] == [1] db.close() + + +def test_memory_attribute_for_memory_true(): + db = Database(memory=True) + assert db.memory is True + assert db.memory_name is None + + +def test_memory_attribute_for_memory_name(): + db = Database(memory_name="shared_attr") + assert db.memory is True + assert db.memory_name == "shared_attr" + + +def test_memory_attribute_for_memory_string_path(): + db = Database(":memory:") + assert db.memory is True + assert db.memory_name is None + + +def test_memory_attribute_for_file_path(tmpdir): + db = Database(str(tmpdir / "file.db")) + assert db.memory is False + assert db.memory_name is None + + +def test_memory_attribute_for_existing_connection(): + conn = sqlite3.connect(":memory:") + db = Database(conn) + assert db.memory is False + assert db.memory_name is None From 7a52214624ae0e2c3fdf07215c1bcfc1393dbd93 Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Tue, 7 Jul 2026 21:00:43 -0700 Subject: [PATCH 75/86] drop-index command and table.drop_index(index_name) Closes #626 --- docs/changelog.rst | 1 + docs/cli-reference.rst | 26 +++++++++++++++++++++++++- docs/cli.rst | 13 +++++++++++++ docs/python-api.rst | 8 ++++++++ sqlite_utils/cli.py | 28 ++++++++++++++++++++++++++++ sqlite_utils/db.py | 16 ++++++++++++++++ tests/test_cli.py | 18 ++++++++++++++++++ tests/test_create.py | 28 ++++++++++++++++++++++++++++ 8 files changed, 137 insertions(+), 1 deletion(-) diff --git a/docs/changelog.rst b/docs/changelog.rst index 9e03f39..d2c07da 100644 --- a/docs/changelog.rst +++ b/docs/changelog.rst @@ -12,6 +12,7 @@ Unreleased - ``sqlite-utils query`` can now read the SQL query from standard input by passing ``-`` in place of the query, for example ``echo "select * from dogs" | sqlite-utils query dogs.db -``. (:issue:`765`) - ``sqlite-utils insert`` and ``sqlite-utils upsert`` now accept a ``--code`` option for :ref:`providing a block of Python code ` (or a path to a ``.py`` file) that defines a ``rows()`` function or ``rows`` iterable of rows to insert, as an alternative to importing from a file. (:issue:`684`) - ``sqlite-utils insert`` and ``sqlite-utils upsert`` now accept ``--type column-name type`` to :ref:`override the type automatically chosen when the table is created `. This is useful for CSV or TSV columns such as ZIP codes that look like integers but should be stored as ``TEXT`` to preserve leading zeros. (:issue:`131`) +- New ``table.drop_index(name)`` method and ``sqlite-utils drop-index`` command for dropping an index by name. Both accept ``ignore=True``/``--ignore`` to ignore a missing index. (:issue:`626`) .. _v4_0: diff --git a/docs/cli-reference.rst b/docs/cli-reference.rst index 39226ac..71e8377 100644 --- a/docs/cli-reference.rst +++ b/docs/cli-reference.rst @@ -19,7 +19,7 @@ This page lists the ``--help`` for every ``sqlite-utils`` CLI sub-command. go_first = [ "query", "memory", "insert", "upsert", "bulk", "search", "transform", "extract", "schema", "insert-files", "analyze-tables", "convert", "tables", "views", "rows", - "triggers", "indexes", "create-database", "create-table", "create-index", + "triggers", "indexes", "create-database", "create-table", "create-index", "drop-index", "migrate", "enable-fts", "populate-fts", "rebuild-fts", "disable-fts" ] refs = { @@ -46,6 +46,7 @@ This page lists the ``--help`` for every ``sqlite-utils`` CLI sub-command. "add-foreign-keys": "cli_add_foreign_keys", "index-foreign-keys": "cli_index_foreign_keys", "create-index": "cli_create_index", + "drop-index": "cli_drop_index", "enable-wal": "cli_wal", "enable-counts": "cli_enable_counts", "bulk": "cli_bulk", @@ -1006,6 +1007,29 @@ See :ref:`cli_create_index`. -h, --help Show this message and exit. +.. _cli_ref_drop_index: + +drop-index +========== + +See :ref:`cli_drop_index`. + +:: + + Usage: sqlite-utils drop-index [OPTIONS] PATH TABLE INDEX + + Drop an index by index name from the specified table + + Example: + + sqlite-utils drop-index chickens.db chickens idx_chickens_name + + Options: + --ignore Ignore if index does not exist + --load-extension TEXT Path to SQLite extension, with optional :entrypoint + -h, --help Show this message and exit. + + .. _cli_ref_migrate: migrate diff --git a/docs/cli.rst b/docs/cli.rst index 731e93b..c446a72 100644 --- a/docs/cli.rst +++ b/docs/cli.rst @@ -2615,6 +2615,19 @@ If your column names are already prefixed with a hyphen you'll need to manually Add the ``--analyze`` option to run ``ANALYZE`` against the index after it has been created. +.. _cli_drop_index: + +Dropping indexes +================ + +You can drop an index from an existing table using the ``drop-index`` command: + +.. code-block:: bash + + sqlite-utils drop-index mydb.db mytable idx_mytable_col1 + +Use ``--ignore`` to ignore the error if the index does not exist on that table. + .. _cli_fts: Configuring full-text search diff --git a/docs/python-api.rst b/docs/python-api.rst index f8e0787..fed617e 100644 --- a/docs/python-api.rst +++ b/docs/python-api.rst @@ -2877,6 +2877,14 @@ Use ``if_not_exists=True`` to do nothing if an index with that name already exis Pass ``analyze=True`` to run ``ANALYZE`` against the new index after creating it. +You can drop an index from a table using ``.drop_index(index_name)``: + +.. code-block:: python + + db.table("dogs").drop_index("idx_dogs_name") + +Use ``ignore=True`` to ignore the error if the index does not exist. + .. _python_api_analyze: Optimizing index usage with ANALYZE diff --git a/sqlite_utils/cli.py b/sqlite_utils/cli.py index cf39ff8..7fab72b 100644 --- a/sqlite_utils/cli.py +++ b/sqlite_utils/cli.py @@ -692,6 +692,34 @@ def create_index( ) +@cli.command(name="drop-index") +@click.argument( + "path", + type=click.Path(exists=True, file_okay=True, dir_okay=False, allow_dash=False), + required=True, +) +@click.argument("table") +@click.argument("index") +@click.option("--ignore", help="Ignore if index does not exist", is_flag=True) +@load_extension_option +def drop_index(path, table, index, ignore, load_extension): + """ + Drop an index by index name from the specified table + + Example: + + \b + sqlite-utils drop-index chickens.db chickens idx_chickens_name + """ + db = sqlite_utils.Database(path) + _register_db_for_cleanup(db) + _load_extensions(db, load_extension) + try: + db.table(table).drop_index(index, ignore=ignore) + except OperationalError as ex: + raise click.ClickException(str(ex)) + + @cli.command(name="enable-fts") @click.argument( "path", diff --git a/sqlite_utils/db.py b/sqlite_utils/db.py index 3033a36..62e3656 100644 --- a/sqlite_utils/db.py +++ b/sqlite_utils/db.py @@ -3080,6 +3080,22 @@ class Table(Queryable): self.db.analyze(created_index_name) return self + def drop_index(self, index_name: str, ignore: bool = False): + """ + Drop an index on this table. + + :param index_name: Name of the index to drop + :param ignore: Set to ``True`` to ignore the error if the index does not exist + """ + if index_name not in {index.name for index in self.indexes}: + if ignore: + return self + raise OperationalError( + "No index named {} on table {}".format(index_name, self.name) + ) + self.db.execute("DROP INDEX {}".format(quote_identifier(index_name))) + return self + def add_column( self, col_name: str, diff --git a/tests/test_cli.py b/tests/test_cli.py index bc5d492..a828c70 100644 --- a/tests/test_cli.py +++ b/tests/test_cli.py @@ -291,6 +291,24 @@ def test_create_index(db_path): ) +def test_drop_index(db_path): + db = Database(db_path) + db["Gosh"].create_index(["c1"]) + assert [index.name for index in db["Gosh"].indexes] == ["idx_Gosh_c1"] + result = CliRunner().invoke(cli.cli, ["drop-index", db_path, "Gosh", "idx_Gosh_c1"]) + assert result.exit_code == 0 + assert db["Gosh"].indexes == [] + + result = CliRunner().invoke(cli.cli, ["drop-index", db_path, "Gosh", "idx_Gosh_c1"]) + assert result.exit_code == 1 + assert "No index named idx_Gosh_c1" in result.output + + result = CliRunner().invoke( + cli.cli, ["drop-index", db_path, "Gosh", "idx_Gosh_c1", "--ignore"] + ) + assert result.exit_code == 0 + + def test_create_index_analyze(db_path): db = Database(db_path) assert "sqlite_stat1" not in db.table_names() diff --git a/tests/test_create.py b/tests/test_create.py index 7fab5e6..d281eb4 100644 --- a/tests/test_create.py +++ b/tests/test_create.py @@ -809,6 +809,34 @@ def test_create_index_if_not_exists(fresh_db): dogs.create_index(["name"], if_not_exists=True) +def test_drop_index(fresh_db): + dogs = fresh_db["dogs"] + dogs.insert({"name": "Cleo", "twitter": "cleopaws", "age": 3, "is_good_dog": True}) + dogs.create_index(["name"]) + assert [index.name for index in dogs.indexes] == ["idx_dogs_name"] + dogs.drop_index("idx_dogs_name") + assert dogs.indexes == [] + + +def test_drop_index_ignore(fresh_db): + dogs = fresh_db["dogs"] + dogs.insert({"name": "Cleo"}) + with pytest.raises(OperationalError, match="No index named idx_dogs_name"): + dogs.drop_index("idx_dogs_name") + dogs.drop_index("idx_dogs_name", ignore=True) + + +def test_drop_index_wrong_table(fresh_db): + dogs = fresh_db["dogs"] + cats = fresh_db["cats"] + dogs.insert({"name": "Cleo"}) + cats.insert({"name": "Misty"}) + dogs.create_index(["name"]) + with pytest.raises(OperationalError, match="No index named idx_dogs_name"): + cats.drop_index("idx_dogs_name") + assert [index.name for index in dogs.indexes] == ["idx_dogs_name"] + + def test_create_index_desc(fresh_db): dogs = fresh_db["dogs"] dogs.insert({"name": "Cleo", "twitter": "cleopaws", "age": 3, "is good dog": True}) From 0f2d525d0686c7cafcd819ab20bf0f89c6bf9f69 Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Sat, 11 Jul 2026 16:20:37 -0700 Subject: [PATCH 76/86] Clarify transaction documentation Based on extensive digging into how this stuff all works. --- docs/python-api.rst | 35 +++++++++++++++++++++++++++++++++-- 1 file changed, 33 insertions(+), 2 deletions(-) diff --git a/docs/python-api.rst b/docs/python-api.rst index fed617e..d444806 100644 --- a/docs/python-api.rst +++ b/docs/python-api.rst @@ -320,11 +320,15 @@ Every method in this library that writes to the database - ``insert()``, ``upser The same applies to raw SQL executed with :ref:`db.execute() ` - a write statement is committed as soon as it has run. +Another way to think about this is that each sqlite-utils method call is its own unit of work. If several method calls must either all succeed or all fail, use ``db.atomic()`` to turn them into a single unit of work. + You never need to call ``commit()``, and you do not need to close the database to persist your changes. There are exactly two situations where you need to think about transactions: 1. You want to group several write operations together, so they either all succeed or all fail - use :ref:`db.atomic() `. 