The inserted rows are saved to the database file straight away - methods like ``insert_all()`` commit their own changes, so no ``commit()`` call is needed. See :ref:`python_api_transactions` for how this works.
This will create ``my_database.db`` if it does not already exist.
If you want to recreate a database from scratch (first removing the existing file from disk if it already exists) you can use the ``recreate=True`` argument:
..code-block:: python
db = Database("my_database.db", recreate=True)
Instead of a file path you can pass in an existing SQLite connection:
The connection must use Python's default transaction handling. Connections created with the Python 3.12+ ``sqlite3.connect(..., autocommit=True)`` or ``autocommit=False`` options are rejected with a ``sqlite_utils.db.TransactionError`` - see :ref:`python_api_transactions_modes`.
You can also create a named in-memory database. Unlike regular memory databases these can be accessed by multiple threads, provided at least one reference to the database still exists. `del db` will clear the database from memory.
Connections use ``PRAGMA recursive_triggers=on`` by default. If you don't want to use `recursive triggers <https://www.sqlite.org/pragma.html#pragma_recursive_triggers>`__ you can turn them off using:
..code-block:: python
db = Database(memory=True, recursive_triggers=False)
By default, any :ref:`sqlite-utils plugins <plugins>` that implement the :ref:`plugins_hooks_prepare_connection` hook will be executed against the connection when you create the ``Database`` object. You can opt out of executing plugins using ``execute_plugins=False`` like this:
You can pass ``strict=True`` to enable `SQLite STRICT mode <https://www.sqlite.org/stricttables.html>`__ for all tables created using this database object:
Exiting the block is equivalent to calling ``db.close()``: the connection is closed and any transaction still open at that point is rolled back. This matches SQLite's own behavior when a connection closes.
This rarely matters in practice. Everything that writes to the database - including raw ``db.execute()`` statements - commits automatically, so a transaction can only be open here if you explicitly started one with ``db.begin()`` and have not yet committed it. In that case the decision to commit stays with you: committing automatically on exit could silently persist half-finished work, for example if your code returned early from the block. Call ``db.commit()`` when the work is complete.
Note this differs from the ``sqlite3.Connection`` context manager in the standard library, which commits on success but does not close the connection. See :ref:`python_api_transactions` for the full transaction model.
SQLite supports cross-database SQL queries, which can join data from tables in more than one database file.
You can attach an additional database using the ``.attach()`` method, providing an alias to use for that database and the path to the SQLite file on disk.
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.
You can use the ``tracer`` mechanism to see SQL queries that are being executed by SQLite. A tracer is a function that you provide which will be called with ``sql`` and ``params`` arguments every time SQL is executed, for example:
..code-block:: python
def tracer(sql, params):
print("SQL: {} - params: {}".format(sql, params))
You can pass this function to the ``Database()`` constructor like so:
..code-block:: python
db = Database(memory=True, tracer=tracer)
You can also turn on a tracer function temporarily for a block of code using the ``with db.tracer(...)`` context manager:
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``. The rejected statement is rolled back, so it has no effect on the database. Use :ref:`db.execute() <python_api_execute>` for those statements instead.
The ``db.execute()`` and ``db.executescript()`` methods provide wrappers around ``.execute()`` and ``.executescript()`` on the underlying SQLite connection. These wrappers log to the :ref:`tracer function <python_api_tracing>` if one has been registered.
``db.execute(sql)`` returns a `sqlite3.Cursor <https://docs.python.org/3/library/sqlite3.html#sqlite3.Cursor>`__ that was used to execute the SQL.
cursor = db.execute("update dogs set name = 'Cleopaws'")
print(cursor.rowcount)
# Outputs the number of rows affected by the update
# In this case 2
Other cursor methods such as ``.fetchone()`` and ``.fetchall()`` are also available, see the `standard library documentation <https://docs.python.org/3/library/sqlite3.html#sqlite3.Cursor>`__.
Write statements executed this way are committed automatically, unless a transaction is already open in which case they become part of it - see :ref:`python_api_transactions_execute`.
Both ``db.query()`` and ``db.execute()`` accept an optional second argument for parameters to be passed to the SQL query.
This can take the form of either a tuple/list or a dictionary, depending on the type of parameters used in the query. Values passed in this way will be correctly quoted and escaped, helping avoid SQL injection vulnerabilities.
``?`` parameters in the SQL query can be filled in using a list:
..code-block:: python
db.execute("update dogs set name = ?", ["Cleopaws"])
# This will rename ALL dogs to be called "Cleopaws"
Named parameters using ``:name`` can be filled using a dictionary:
..code-block:: python
dog = next(db.query(
"select rowid, name from dogs where name = :name",
{"name": "Cleopaws"}
))
# dog is now {'rowid': 1, 'name': 'Cleopaws'}
In this example ``next()`` is used to retrieve the first result in the iterator returned by the ``db.query()`` method.
