Improve performance of WebSocketReader

[bdraco]

Optimization

• Replace bytearray.extend with + and += https://stackoverflow.com/questions/40004517/why-does-bytearray-obj-extendbytes-differ-from-bytearray-obj-bytes
• Avoid making copies of the incoming data when there is no continuation. self._partial is never needed for the non-continuation case and we can avoid having to extend and clear it for this case, which is the common case for many use cases.

This is a paired down version of #8735 to better optimize for both the non-continuation and continuation case

Testing feed_data on a live HA instance. Testing was done on an Odroid N2+ SBC. Performance numbers will vary based on CPU and memory performance.

~6% faster for the continuation case (real world use case is listening to a unifi instance)
~34% faster for the non-continuation case (real world use case is reading Bluetooth data from a shelly device, web based terminal shows a similar improvement)

Improvement percentages will be less impressive on systems with faster CPU/memory performance.

related discussion #8258

benchmark for non-continuation case

import random
import zlib
from typing import Any
import timeit
import asyncio
import aiohttp
from aiohttp.http import WSMsgType
from aiohttp.http_websocket import (
    _WS_DEFLATE_TRAILING,
    PACK_LEN1,
    PACK_LEN2,
    PACK_LEN3,
    WebSocketReader,
    _websocket_mask,
)


def build_frame(
    message: Any,
    opcode: Any,
    use_mask: bool = False,
    noheader: bool = False,
    is_fin: bool = True,
    compress: bool = False,
):
    # Send a frame over the websocket with message as its payload.
    if compress:
        compressobj = zlib.compressobj(wbits=-9)
        message = compressobj.compress(message)
        message = message + compressobj.flush(zlib.Z_SYNC_FLUSH)
        if message.endswith(_WS_DEFLATE_TRAILING):
            message = message[:-4]
    msg_length = len(message)
    if use_mask:  # pragma: no cover
        mask_bit = 0x80
    else:
        mask_bit = 0

    if is_fin:
        header_first_byte = 0x80 | opcode
    else:
        header_first_byte = opcode

    if compress:
        header_first_byte |= 0x40

    if msg_length < 126:
        header = PACK_LEN1(header_first_byte, msg_length | mask_bit)
    elif msg_length < (1 << 16):  # pragma: no cover
        header = PACK_LEN2(header_first_byte, 126 | mask_bit, msg_length)
    else:
        header = PACK_LEN3(header_first_byte, 127 | mask_bit, msg_length)

    if use_mask:  # pragma: no cover
        mask = random.randrange(0, 0xFFFFFFFF)
        mask = mask.to_bytes(4, "big")
        message = bytearray(message)
        _websocket_mask(mask, message)
        if noheader:
            return message
        else:
            return header + mask + message
    else:
        if noheader:
            return message
        else:
            return header + message


def out(loop):
    return aiohttp.DataQueue(loop)


def parser(out):
    return WebSocketReader(out, 4 * 1024 * 1024)


async def bench():
    data = build_frame(b"thisisthebinaryframe" * 256, WSMsgType.BINARY)
    loop = asyncio.get_running_loop()
    out_queue = out(loop)
    parser_instance = parser(out_queue)
    print(
        timeit.timeit(
            "parser_instance.feed_data(data)", globals=locals(), number=1000000
        )
    )


asyncio.run(bench())

profiler

import random
import zlib
from typing import Any
import asyncio
import aiohttp
from aiohttp.http import WSMsgType
import cProfile
from aiohttp.http_websocket import (
    _WS_DEFLATE_TRAILING,
    PACK_LEN1,
    PACK_LEN2,
    PACK_LEN3,
    WebSocketReader,
    _websocket_mask,
)


def build_frame(
    message: Any,
    opcode: Any,
    use_mask: bool = False,
    noheader: bool = False,
    is_fin: bool = True,
    compress: bool = False,
):
    # Send a frame over the websocket with message as its payload.
    if compress:
        compressobj = zlib.compressobj(wbits=-9)
        message = compressobj.compress(message)
        message = message + compressobj.flush(zlib.Z_SYNC_FLUSH)
        if message.endswith(_WS_DEFLATE_TRAILING):
            message = message[:-4]
    msg_length = len(message)
    if use_mask:  # pragma: no cover
        mask_bit = 0x80
    else:
        mask_bit = 0

    if is_fin:
        header_first_byte = 0x80 | opcode
    else:
        header_first_byte = opcode

    if compress:
        header_first_byte |= 0x40

    if msg_length < 126:
        header = PACK_LEN1(header_first_byte, msg_length | mask_bit)
    elif msg_length < (1 << 16):  # pragma: no cover
        header = PACK_LEN2(header_first_byte, 126 | mask_bit, msg_length)
    else:
        header = PACK_LEN3(header_first_byte, 127 | mask_bit, msg_length)

    if use_mask:  # pragma: no cover
        mask = random.randrange(0, 0xFFFFFFFF)
        mask = mask.to_bytes(4, "big")
        message = bytearray(message)
        _websocket_mask(mask, message)
        if noheader:
            return message
        else:
            return header + mask + message
    else:
        if noheader:
            return message
        else:
            return header + message