2. You are :ref:`managing a transaction yourself ` with ``db.begin()``, in which case nothing is committed until you commit - the library will never commit a transaction you opened. +``with Database(...) as db:`` is not a transaction block. It manages the lifetime of the database connection and closes it on exit. Use ``with db.atomic():`` for a transaction. + .. _python_api_atomic: Grouping changes with db.atomic() @@ -340,6 +344,27 @@ Use ``db.atomic()`` to group multiple operations in a single transaction: The transaction commits when the block exits. If an exception is raised, changes made inside the block will be rolled back. +This matters when several operations represent a single logical change. Without ``db.atomic()``, an earlier method call remains committed if a later one fails: + +.. code-block:: python + + # These are two separate transactions + db.table("accounts").update(1, {"balance": 90}) + db.table("accounts").update(2, {"balance": 110}) + + # These updates either both succeed or both fail + with db.atomic(): + db.table("accounts").update(1, {"balance": 90}) + db.table("accounts").update(2, {"balance": 110}) + +Transactions can also improve performance. Calling ``insert()`` repeatedly outside ``db.atomic()`` creates and commits a separate transaction for every call. For bulk inserts, prefer :ref:`insert_all() `. If you need to call several different methods in a loop, wrap the loop in ``db.atomic()``: + +.. code-block:: python + + with db.atomic(): + for row in rows: + db.table("events").insert(row) + ``db.atomic()`` can be nested. Nested blocks use SQLite savepoints, so an exception in an inner block can roll back to that savepoint without rolling back the entire outer transaction: .. code-block:: python @@ -368,6 +393,8 @@ Write statements executed with :ref:`db.execute() ` follow t db.execute("insert into news (headline) values (?)", ["Dog wins award"]) # Already committed +``db.execute()`` participates in sqlite-utils transaction handling. Calling ``db.conn.execute()`` directly bypasses that policy and leaves transaction handling to Python's underlying ``sqlite3.Connection``. Prefer ``db.execute()`` unless you deliberately need the lower-level API. + If a transaction is open - because the call happens inside a ``db.atomic()`` block, or after ``db.begin()`` - the statement becomes part of that transaction instead, and commits when the transaction commits: .. code-block:: python @@ -399,6 +426,8 @@ You can take full manual control using the ``db.begin()``, ``db.commit()`` and ` The library will never commit a transaction you opened. If you call write methods such as ``insert()`` - or use ``db.atomic()`` - while your transaction is open, they participate in it using SQLite savepoints instead of committing: exiting an ``atomic()`` block releases its savepoint, but nothing is saved to disk until you commit the outer transaction yourself. If you roll back, their changes are rolled back too. +Prefer ``db.atomic()`` or ``db.begin()``, ``db.commit()`` and ``db.rollback()`` over mixing sqlite-utils transaction methods with calls to ``db.conn.commit()``, ``db.conn.rollback()`` or raw transaction-control SQL. Mixing the two layers makes it much harder to tell which layer owns the current transaction. + Two related safeguards to be aware of: - ``db.enable_wal()`` and ``db.disable_wal()`` raise a ``sqlite_utils.db.TransactionError`` if called while a transaction is open, because changing the journal mode would commit it as a side effect. @@ -409,9 +438,11 @@ Two related safeguards to be aware of: Supported connection modes -------------------------- -``db.atomic()`` and the automatic per-method transactions require a connection in Python's default transaction handling mode. Passing a connection created with the Python 3.12+ ``sqlite3.connect(..., autocommit=True)`` or ``autocommit=False`` options to ``Database()`` raises a ``sqlite_utils.db.TransactionError``. +``db.atomic()`` and the automatic per-method transactions currently require a connection using Python's legacy transaction control mode (``sqlite3.LEGACY_TRANSACTION_CONTROL`` on Python 3.12 and later). Passing a connection created with the Python 3.12+ ``sqlite3.connect(..., autocommit=True)`` or ``autocommit=False`` options to ``Database()`` raises a ``sqlite_utils.db.TransactionError``. -This is because ``commit()`` and ``rollback()`` behave differently on those connections - under ``autocommit=True`` they are documented no-ops - which would cause every write made by this library to be silently discarded when the connection closed, rather than failing loudly. +Connections using ``autocommit=False`` are not supported because Python keeps a transaction open continuously. sqlite-utils uses ``Connection.in_transaction`` to distinguish its own transactions from transactions opened by its caller, and that distinction is not available in this mode. + +Connections using ``autocommit=True`` are also currently rejected because sqlite-utils has not formally exposed that as a supported configuration. .. _python_api_table: From 6531a57863ce23d502e504fd8fcd375fbe5cbb7f Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Sat, 11 Jul 2026 16:28:36 -0700 Subject: [PATCH 77/86] Delete obsolete test_memory_attribute_for_existing_connection test Refs https://github.com/simonw/sqlite-utils/issues/789#issuecomment-4949146177 Closes #789 --- tests/test_constructor.py | 7 ------- 1 file changed, 7 deletions(-) diff --git a/tests/test_constructor.py b/tests/test_constructor.py index 7428dcf..a619fba 100644 --- a/tests/test_constructor.py +++ b/tests/test_constructor.py @@ -111,10 +111,3 @@ def test_memory_attribute_for_file_path(tmpdir): db = Database(str(tmpdir / "file.db")) assert db.memory is False assert db.memory_name is None - - -def test_memory_attribute_for_existing_connection(): - conn = sqlite3.connect(":memory:") - db = Database(conn) - assert db.memory is False - assert db.memory_name is None From dc61f75a0ba84bca00f88da374df29610e8164cb Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Sat, 11 Jul 2026 16:42:59 -0700 Subject: [PATCH 78/86] sqlite-utils upsert optional --pk in changelog --- docs/changelog.rst | 1 + 1 file changed, 1 insertion(+) diff --git a/docs/changelog.rst b/docs/changelog.rst index d2c07da..7333b26 100644 --- a/docs/changelog.rst +++ b/docs/changelog.rst @@ -13,6 +13,7 @@ Unreleased - ``sqlite-utils insert`` and ``sqlite-utils upsert`` now accept a ``--code`` option for :ref:`providing a block of Python code ` (or a path to a ``.py`` file) that defines a ``rows()`` function or ``rows`` iterable of rows to insert, as an alternative to importing from a file. (:issue:`684`) - ``sqlite-utils insert`` and ``sqlite-utils upsert`` now accept ``--type column-name type`` to :ref:`override the type automatically chosen when the table is created `. This is useful for CSV or TSV columns such as ZIP codes that look like integers but should be stored as ``TEXT`` to preserve leading zeros. (:issue:`131`) - New ``table.drop_index(name)`` method and ``sqlite-utils drop-index`` command for dropping an index by name. Both accept ``ignore=True``/``--ignore`` to ignore a missing index. (:issue:`626`) +- ``sqlite-utils upsert`` can now infer the primary key of an existing table, so ``--pk`` can be omitted when upserting into a table that already has a primary key. .. _v4_0: From b74b72703588863464880232e596001d958df180 Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Sat, 11 Jul 2026 16:43:37 -0700 Subject: [PATCH 79/86] .transform(strict=) and sqlite-utils transform --strict/--no-strict (#788) * .transform(strict=) and sqlite-utils transform --strict/--no-strict Closes #787 --- docs/changelog.rst | 2 ++ docs/cli-reference.rst | 2 ++ docs/cli.rst | 8 +++++- docs/python-api.rst | 23 +++++++++++++++ sqlite_utils/cli.py | 8 ++++++ sqlite_utils/db.py | 13 ++++++++- tests/test_cli.py | 59 +++++++++++++++++++++++++++++++++++++ tests/test_transform.py | 64 +++++++++++++++++++++++++++++++++++++---- 8 files changed, 171 insertions(+), 8 deletions(-) diff --git a/docs/changelog.rst b/docs/changelog.rst index 7333b26..61f8381 100644 --- a/docs/changelog.rst +++ b/docs/changelog.rst @@ -9,6 +9,8 @@ Unreleased ---------- +- ``table.transform()`` and ``table.transform_sql()`` now