.._python_api_transactions:
Transactions and saving your changes
====================================
Every method in this library that writes to the database - ``insert()``, ``upsert()``, ``update()``, ``delete()``, ``delete_where()``, ``transform()``, ``create_table()``, ``create_index()``, ``enable_fts()`` and the rest - runs inside its own transaction and commits it before returning. Your changes are saved to disk as soon as the method call finishes:
The same applies to raw SQL executed with :ref:`db.execute() <python_api_transactions_execute>` - a write statement is committed as soon as it has run.
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:
2. You are :ref:`managing a transaction yourself <python_api_transactions_manual>` with ``db.begin()``, in which case nothing is committed until you commit - the library will never commit a transaction you opened.
``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:
Write statements executed with :ref:`db.execute() <python_api_execute>` follow the same rule as everything else: they are committed automatically as soon as they have run.
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:
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() <python_api_query>` for those statements, which executes the write and commits it immediately.
``db.begin()`` raises ``sqlite3.OperationalError`` if a transaction is already open. ``db.commit()`` and ``db.rollback()`` do nothing if there is no open transaction.
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.
-``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.
- 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`.
``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``.
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.
Using this factory function allows you to set :ref:`python_api_table_configuration`. Additional keyword arguments to ``db.table()`` will be used if a further method call causes the table to be created.
The ``db.table()`` method will always return a :ref:`reference_db_table` instance, or raise a ``sqlite_utils.db.NoTable`` exception if the table name is actually a SQL view.
You can iterate through view objects using the ``.views`` property::
>>> db.views
[<View good_dogs>]
View objects are similar to Table objects, except that any attempts to insert or update data will throw an error. The full list of methods and properties available on a view object is as follows:
The first argument is a fragment of SQL. The second, optional argument is values to be passed to that fragment - you can use ``?`` placeholders and pass an array, or you can use ``:named`` parameters and pass a dictionary, like this::
Sometimes it can be useful to retrieve the primary key along with each row, in order to pass that key (or primary key tuple) to the ``.get()`` or ``.update()`` methods.
The ``.pks_and_rows_where()`` method takes the same signature as ``.rows_where()`` (with the exception of the ``select=`` parameter) but returns a generator that yields pairs of ``(primary key, row dictionary)``.
The primary key value will usually be a single value but can also be a tuple if the table has a compound primary key.
If the table is a ``rowid`` table (with no explicit primary key column) then that ID will be returned.
This will automatically create a new table called "dogs" with the following schema::
CREATE TABLE dogs (
name TEXT,
twitter TEXT,
age INTEGER,
is_good_dog INTEGER
)
You can also specify a primary key by passing the ``pk=`` parameter to the ``.insert()`` call. This will only be obeyed if the record being inserted causes the table to be created:
After inserting a row like this, the ``dogs.last_rowid`` property will return the SQLite ``rowid`` assigned to the most recently inserted record.
The ``dogs.last_pk`` property will return the last inserted primary key value, if you specified one. This can be very useful when writing code that creates foreign keys or many-to-many relationships.
You don't need to pass all of the columns to the ``column_order`` parameter. If you only pass a subset of the columns the remaining columns will be ordered based on the key order of the dictionary.
Column types are detected based on the example data provided. Sometimes you may find you need to over-ride these detected types - to create an integer column for data that was provided as a string for example, or to ensure that a table where the first example was ``None`` is created as an ``INTEGER`` rather than a ``TEXT`` column. You can do this using the ``columns=`` parameter:
The first argument here is a dictionary specifying the columns you would like to create. Each column is paired with a Python type indicating the type of column. See :ref:`python_api_add_column` for full details on how these types work.
You can pass ``ignore=True`` to ignore that error. You can also use ``if_not_exists=True`` to use the SQL ``CREATE TABLE IF NOT EXISTS`` pattern to achieve the same effect:
To drop and replace any existing table of that name, pass ``replace=True``. This is a **dangerous operation** that will result in loss of existing data in the table.
You can also pass ``transform=True`` to have any existing tables :ref:`transformed <python_api_transform>` to match your new table specification. This is a **dangerous operation** as it will drop columns that are no longer listed in your call to ``.create()``, so be careful when running this.
If you want to create a table with a compound primary key that spans multiple columns, you can do so by passing a tuple of column names to any of the methods that accept a ``pk=`` parameter. For example:
Any operation that can create a table (``.create()``, ``.insert()``, ``.insert_all()``, ``.upsert()`` and ``.upsert_all()``) accepts an optional ``foreign_keys=`` argument which can be used to set up foreign key constraints for the table that is being created.
