sqlite3 --- SQLite 数据库 DB-API 2.0 接口模块


SQLite 是一个C语言库,它可以提供一种轻量级的基于磁盘的数据库,这种数据库不需要独立的服务器进程,也允许需要使用一种非标准的 SQL 查询语言来访问它。一些应用程序可以使用 SQLite 作为内部数据存储。可以用它来创建一个应用程序原型,然后再迁移到更大的数据库,比如 PostgreSQL 或 Oracle。

The sqlite3 module was written by Gerhard Häring. It provides an SQL interface compliant with the DB-API 2.0 specification described by PEP 249, and requires SQLite 3.7.15 or newer.

This document includes four main sections:

  • 教程 teaches how to use the sqlite3 module.

  • 参考 describes the classes and functions this module defines.

  • How-to guides details how to handle specific tasks.

  • 解釋 provides in-depth background on transaction control.



SQLite的主页;它的文档详细描述了它所支持的 SQL 方言的语法和可用的数据类型。


学习 SQL 语法的教程、参考和例子。

PEP 249 - DB-API 2.0 规范

PEP 由 Marc-André Lemburg 撰寫。


In this tutorial, you will create a database of Monty Python movies using basic sqlite3 functionality. It assumes a fundamental understanding of database concepts, including cursors and transactions.

First, we need to create a new database and open a database connection to allow sqlite3 to work with it. Call sqlite3.connect() to create a connection to the database tutorial.db in the current working directory, implicitly creating it if it does not exist:

import sqlite3
con = sqlite3.connect("tutorial.db")

The returned Connection object con represents the connection to the on-disk database.

In order to execute SQL statements and fetch results from SQL queries, we will need to use a database cursor. Call con.cursor() to create the Cursor:

cur = con.cursor()

Now that we've got a database connection and a cursor, we can create a database table movie with columns for title, release year, and review score. For simplicity, we can just use column names in the table declaration -- thanks to the flexible typing feature of SQLite, specifying the data types is optional. Execute the CREATE TABLE statement by calling cur.execute(...):

cur.execute("CREATE TABLE movie(title, year, score)")

We can verify that the new table has been created by querying the sqlite_master table built-in to SQLite, which should now contain an entry for the movie table definition (see The Schema Table for details). Execute that query by calling cur.execute(...), assign the result to res, and call res.fetchone() to fetch the resulting row:

>>> res = cur.execute("SELECT name FROM sqlite_master")
>>> res.fetchone()

We can see that the table has been created, as the query returns a tuple containing the table's name. If we query sqlite_master for a non-existent table spam, res.fetchone() will return None:

>>> res = cur.execute("SELECT name FROM sqlite_master WHERE name='spam'")
>>> res.fetchone() is None

Now, add two rows of data supplied as SQL literals by executing an INSERT statement, once again by calling cur.execute(...):

        ('Monty Python and the Holy Grail', 1975, 8.2),
        ('And Now for Something Completely Different', 1971, 7.5)

The INSERT statement implicitly opens a transaction, which needs to be committed before changes are saved in the database (see Transaction control for details). Call con.commit() on the connection object to commit the transaction:


We can verify that the data was inserted correctly by executing a SELECT query. Use the now-familiar cur.execute(...) to assign the result to res, and call res.fetchall() to return all resulting rows:

>>> res = cur.execute("SELECT score FROM movie")
>>> res.fetchall()
[(8.2,), (7.5,)]

The result is a list of two tuples, one per row, each containing that row's score value.

Now, insert three more rows by calling cur.executemany(...):

data = [
    ("Monty Python Live at the Hollywood Bowl", 1982, 7.9),
    ("Monty Python's The Meaning of Life", 1983, 7.5),
    ("Monty Python's Life of Brian", 1979, 8.0),
cur.executemany("INSERT INTO movie VALUES(?, ?, ?)", data)
con.commit()  # Remember to commit the transaction after executing INSERT.

Notice that ? placeholders are used to bind data to the query. Always use placeholders instead of string formatting to bind Python values to SQL statements, to avoid SQL injection attacks (see How to use placeholders to bind values in SQL queries for more details).

We can verify that the new rows were inserted by executing a SELECT query, this time iterating over the results of the query:

>>> for row in cur.execute("SELECT year, title FROM movie ORDER BY year"):
...     print(row)
(1971, 'And Now for Something Completely Different')
(1975, 'Monty Python and the Holy Grail')
(1979, "Monty Python's Life of Brian")
(1982, 'Monty Python Live at the Hollywood Bowl')
(1983, "Monty Python's The Meaning of Life")

Each row is a two-item tuple of (year, title), matching the columns selected in the query.

Finally, verify that the database has been written to disk by calling con.close() to close the existing connection, opening a new one, creating a new cursor, then querying the database:

>>> con.close()
>>> new_con = sqlite3.connect("tutorial.db")
>>> new_cur = new_con.cursor()
>>> res = new_cur.execute("SELECT title, year FROM movie ORDER BY score DESC")
>>> title, year = res.fetchone()
>>> print(f'The highest scoring Monty Python movie is {title!r}, released in {year}')
The highest scoring Monty Python movie is 'Monty Python and the Holy Grail', released in 1975

You've now created an SQLite database using the sqlite3 module, inserted data and retrieved values from it in multiple ways.


Module functions

sqlite3.connect(database, timeout=5.0, detect_types=0, isolation_level='DEFERRED', check_same_thread=True, factory=sqlite3.Connection, cached_statements=100, uri=False)

Open a connection to an SQLite database.