accept ``strict=True`` or ``strict=False`` to change a table's SQLite strict mode. Omitting the option, or passing ``strict=None``, preserves the existing mode. (:issue:`787`) +- The ``sqlite-utils transform`` command now accepts ``--strict`` and ``--no-strict`` to change a table's SQLite strict mode. Omitting both options preserves the existing mode. (:issue:`787`) - ``sqlite-utils query`` can now read the SQL query from standard input by passing ``-`` in place of the query, for example ``echo "select * from dogs" | sqlite-utils query dogs.db -``. (:issue:`765`) - ``sqlite-utils insert`` and ``sqlite-utils upsert`` now accept a ``--code`` option for :ref:`providing a block of Python code ` (or a path to a ``.py`` file) that defines a ``rows()`` function or ``rows`` iterable of rows to insert, as an alternative to importing from a file. (:issue:`684`) - ``sqlite-utils insert`` and ``sqlite-utils upsert`` now accept ``--type column-name type`` to :ref:`override the type automatically chosen when the table is created `. This is useful for CSV or TSV columns such as ZIP codes that look like integers but should be stored as ``TEXT`` to preserve leading zeros. (:issue:`131`) diff --git a/docs/cli-reference.rst b/docs/cli-reference.rst index 71e8377..9fafe28 100644 --- a/docs/cli-reference.rst +++ b/docs/cli-reference.rst @@ -508,6 +508,8 @@ See :ref:`cli_transform_table`. Add a foreign key constraint from a column to another table with another column --drop-foreign-key TEXT Drop foreign key constraint for this column + --strict / --no-strict Enable or disable STRICT mode (default: + preserve current mode) --sql Output SQL without executing it --load-extension TEXT Path to SQLite extension, with optional :entrypoint diff --git a/docs/cli.rst b/docs/cli.rst index c446a72..dce36d9 100644 --- a/docs/cli.rst +++ b/docs/cli.rst @@ -2182,7 +2182,7 @@ Use ``--ignore`` to ignore the error if the table does not exist. Transforming tables =================== -The ``transform`` command allows you to apply complex transformations to a table that cannot be implemented using a regular SQLite ``ALTER TABLE`` command. See :ref:`python_api_transform` for details of how this works. The ``transform`` command preserves a table's ``STRICT`` mode. +The ``transform`` command allows you to apply complex transformations to a table that cannot be implemented using a regular SQLite ``ALTER TABLE`` command. See :ref:`python_api_transform` for details of how this works. By default, the ``transform`` command preserves a table's ``STRICT`` mode. .. code-block:: bash @@ -2228,6 +2228,12 @@ Every option for this table (with the exception of ``--pk-none``) can be specifi ``--add-foreign-key column other_table other_column`` Add a foreign key constraint to ``column`` pointing to ``other_table.other_column``. +``--strict`` + Convert the table to a `SQLite STRICT table `__. The command fails if the available SQLite version does not support strict tables. If existing rows contain values that are incompatible with their declared column types the transformation fails and the original table is left unchanged. + +``--no-strict`` + Convert a strict table back to a regular non-strict table. + If you want to see the SQL that will be executed to make the change without actually executing it, add the ``--sql`` flag. For example: .. code-block:: bash diff --git a/docs/python-api.rst b/docs/python-api.rst index d444806..8b07677 100644 --- a/docs/python-api.rst +++ b/docs/python-api.rst @@ -1753,6 +1753,29 @@ To alter the type of a column, use the ``types=`` argument: See :ref:`python_api_add_column` for a list of available types. +.. _python_api_transform_strict: + +Changing strict mode +-------------------- + +The optional ``strict=`` parameter can change whether a table uses `SQLite STRICT mode `__. Pass ``strict=True`` to convert a regular table to a strict table: + +.. code-block:: python + + table.transform(strict=True) + +Pass ``strict=False`` to convert a strict table back to a regular non-strict table: + +.. code-block:: python + + table.transform(strict=False) + +The default is ``strict=None``, which preserves the table's existing strict mode. + +Passing ``strict=True`` raises ``sqlite_utils.db.TransformError`` if the available SQLite version does not support strict tables. + +Converting to a strict table validates all existing rows as they are copied into the replacement table. If a value is incompatible with its declared column type, SQLite raises ``sqlite3.IntegrityError`` and the transformation is rolled back, leaving the original table and its data unchanged. + .. _python_api_transform_rename_columns: Renaming columns diff --git a/sqlite_utils/cli.py b/sqlite_utils/cli.py index 7fab72b..e0b8969 100644 --- a/sqlite_utils/cli.py +++ b/sqlite_utils/cli.py @@ -2718,6 +2718,11 @@ def schema( multiple=True, help="Drop foreign key constraint for this column", ) +@click.option( + "--strict/--no-strict", + default=None, + help="Enable or disable STRICT mode (default: preserve current mode)", +) @click.option("--sql", is_flag=True, help="Output SQL without executing it") @load_extension_option def transform( @@ -2735,6 +2740,7 @@ def transform( default_none, add_foreign_keys, drop_foreign_keys, + strict, sql, load_extension, ): @@ -2796,6 +2802,7 @@ def transform( defaults=default_dict, drop_foreign_keys=drop_foreign_keys_value, add_foreign_keys=add_foreign_keys_value, + strict=strict, ): click.echo(line) else: @@ -2809,6 +2816,7 @@ def transform( defaults=default_dict, drop_foreign_keys=drop_foreign_keys_value, add_foreign_keys=add_foreign_keys_value, + strict=strict, ) diff --git a/sqlite_utils/db.py b/sqlite_utils/db.py index 62e3656..d709fb9 100644 --- a/sqlite_utils/db.py +++ b/sqlite_utils/db.py @@ -2514,6 +2514,7 @@ class Table(Queryable): foreign_keys: Optional[ForeignKeysType] = None, column_order: Optional[List[str]] = None, keep_table: Optional[str] = None, + strict: Optional[bool] = None, ) -> "Table": """ Apply an advanced alter table, including operations that are not supported by @@ -2536,6 +2537,8 @@ class Table(Queryable): to use when creating the table :param keep_table: If specified, the existing table will be renamed to this and will not be dropped + :param strict: Set to ``True`` to make the table strict or ``False`` to make it + non-strict. Defaults to ``None``, which preserves the existing strict mode. """ if not self.exists(): raise ValueError("Cannot transform a table that doesn't exist yet") @@ -2551,6 +2554,7 @@ class Table(Queryable): foreign_keys=foreign_keys, column_order=column_order, keep_table=keep_table, + strict=strict, ) pragma_foreign_keys_was_on = bool( self.db.execute("PRAGMA foreign_keys").fetchone()[0] @@ -2587,6 +2591,8 @@ class Table(Queryable): self.db.execute("PRAGMA defer_foreign_keys=OFF;") if should_disable_foreign_keys: self.db.execute("PRAGMA foreign_keys=1;") + if strict is not None: + self._defaults["strict"] = strict return self def transform_sql( @@ -2604,6 +2610,7 @@ class Table(Queryable): column_order: Optional[List[str]] = None, tmp_suffix: Optional[str] = None, keep_table: Optional[str] = None, + strict: Optional[bool] = None, ) -> List[str]: """ Return a list of SQL statements that should be executed in order to apply this transformation. @@ -2624,7 +2631,11 @@ class Table(Queryable): :param tmp_suffix: Suffix to use for the temporary table name :param keep_table: If specified, the existing table will be renamed to this and will not be dropped + :param strict: Set to ``True`` to make the table strict or ``False`` to make it + non-strict. Defaults to ``None``, which preserves the existing strict mode. """ + if strict is True and not self.db.supports_strict: + raise TransformError("SQLite does not support STRICT tables") types = types or {} rename = rename or {} drop = drop or set() @@ -2806,7 +2817,7 @@ class Table(Queryable): defaults=create_table_defaults, foreign_keys=create_table_foreign_keys, column_order=column_order, - strict=self.strict, + strict=self.strict if strict is None else strict, ).strip() ) diff --git a/tests/test_cli.py b/tests/test_cli.py index a828c70..a2135b0 100644 --- a/tests/test_cli.py +++ b/tests/test_cli.py @@ -3,6 +3,7 @@ from sqlite_utils.db import Index, ForeignKey from click.testing import CliRunner from pathlib import Path import subprocess +import sqlite3 import sys import json import os @@ -1939,6 +1940,64 @@ def test_transform_sql(db_path): assert db["dogs"].schema == original_schema +@pytest.mark.parametrize( + "initial_strict,args,expected_strict", + ( + (False, [], False), + (True, [], True), + (False, ["--strict"], True), + (True, ["--no-strict"], False), + ), +) +def test_transform_strict_option(db_path, initial_strict, args, expected_strict): + db = Database(db_path) + if not db.supports_strict: + pytest.skip("SQLite version does not support strict tables") + db["dogs"].create({"id": int}, strict=initial_strict) + + result = CliRunner().invoke(cli.cli, ["transform", db_path, "dogs"] + args) + + assert result.exit_code == 0, result.output + assert db["dogs"].strict is expected_strict + + +@pytest.mark.parametrize( + "initial_strict,flag,sql_is_strict", + ( + (False, "--strict", True), + (True, "--no-strict", False), + ), +) +def test_transform_strict_option_sql(db_path, initial_strict, flag, sql_is_strict): + db = Database(db_path) + if not db.supports_strict: + pytest.skip("SQLite version does not support strict tables") + db["dogs"].create({"id": int}, strict=initial_strict) + + result = CliRunner().invoke(cli.cli, ["transform", db_path, "dogs", flag, "--sql"]) + + assert result.exit_code == 0, result.output + assert (") STRICT;" in result.output) is sql_is_strict + assert db["dogs"].strict is initial_strict + + +def test_transform_strict_option_with_invalid_data(db_path): + db = Database(db_path) + if not db.supports_strict: + pytest.skip("SQLite version does not support strict tables") + dogs = db["dogs"] + dogs.create({"id": int}) + dogs.insert({"id": "not-an-integer"}) + + result = CliRunner().invoke(cli.cli, ["transform", db_path, "dogs", "--strict"]) + + assert result.exit_code == 1 + assert isinstance(result.exception, sqlite3.IntegrityError) + assert dogs.strict is False + assert list(dogs.rows) == [{"id": "not-an-integer"}] + assert not any(name.startswith("dogs_new_") for name in db.table_names()) + + @pytest.mark.parametrize( "extra_args,expected_schema", ( diff --git a/tests/test_transform.py b/tests/test_transform.py index 71518be..f0f5019 100644 --- a/tests/test_transform.py +++ b/tests/test_transform.py @@ -1,3 +1,5 @@ +import sqlite3 + from sqlite_utils.db import ForeignKey, TransformError from sqlite_utils.utils import OperationalError import pytest @@ -566,13 +568,63 @@ def test_transform_preserves_rowids(fresh_db, table_type): assert previous_rows == next_rows -@pytest.mark.parametrize("strict", (False, True)) -def test_transform_strict(fresh_db, strict): - dogs = fresh_db.table("dogs", strict=strict) +@pytest.mark.parametrize( + "initial_strict,transform_strict,expected_strict", + ( + (False, None, False), + (True, None, True), + (False, True, True), + (True, False, False), + ), +) +def test_transform_strict(fresh_db, initial_strict, transform_strict, expected_strict): + if not fresh_db.supports_strict: + pytest.skip("SQLite version does not support strict tables") + dogs = fresh_db.table("dogs", strict=initial_strict) dogs.insert({"id": 1, "name": "Cleo"}) - assert dogs.strict == strict or not fresh_db.supports_strict - dogs.transform(not_null={"name"}) - assert dogs.strict == strict or not fresh_db.supports_strict + assert dogs.strict is initial_strict + dogs.transform(strict=transform_strict) + assert dogs.strict is expected_strict + + +def test_transform_to_strict_with_invalid_data(fresh_db): + if not fresh_db.supports_strict: + pytest.skip("SQLite version does not support strict tables") + dogs = fresh_db["dogs"] + dogs.create({"id": int}) + dogs.insert({"id": "not-an-integer"}) + + with pytest.raises(sqlite3.IntegrityError): + dogs.transform(strict=True) + + assert dogs.strict is False + assert list(dogs.rows) == [{"id": "not-an-integer"}] + assert fresh_db.table_names() == ["dogs"] + + +def test_transform_strict_updates_default(fresh_db): + if not fresh_db.supports_strict: + pytest.skip("SQLite version does not support strict tables") + table = fresh_db.table("items", strict=True) + table.create({"id": int}) + + table.transform(strict=False) + assert table.strict is False + + table.create({"id": int}, replace=True) + assert table.strict is False + + +@pytest.mark.parametrize("method_name", ("transform", "transform_sql")) +def test_transform_to_strict_not_supported(fresh_db, method_name): + table = fresh_db["items"] + table.create({"id": int}) + fresh_db._supports_strict = False + + with pytest.raises(TransformError, match="SQLite does not support STRICT tables"): + getattr(table, method_name)(strict=True) + + assert table.strict is False @pytest.mark.parametrize( From 57c16173912e00c639b9620cadc4dfe475ddb273 Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Sat, 11 Jul 2026 16:48:41 -0700 Subject: [PATCH 80/86] Release 4.1 Refs #131, #626, #684, #765, #787 --- docs/changelog.rst | 12 ++++++------ pyproject.toml | 2 +- 2 files changed, 7 insertions(+), 7 deletions(-) diff --git a/docs/changelog.rst b/docs/changelog.rst index 61f8381..5b9355f 100644 --- a/docs/changelog.rst +++ b/docs/changelog.rst @@ -4,18 +4,18 @@ Changelog =========== -.. _v_unreleased: +.. _v4_1: -Unreleased ----------- +4.1 (2026-07-11) +---------------- -- ``table.transform()`` and ``table.transform_sql()`` now accept ``strict=True`` or ``strict=False`` to change a table's SQLite strict mode. Omitting the option, or passing ``strict=None``, preserves the existing mode. (:issue:`787`) -- The ``sqlite-utils transform`` command now accepts ``--strict`` and ``--no-strict`` to change a table's SQLite strict mode. Omitting both options preserves the existing mode. (:issue:`787`) -- ``sqlite-utils query`` can now read the SQL query from standard input by passing ``-`` in place of the query, for example ``echo "select * from dogs" | sqlite-utils query dogs.db -``. (:issue:`765`) - ``sqlite-utils insert`` and ``sqlite-utils upsert`` now accept a ``--code`` option for :ref:`providing a block of Python code ` (or a path to a ``.py`` file) that defines a ``rows()`` function or ``rows`` iterable of rows to insert, as an alternative to importing from a file. (:issue:`684`) - ``sqlite-utils insert`` and ``sqlite-utils upsert`` now accept ``--type column-name type`` to :ref:`override the type automatically chosen when the table is created `. This is useful for CSV or TSV columns such as ZIP codes that look like integers but should be stored as ``TEXT`` to preserve leading zeros. (:issue:`131`) - New ``table.drop_index(name)`` method and ``sqlite-utils drop-index`` command for dropping an index by name. Both accept ``ignore=True``/``--ignore`` to ignore a missing index. (:issue:`626`) +- ``sqlite-utils query`` can now read the SQL query from standard input by passing ``-`` in place of the query, for example ``echo "select * from dogs" | sqlite-utils query dogs.db -``. (:issue:`765`) - ``sqlite-utils upsert`` can now infer the primary key of an existing table, so ``--pk`` can be omitted when upserting into a table that already has a primary key. +- ``table.transform()`` and ``table.transform_sql()`` now accept ``strict=True`` or ``strict=False`` to change a table's `SQLite strict mode `__. Omitting the option preserves the existing mode. (:issue:`787`) +- The ``sqlite-utils transform`` command now accepts ``--strict`` and ``--no-strict`` to change a table's strict mode. (:issue:`787`) .. _v4_0: diff --git a/pyproject.toml b/pyproject.toml index a3a42a9..971f5a0 100644 --- a/pyproject.toml +++ b/pyproject.toml @@ -1,6 +1,6 @@ [project] name = "sqlite-utils" -version = "4.0" +version = "4.1" description = "CLI tool and Python library for manipulating SQLite databases" readme = { file = "README.md", content-type = "text/markdown" } authors = [ From 5e8822efd27da9bc8495990f1decadea5f2b3167 Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Sat, 11 Jul 2026 18:52:09 -0700 Subject: [PATCH 81/86] Clarify usage of named parameters in CLI documentation --- docs/cli.rst | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/docs/cli.rst b/docs/cli.rst index dce36d9..8f65f45 100644 --- a/docs/cli.rst +++ b/docs/cli.rst @@ -361,7 +361,7 @@ To return the first column of each result as raw data, separated by newlines, us Using named parameters ---------------------- -You can pass named parameters to the query using ``-p``: +You can pass named parameters to the query using ``-p name value``: .. code-block:: bash From 3f0471701b5f8c7de888a467020e5dd34310ce9a Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Sat, 11 Jul 2026 21:45:21 -0700 Subject: [PATCH 82/86] Add cross-reference notes between CLI and Python API documentation (#791) --- docs/cli.rst | 96 +++++++++++++++++++++++++++++++++++++ docs/python-api.rst | 114 ++++++++++++++++++++++++++++++++++++++++++++ 2 files changed, 210 insertions(+) diff --git a/docs/cli.rst b/docs/cli.rst index 8f65f45..2e506dd 100644 --- a/docs/cli.rst +++ b/docs/cli.rst @@ -424,6 +424,9 @@ The ``--functions`` option can be used multiple times to load functions from mul from urllib.parse import urlparse return urlparse(url).path' +.. note:: + In Python: :ref:`db.register_function() ` + .. _cli_query_extensions: SQLite extensions @@ -1024,6 +1027,9 @@ To show more than 10 common values, use ``--common-limit 20``. To skip the most sqlite-utils analyze-tables github.db tags --common-limit 20 --no-least +.. note:: + In Python: :ref:`table.analyze_column() ` CLI reference: :ref:`sqlite-utils analyze-tables ` + .. _cli_analyze_tables_save: Saving the analyzed table details @@ -1191,6 +1197,9 @@ You can delete all the existing rows in the table before inserting the new recor You can add the ``--analyze`` option to run ``ANALYZE`` against the table after the rows have been inserted. +.. note:: + In Python: :ref:`table.insert_all() ` CLI reference: :ref:`sqlite-utils insert ` + .. _cli_inserting_data_binary: Inserting binary data @@ -1615,6 +1624,9 @@ To replace a dog with in ID of 2 with a new record, run the following: echo '{"id": 2, "name": "Pancakes", "age": 3}' | \ sqlite-utils insert dogs.db dogs - --pk=id --replace +.. note:: + In Python: :ref:`table.insert(..., replace=True) ` CLI reference: :ref:`sqlite-utils insert ` + .. _cli_upsert: Upserting data @@ -1641,6 +1653,9 @@ The command will fail if you reference columns that do not exist on the table. T ``upsert`` in sqlite-utils 1.x worked like ``insert ... --replace`` does in 2.x. See `issue #66 `__ for details of this change. +.. note:: + In Python: :ref:`table.upsert() ` CLI reference: :ref:`sqlite-utils upsert ` + .. _cli_bulk: Executing SQL in bulk @@ -1842,6 +1857,9 @@ You can include named parameters in your where clause and