If you are using your database with `Datasette <https://datasette.io/>`__, Datasette will detect these constraints and use them to generate hyperlinks to associated records.
The ``foreign_keys`` argument takes a list that indicates which foreign keys should be created. The list can take several forms. The simplest is a list of columns:
..code-block:: python
foreign_keys=["author_id"]
The library will guess which tables you wish to reference based on the column names using the rules described in :ref:`python_api_add_foreign_key`.
You can also be more explicit, by passing in a list of tuples:
..code-block:: python
foreign_keys=[
("author_id", "authors", "id")
]
This means that the ``author_id`` column should be a foreign key that references the ``id`` column in the ``authors`` table.
You can leave off the third item in the tuple to have the referenced column automatically set to the primary key of that table. A full example:
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() <python_api_transform>` - prior to sqlite-utils 4.0 they were silently dropped when a table was transformed.
The ``.insert()``, ``.upsert()``, ``.insert_all()`` and ``.upsert_all()`` methods each take a number of keyword arguments, some of which influence what happens should they cause a table to be created and some of which affect the behavior of those methods.
You can set default values for these methods by accessing the table through the ``db.table(...)`` method (instead of using ``db.table("table_name")``), like so:
The configuration options that can be specified in this way are ``pk``, ``foreign_keys``, ``column_order``, ``not_null``, ``defaults``, ``batch_size``, ``hash_id``, ``hash_id_columns``, ``alter``, ``ignore``, ``replace``, ``extracts``, ``conversions``, ``columns``, ``strict``. These are all documented below.
Each of the methods that can cause a table to be created take optional arguments ``not_null=set()`` and ``defaults=dict()``. The methods that take these optional arguments are:
*``db.create_table(...)``
*``table.create(...)``
*``table.insert(...)``
*``table.insert_all(...)``
*``table.upsert(...)``
*``table.upsert_all(...)``
You can use ``not_null=`` to pass a set of column names that should have a ``NOT NULL`` constraint set on them when they are created.
You can use ``defaults=`` to pass a dictionary mapping columns to the default value that should be specified in the ``CREATE TABLE`` statement.
If you have more than one record to insert, the ``insert_all()`` method is a much more efficient way of inserting them. Just like ``insert()`` it will automatically detect the columns that should be created, but it will inspect the first batch of 100 items to help decide what those column types should be.
The column types used in the ``CREATE TABLE`` statement are automatically derived from the types of data in that first batch of rows. Any additional columns in subsequent batches will cause a ``sqlite3.OperationalError`` exception to be raised unless the ``alter=True`` argument is supplied, in which case the new columns will be created.
The function can accept an iterator or generator of rows and will commit them according to the batch size. The default batch size is 100, but you can specify a different size using the ``batch_size`` parameter:
You can skip inserting any records that have a primary key that already exists using ``ignore=True``. This works with both ``.insert({...}, ignore=True)`` and ``.insert_all([...], ignore=True)``.
You can delete all the existing rows in the table before inserting the new records using ``truncate=True``. This is useful if you want to replace the data in the table.
Pass ``analyze=True`` to run ``ANALYZE`` against the table after inserting the new records.
As an alternative to passing an iterator of dictionaries, you can pass an iterator of lists or tuples. The first item yielded by the iterator must be a list or tuple of string column names, and subsequent items should be lists or tuples of values:
If you try to insert data using a primary key that already exists, the ``.insert()`` or ``.insert_all()`` method will raise a ``sqlite3.IntegrityError`` exception.
print("Record already exists with that primary key")
Importing from ``sqlite_utils.utils.sqlite3`` ensures your code continues to work even if you are using the ``pysqlite3`` library instead of the Python standard library ``sqlite3`` module.
Use the ``ignore=True`` parameter to ignore this error:
..code-block:: python
# This fails silently if a record with id=1 already exists
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.
Any existing columns that are not referenced in the dictionary passed to ``.upsert()`` will be unchanged. If you want to replace a record entirely, use ``.insert(doc, replace=True)`` instead.
Note that the ``pk`` and ``column_order`` parameters here are optional if you are certain that the table has already been created. You should pass them if the table may not exist at the time the first upsert is performed.
An ``upsert_all()`` method is also available, which behaves like ``insert_all()`` but performs upserts instead.
Every record passed to ``upsert()`` or ``upsert_all()`` must include a value for each primary key column - a record without one could never match an existing row, so a ``sqlite_utils.db.PrimaryKeyRequired`` exception is raised instead of quietly inserting a new row.
``.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 <https://github.com/simonw/sqlite-utils/issues/66>`__ for details of this change.
