An executor-based async sqlite wrapper
Project description
easqlite
A simple Executor-based async sqlite wrapper.
This is used very similarly to the standard sqlite3
module.
By default, ThreadPoolExecutor(max_workers=1)
is used as the executor. If you
pass your own executor, you are responsible for shutting it down and ensuring it
only uses one thread.
Differences from sqlite3
:
connect
is not a function, but just an alias to theConnection
class.Connect
check_same_thread
defaults to__debug__
instead ofTrue
Connection
's constructor takes an optionalexecutor
argument.- Every method, function, context manager, property accessor, and iterator is
awaitable.
Connection.interrupt
operates immediately without beingawaited
, and its returned coroutine is actually a no-op.
- Every call that takes a factory uses the factory for the internal calls, and defers to a statically defined wrapper class. The internal calls will still use the factories.
- All objects with a
close
method are async context managers. - All properties are now methods with an optional setter parameter, so they
are properly awaitable and set and get on the same thread.
- An exception to this is
Cursor.connection
, which is still a property.
- An exception to this is
Blob.__getitem__
is async, butBlob.__setitem__
can not be.Blob.set
is provided instead, with the exact same semantics (it can be passed aslice
). You can useBlob.__setitem__
, but it doesn't actually directly set the blob, but rather queues a set to be run on flush. Any other coroutine flushes the blob, or you can use an explicitBlob.flush
, or just let the blob exit its context manager.Blob.__len__
andBlob.__bool__
block on the executor.Blob.len
andblob.bool
are async replacements for these.
Connection
, Cursor
, and Blob
are lazy. They will not open on
construction, but will open when awaited, or when entered as an asynchronous
context manager.
All these objects must be either awaited or used as an asynchronous context manager:
# OK: The cursor was awaited
cursor = await connection.cursor()
await cursor.execute(sql)
async for row in cursor:
do_something_with(row)
# ERROR: the cursor is not opened
cursor = connection.cursor()
await cursor.execute(sql)
async for row in cursor:
do_something_with(row)
# OK: The cursor was entered (the preferred style)
async with connection.cursor() as cursor:
await cursor.execute(sql)
async for row in cursor:
do_something_with(row)
# The connection.execute family are coroutines, and must be awaited.
# OK: The coroutine was awaited, and returns an open cursor.
async with await connection.execute(sql) as cursor:
async for row in cursor:
do_something_with(row)
# ERROR: The connection.execute result is a coroutine. It is not a context manager
async with connection.execute(sql) as cursor:
async for row in cursor:
do_something_with(row)
# OK: connection.execute opens its cursor, it can just be iterated.
async for row in await connection.execute(sql):
do_something_with(row)
# ERROR: __aiter__ is not defined on a running coroutine. It must be awaited.
async for row in connection.execute(sql):
do_something_with(row)
This can be used nearly identically to the regular sqlite module if you sprinkle
an await
on every function call, but it is preferred to use the async context
managers everywhere possible. You can't easily go wrong when using the context
managers.
Constants are not re-exported, so this library should usually be used in conjunction with the core sqlite3 library.
This is very similar in spirit to the aiosqlite project, but this one takes a more earnest attempt at deferring responsibility to other components. This one also should be more responsive on close, because it doesn't rely on a timeout to shut itself off.
This one also pushes much more extremely on async use, and defers everything it can to the executor thread, even properties.
If you want a more mature and battle-tested package, use aiosqlite
. In my
rough tests, it performs better than this package as well.
Project details
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
File details
Details for the file easqlite-1.1.0.tar.gz
.
File metadata
- Download URL: easqlite-1.1.0.tar.gz
- Upload date:
- Size: 12.3 kB
- Tags: Source
- Uploaded using Trusted Publishing? No
- Uploaded via: python-requests/2.28.2
File hashes
Algorithm | Hash digest | |
---|---|---|
SHA256 | 1587aae4b6c720db374e2f49d3e182db2de72bbcc514e600cbf854a6c6faac2a |
|
MD5 | 3f9ef109654b1267b2fdf591b602b275 |
|
BLAKE2b-256 | 2301674e43e0c5f0d8f656bae59b8a434f49946b48630bc877498ad0924cc576 |
File details
Details for the file easqlite-1.1.0-py3-none-any.whl
.
File metadata
- Download URL: easqlite-1.1.0-py3-none-any.whl
- Upload date:
- Size: 10.8 kB
- Tags: Python 3
- Uploaded using Trusted Publishing? No
- Uploaded via: python-requests/2.28.2
File hashes
Algorithm | Hash digest | |
---|---|---|
SHA256 | 080d7e8f62733a23114b59cb840dc1902298ad9c0ed2f97ca2385f150cf8db24 |
|
MD5 | 0b137ef4a21d77f3ec877e8981e667b8 |
|
BLAKE2b-256 | dabd0a1d88c2d2e5c58c0585a20f5a28009438ab7c13d50c646bf53867b11ff7 |