populate them using on The ``--dry-run`` option will output a preview of the conversion against the first ten rows, without modifying the database. +.. note:: + In Python: :ref:`table.convert() ` CLI reference: :ref:`sqlite-utils convert ` + .. _cli_convert_import: Importing additional modules @@ -2140,6 +2158,9 @@ If a table with the same name already exists, you will get an error. You can cho You can also pass ``--transform`` to transform the existing table to match the new schema. See :ref:`python_api_explicit_create` in the Python library documentation for details of how this option works. +.. note:: + In Python: :ref:`table.create() ` CLI reference: :ref:`sqlite-utils create-table ` + .. _cli_renaming_tables: Renaming a table @@ -2153,6 +2174,9 @@ Yo ucan rename a table using the ``rename-table`` command: Pass ``--ignore`` to ignore any errors caused by the table not existing, or the new name already being in use. +.. note:: + In Python: :ref:`db.rename_table() ` CLI reference: :ref:`sqlite-utils rename-table ` + .. _cli_duplicate_table: Duplicating tables @@ -2164,6 +2188,9 @@ The ``duplicate`` command duplicates a table - creating a new table with the sam sqlite-utils duplicate books.db authors authors_copy +.. note:: + In Python: :ref:`table.duplicate() ` CLI reference: :ref:`sqlite-utils duplicate ` + .. _cli_drop_table: Dropping tables @@ -2177,6 +2204,9 @@ You can drop a table using the ``drop-table`` command: Use ``--ignore`` to ignore the error if the table does not exist. +.. note:: + In Python: :ref:`table.drop() ` CLI reference: :ref:`sqlite-utils drop-table ` + .. _cli_transform_table: Transforming tables @@ -2260,6 +2290,9 @@ If you want to see the SQL that will be executed to make the change without actu DROP TABLE "roadside_attractions"; ALTER TABLE "roadside_attractions_new_4033a60276b9" RENAME TO "roadside_attractions"; +.. note:: + In Python: :ref:`table.transform() ` CLI reference: :ref:`sqlite-utils transform ` + .. _cli_transform_table_add_primary_key_to_rowid: Adding a primary key to a rowid table @@ -2437,6 +2470,9 @@ After running the above, the command ``sqlite-utils schema global.db`` reveals t CREATE UNIQUE INDEX "idx_countries_country_name" ON "countries" ("country", "name"); +.. note:: + In Python: :ref:`table.extract() ` CLI reference: :ref:`sqlite-utils extract ` + .. _cli_create_view: Creating views @@ -2458,6 +2494,9 @@ You can create a view using the ``create-view`` command: Use ``--replace`` to replace an existing view of the same name, and ``--ignore`` to do nothing if a view already exists. +.. note:: + In Python: :ref:`db.create_view() ` CLI reference: :ref:`sqlite-utils create-view ` + .. _cli_drop_view: Dropping views @@ -2471,6 +2510,9 @@ You can drop a view using the ``drop-view`` command: Use ``--ignore`` to ignore the error if the view does not exist. +.. note:: + In Python: :ref:`view.drop() ` CLI reference: :ref:`sqlite-utils drop-view ` + .. _cli_add_column: Adding columns @@ -2511,6 +2553,9 @@ You can set a ``NOT NULL DEFAULT 'x'`` constraint on the new column using ``--no sqlite-utils add-column mydb.db dogs friends_count integer --not-null-default 0 +.. note:: + In Python: :ref:`table.add_column() ` CLI reference: :ref:`sqlite-utils add-column ` + .. _cli_add_column_alter: Adding columns automatically on insert/update @@ -2522,6 +2567,9 @@ You can use the ``--alter`` option to automatically add new columns if the data sqlite-utils insert dogs.db dogs new-dogs.json --pk=id --alter +.. note:: + In Python: :ref:`table.insert(..., alter=True) ` + .. _cli_add_foreign_key: Adding foreign key constraints @@ -2549,6 +2597,9 @@ Add ``--ignore`` to ignore an existing foreign key (as opposed to returning an e See :ref:`python_api_add_foreign_key` in the Python API documentation for further details, including how the automatic table guessing mechanism works. +.. note:: + In Python: :ref:`table.add_foreign_key() ` CLI reference: :ref:`sqlite-utils add-foreign-key ` + .. _cli_add_foreign_keys: Adding multiple foreign keys at once @@ -2564,6 +2615,9 @@ Adding a foreign key requires a ``VACUUM``. On large databases this can be an ex When you are using this command each foreign key needs to be defined in full, as four arguments - the table, column, other table and other column. +.. note:: + In Python: :ref:`db.add_foreign_keys() ` CLI reference: :ref:`sqlite-utils add-foreign-keys ` + .. _cli_index_foreign_keys: Adding indexes for all foreign keys @@ -2575,6 +2629,9 @@ If you want to ensure that every foreign key column in your database has a corre sqlite-utils index-foreign-keys books.db +.. note:: + In Python: :ref:`db.index_foreign_keys() ` CLI reference: :ref:`sqlite-utils index-foreign-keys ` + .. _cli_defaults_not_null: Setting defaults and not null constraints @@ -2590,6 +2647,9 @@ You can use the ``--not-null`` and ``--default`` options (to both ``insert`` and --default age 2 \ --default score 5 +.. note:: + In Python: :ref:`not_null= and defaults= arguments ` + .. _cli_create_index: Creating indexes @@ -2621,6 +2681,9 @@ If your column names are already prefixed with a hyphen you'll need to manually Add the ``--analyze`` option to run ``ANALYZE`` against the index after it has been created. +.. note:: + In Python: :ref:`table.create_index() ` CLI reference: :ref:`sqlite-utils create-index ` + .. _cli_drop_index: Dropping indexes @@ -2634,6 +2697,9 @@ You can drop an index from an existing table using the ``drop-index`` command: Use ``--ignore`` to ignore the error if the index does not exist on that table. +.. note:: + In Python: :ref:`table.drop_index() ` CLI reference: :ref:`sqlite-utils drop-index ` + .. _cli_fts: Configuring full-text search @@ -2687,6 +2753,9 @@ You can rebuild every FTS table by running ``rebuild-fts`` without passing any t sqlite-utils rebuild-fts mydb.db +.. note:: + In Python: :ref:`table.enable_fts() ` CLI reference: :ref:`sqlite-utils enable-fts ` + .. _cli_search: Executing searches @@ -2743,6 +2812,9 @@ Use the ``--sql`` option to output the SQL that would be executed, rather than r order by "documents_fts".rank +.. note:: + In Python: :ref:`table.search() ` CLI reference: :ref:`sqlite-utils search ` + .. _cli_enable_counts: Enabling cached counts @@ -2766,6 +2838,9 @@ If the ``_counts`` table ever becomes out-of-sync with the actual table counts y sqlite-utils reset-counts mydb.db +.. note:: + In Python: :ref:`table.enable_counts() ` CLI reference: :ref:`sqlite-utils enable-counts ` + .. _cli_analyze: Optimizing index usage with ANALYZE @@ -2789,6 +2864,9 @@ You can run it against specific tables, or against specific named indexes, by pa You can also run ``ANALYZE`` as part of another command using the ``--analyze`` option. This is supported by the ``create-index``, ``insert`` and ``upsert`` commands. +.. note:: + In Python: :ref:`db.analyze() ` CLI reference: :ref:`sqlite-utils analyze ` + .. _cli_vacuum: Vacuum @@ -2800,6 +2878,9 @@ You can run VACUUM to optimize your database like so: sqlite-utils vacuum mydb.db +.. note:: + In Python: :ref:`db.vacuum() ` CLI reference: :ref:`sqlite-utils vacuum ` + .. _cli_optimize: Optimize @@ -2823,6 +2904,9 @@ To optimize specific tables rather than every FTS table, pass those tables as ex sqlite-utils optimize mydb.db table_1 table_2 +.. note:: + In Python: :ref:`table.optimize() ` CLI reference: :ref:`sqlite-utils optimize ` + .. _cli_wal: WAL mode @@ -2842,6 +2926,9 @@ You can disable WAL mode using ``disable-wal``: Both of these commands accept one or more database files as arguments. +.. note:: + In Python: :ref:`db.enable_wal() and db.disable_wal() ` CLI reference: :ref:`sqlite-utils enable-wal ` + .. _cli_dump: Dumping the database to SQL @@ -2857,6 +2944,9 @@ The ``dump`` command outputs a SQL dump of the schema and full contents of the s ... COMMIT; +.. note:: + In Python: :ref:`db.iterdump() ` CLI reference: :ref:`sqlite-utils dump ` + .. _cli_load_extension: Loading SQLite extensions @@ -2904,6 +2994,9 @@ Eight (case-insensitive) types are allowed: * GEOMETRYCOLLECTION * GEOMETRY +.. note:: + In Python: :ref:`table.add_geometry_column() ` CLI reference: :ref:`sqlite-utils add-geometry-column ` + .. _cli_spatialite_indexes: Adding spatial indexes @@ -2917,6 +3010,9 @@ Once you have a geometry column, you can speed up bounding box queries by adding See this `SpatiaLite Cookbook recipe `__ for examples of how to use a spatial index. +.. note:: + In Python: :ref:`table.create_spatial_index() ` CLI reference: :ref:`sqlite-utils create-spatial-index ` + .. _cli_install: Installing packages diff --git a/docs/python-api.rst b/docs/python-api.rst index 8b07677..1ed238e 100644 --- a/docs/python-api.rst +++ b/docs/python-api.rst @@ -184,6 +184,9 @@ You can attach an additional database using the ``.attach()`` method, providing You can reference tables in the attached database using the alias value you passed to ``db.attach(alias, filepath)`` as a prefix, for example the ``second.table_in_second`` reference in the SQL query above. +.. note:: + In the CLI: :ref:`sqlite-utils --attach ` + .. _python_api_tracing: Tracing queries @@ -254,6 +257,9 @@ If a query returns more than one column with the same name - a join between two A suffix that would collide with another column in the query is skipped - ``select 1 as id, 2 as id, 3 as id_2`` returns ``{'id': 1, 'id_3': 2, 'id_2': 3}``. The same renaming is applied by ``table.rows_where()`` and ``table.search()``. +.. note:: + In the CLI: :ref:`sqlite-utils query ` + .. _python_api_execute: db.execute(sql, params) @@ -497,6 +503,9 @@ You can also iterate through the table objects themselves using the ``.tables`` >>> db.tables [] +.. note:: + In the CLI: :ref:`sqlite-utils tables ` + .. _python_api_views: Listing views @@ -522,6 +531,9 @@ View objects are similar to Table objects, except that any attempts to insert or * ``rows_where(where, where_args, order_by, select)`` * ``drop()`` +.. note:: + In the CLI: :ref:`sqlite-utils views ` + .. _python_api_rows: Listing rows @@ -577,6 +589,9 @@ This method also accepts ``offset=`` and ``limit=`` arguments, for specifying an ... print(row) {'id': 1, 'age': 4, 'name': 'Cleo'} +.. note:: + In the CLI: :ref:`sqlite-utils rows ` + .. _python_api_rows_count_where: Counting rows @@ -661,6 +676,9 @@ The ``db.schema`` property returns the full SQL schema for the database as a str "name" TEXT ); +.. note:: + In the CLI: :ref:`sqlite-utils schema ` + .. _python_api_creating_tables: Creating tables @@ -809,6 +827,9 @@ You can pass ``strict=True`` to create a table in ``STRICT`` mode: "name": str, }, strict=True) +.. note:: + In the CLI: :ref:`sqlite-utils create-table ` + .. _python_api_compound_primary_keys: Compound primary keys @@ -989,6 +1010,9 @@ Here's an example that uses these features: # ) +.. note:: + In the CLI: :ref:`sqlite-utils insert --not-null and --default ` + .. _python_api_rename_table: Renaming a table @@ -1006,6 +1030,9 @@ This executes the following SQL: ALTER TABLE [my_table] RENAME TO [new_name_for_my_table] +.. note:: + In the CLI: :ref:`sqlite-utils rename-table ` + .. _python_api_duplicate: Duplicating tables @@ -1021,6 +1048,9 @@ The new ``authors_copy`` table will now contain a duplicate copy of the data fro This method raises ``sqlite_utils.db.NoTable`` if the table does not exist. +.. note:: + In the CLI: :ref:`sqlite-utils duplicate ` + .. _python_api_bulk_inserts: Bulk inserts @@ -1063,6 +1093,9 @@ You can delete all the existing rows in the table before inserting the new recor Pass ``analyze=True`` to run ``ANALYZE`` against the table after inserting the new records. +.. note:: + In the CLI: :ref:`sqlite-utils insert ` + .. _python_api_insert_lists: Inserting data from a list or tuple iterator @@ -1140,6 +1173,9 @@ To replace any existing records that have a matching primary key, use the ``repl .. note:: Prior to sqlite-utils 2.0 the ``.upsert()`` and ``.upsert_all()`` methods worked the same way as ``.insert(replace=True)`` does today. See :ref:`python_api_upsert` for the new behaviour of those methods introduced in 2.0. +.. note:: + In the CLI: :ref:`sqlite-utils insert --replace ` + .. _python_api_update: Updating a specific record @@ -1225,6 +1261,9 @@ Every record passed to ``upsert()`` or ``upsert_all()`` must include a value for .. note:: ``.upsert()`` and ``.upsert_all()`` in sqlite-utils 1.x worked like ``.insert(..., replace=True)`` and ``.insert_all(..., replace=True)`` do in 2.x. See `issue #66 `__ for details of this change. +.. note:: + In the CLI: :ref:`sqlite-utils upsert ` + .. _python_api_old_upsert: Alternative upserts using INSERT OR IGNORE @@ -1576,6 +1615,9 @@ You can set a ``NOT NULL DEFAULT 'x'`` constraint on the new column using ``not_ db.table("dogs").add_column("friends_count", int, not_null_default=0) +.. note:: + In the CLI: :ref:`sqlite-utils add-column ` + .. _python_api_add_column_alter: Adding columns automatically on insert/update @@ -1599,6 +1641,9 @@ You can insert or update data that includes new columns and have the table autom new_table = db.table("new_table", alter=True) new_table.insert({"name": "Gareth", "age": 32, "shoe_size": 11}) +.. note:: + In the CLI: :ref:`sqlite-utils insert --alter ` + .. _python_api_add_foreign_key: Adding foreign key constraints @@ -1657,6 +1702,9 @@ Use ``on_delete=`` and ``on_update=`` to specify ``ON DELETE`` and ``ON UPDATE`` This creates a foreign key with an ``ON DELETE CASCADE`` clause, so deleting an author will also delete their books (provided foreign key enforcement is enabled with ``PRAGMA foreign_keys = ON``). Valid actions are ``"SET NULL"``, ``"SET DEFAULT"``, ``"CASCADE"``, ``"RESTRICT"`` and the default ``"NO ACTION"``. +.. note:: + In the CLI: :ref:`sqlite-utils add-foreign-key ` + .. _python_api_add_foreign_keys: Adding multiple foreign key constraints at once @@ -1677,6 +1725,9 @@ This method runs the same checks as ``.add_foreign_keys()`` and will raise ``sql Foreign keys that already exist are silently skipped, so repeated calls are idempotent - but only if they match exactly. Requesting a foreign key that exists with different ``ON DELETE``/``ON UPDATE`` actions raises ``AlterError``: use ``table.transform()`` to change the actions of an existing foreign key. +.. note:: + In the CLI: :ref:`sqlite-utils add-foreign-keys ` + .. _python_api_index_foreign_keys: Adding indexes for all foreign keys @@ -1690,6 +1741,9 @@ If you want to ensure that every foreign key column in your database has a corre Compound foreign keys get a single composite index across their columns. +.. note:: + In the CLI: :ref:`sqlite-utils index-foreign-keys ` + .. _python_api_drop: Dropping a table or view @@ -1711,6 +1765,9 @@ Pass ``ignore=True`` if you want to ignore the error caused by the table or view db.table("my_table").drop(ignore=True) +.. note:: + In the CLI: :ref:`sqlite-utils drop-table ` and :ref:`sqlite-utils drop-view ` + .. _python_api_transform: Transforming a table @@ -1739,6 +1796,9 @@ To keep the original table around instead of dropping it, pass the ``keep_table= This method raises a ``sqlite_utils.db.TransformError`` exception if the table cannot be transformed, usually because there are existing constraints or indexes that are incompatible with modifications to the columns. +.. note:: + In the CLI: :ref:`sqlite-utils transform ` + .. _python_api_transform_alter_column_types: Altering column types @@ -2094,6 +2154,9 @@ This produces a lookup table like so: Rows where every extracted column is ``null`` are not extracted: no record is created for them in the lookup table and their foreign key column is left as ``null``. When extracting multiple columns, rows where at least one of the extracted columns has a value will be extracted as usual. +.. note:: + In the CLI: :ref:`sqlite-utils extract ` + .. _python_api_hash: Setting an ID based on the hash of the row contents @@ -2155,6 +2218,9 @@ You can pass ``ignore=True`` to silently ignore an existing view and do nothing, select * from dogs where is_good_dog = 1 """, replace=True) +.. note:: + In the CLI: :ref:`sqlite-utils create-view ` + Storing JSON ============ @@ -2273,6 +2339,9 @@ If you are using ``pysqlite3`` the underlying method may be missing. If you inst pip install sqlite-dump +.. note:: + In the CLI: :ref:`sqlite-utils dump ` + .. _python_api_introspection: Introspecting tables and views @@ -2472,6 +2541,9 @@ The ``.indexes`` property returns all indexes created for a table, as a list of Index(seq=4, name='"Street_Tree_List_qCaretaker"', unique=0, origin='c', partial=0, columns=['qCaretaker']), Index(seq=5, name='"Street_Tree_List_PlantType"', unique=0, origin='c', partial=0, columns=['PlantType'])] +.. note:: + In the CLI: :ref:`sqlite-utils indexes ` + .. _python_api_introspection_xindexes: .xindexes @@ -2515,6 +2587,9 @@ The ``.triggers`` property lists database triggers. It can be used on both datab >>> db.triggers ... similar output to db.table("authors").triggers +.. note:: + In the CLI: :ref:`sqlite-utils triggers ` + .. _python_api_introspection_triggers_dict: .triggers_dict @@ -2663,6 +2738,9 @@ To remove the FTS tables and triggers you created, use the ``disable_fts()`` tab db.table("dogs").disable_fts() +.. note:: + In the CLI: :ref:`sqlite-utils enable-fts ` + .. _python_api_quote_fts: Quoting characters for use in search @@ -2727,6 +2805,9 @@ To return just the title and published columns for three matches for ``"dog"`` w ): print(article) +.. note:: + In the CLI: :ref:`sqlite-utils search ` + .. _python_api_fts_search_sql: Building SQL queries with table.search_sql() @@ -2811,6 +2892,9 @@ This runs the following SQL:: INSERT INTO dogs_fts (dogs_fts) VALUES ("rebuild"); +.. note:: + In the CLI: :ref:`sqlite-utils rebuild-fts ` + .. _python_api_fts_optimize: Optimizing a full-text search table @@ -2826,6 +2910,9 @@ This runs the following SQL:: INSERT INTO dogs_fts (dogs_fts) VALUES ("optimize"); +.. note:: + In the CLI: :ref:`sqlite-utils optimize ` + .. _python_api_cached_table_counts: Cached table counts using triggers @@ -2888,6 +2975,9 @@ If the ``_counts`` table ever becomes out-of-sync with the actual table counts y db.reset_counts() +.. note:: + In the CLI: :ref:`sqlite-utils enable-counts ` + .. _python_api_create_index: Creating indexes @@ -2939,6 +3029,9 @@ You can drop an index from a table using ``.drop_index(index_name)``: Use ``ignore=True`` to ignore the error if the index does not exist. +.. note:: + In the CLI: :ref:`sqlite-utils create-index ` and :ref:`sqlite-utils drop-index ` + .. _python_api_analyze: Optimizing index usage with ANALYZE @@ -2966,6 +3059,9 @@ To run against all indexes attached to a specific table, you can either pass the db.table("dogs").analyze() +.. note:: + In the CLI: :ref:`sqlite-utils analyze ` + .. _python_api_vacuum: Vacuum @@ -2977,6 +3073,9 @@ You can optimize your database by running VACUUM against it like so: Database("my_database.db").vacuum() +.. note:: + In the CLI: :ref:`sqlite-utils vacuum ` + .. _python_api_wal: WAL mode @@ -3004,6 +3103,9 @@ You can check the current journal mode for a database using the ``journal_mode`` This will usually be ``wal`` or ``delete`` (meaning WAL is disabled), but can have other values - see the `PRAGMA journal_mode `__ documentation. +.. note:: + In the CLI: :ref:`sqlite-utils enable-wal and disable-wal ` + .. _python_api_suggest_column_types: Suggesting column types @@ -3144,6 +3246,9 @@ You can cause ``sqlite3`` to return more useful errors, including the traceback sqlite3.enable_callback_tracebacks(True) +.. note:: + In the CLI: :ref:`sqlite-utils query --functions ` + .. _python_api_quote: Quoting strings for use in SQL @@ -3276,6 +3381,9 @@ Initialize SpatiaLite .. automethod:: sqlite_utils.db.Database.init_spatialite :noindex: +.. note:: + In the CLI: :ref:`sqlite-utils create-database --init-spatialite ` + .. _python_api_gis_find_spatialite: Finding SpatiaLite @@ -3291,6 +3399,9 @@ Adding geometry columns .. automethod:: sqlite_utils.db.Table.add_geometry_column :noindex: +.. note:: + In the CLI: :ref:`sqlite-utils add-geometry-column ` + .. _python_api_gis_create_spatial_index: Creating a spatial index @@ -3298,3 +3409,6 @@ Creating a spatial index .. automethod:: sqlite_utils.db.Table.create_spatial_index :noindex: + +.. note:: + In the CLI: :ref:`sqlite-utils create-spatial-index ` From d71420065903ff54247b5062b8c6af6165b7e638 Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Sun, 12 Jul 2026 05:00:32 -0700 Subject: [PATCH 83/86] Add test: transform does not cascade-delete referencing records (#792) > Add a test that covers what happens if you run transform against a table with ON CASCADE DELETE for one of its foreign keys - those records should not be deleted during the transform even though the table is dropped as part of that procedure --- tests/test_transform.py | 37 +++++++++++++++++++++++++++++++++++++ 1 file changed, 37 insertions(+) diff --git a/tests/test_transform.py b/tests/test_transform.py index f0f5019..3cc5ba6 100644 --- a/tests/test_transform.py +++ b/tests/test_transform.py @@ -432,6 +432,43 @@ def test_transform_verify_foreign_keys(fresh_db): assert fresh_db.conn.execute("PRAGMA foreign_keys").fetchone()[0] +@pytest.mark.parametrize("use_pragma_foreign_keys", [False, True]) +def test_transform_on_delete_cascade_does_not_delete_records( + fresh_db, use_pragma_foreign_keys +): + # Transforming a table drops and recreates it - if another table references + # it with ON DELETE CASCADE and PRAGMA foreign_keys is on, that drop must + # not cascade and delete the referencing records + if use_pragma_foreign_keys: + fresh_db.conn.execute("PRAGMA foreign_keys=ON") + fresh_db.executescript(""" + CREATE TABLE authors (id INTEGER PRIMARY KEY, name TEXT); + CREATE TABLE books ( + id INTEGER PRIMARY KEY, + title TEXT, + author_id INTEGER REFERENCES authors(id) ON DELETE CASCADE + ); + """) + fresh_db["authors"].insert({"id": 1, "name": "Ursula K. Le Guin"}) + fresh_db["books"].insert({"id": 1, "title": "The Dispossessed", "author_id": 1}) + # Transform the table on the other end of the cascading foreign key + fresh_db["authors"].transform(rename={"name": "author_name"}) + assert list(fresh_db["authors"].rows) == [ + {"id": 1, "author_name": "Ursula K. Le Guin"} + ] + assert list(fresh_db["books"].rows) == [ + {"id": 1, "title": "The Dispossessed", "author_id": 1} + ] + # Transforming the table with the cascading foreign key should not + # delete its records either + fresh_db["books"].transform(rename={"title": "book_title"}) + assert list(fresh_db["books"].rows) == [ + {"id": 1, "book_title": "The Dispossessed", "author_id": 1} + ] + if use_pragma_foreign_keys: + assert fresh_db.conn.execute("PRAGMA foreign_keys").fetchone()[0] + + def test_transform_add_foreign_keys_from_scratch(fresh_db): _add_country_city_continent(fresh_db) fresh_db["places"].insert(_CAVEAU) From f66ddcb215e76dcbc1fa1dff4359f2ac3dc702e5 Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Sun, 12 Jul 2026 08:43:51 -0700 Subject: [PATCH 84/86] Transform now refuses to run inside a transaction if destructive foreign keys exist (#795) * Transform now refuses to run inside a transaction if destructive foreign keys exist Closes #794 Co-Authored-By: Claude Fable 5 Claude-Session: https://claude.ai/code/session_014StVTWQJpFhfZJK2CYVBwv --- docs/python-api.rst | 33 ++++++++++- sqlite_utils/db.py | 35 ++++++++++++ tests/test_transform.py | 124 +++++++++++++++++++++++++++++++++++++++- 3 files changed, 190 insertions(+), 2 deletions(-) diff --git a/docs/python-api.rst b/docs/python-api.rst index 1ed238e..43b734d 100644 --- a/docs/python-api.rst +++ b/docs/python-api.rst @@ -434,9 +434,10 @@ The library will never commit a transaction you opened. If you call write method Prefer ``db.atomic()`` or ``db.begin()``, ``db.commit()`` and ``db.rollback()`` over mixing sqlite-utils transaction methods with calls to ``db.conn.commit()``, ``db.conn.rollback()`` or raw transaction-control SQL. Mixing the two layers makes it much harder to tell which layer owns the current transaction. -Two related safeguards to be aware of: +Some related safeguards to be aware of: - ``db.enable_wal()`` and ``db.disable_wal()`` raise a ``sqlite_utils.db.TransactionError`` if called while a transaction is open, because changing the journal mode would commit it as a side effect. +- ``table.transform()`` raises a ``sqlite_utils.db.TransactionError`` if called while a transaction is open with ``PRAGMA foreign_keys`` enabled and the table is referenced by foreign keys with destructive ``ON DELETE`` actions, because the pragma cannot be turned off mid-transaction to protect those referencing rows - see :ref:`python_api_transform_foreign_keys_transactions`. - Closing the database - explicitly with ``db.close()``, or by exiting a ``with Database(...) as db:`` block - rolls back any transaction that is still open, see :ref:`python_api_close`. .. _python_api_transactions_modes: @@ -1996,6 +1997,36 @@ If you want to do something more advanced, you can call the ``table.transform_sq This method will return a list of SQL statements that should be executed to implement the change. You can then make modifications to that SQL - or add additional SQL statements - before executing it yourself. +.. _python_api_transform_foreign_keys_transactions: + +Foreign keys and transactions +----------------------------- + +Because ``.transform()`` drops the old table, running it with ``PRAGMA foreign_keys`` enabled could fire ``ON DELETE`` actions on any tables that reference it - an inbound ``ON DELETE CASCADE`` foreign key would silently delete those referencing rows. To prevent this, ``.transform()`` turns ``PRAGMA foreign_keys`` off for the duration of the operation and restores it afterwards, running ``PRAGMA foreign_key_check`` before committing. + +``PRAGMA foreign_keys`` cannot be changed inside a transaction, so this protection is impossible if you call ``.transform()`` while a transaction is already open - for example inside a ``with db.atomic():`` block or after ``db.begin()``. If ``PRAGMA foreign_keys`` is on and another table references the table being transformed with a destructive ``ON DELETE`` action - ``CASCADE``, ``SET NULL`` or ``SET DEFAULT`` - the method will refuse to run and raise a ``sqlite_utils.db.TransactionError``: + +.. code-block:: python + + from sqlite_utils.db import TransactionError + + try: + with db.atomic(): + db["authors"].transform(types={"id": str}) + except TransactionError as ex: + print("Could not transform in transaction:", ex) + +To transform such a table either call ``.transform()`` outside of the transaction, or execute ``PRAGMA foreign_keys = off`` before opening it: + +.. code-block:: python + + db.execute("PRAGMA foreign_keys = off") + with db.atomic(): + db["authors"].transform(types={"id": str}) + db.execute("PRAGMA foreign_keys = on") + +Tables referenced by foreign keys without a destructive action (the default ``NO ACTION``, or ``RESTRICT``) can still be transformed inside a transaction - sqlite-utils uses ``PRAGMA defer_foreign_keys`` to postpone the foreign key checks until the transaction commits. + .. _python_api_extract: Extracting columns into a separate table diff --git a/sqlite_utils/db.py b/sqlite_utils/db.py index d709fb9..e97b7d9 100644 --- a/sqlite_utils/db.py +++ b/sqlite_utils/db.py @@ -2522,6 +2522,11 @@ class Table(Queryable): See :ref:`python_api_transform` for full details. + Raises :py:class:`sqlite_utils.db.TransactionError` if called while a + transaction is open with ``PRAGMA foreign_keys`` enabled and the table + is referenced by foreign keys with destructive ``ON DELETE`` actions - + see :ref:`python_api_transform_foreign_keys_transactions`. + :param types: Columns that should have their type changed, for example ``{"weight": float}`` :param rename: Columns to rename, for example ``{"headline": "title"}`` :param drop: Columns to drop @@ -2566,6 +2571,36 @@ class Table(Queryable): should_defer_foreign_keys = ( pragma_foreign_keys_was_on and already_in_transaction ) + if should_defer_foreign_keys: + # PRAGMA foreign_keys is a no-op inside a transaction, and + # defer_foreign_keys only defers violation checks, not ON DELETE + # actions - so dropping the old table would still fire destructive + # actions on any tables that reference it. Refuse rather than + # silently modify or delete those rows. + destructive_fks = [ + (table.name, fk) + for table in self.db.tables + for fk in table.foreign_keys + if fk.other_table == self.name + and fk.on_delete in ("CASCADE", "SET NULL", "SET DEFAULT") + ] + if destructive_fks: + raise TransactionError( + "Cannot transform table {table} while a transaction is open: " + "PRAGMA foreign_keys cannot be changed inside a transaction, " + "and the table is referenced by foreign keys with ON DELETE " + "actions that would fire when the old table is dropped: " + "{fks}. Call transform() outside of the transaction, or " + 'execute "PRAGMA foreign_keys = off" before opening it.'