Upserts use ``INSERT INTO ... ON CONFLICT SET``. Prior to ``sqlite-utils 4.0`` these used a sequence of ``INSERT OR IGNORE`` followed by an ``UPDATE``. This older method is still used for SQLite 3.23.1 and earlier. You can force the older implementation by passing ``use_old_upsert=True`` to the ``Database()`` constructor.
The ``table.convert(...)`` method can be used to apply a conversion function to the values in a column, either to update that column or to populate new columns. It is the Python library equivalent of the :ref:`sqlite-utils convert <cli_convert>` command.
This feature works by registering a custom SQLite function that applies a Python transformation, then running a SQL query equivalent to ``UPDATE table SET column = convert_value(column);``
To transform a specific column to uppercase, you would use the following:
This will add the new column, if it does not already exist. You can pass ``output_type=int`` or some other type to control the type of the new column - otherwise it will default to text.
If you want to drop the original column after saving the results in a separate output column, pass ``drop=True``.
You can create multiple new columns from a single input column by passing ``multi=True`` and a conversion function that returns a Python dictionary. This example creates new ``upper`` and ``lower`` columns populated from the single ``title`` column:
The ``.convert()`` method accepts optional ``where=`` and ``where_args=`` parameters which can be used to apply the conversion to a subset of rows specified by a where clause. Here's how to apply the conversion only to rows with an ``id`` that is higher than 20:
These behave the same as the corresponding parameters to the :ref:`.rows_where() <python_api_rows>` method, so you can use ``?`` placeholders and a list of values instead of ``:named`` placeholders with a dictionary.
A useful pattern when populating large tables in to break common values out into lookup tables. Consider a table of ``Trees``, where each tree has a species. Ideally these species would be split out into a separate ``Species`` table, with each one assigned an integer primary key that can be referenced from the ``Trees`` table ``species_id`` column.
Calling ``db.table("Species").lookup({"name": "Palm"})`` creates a table called ``Species`` (if one does not already exist) with two columns: ``id`` and ``name``. It sets up a unique constraint on the ``name`` column to guarantee it will not contain duplicate rows. It then inserts a new row with the ``name`` set to ``Palm`` and returns the new integer primary key value.
If the ``Species`` table already exists, it will insert the new row and return the primary key. If a row with that ``name`` already exists, it will return the corresponding primary key value directly.
If you call ``.lookup()`` against an existing table without the unique constraint it will attempt to add the constraint, raising an ``IntegrityError`` if the constraint cannot be created.
If you pass in a dictionary with multiple values, both values will be used to insert or retrieve the corresponding ID and any unique constraint that is created will cover all of those columns, for example:
The ``.lookup()`` method has an optional second argument which can be used to populate other columns in the table but only if the row does not exist yet. These columns will not be included in the unique index.
To create a species record with a note on when it was first seen, you can use this:
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.
``.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:
A more efficient way to work with lookup tables is to define them using the ``extracts=`` parameter, which is accepted by ``.insert()``, ``.upsert()``, ``.insert_all()``, ``.upsert_all()`` and by the ``.table(...)`` factory function.
``extracts=`` specifies columns which should be "extracted" out into a separate lookup table during the data insertion.
It can be either a list of column names, in which case the extracted table names will match the column names exactly, or it can be a dictionary mapping column names to the desired name of the extracted table.
To extract the ``species`` column out to a separate ``Species`` table, you can do this:
..code-block:: python
# Using the table factory
trees = db.table("Trees", extracts={"species": "Species"})
trees.insert({
"latitude": 49.1265976,
"longitude": 2.5496218,
"species": "Common Juniper"
})
# If you want the table to be called 'species', you can do this:
Running this example actually creates three tables: ``dogs``, ``humans`` and a many-to-many ``dogs_humans`` table. It will insert a record into each of those tables.
The ``.m2m()`` method executes against the last record that was affected by ``.insert()`` or ``.update()`` - the record identified by the ``table.last_pk`` property. To execute ``.m2m()`` against a specific record you can first select it by passing its primary key to ``.update()``:
The second argument can be a single dictionary record or a list of dictionaries. These dictionaries will be passed to ``.upsert()`` against the specified table.
Here's alternative code that creates the dog record and adds two people to it:
..code-block:: python
db = Database(memory=True)
dogs = db.table("dogs", pk="id")
humans = db.table("humans", pk="id")
dogs.insert({"id": 1, "name": "Cleo"}).m2m(
humans, [
{"id": 1, "name": "Natalie"},
{"id": 2, "name": "Simon"}
]
)
The method will attempt to find an existing many-to-many table by looking for a table that has foreign key relationships against both of the tables in the relationship.
If it cannot find such a table, it will create a new one using the names of the two tables - ``dogs_humans`` in this example. You can customize the name of this table using the ``m2m_table=`` argument to ``.m2m()``.