  • database (path-like object) -- The path to the database file to be opened. Pass ":memory:" to open a connection to a database that is in RAM instead of on disk.

  • timeout (float) -- How many seconds the connection should wait before raising an OperationalError when a table is locked. If another connection opens a transaction to modify a table, that table will be locked until the transaction is committed. Default five seconds.

  • detect_types (int) -- Control whether and how data types not natively supported by SQLite are looked up to be converted to Python types, using the converters registered with register_converter(). Set it to any combination (using |, bitwise or) of PARSE_DECLTYPES and PARSE_COLNAMES to enable this. Column names takes precedence over declared types if both flags are set. Types cannot be detected for generated fields (for example max(data)), even when the detect_types parameter is set; str will be returned instead. By default (0), type detection is disabled.

  • isolation_level (str | None) -- The isolation_level of the connection, controlling whether and how transactions are implicitly opened. Can be "DEFERRED" (default), "EXCLUSIVE" or "IMMEDIATE"; or None to disable opening transactions implicitly. See Transaction control for more.

  • check_same_thread (bool) -- If True (default), ProgrammingError will be raised if the database connection is used by a thread other than the one that created it. If False, the connection may be accessed in multiple threads; write operations may need to be serialized by the user to avoid data corruption. See threadsafety for more information.

  • factory (Connection) -- A custom subclass of Connection to create the connection with, if not the default Connection class.

  • cached_statements (int) -- The number of statements that sqlite3 should internally cache for this connection, to avoid parsing overhead. By default, 100 statements.

  • uri (bool) -- If set to True, database is interpreted as a URI with a file path and an optional query string. The scheme part must be "file:", and the path can be relative or absolute. The query string allows passing parameters to SQLite, enabling various How to work with SQLite URIs.



引发一个 审计事件 sqlite3.connect,附带参数 database

引发一个 审计事件 sqlite3.connect/handle,附带参数 connection_handle

3.4 版新加入: uri 參數。

3.7 版更變: database 现在可以是一个 path-like object 对象了,不仅仅是字符串。

3.10 版新加入: The sqlite3.connect/handle auditing event.


Return True if the string statement appears to contain one or more complete SQL statements. No syntactic verification or parsing of any kind is performed, other than checking that there are no unclosed string literals and the statement is terminated by a semicolon.


>>> sqlite3.complete_statement("SELECT foo FROM bar;")
>>> sqlite3.complete_statement("SELECT foo")

This function may be useful during command-line input to determine if the entered text seems to form a complete SQL statement, or if additional input is needed before calling execute().

sqlite3.enable_callback_tracebacks(flag, /)

Enable or disable callback tracebacks. By default you will not get any tracebacks in user-defined functions, aggregates, converters, authorizer callbacks etc. If you want to debug them, you can call this function with flag set to True. Afterwards, you will get tracebacks from callbacks on sys.stderr. Use False to disable the feature again.

sqlite3.register_adapter(type, adapter, /)

Register an adapter callable to adapt the Python type type into an SQLite type. The adapter is called with a Python object of type type as its sole argument, and must return a value of a type that SQLite natively understands.

sqlite3.register_converter(typename, converter, /)

Register the converter callable to convert SQLite objects of type typename into a Python object of a specific type. The converter is invoked for all SQLite values of type typename; it is passed a bytes object and should return an object of the desired Python type. Consult the parameter detect_types of connect() for information regarding how type detection works.

Note: typename and the name of the type in your query are matched case-insensitively.

Module constants


Pass this flag value to the detect_types parameter of connect() to look up a converter function by using the type name, parsed from the query column name, as the converter dictionary key. The type name must be wrapped in square brackets ([]).

SELECT p as "p [point]" FROM test;  ! will look up converter "point"

This flag may be combined with PARSE_DECLTYPES using the | (bitwise or) operator.


Pass this flag value to the detect_types parameter of connect() to look up a converter function using the declared types for each column. The types are declared when the database table is created. sqlite3 will look up a converter function using the first word of the declared type as the converter dictionary key. For example:

   i integer primary key,  ! will look up a converter named "integer"
   p point,                ! will look up a converter named "point"
   n number(10)            ! will look up a converter named "number"

This flag may be combined with PARSE_COLNAMES using the | (bitwise or) operator.


Flags that should be returned by the authorizer_callback callable passed to Connection.set_authorizer(), to indicate whether:

  • Access is allowed (SQLITE_OK),

  • The SQL statement should be aborted with an error (SQLITE_DENY)

  • The column should be treated as a NULL value (SQLITE_IGNORE)


指明所支持的 DB-API 级别的字符串常量。 根据 DB-API 的需要设置。 硬编码为 "2.0"


String constant stating the type of parameter marker formatting expected by the sqlite3 module. Required by the DB-API. Hard-coded to "qmark".


The named DB-API parameter style is also supported.


Version number of the runtime SQLite library as a string.


Version number of the runtime SQLite library as a tuple of integers.


Integer constant required by the DB-API, stating the level of thread safety the sqlite3 module supports. Currently hard-coded to 1, meaning "Threads may share the module, but not connections." However, this may not always be true. You can check the underlying SQLite library's compile-time threaded mode using the following query:

import sqlite3
con = sqlite3.connect(":memory:")
    select * from pragma_compile_options
    where compile_options like 'THREADSAFE=%'

注意 SQLITE_THREADSAFE 级别 与 DB-API 2.0 threadsafety 级别不一致。


Version number of this module as a string. This is not the version of the SQLite library.