.format( + table=self.name, + fks=", ".join( + "{}.{} (ON DELETE {})".format( + table_name, ", ".join(fk.columns), fk.on_delete + ) + for table_name, fk in destructive_fks + ), + ) + ) defer_foreign_keys_was_on = False try: if should_disable_foreign_keys: diff --git a/tests/test_transform.py b/tests/test_transform.py index 3cc5ba6..362f1ca 100644 --- a/tests/test_transform.py +++ b/tests/test_transform.py @@ -1,6 +1,6 @@ import sqlite3 -from sqlite_utils.db import ForeignKey, TransformError +from sqlite_utils.db import ForeignKey, TransactionError, TransformError from sqlite_utils.utils import OperationalError import pytest @@ -469,6 +469,128 @@ def test_transform_on_delete_cascade_does_not_delete_records( assert fresh_db.conn.execute("PRAGMA foreign_keys").fetchone()[0] +@pytest.mark.parametrize("on_delete", ["CASCADE", "SET NULL", "SET DEFAULT", "cascade"]) +def test_transform_in_transaction_refuses_destructive_on_delete(fresh_db, on_delete): + # PRAGMA foreign_keys is a no-op inside a transaction, so transforming a + # table referenced by ON DELETE CASCADE / SET NULL / SET DEFAULT foreign + # keys inside an open transaction would fire those actions when the old + # table is dropped - transform() should refuse instead + fresh_db.conn.execute("PRAGMA foreign_keys=ON") + fresh_db.executescript(""" + CREATE TABLE authors (id INTEGER PRIMARY KEY, name TEXT); + CREATE TABLE books ( + id INTEGER PRIMARY KEY, + title TEXT, + author_id INTEGER REFERENCES authors(id) ON DELETE {} + ); + """.format(on_delete)) + fresh_db["authors"].insert({"id": 1, "name": "Ursula K. Le Guin"}) + fresh_db["books"].insert({"id": 1, "title": "The Dispossessed", "author_id": 1}) + previous_schema = fresh_db["authors"].schema + with fresh_db.atomic(): + with pytest.raises(TransactionError) as excinfo: + fresh_db["authors"].transform(rename={"name": "author_name"}) + message = str(excinfo.value) + assert "books" in message + assert "ON DELETE {}".format(on_delete.upper()) in message + # Nothing should have changed + assert fresh_db["authors"].schema == previous_schema + assert list(fresh_db["books"].rows) == [ + {"id": 1, "title": "The Dispossessed", "author_id": 1} + ] + assert fresh_db.conn.execute("PRAGMA foreign_keys").fetchone()[0] + + +def test_transform_in_transaction_refuses_self_referential_cascade(fresh_db): + # The copied table carries a foreign key referencing the original table + # name, so a self-referential cascade would wipe the copy too + fresh_db.conn.execute("PRAGMA foreign_keys=ON") + fresh_db.executescript(""" + CREATE TABLE categories ( + id INTEGER PRIMARY KEY, + name TEXT, + parent_id INTEGER REFERENCES categories(id) ON DELETE CASCADE + ); + """) + fresh_db["categories"].insert_all( + [ + {"id": 1, "name": "Fiction", "parent_id": None}, + {"id": 2, "name": "Science Fiction", "parent_id": 1}, + ] + ) + with fresh_db.atomic(): + with pytest.raises(TransactionError) as excinfo: + fresh_db["categories"].transform(rename={"name": "title"}) + assert "categories" in str(excinfo.value) + assert fresh_db["categories"].count == 2 + + +def test_transform_in_transaction_allowed_with_no_action_foreign_key(fresh_db): + # An inbound foreign key without a destructive ON DELETE action is safe + # inside a transaction thanks to PRAGMA defer_foreign_keys + fresh_db.conn.execute("PRAGMA foreign_keys=ON") + fresh_db.executescript(""" + CREATE TABLE authors (id INTEGER PRIMARY KEY, name TEXT); + CREATE TABLE books ( + id INTEGER PRIMARY KEY, + title TEXT, + author_id INTEGER REFERENCES authors(id) + ); + """) + fresh_db["authors"].insert({"id": 1, "name": "Ursula K. Le Guin"}) + fresh_db["books"].insert({"id": 1, "title": "The Dispossessed", "author_id": 1}) + with fresh_db.atomic(): + fresh_db["authors"].transform(rename={"name": "author_name"}) + assert list(fresh_db["authors"].rows) == [ + {"id": 1, "author_name": "Ursula K. Le Guin"} + ] + assert list(fresh_db["books"].rows) == [ + {"id": 1, "title": "The Dispossessed", "author_id": 1} + ] + assert fresh_db.conn.execute("PRAGMA foreign_keys").fetchone()[0] + + +def test_transform_in_transaction_allowed_for_child_table(fresh_db): + # The table being transformed only has an outbound foreign key - dropping + # it fires no ON DELETE actions, so this is allowed inside a transaction + fresh_db.conn.execute("PRAGMA foreign_keys=ON") + fresh_db.executescript(""" + CREATE TABLE authors (id INTEGER PRIMARY KEY, name TEXT); + CREATE TABLE books ( + id INTEGER PRIMARY KEY, + title TEXT, + author_id INTEGER REFERENCES authors(id) ON DELETE CASCADE + ); + """) + fresh_db["authors"].insert({"id": 1, "name": "Ursula K. Le Guin"}) + fresh_db["books"].insert({"id": 1, "title": "The Dispossessed", "author_id": 1}) + with fresh_db.atomic(): + fresh_db["books"].transform(rename={"title": "book_title"}) + assert list(fresh_db["books"].rows) == [ + {"id": 1, "book_title": "The Dispossessed", "author_id": 1} + ] + + +def test_transform_in_transaction_allowed_with_foreign_keys_off(fresh_db): + # With PRAGMA foreign_keys off (the default) no cascades can fire, so + # transform inside a transaction is safe even with a CASCADE schema + fresh_db.executescript(""" + CREATE TABLE authors (id INTEGER PRIMARY KEY, name TEXT); + CREATE TABLE books ( + id INTEGER PRIMARY KEY, + title TEXT, + author_id INTEGER REFERENCES authors(id) ON DELETE CASCADE + ); + """) + fresh_db["authors"].insert({"id": 1, "name": "Ursula K. Le Guin"}) + fresh_db["books"].insert({"id": 1, "title": "The Dispossessed", "author_id": 1}) + with fresh_db.atomic(): + fresh_db["authors"].transform(rename={"name": "author_name"}) + assert list(fresh_db["books"].rows) == [ + {"id": 1, "title": "The Dispossessed", "author_id": 1} + ] + + def test_transform_add_foreign_keys_from_scratch(fresh_db): _add_country_city_continent(fresh_db) fresh_db["places"].insert(_CAVEAU) From 458b3ab5b169eff1f8319c44a7c320c68f54d28b Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Sun, 12 Jul 2026 13:52:14 -0700 Subject: [PATCH 85/86] Release 4.1.1 Refs #791, #792, #794, #795 --- docs/changelog.rst | 7 +++++++ pyproject.toml | 2 +- 2 files changed, 8 insertions(+), 1 deletion(-) diff --git a/docs/changelog.rst b/docs/changelog.rst index 5b9355f..a8006c3 100644 --- a/docs/changelog.rst +++ b/docs/changelog.rst @@ -4,6 +4,13 @@ Changelog =========== +.. _v4_1_1: + +4.1.1 (2026-07-12) +------------------ + +- ``table.transform()`` now raises a ``TransactionError`` if called while a transaction is open with ``PRAGMA foreign_keys`` enabled and the table is referenced by foreign keys with destructive ``ON DELETE`` actions - ``CASCADE``, ``SET NULL`` or ``SET DEFAULT``. The pragma cannot be changed inside a transaction, so previously dropping the old table as part of the transform could fire those actions and silently delete or modify referencing rows. See :ref:`python_api_transform_foreign_keys_transactions` for details and workarounds. (:issue:`794`) +- The CLI and Python API documentation now cross-reference each other: CLI sections link to the equivalent Python API functionality and Python API sections link back to the corresponding CLI command. (:issue:`791`) .. _v4_1: 4.1 (2026-07-11) diff --git a/pyproject.toml b/pyproject.toml index 971f5a0..003322c 100644 --- a/pyproject.toml +++ b/pyproject.toml @@ -1,6 +1,6 @@ [project] name = "sqlite-utils" -version = "4.1" +version = "4.1.1" description = "CLI tool and Python library for manipulating SQLite databases" readme = { file = "README.md", content-type = "text/markdown" } authors = [ From a947dc673923ff6e95b41d3dfabe1cbd95e6de86 Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Sun, 12 Jul 2026 17:14:01 -0700 Subject: [PATCH 86/86] Changelog now links to CLI and Python API in most recent entry --- docs/changelog.rst | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/docs/changelog.rst b/docs/changelog.rst index a8006c3..4c868f4 100644 --- a/docs/changelog.rst +++ b/docs/changelog.rst @@ -10,7 +10,7 @@ ------------------ - ``table.transform()`` now raises a ``TransactionError`` if called while a transaction is open with ``PRAGMA foreign_keys`` enabled and the table is referenced by foreign keys with destructive ``ON DELETE`` actions - ``CASCADE``, ``SET NULL`` or ``SET DEFAULT``. The pragma cannot be changed inside a transaction, so previously dropping the old table as part of the transform could fire those actions and silently delete or modify referencing rows. See :ref:`python_api_transform_foreign_keys_transactions` for details and workarounds. (:issue:`794`) -- The CLI and Python API documentation now cross-reference each other: CLI sections link to the equivalent Python API functionality and Python API sections link back to the corresponding CLI command. (:issue:`791`) +- The :ref:`CLI ` and :ref:`Python API ` documentation now cross-reference each other: CLI sections link to the equivalent Python API functionality and Python API sections link back to the corresponding CLI command. (:issue:`791`) .. _v4_1: 4.1 (2026-07-11)