It it finds multiple candidate tables with foreign keys to both of the specified tables it will raise a ``sqlite_utils.db.NoObviousTable`` exception. You can avoid this error by specifying the correct table using ``m2m_table=``.
The ``.m2m()`` method also takes an optional ``pk=`` argument to specify the primary key that should be used if the table is created, and an optional ``alter=True`` argument to specify that any missing columns of an existing table should be added if they are needed.
You can work with (or create) lookup tables as part of a call to ``.m2m()`` using the ``lookup=`` parameter. This accepts the same argument as ``table.lookup()`` does - a dictionary of values that should be used to lookup or create a row in the lookup table.
This example creates a dogs table, populates it, creates a characteristics table, populates that and sets up a many-to-many relationship between the two. It chains ``.m2m()`` twice to create two associated characteristics:
..code-block:: python
db = Database(memory=True)
dogs = db.table("dogs", pk="id")
dogs.insert({"id": 1, "name": "Cleo"}).m2m(
"characteristics", lookup={
"name": "Playful"
}
).m2m(
"characteristics", lookup={
"name": "Opinionated"
}
)
You can inspect the database to see the results like this::
The number of rows for which this column is blank (the empty string)
``num_distinct``
The number of distinct values in this column
``most_common``
The ``N`` most common values as a list of ``(value, count)`` tuples`, or ``None`` if the table consists entirely of distinct values
``least_common``
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``)
You can insert or update data that includes new columns and have the table automatically altered to fit the new schema using the ``alter=True`` argument. This can be passed to all four of ``.insert()``, ``.upsert()``, ``.insert_all()`` and ``.upsert_all()``, or it can be passed to ``db.table(table_name, alter=True)`` to enable it by default for all method calls against that table instance.
It's also possible to add foreign keys by directly updating the `sqlite_master` table. The `sqlite-utils-fast-fks <https://github.com/simonw/sqlite-utils-fast-fks>`__ plugin implements this pattern, using code that was included with ``sqlite-utils`` prior to version 3.35.
The ``table.add_foreign_key(column, other_table, other_column)`` method takes the name of the column, the table that is being referenced and the key column within that other table. If you omit the ``other_column`` argument the primary key from that table will be used automatically. If you omit the ``other_table`` argument the table will be guessed based on some simple rules:
This method first checks that the specified foreign key references tables and columns that exist and does not clash with an existing foreign key. It will raise a ``sqlite_utils.db.AlterError`` exception if these checks fail.
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.
You can use ``db.add_foreign_keys(...)`` to add multiple foreign keys in one go. This method takes a list of four-tuples, each one specifying a ``table``, ``column``, ``other_table`` and ``other_column``.
The SQLite ``ALTER TABLE`` statement is limited. It can add and drop columns and rename tables, but it cannot change column types, change ``NOT NULL`` status or change the primary key for a table.
The ``table.transform()`` method can do all of these things, by implementing a multi-step pattern `described in the SQLite documentation <https://www.sqlite.org/lang_altertable.html#otheralter>`__:
1. Start a transaction
2.``CREATE TABLE tablename_new_x123`` with the required changes
3. Copy the old data into the new table using ``INSERT INTO tablename_new_x123 SELECT * FROM tablename;``
4.``DROP TABLE tablename;``
5.``ALTER TABLE tablename_new_x123 RENAME TO tablename;``
6. Commit the transaction
The ``.transform()`` method takes a number of parameters, all of which are optional.
As a bonus, calling ``.transform()`` will reformat the schema for the table that is stored in SQLite to make it more readable. This works even if you call it without any arguments.
To keep the original table around instead of dropping it, pass the ``keep_table=`` option and specify the name of the table you would like it to be renamed to:
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.
To change the primary key for a table, use ``pk=``. This can be passed a single column for a regular primary key, or a tuple of columns to create a compound primary key. Passing ``pk=None`` will remove the primary key and convert the table into a ``rowid`` table.
You can change the ``NOT NULL`` status of columns by using ``not_null=``. You can pass this a set of columns to make those columns ``NOT NULL``:
..code-block:: python
# Make the 'age' and 'weight' columns NOT NULL
table.transform(not_null={"age", "weight"})
If you want to take existing ``NOT NULL`` columns and change them to allow null values, you can do so by passing a dictionary of true/false values instead:
..code-block:: python
# 'age' is NOT NULL but we want to allow NULL:
table.transform(not_null={"age": False})
# Make age allow NULL and switch weight to being NOT NULL:
The ``column_order=`` parameter can be used to change the order of the columns. If you pass the names of a subset of the columns those will go first and columns you omitted will appear in their existing order after them.