Version number of this module as a tuple of integers. This is not the version of the SQLite library.

Connection objects

class sqlite3.Connection

Each open SQLite database is represented by a Connection object, which is created using sqlite3.connect(). Their main purpose is creating Cursor objects, and Transaction control.

SQLite 数据库连接对象有如下的属性和方法:


Create and return a Cursor object. The cursor method accepts a single optional parameter factory. If supplied, this must be a callable returning an instance of Cursor or its subclasses.


Commit any pending transaction to the database. If there is no open transaction, this method is a no-op.


Roll back to the start of any pending transaction. If there is no open transaction, this method is a no-op.


Close the database connection. Any pending transaction is not committed implicitly; make sure to commit() before closing to avoid losing pending changes.

execute(sql, parameters=(), /)

创建一个新的 Cursor 对象,并在其上使用给出的 sqlparameters 调用 execute()。 返回新的游标对象。

executemany(sql, parameters, /)

创建一个新的 Cursor 对象,并在其上使用给出的 sqlparameters 调用 executemany()。 返回新的游标对象。

executescript(sql_script, /)

创建一个新的 Cursor 对象,并在其上使用给出的 sql_script 调用 executescript()。 返回新的游标对象。

create_function(name, narg, func, *, deterministic=False)

Create or remove a user-defined SQL function.

  • name (str) -- The name of the SQL function.

  • narg (int) -- The number of arguments the SQL function can accept. If -1, it may take any number of arguments.

  • func (callback | None) -- A callable that is called when the SQL function is invoked. The callable must return a type natively supported by SQLite. Set to None to remove an existing SQL function.

  • deterministic (bool) -- If True, the created SQL function is marked as deterministic, which allows SQLite to perform additional optimizations.


NotSupportedError -- If deterministic is used with SQLite versions older than 3.8.3.

3.8 版新加入: 新增 deterministic 參數。


>>> import hashlib
>>> def md5sum(t):
...     return hashlib.md5(t).hexdigest()
>>> con = sqlite3.connect(":memory:")
>>> con.create_function("md5", 1, md5sum)
>>> for row in con.execute("SELECT md5(?)", (b"foo",)):
...     print(row)
create_aggregate(name, /, n_arg, aggregate_class)

Create or remove a user-defined SQL aggregate function.

  • name (str) -- The name of the SQL aggregate function.

  • n_arg (int) -- The number of arguments the SQL aggregate function can accept. If -1, it may take any number of arguments.

  • aggregate_class (class | None) --

    A class must implement the following methods:

    The number of arguments that the step() method must accept is controlled by n_arg.

    Set to None to remove an existing SQL aggregate function.


class MySum:
    def __init__(self):
        self.count = 0

    def step(self, value):
        self.count += value

    def finalize(self):
        return self.count

con = sqlite3.connect(":memory:")
con.create_aggregate("mysum", 1, MySum)
cur = con.execute("CREATE TABLE test(i)")
cur.execute("INSERT INTO test(i) VALUES(1)")
cur.execute("INSERT INTO test(i) VALUES(2)")
cur.execute("SELECT mysum(i) FROM test")

create_collation(name, callable)

使用排序函数 callable 创建一个名为 name 的排序规则。 callable 被传递给两个 字符串 参数,并且它应该返回一个 整数

  • 如果前者的排序高于后者则为 1

  • 如果前者的排序低于于后者则为 -1

  • 如果它们的顺序相同则为 0


def collate_reverse(string1, string2):
    if string1 == string2:
        return 0
    elif string1 < string2:
        return 1
        return -1

con = sqlite3.connect(":memory:")
con.create_collation("reverse", collate_reverse)

cur = con.execute("CREATE TABLE test(x)")
cur.executemany("INSERT INTO test(x) VALUES(?)", [("a",), ("b",)])
cur.execute("SELECT x FROM test ORDER BY x COLLATE reverse")
for row in cur:

Remove a collation function by setting callable to None.


Call this method from a different thread to abort any queries that might be executing on the connection. Aborted queries will raise an OperationalError.


Register callable authorizer_callback to be invoked for each attempt to access a column of a table in the database. The callback should return one of SQLITE_OK, SQLITE_DENY, or SQLITE_IGNORE to signal how access to the column should be handled by the underlying SQLite library.

The first argument to the callback signifies what kind of operation is to be authorized. The second and third argument will be arguments or None depending on the first argument. The 4th argument is the name of the database ("main", "temp", etc.) if applicable. The 5th argument is the name of the inner-most trigger or view that is responsible for the access attempt or None if this access attempt is directly from input SQL code.

Please consult the SQLite documentation about the possible values for the first argument and the meaning of the second and third argument depending on the first one. All necessary constants are available in the sqlite3 module.

set_progress_handler(progress_handler, n)

Register callable progress_handler to be invoked for every n instructions of the SQLite virtual machine. This is useful if you want to get called from SQLite during long-running operations, for example to update a GUI.

If you want to clear any previously installed progress handler, call the method with None for progress_handler.

从处理函数返回非零值将终止当前正在执行的查询并导致它引发 OperationalError 异常。


Register callable trace_callback to be invoked for each SQL statement that is actually executed by the SQLite backend.

The only argument passed to the callback is the statement (as str) that is being executed. The return value of the callback is ignored. Note that the backend does not only run statements passed to the Cursor.execute() methods. Other sources include the transaction management of the sqlite3 module and the execution of triggers defined in the current database.

