.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 <noreply@anthropic.com>

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`)

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

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 <python_api_creating_tables>` 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

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.

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(

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},
     ]

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},
+    ]

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"},
+    ]