This accepts the same arguments described in :ref:`specifying foreign keys <python_api_foreign_keys>` - so you can specify them as a full tuple of ``(column, other_table, other_column)``, or you can take a shortcut and pass just the name of the column, provided the table can be automatically derived from the column name:
The ``foreign_keys=`` parameter is similar to to ``add_foreign_keys=`` but can be be used to replace all foreign key constraints on a table, dropping any that are not explicitly mentioned:
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.
The ``.transform()`` method can handle most cases, but it does not automatically upgrade indexes, views or triggers associated with the table that is being transformed.
If you want to do something more advanced, you can call the ``table.transform_sql(...)`` method with the same arguments that you would have passed to ``table.transform(...)``.
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.
The ``table.extract()`` method can be used to extract specified columns into a separate table.
Imagine a ``Trees`` table that looks like this:
=== ============ =======
id TreeAddress Species
=== ============ =======
1 52 Vine St Palm
2 12 Draft St Oak
3 51 Dark Ave Palm
4 1252 Left St Palm
=== ============ =======
The ``Species`` column contains duplicate values. This database could be improved by extracting that column out into a separate ``Species`` table and pointing to it using a foreign key column.
The ``.extract()`` method defaults to creating a table with the same name as the column that was extracted, and adding a foreign key column called ``tablename_id``.
You can specify a custom table name using ``table=``, and a custom foreign key name using ``fk_column=``. This example creates a table called ``tree_species`` and a foreign key column called ``tree_species_id``:
The table name ``CommonName_LatinName`` is derived from the extract columns. You can use ``table=`` and ``fk_column=`` to specify custom names like this:
You can use the ``rename=`` argument to rename columns in the lookup table. To create a ``Species`` table with columns called ``name`` and ``latin`` you can do this:
Sometimes you will find yourself working with a dataset that includes rows that do not have a provided obvious ID, but where you would like to assign one so that you can later upsert into that table without creating duplicate records.
In these cases, a useful technique is to create an ID that is derived from the sha1 hash of the row contents.
``sqlite-utils`` can do this for you using the ``hash_id=`` option. For example::
This will raise a ``sqlite_utils.utils.OperationalError`` if a view with that name already exists.
You can pass ``ignore=True`` to silently ignore an existing view and do nothing, or ``replace=True`` to replace an existing view with a new definition if your select statement differs from the current view:
SQLite has `excellent JSON support <https://www.sqlite.org/json1.html>`_, and ``sqlite-utils`` can help you take advantage of this: if you attempt to insert a value that can be represented as a JSON list or dictionary, ``sqlite-utils`` will create TEXT column and store your data as serialized JSON. This means you can quickly store even complex data structures in SQLite and query them using JSON features.
Sometimes it can be useful to run values through a SQL function prior to inserting them. A simple example might be converting a value to upper case while it is being inserted.
The ``conversions={...}`` parameter can be used to specify custom SQL to be used as part of a ``INSERT`` or ``UPDATE`` SQL statement.
You can specify an upper case conversion for a specific column like so:
The dictionary key is the column name to be converted. The value is the SQL fragment to use, with a ``?`` placeholder for the original value.
A more useful example: if you are working with `SpatiaLite <https://www.gaia-gis.it/fossil/libspatialite/index>`__ you may find yourself wanting to create geometry values from a WKT value. Code to do that could look like this:
This example uses gographical data from `Who's On First <https://whosonfirst.org/>`__ and depends on the `Shapely <https://shapely.readthedocs.io/en/stable/manual.html>`__ and `HTTPX <https://www.python-httpx.org/>`__ Python libraries.
The ``db.iterdump()`` method returns a sequence of SQL strings representing a complete dump of the database. Use it like this:
..code-block:: python
full_sql = "".join(db.iterdump())
This uses the `sqlite3.Connection.iterdump() <https://docs.python.org/3/library/sqlite3.html#sqlite3.Connection.iterdump>`__ method.
If you are using ``pysqlite3`` or ``sqlean.py`` the underlying method may be missing. If you install the `sqlite-dump <https://pypi.org/project/sqlite-dump/>`__ package then the ``db.iterdump()`` method will use that implementation instead:
This property will take advantage of :ref:`python_api_cached_table_counts` if the ``use_counts_table`` property is set on the database. You can avoid that optimization entirely by calling ``table.count_where()`` instead of accessing the property.
The ``.columns`` property shows the columns in the table or view. It returns a list of ``Column(cid, name, type, notnull, default_value, is_pk)`` named tuples.
If a table has no primary keys but is a `rowid table <https://www.sqlite.org/rowidtable.html>`__, this property will return ``['rowid']``.
.._python_api_introspection_use_rowid:
.use_rowid
----------
Almost all SQLite tables have a ``rowid`` column, but a table with no explicitly defined primary keys must use that ``rowid`` as the primary key for identifying individual rows. The ``.use_rowid`` property checks to see if a table needs to use the ``rowid`` in this way - it returns ``True`` if the table has no explicitly defined primary keys and ``False`` otherwise.