Passing None as trace_callback will disable the trace callback.


在跟踪回调中产生的异常不会被传播。作为开发和调试的辅助手段,使用 enable_callback_tracebacks() 来启用打印跟踪回调中产生的异常的回调。

3.3 版新加入.

enable_load_extension(enabled, /)

Enable the SQLite engine to load SQLite extensions from shared libraries if enabled is True; else, disallow loading SQLite extensions. SQLite extensions can define new functions, aggregates or whole new virtual table implementations. One well-known extension is the fulltext-search extension distributed with SQLite.


The sqlite3 module is not built with loadable extension support by default, because some platforms (notably macOS) have SQLite libraries which are compiled without this feature. To get loadable extension support, you must pass the --enable-loadable-sqlite-extensions option to configure.

引发一个 审计事件 sqlite3.enable_load_extension,附带参数 connection, enabled

3.2 版新加入.

3.10 版更變: 增加了 sqlite3.enable_load_extension 审计事件。


# Load the fulltext search extension
con.execute("select load_extension('./fts3.so')")

# alternatively you can load the extension using an API call:
# con.load_extension("./fts3.so")

# disable extension loading again

# example from SQLite wiki
con.execute("CREATE VIRTUAL TABLE recipe USING fts3(name, ingredients)")
    INSERT INTO recipe (name, ingredients) VALUES('broccoli stew', 'broccoli peppers cheese tomatoes');
    INSERT INTO recipe (name, ingredients) VALUES('pumpkin stew', 'pumpkin onions garlic celery');
    INSERT INTO recipe (name, ingredients) VALUES('broccoli pie', 'broccoli cheese onions flour');
    INSERT INTO recipe (name, ingredients) VALUES('pumpkin pie', 'pumpkin sugar flour butter');
for row in con.execute("SELECT rowid, name, ingredients FROM recipe WHERE name MATCH 'pie'"):

load_extension(path, /)

Load an SQLite extension from a shared library located at path. Enable extension loading with enable_load_extension() before calling this method.

引发一个 审计事件 sqlite3.load_extension,附带参数 connection, path

3.2 版新加入.

3.10 版更變: 增加了 sqlite3.load_extension 审计事件。


Return an iterator to dump the database as SQL source code. Useful when saving an in-memory database for later restoration. Similar to the .dump command in the sqlite3 shell.


# Convert file example.db to SQL dump file dump.sql
con = sqlite3.connect('example.db')
with open('dump.sql', 'w') as f:
    for line in con.iterdump():
        f.write('%s\n' % line)
backup(target, *, pages=- 1, progress=None, name='main', sleep=0.250)

Create a backup of an SQLite database.

Works even if the database is being accessed by other clients or concurrently by the same connection.

  • target (Connection) -- The database connection to save the backup to.

  • pages (int) -- The number of pages to copy at a time. If equal to or less than 0, the entire database is copied in a single step. Defaults to -1.

  • progress (callback | None) -- If set to a callable, it is invoked with three integer arguments for every backup iteration: the status of the last iteration, the remaining number of pages still to be copied, and the total number of pages. Defaults to None.

  • name (str) -- The name of the database to back up. Either "main" (the default) for the main database, "temp" for the temporary database, or the name of a custom database as attached using the ATTACH DATABASE SQL statement.

  • sleep (float) -- The number of seconds to sleep between successive attempts to back up remaining pages.

Example 1, copy an existing database into another:

def progress(status, remaining, total):
    print(f'Copied {total-remaining} of {total} pages...')

src = sqlite3.connect('example.db')
dst = sqlite3.connect('backup.db')
with dst:
    src.backup(dst, pages=1, progress=progress)

Example 2, copy an existing database into a transient copy:

src = sqlite3.connect('example.db')
dst = sqlite3.connect(':memory:')

3.7 版新加入.


This read-only attribute corresponds to the low-level SQLite autocommit mode.

True if a transaction is active (there are uncommitted changes), False otherwise.

3.2 版新加入.


This attribute controls the transaction handling performed by sqlite3. If set to None, transactions are never implicitly opened. If set to one of "DEFERRED", "IMMEDIATE", or "EXCLUSIVE", corresponding to the underlying SQLite transaction behaviour, implicit transaction management is performed.

If not overridden by the isolation_level parameter of connect(), the default is "", which is an alias for "DEFERRED".


The initial row_factory for Cursor objects created from this connection. Assigning to this attribute does not affect the row_factory of existing cursors belonging to this connection, only new ones. Is None by default, meaning each row is returned as a tuple.

See How to create and use row factories for more details.


A callable that accepts a bytes parameter and returns a text representation of it. The callable is invoked for SQLite values with the TEXT data type. By default, this attribute is set to str. If you want to return bytes instead, set text_factory to bytes.


con = sqlite3.connect(":memory:")
cur = con.cursor()

AUSTRIA = "Österreich"

# by default, rows are returned as str
cur.execute("SELECT ?", (AUSTRIA,))
row = cur.fetchone()
assert row[0] == AUSTRIA

# but we can make sqlite3 always return bytestrings ...
con.text_factory = bytes
cur.execute("SELECT ?", (AUSTRIA,))
row = cur.fetchone()
assert type(row[0]) is bytes
# the bytestrings will be encoded in UTF-8, unless you stored garbage in the
# database ...
assert row[0] == AUSTRIA.encode("utf-8")

# we can also implement a custom text_factory ...
# here we implement one that appends "foo" to all strings
con.text_factory = lambda x: x.decode("utf-8") + "foo"
cur.execute("SELECT ?", ("bar",))
row = cur.fetchone()
assert row[0] == "barfoo"


Return the total number of database rows that have been modified, inserted, or deleted since the database connection was opened.