``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.
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.
The ``.indexes`` property returns all indexes created for a table, as a list of ``Index(seq, name, unique, origin, partial, columns)`` named tuples. It is not available on views.
The ``.xindexes`` property returns more detailed information about the indexes on the table, using the SQLite `PRAGMA index_xinfo() <https://sqlite.org/pragma.html#pragma_index_xinfo>`__ mechanism. It returns a list of ``XIndex(name, columns)`` named tuples, where ``columns`` is a list of ``XIndexColumn(seqno, cid, name, desc, coll, key)`` named tuples.
The ``.triggers`` property lists database triggers. It can be used on both database and table objects. It returns a list of ``Trigger(name, table, sql)`` named tuples.
The ``detect_fts()`` method returns the associated SQLite FTS table name, if one exists for this table. If the table has not been configured for full-text search it returns ``None``.
The ``.virtual_table_using`` property reveals if a table is a virtual table. It returns ``None`` for regular tables and the upper case version of the type of virtual table otherwise. For example::
The ``.has_counts_triggers`` property shows if a table has been configured with triggers for updating a ``_counts`` table, as described in :ref:`python_api_cached_table_counts`.
This property on the database object returns ``True`` if the available SQLite version supports `STRICT mode <https://www.sqlite.org/stricttables.html>`__, which was added in SQLite 3.37.0 (on 2021-11-27).
A better solution is to use database triggers. You can set up database triggers to automatically update the full-text index using ``create_triggers=True``:
``.enable_fts()`` defaults to using `FTS5 <https://www.sqlite.org/fts5.html>`__. If you wish to use `FTS4 <https://www.sqlite.org/fts3.html>`__ instead, use the following:
You can customize the tokenizer configured for the table using the ``tokenize=`` parameter. For example, to enable Porter stemming, where English words like "running" will match stemmed alternatives such as "run", use ``tokenize="porter"``:
The SQLite documentation has more on `FTS5 tokenizers <https://www.sqlite.org/fts5.html#tokenizers>`__ and `FTS4 tokenizers <https://www.sqlite.org/fts3.html#tokenizer>`__. ``porter`` is a valid option for both.
This will have no effect if the FTS table already exists, otherwise it will drop and recreate the table with the new settings. This takes into consideration the columns, the tokenizer, the FTS version used and whether or not the table has triggers.
SQLite supports `advanced search query syntax <https://www.sqlite.org/fts3.html#full_text_index_queries>`__. In some situations you may wish to disable this, since characters such as ``.`` may have special meaning that causes errors when searching for strings provided by your users.
The ``db.quote_fts(query)`` method returns the query with SQLite full-text search quoting applied such that the query should be safe to use in a search::
The ``table.search(q)`` method returns a generator over Python dictionaries representing rows that match the search phrase ``q``, ordered by relevance with the most relevant results first.
To return just the title and published columns for three matches for ``"dog"`` where the ``id`` is greater than 10 ordered by ``published`` with the most recent first, use the following:
You can generate the SQL query that would be used for a search using the ``table.search_sql()`` method. It takes the same arguments as ``table.search()``, with the exception of the search query and the ``where_args`` parameter, since those should be provided when the returned SQL is executed.
This uses the ``rank_bm25()`` custom SQL function from `sqlite-fts4 <https://github.com/simonw/sqlite-fts4>`__. You can register that custom function against a ``Database`` connection using this method:
You can rebuild a table using the ``table.rebuild_fts()`` method. This is useful for if the table configuration changes or the indexed data has become corrupted in some way.
The ``select count(*)`` query in SQLite requires a full scan of the primary key index, and can take an increasingly long time as the table grows larger.
The ``table.enable_counts()`` method can be used to configure triggers to continuously update a record in a ``_counts`` table. This value can then be used to quickly retrieve the count of rows in the associated table.
You can enable cached counts for every table in a database (except for virtual tables and the ``_counts`` table itself) using the database ``enable_counts()`` method:
Once enabled, table counts will be stored in the ``_counts`` table. The count records will be automatically kept up-to-date by the triggers when rows are added or deleted to the table.
To access these counts you can query the ``_counts`` table directly or you can use the ``db.cached_counts()`` method. This method returns a dictionary mapping tables to their counts::
>>> db.cached_counts()
{'global-power-plants': 33643,
'global-power-plants_fts_data': 136,
'global-power-plants_fts_idx': 199,
'global-power-plants_fts_docsize': 33643,
'global-power-plants_fts_config': 1}
You can pass a list of table names to this method to retrieve just those counts::
>>> db.cached_counts(["global-power-plants"])
{'global-power-plants': 33643}
The ``table.count`` property executes a ``select count(*)`` query by default, unless the ``db.use_counts_table`` property is set to ``True``.