Cursor objects

A Cursor object represents a database cursor which is used to execute SQL statements, and manage the context of a fetch operation. Cursors are created using Connection.cursor(), or by using any of the connection shortcut methods.

Cursor objects are iterators, meaning that if you execute() a SELECT query, you can simply iterate over the cursor to fetch the resulting rows:

for row in cur.execute("SELECT t FROM data"):
class sqlite3.Cursor

Cursor 游标实例具有以下属性和方法。

execute(sql, parameters=(), /)

Execute SQL a single SQL statement, optionally binding Python values using placeholders.


Warning -- If sql contains more than one SQL statement.

If isolation_level is not None, sql is an INSERT, UPDATE, DELETE, or REPLACE statement, and there is no open transaction, a transaction is implicitly opened before executing sql.

Use executescript() to execute multiple SQL statements.

executemany(sql, parameters, /)

For every item in parameters, repeatedly execute the parameterized SQL statement sql.

Uses the same implicit transaction handling as execute().



rows = [
# cur is an sqlite3.Cursor object
cur.executemany("INSERT INTO data VALUES(?)", rows)
executescript(sql_script, /)

Execute the SQL statements in sql_script. If there is a pending transaction, an implicit COMMIT statement is executed first. No other implicit transaction control is performed; any transaction control must be added to sql_script.

sql_script must be a string.


# cur is an sqlite3.Cursor object
    CREATE TABLE person(firstname, lastname, age);
    CREATE TABLE book(title, author, published);
    CREATE TABLE publisher(name, address);

If row_factory is None, return the next row query result set as a tuple. Else, pass it to the row factory and return its result. Return None if no more data is available.


Return the next set of rows of a query result as a list. Return an empty list if no more rows are available.

The number of rows to fetch per call is specified by the size parameter. If size is not given, arraysize determines the number of rows to be fetched. If fewer than size rows are available, as many rows as are available are returned.

请注意 size 形参会涉及到性能方面的考虑。为了获得优化的性能,通常最好是使用 arraysize 属性。 如果使用 size 形参,则最好在从一个 fetchmany() 调用到下一个调用之间保持相同的值。


Return all (remaining) rows of a query result as a list. Return an empty list if no rows are available. Note that the arraysize attribute can affect the performance of this operation.


立即关闭 cursor(而不是在当 __del__ 被调用的时候)。

从这一时刻起该 cursor 将不再可用,如果再尝试用该 cursor 执行任何操作将引发 ProgrammingError 异常。

setinputsizes(sizes, /)

Required by the DB-API. Does nothing in sqlite3.

setoutputsize(size, column=None, /)

Required by the DB-API. Does nothing in sqlite3.


用于控制 fetchmany() 返回行数的可读取/写入属性。 该属性的默认值为 1,表示每次调用将获取单独一行。


Read-only attribute that provides the SQLite database Connection belonging to the cursor. A Cursor object created by calling con.cursor() will have a connection attribute that refers to con:

>>> con = sqlite3.connect(":memory:")
>>> cur = con.cursor()
>>> cur.connection == con

Read-only attribute that provides the column names of the last query. To remain compatible with the Python DB API, it returns a 7-tuple for each column where the last six items of each tuple are None.

对于没有任何匹配行的 SELECT 语句同样会设置该属性。


Read-only attribute that provides the row id of the last inserted row. It is only updated after successful INSERT or REPLACE statements using the execute() method. For other statements, after executemany() or executescript(), or if the insertion failed, the value of lastrowid is left unchanged. The initial value of lastrowid is None.



3.6 版更變: 新增 REPLACE 陳述式的支援。


Read-only attribute that provides the number of modified rows for INSERT, UPDATE, DELETE, and REPLACE statements; is -1 for other statements, including CTE queries. It is only updated by the execute() and executemany() methods.


Control how a row fetched from this Cursor is represented. If None, a row is represented as a tuple. Can be set to the included sqlite3.Row; or a callable that accepts two arguments, a Cursor object and the tuple of row values, and returns a custom object representing an SQLite row.

Defaults to what Connection.row_factory was set to when the Cursor was created. Assigning to this attribute does not affect Connection.row_factory of the parent connection.

See How to create and use row factories for more details.

Row objects

class sqlite3.Row

A Row instance serves as a highly optimized row_factory for Connection objects. It supports iteration, equality testing, len(), and mapping access by column name and index.

Two Row objects compare equal if they have identical column names and values.

See How to create and use row factories for more details.


Return a list of column names as strings. Immediately after a query, it is the first member of each tuple in Cursor.description.

3.5 版更變: 添加了对切片操作的支持。

PrepareProtocol objects

class sqlite3.PrepareProtocol

The PrepareProtocol type's single purpose is to act as a PEP 246 style adaption protocol for objects that can adapt themselves to native SQLite types.