You can set ``use_counts_table`` to ``True`` when you instantiate the database object:
..code-block:: python
db = Database("global-power-plants.db", use_counts_table=True)
If the property is ``True`` any calls to the ``table.count`` property will first attempt to find the cached count in the ``_counts`` table, and fall back on a ``count(*)`` query if the value is not available or the table is missing.
Calling the ``.enable_counts()`` method on a database or table object will set ``use_counts_table`` to ``True`` for the lifetime of that database object.
By default the index will be named ``idx_{table-name}_{columns}``. If you pass ``find_unique_name=True`` and the automatically derived name already exists, an available name will be found by incrementing a suffix number, for example ``idx_items_title_2``.
You can customize the name of the created index by passing the ``index_name`` parameter:
The `SQLite ANALYZE command <https://www.sqlite.org/lang_analyze.html>`__ builds a table of statistics which the query planner can use to make better decisions about which indexes to use for a given query.
You should run ``ANALYZE`` if your database is large and you do not think your indexes are being efficiently used.
To run ``ANALYZE`` against every index in a database, use this:
..code-block:: python
db.analyze()
To run it just against a specific named index, pass the name of the index to that method:
..code-block:: python
db.analyze("idx_countries_country_name")
To run against all indexes attached to a specific table, you can either pass the table name to ``db.analyze(...)`` or you can call the method directly on the table, like this:
The journal mode can only be changed outside of a transaction. Calling either method while a transaction is open - inside a ``db.atomic()`` block, for example - raises a ``sqlite_utils.db.TransactionError``, unless the database is already in the requested mode in which case the call is a no-op.
This will usually be ``wal`` or ``delete`` (meaning WAL is disabled), but can have other values - see the `PRAGMA journal_mode <https://www.sqlite.org/pragma.html#pragma_journal_mode>`__ documentation.
When you create a new table for a list of inserted or upserted Python dictionaries, those methods detect the correct types for the database columns based on the data you pass in.
In some situations you may need to intervene in this process, to customize the columns that are being created in some way - see :ref:`python_api_explicit_create`.
That table ``.create()`` method takes a dictionary mapping column names to the Python type they should store:
You can use the ``suggest_column_types()`` helper function to derive a dictionary of column names and types from a list of records, suitable to be passed to ``table.create()``.
For example:
..code-block:: python
from sqlite_utils import Database, suggest_column_types
SQLite supports registering custom SQL functions written in Python. The ``db.register_function()`` method lets you register these functions, and keeps track of functions that have already been registered.
If you use it as a method it will automatically detect the name and number of arguments needed by the function:
If a function will return the exact same result for any given inputs you can register it as a `deterministic SQLite function <https://sqlite.org/deterministic.html>`__ allowing SQLite to apply some performance optimizations:
By default registering a function with the same name and number of arguments will have no effect - the ``Database`` instance keeps track of functions that have already been registered and skips registering them if ``@db.register_function`` is called a second time.
If you want to deliberately replace the registered function with a new implementation, use the ``replace=True`` argument:
You can cause ``sqlite3`` to return more useful errors, including the traceback from the custom function, by executing the following before your custom functions are executed:
In almost all cases you should pass values to your SQL queries using the optional ``parameters`` argument to ``db.query()``, as described in :ref:`python_api_parameters`.
The ``sqlite_utils.utils.rows_from_file()`` helper function can read rows (a sequence of dictionaries) from CSV, TSV, JSON or newline-delimited JSON files.
Sometimes when working with CSV files that include extremely long fields you may see an error that looks like this::
_csv.Error: field larger than field limit (131072)
The Python standard library ``csv`` module enforces a field size limit. You can increase that limit using the ``csv.field_size_limit(new_limit)`` method (`documented here <https://docs.python.org/3/library/csv.html#csv.field_size_limit>`__) but if you don't want to pick a new level you may instead want to increase it to the maximum possible.
The maximum possible value for this is not documented, and varies between systems.
Calling ``sqlite_utils.utils.maximize_csv_field_size_limit()`` will set the value to the highest possible for the current system:
..code-block:: python
from sqlite_utils.utils import maximize_csv_field_size_limit
maximize_csv_field_size_limit()
If you need to reset to the original value after calling this function you can do so like this:
..code-block:: python
from sqlite_utils.utils import ORIGINAL_CSV_FIELD_SIZE_LIMIT
`SpatiaLite <https://www.gaia-gis.it/fossil/libspatialite/index>`__ is a geographic extension to SQLite (similar to PostgreSQL + PostGIS). Using requires finding, loading and initializing the extension, adding geometry columns to existing tables and optionally creating spatial indexes. The utilities here help streamline that setup.