异常层次是由 DB-API 2.0 (PEP 249) 定义的。

exception sqlite3.Warning

This exception is raised by sqlite3 if an SQL query is not a string, or if multiple statements are passed to execute() or executemany(). Warning is a subclass of Exception.

exception sqlite3.Error

本模块中其他异常的基类。使用它来捕捉所有的错误,只需一条 except 语句。 ErrorException 的子类。

exception sqlite3.InterfaceError

This exception is raised by sqlite3 for fetch across rollback, or if sqlite3 is unable to bind parameters. InterfaceError is a subclass of Error.

exception sqlite3.DatabaseError

对与数据库有关的错误引发的异常。它作为几种数据库错误的基础异常。它只通过专门的子类隐式引发。 DatabaseErrorError 的一个子类。

exception sqlite3.DataError

由于处理的数据有问题而产生的异常,比如数字值超出范围,字符串太长。 DataErrorDatabaseError 的子类。

exception sqlite3.OperationalError

与数据库操作有关的错误而引发的异常,不一定在程序员的控制之下。例如,数据库路径没有找到,或者一个事务无法被处理。 OperationalErrorDatabaseError 的子类。

exception sqlite3.IntegrityError

当数据库的关系一致性受到影响时引发的异常。 例如外键检查失败等。 它是 DatabaseError 的子类。

exception sqlite3.InternalError

当 SQLite 遇到一个内部错误时引发的异常。如果它被引发,可能表明运行中的 SQLite 库有问题。 InternalErrorDatabaseError 的子类。

exception sqlite3.ProgrammingError

Exception raised for sqlite3 API programming errors, for example trying to operate on a closed Connection, or trying to execute non-DML statements with executemany(). ProgrammingError is a subclass of DatabaseError.

exception sqlite3.NotSupportedError

Exception raised in case a method or database API is not supported by the underlying SQLite library. For example, setting deterministic to True in create_function(), if the underlying SQLite library does not support deterministic functions. NotSupportedError is a subclass of DatabaseError.

SQLite 与 Python 类型



Python 类型

SQLite 类型












SQLite 类型

Python 类型








取决于 text_factory , 默认为 str



The type system of the sqlite3 module is extensible in two ways: you can store additional Python types in an SQLite database via object adapters, and you can let the sqlite3 module convert SQLite types to Python types via converters.


对于 datetime 模块中的 date 和 datetime 类型已提供了默认的适配器。 它们将会以 ISO 日期/ISO 时间戳的形式发给 SQLite。

默认转换器使用的注册名称是针对 datetime.date 的 "date" 和针对 datetime.datetime 的 "timestamp"。

通过这种方式,你可以在大多数情况下使用 Python 的 date/timestamp 对象而无须任何额外处理。 适配器的格式还与实验性的 SQLite date/time 函数兼容。


import sqlite3
import datetime

con = sqlite3.connect(":memory:", detect_types=sqlite3.PARSE_DECLTYPES|sqlite3.PARSE_COLNAMES)
cur = con.cursor()
cur.execute("create table test(d date, ts timestamp)")

today = datetime.date.today()
now = datetime.datetime.now()

cur.execute("insert into test(d, ts) values (?, ?)", (today, now))
cur.execute("select d, ts from test")
row = cur.fetchone()
print(today, "=>", row[0], type(row[0]))
print(now, "=>", row[1], type(row[1]))

cur.execute('select current_date as "d [date]", current_timestamp as "ts [timestamp]"')
row = cur.fetchone()
print("current_date", row[0], type(row[0]))
print("current_timestamp", row[1], type(row[1]))


如果存储在 SQLite 中的时间戳的小数位多于 6 个数字,则时间戳转换器会将该值截断至微秒精度。


默认的 "时间戳" 转换器忽略了数据库中的 UTC 偏移,总是返回一个原生的 datetime.datetime 对象。要在时间戳中保留 UTC 偏移,可以不使用转换器,或者用 register_converter() 注册一个偏移感知的转换器。

How-to guides

How to use placeholders to bind values in SQL queries

SQL operations usually need to use values from Python variables. However, beware of using Python's string operations to assemble queries, as they are vulnerable to SQL injection attacks. For example, an attacker can simply close the single quote and inject OR TRUE to select all rows:

>>> # Never do this -- insecure!
>>> symbol = input()
' OR TRUE; --
>>> sql = "SELECT * FROM stocks WHERE symbol = '%s'" % symbol
>>> print(sql)
SELECT * FROM stocks WHERE symbol = '' OR TRUE; --'
>>> cur.execute(sql)

Instead, use the DB-API's parameter substitution. To insert a variable into a query string, use a placeholder in the string, and substitute the actual values into the query by providing them as a tuple of values to the second argument of the cursor's execute() method.

An SQL statement may use one of two kinds of placeholders: question marks (qmark style) or named placeholders (named style). For the qmark style, parameters must be a sequence whose length must match the number of placeholders, or a ProgrammingError is raised. For the named style, parameters should be an instance of a dict (or a subclass), which must contain keys for all named parameters; any extra items are ignored. Here's an example of both styles:

con = sqlite3.connect(":memory:")
cur = con.execute("CREATE TABLE lang(name, first_appeared)")

# This is the named style used with executemany():
data = (
    {"name": "C", "year": 1972},
    {"name": "Fortran", "year": 1957},
    {"name": "Python", "year": 1991},
    {"name": "Go", "year": 2009},
cur.executemany("INSERT INTO lang VALUES(:name, :year)", data)

# This is the qmark style used in a SELECT query:
params = (1972,)
cur.execute("SELECT * FROM lang WHERE first_appeared = ?", params)


PEP 249 numeric placeholders are not supported. If used, they will be interpreted as named placeholders.

How to adapt custom Python types to SQLite values

SQLite supports only a limited set of data types natively. To store custom Python types in SQLite databases, adapt them to one of the Python types SQLite natively understands.

There are two ways to adapt Python objects to SQLite types: letting your object adapt itself, or using an adapter callable. The latter will take precedence above the former. For a library that exports a custom type, it may make sense to enable that type to adapt itself. As an application developer, it may make more sense to take direct control by registering custom adapter functions.

How to write adaptable objects

Suppose we have a Point class that represents a pair of coordinates, x and y, in a Cartesian coordinate system. The coordinate pair will be stored as a text string in the database, using a semicolon to separate the coordinates. This can be implemented by adding a __conform__(self, protocol) method which returns the adapted value. The object passed to protocol will be of type PrepareProtocol.

class Point:
    def __init__(self, x, y):
        self.x, self.y = x, y

    def __conform__(self, protocol):
        if protocol is sqlite3.PrepareProtocol:
            return f"{self.x};{self.y}"

con = sqlite3.connect(":memory:")
cur = con.cursor()

cur.execute("SELECT ?", (Point(4.0, -3.2),))

How to register adapter callables

The other possibility is to create a function that converts the Python object to an SQLite-compatible type. This function can then be registered using register_adapter().

class Point:
    def __init__(self, x, y):
        self.x, self.y = x, y

def adapt_point(point):
    return f"{point.x};{point.y}"

sqlite3.register_adapter(Point, adapt_point)

con = sqlite3.connect(":memory:")
cur = con.cursor()

cur.execute("SELECT ?", (Point(1.0, 2.5),))

How to convert SQLite values to custom Python types

Writing an adapter lets you convert from custom Python types to SQLite values. To be able to convert from SQLite values to custom Python types, we use converters.

Let's go back to the Point class. We stored the x and y coordinates separated via semicolons as strings in SQLite.

First, we'll define a converter function that accepts the string as a parameter and constructs a Point object from it.


Converter functions are always passed a bytes object, no matter the underlying SQLite data type.

def convert_point(s):
    x, y = map(float, s.split(b";"))
    return Point(x, y)

We now need to tell sqlite3 when it should convert a given SQLite value. This is done when connecting to a database, using the detect_types parameter of connect(). There are three options:

  • Implicit: set detect_types to PARSE_DECLTYPES

  • Explicit: set detect_types to PARSE_COLNAMES

  • Both: set detect_types to sqlite3.PARSE_DECLTYPES | sqlite3.PARSE_COLNAMES. Column names take precedence over declared types.

The following example illustrates the implicit and explicit approaches:

class Point:
    def __init__(self, x, y):
        self.x, self.y = x, y

    def __repr__(self):
        return f"Point({self.x}, {self.y})"

def adapt_point(point):
    return f"{point.x};{point.y}"

def convert_point(s):
    x, y = list(map(float, s.split(b";")))
    return Point(x, y)

# Register the adapter and converter
sqlite3.register_adapter(Point, adapt_point)
sqlite3.register_converter("point", convert_point)

# 1) Parse using declared types
p = Point(4.0, -3.2)
con = sqlite3.connect(":memory:", detect_types=sqlite3.PARSE_DECLTYPES)
cur = con.execute("CREATE TABLE test(p point)")

cur.execute("INSERT INTO test(p) VALUES(?)", (p,))
cur.execute("SELECT p FROM test")
print("with declared types:", cur.fetchone()[0])

# 2) Parse using column names
con = sqlite3.connect(":memory:", detect_types=sqlite3.PARSE_COLNAMES)
cur = con.execute("CREATE TABLE test(p)")

cur.execute("INSERT INTO test(p) VALUES(?)", (p,))
cur.execute('SELECT p AS "p [point]" FROM test')
print("with column names:", cur.fetchone()[0])

Adapter and converter recipes

This section shows recipes for common adapters and converters.

import datetime
import sqlite3

def adapt_date_iso(val):
    """Adapt datetime.date to ISO 8601 date."""
    return val.isoformat()

def adapt_datetime_iso(val):
    """Adapt datetime.datetime to timezone-naive ISO 8601 date."""
    return val.isoformat()

def adapt_datetime_epoch(val):
    """Adapt datetime.datetime to Unix timestamp."""
    return int(val.timestamp())

sqlite3.register_adapter(datetime.date, adapt_date_iso)
sqlite3.register_adapter(datetime.datetime, adapt_datetime_iso)
sqlite3.register_adapter(datetime.datetime, adapt_datetime_epoch)

def convert_date(val):
    """Convert ISO 8601 date to datetime.date object."""
    return datetime.date.fromisoformat(val.decode())

def convert_datetime(val):
    """Convert ISO 8601 datetime to datetime.datetime object."""
    return datetime.datetime.fromisoformat(val.decode())

def convert_timestamp(val):
    """Convert Unix epoch timestamp to datetime.datetime object."""
    return datetime.datetime.fromtimestamp(int(val))

sqlite3.register_converter("date", convert_date)
sqlite3.register_converter("datetime", convert_datetime)
sqlite3.register_converter("timestamp", convert_timestamp)

How to use connection shortcut methods

Using the execute(), executemany(), and executescript() methods of the Connection class, your code can be written more concisely because you don't have to create the (often superfluous) Cursor objects explicitly. Instead, the Cursor objects are created implicitly and these shortcut methods return the cursor objects. This way, you can execute a SELECT statement and iterate over it directly using only a single call on the Connection object.

# Create and fill the table.
con = sqlite3.connect(":memory:")
con.execute("CREATE TABLE lang(name, first_appeared)")
data = [
    ("C++", 1985),
    ("Objective-C", 1984),
con.executemany("INSERT INTO lang(name, first_appeared) VALUES(?, ?)", data)

# Print the table contents
for row in con.execute("SELECT name, first_appeared FROM lang"):

print("I just deleted", con.execute("DELETE FROM lang").rowcount, "rows")

# close() is not a shortcut method and it's not called automatically;
# the connection object should be closed manually

How to use the connection context manager

A Connection object can be used as a context manager that automatically commits or rolls back open transactions when leaving the body of the context manager. If the body of the with statement finishes without exceptions, the transaction is committed. If this commit fails, or if the body of the with statement raises an uncaught exception, the transaction is rolled back.

If there is no open transaction upon leaving the body of the with statement, the context manager is a no-op.


The context manager neither implicitly opens a new transaction nor closes the connection.

con = sqlite3.connect(":memory:")

# Successful, con.commit() is called automatically afterwards
with con:
    con.execute("INSERT INTO lang(name) VALUES(?)", ("Python",))

# con.rollback() is called after the with block finishes with an exception,
# the exception is still raised and must be caught
    with con:
        con.execute("INSERT INTO lang(name) VALUES(?)", ("Python",))
except sqlite3.IntegrityError:
    print("couldn't add Python twice")

# Connection object used as context manager only commits or rollbacks transactions,
# so the connection object should be closed manually

How to work with SQLite URIs

Some useful URI tricks include:

  • Open a database in read-only mode:

>>> con = sqlite3.connect("file:tutorial.db?mode=ro", uri=True)
>>> con.execute("CREATE TABLE readonly(data)")
Traceback (most recent call last):
OperationalError: attempt to write a readonly database
  • Do not implicitly create a new database file if it does not already exist; will raise OperationalError if unable to create a new file:

>>> con = sqlite3.connect("file:nosuchdb.db?mode=rw", uri=True)
Traceback (most recent call last):
OperationalError: unable to open database file
  • Create a shared named in-memory database:

db = "file:mem1?mode=memory&cache=shared"
con1 = sqlite3.connect(db, uri=True)
con2 = sqlite3.connect(db, uri=True)
with con1:
    con1.execute("CREATE TABLE shared(data)")
    con1.execute("INSERT INTO shared VALUES(28)")
res = con2.execute("SELECT data FROM shared")
assert res.fetchone() == (28,)

More information about this feature, including a list of parameters, can be found in the SQLite URI documentation.

How to create and use row factories

By default, sqlite3 represents each row as a tuple. If a tuple does not suit your needs, you can use the sqlite3.Row class or a custom row_factory.

While row_factory exists as an attribute both on the Cursor and the Connection, it is recommended to set Connection.row_factory, so all cursors created from the connection will use the same row factory.

Row provides indexed and case-insensitive named access to columns, with minimal memory overhead and performance impact over a tuple. To use Row as a row factory, assign it to the row_factory attribute:

>>> con = sqlite3.connect(":memory:")
>>> con.row_factory = sqlite3.Row

Queries now return Row objects:

>>> res = con.execute("SELECT 'Earth' AS name, 6378 AS radius")
>>> row = res.fetchone()
>>> row.keys()
['name', 'radius']
>>> row[0]         # Access by index.
>>> row["name"]    # Access by name.
>>> row["RADIUS"]  # Column names are case-insensitive.

You can create a custom row_factory that returns each row as a dict, with column names mapped to values:

def dict_factory(cursor, row):
    fields = [column[0] for column in cursor.description]
    return {key: value for key, value in zip(fields, row)}

Using it, queries now return a dict instead of a tuple:

>>> con = sqlite3.connect(":memory:")
>>> con.row_factory = dict_factory
>>> for row in con.execute("SELECT 1 AS a, 2 AS b"):
...     print(row)
{'a': 1, 'b': 2}

The following row factory returns a named tuple:

from collections import namedtuple

def namedtuple_factory(cursor, row):
    fields = [column[0] for column in cursor.description]
    cls = namedtuple("Row", fields)
    return cls._make(row)

namedtuple_factory() can be used as follows:

>>> con = sqlite3.connect(":memory:")
>>> con.row_factory = namedtuple_factory
>>> cur = con.execute("SELECT 1 AS a, 2 AS b")
>>> row = cur.fetchone()
>>> row
Row(a=1, b=2)
>>> row[0]  # Indexed access.
>>> row.b   # Attribute access.

With some adjustments, the above recipe can be adapted to use a dataclass, or any other custom class, instead of a namedtuple.


Transaction control

The sqlite3 module does not adhere to the transaction handling recommended by PEP 249.

If the connection attribute isolation_level is not None, new transactions are implicitly opened before execute() and executemany() executes INSERT, UPDATE, DELETE, or REPLACE statements; for other statements, no implicit transaction handling is performed. Use the commit() and rollback() methods to respectively commit and roll back pending transactions. You can choose the underlying SQLite transaction behaviour — that is, whether and what type of BEGIN statements sqlite3 implicitly executes – via the isolation_level attribute.

If isolation_level is set to None, no transactions are implicitly opened at all. This leaves the underlying SQLite library in autocommit mode, but also allows the user to perform their own transaction handling using explicit SQL statements. The underlying SQLite library autocommit mode can be queried using the in_transaction attribute.

The executescript() method implicitly commits any pending transaction before execution of the given SQL script, regardless of the value of isolation_level.

3.6 版更變: sqlite3 used to implicitly commit an open transaction before DDL statements. This is no longer the case.