def out(loop):
    return aiohttp.DataQueue(loop)


def parser(out):
    return WebSocketReader(out, 4 * 1024 * 1024)


async def bench():
    data = build_frame(b"thisisthebinaryframe" * 256, WSMsgType.BINARY)
    loop = asyncio.get_running_loop()
    out_queue = out(loop)
    parser_instance = parser(out_queue)
    pr = cProfile.Profile()
    pr.enable()
    for _ in range(1000000):
        parser_instance.feed_data(data)
    pr.disable()
    pr.create_stats()
    pr.dump_stats("feed.cprof")


asyncio.run(bench())

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 98.13%. Comparing base (1eb6519) to head (7fc3ffb).
Report is 1034 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #8736      +/-   ##
==========================================
- Coverage   98.14%   98.13%   -0.01%     
==========================================
  Files         107      107              
  Lines       34062    34065       +3     
  Branches     4029     4030       +1     
==========================================
+ Hits        33430    33431       +1     
- Misses        456      457       +1     
- Partials      176      177       +1     

Flag	Coverage Δ
CI-GHA	98.03% <100.00%> (-0.01%)	⬇️
OS-Linux	97.69% <100.00%> (-0.01%)	⬇️
OS-Windows	96.06% <98.07%> (-0.01%)	⬇️
OS-macOS	97.34% <98.07%> (-0.02%)	⬇️
Py-3.10.11	97.43% <98.07%> (-0.02%)	⬇️
Py-3.10.14	97.38% <100.00%> (+<0.01%)	⬆️
Py-3.11.9	97.62% <100.00%> (-0.01%)	⬇️
Py-3.12.4	97.62% <100.00%> (-0.12%)	⬇️
Py-3.12.5	97.21% <100.00%> (?)
Py-3.8.10	95.70% <94.23%> (-0.02%)	⬇️
Py-3.8.18	97.20% <96.15%> (-0.01%)	⬇️
Py-3.9.13	97.33% <94.23%> (-0.02%)	⬇️
Py-3.9.19	97.29% <96.15%> (-0.02%)	⬇️
Py-pypy7.3.16	96.90% <96.15%> (-0.01%)	⬇️
VM-macos	97.34% <98.07%> (-0.02%)	⬇️
VM-ubuntu	97.69% <100.00%> (-0.01%)	⬇️
VM-windows	96.06% <98.07%> (-0.01%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[patchback]

Backport to 3.10: 💔 cherry-picking failed — conflicts found

❌ Failed to cleanly apply 1b88af2 on top of patchback/backports/3.10/1b88af2c3f5e5f992d0015f90927fd9c3e00bef0/pr-8736

Backporting merged PR #8736 into master

1. Ensure you have a local repo clone of your fork. Unless you cloned it
   from the upstream, this would be your origin remote.
2. Make sure you have an upstream repo added as a remote too. In these
   instructions you'll refer to it by the name upstream. If you don't
   have it, here's how you can add it:

   $ git remote add upstream https://github.com/aio-libs/aiohttp.git

3. Ensure you have the latest copy of upstream and prepare a branch
   that will hold the backported code:

   $ git fetch upstream
   $ git checkout -b patchback/backports/3.10/1b88af2c3f5e5f992d0015f90927fd9c3e00bef0/pr-8736 upstream/3.10

4. Now, cherry-pick PR Improve performance of WebSocketReader #8736 contents into that branch:

   $ git cherry-pick -x 1b88af2c3f5e5f992d0015f90927fd9c3e00bef0

   If it'll yell at you with something like fatal: Commit 1b88af2c3f5e5f992d0015f90927fd9c3e00bef0 is a merge but no -m option was given., add -m 1 as follows instead:

   $ git cherry-pick -m1 -x 1b88af2c3f5e5f992d0015f90927fd9c3e00bef0

5. At this point, you'll probably encounter some merge conflicts. You must
   resolve them in to preserve the patch from PR Improve performance of WebSocketReader #8736 as close to the
   original as possible.
6. Push this branch to your fork on GitHub:

   $ git push origin patchback/backports/3.10/1b88af2c3f5e5f992d0015f90927fd9c3e00bef0/pr-8736

7. Create a PR, ensure that the CI is green. If it's not — update it so that
   the tests and any other checks pass. This is it!
   Now relax and wait for the maintainers to process your pull request
   when they have some cycles to do reviews. Don't worry — they'll tell you if
   any improvements are necessary when the time comes!

🤖 @patchback
I'm built with octomachinery and
my source is open — https://github.com/sanitizers/patchback-github-app.

[patchback]

Backport to 3.11: 💔 cherry-picking failed — conflicts found

❌ Failed to cleanly apply 1b88af2 on top of patchback/backports/3.11/1b88af2c3f5e5f992d0015f90927fd9c3e00bef0/pr-8736

Backporting merged PR #8736 into master

1. Ensure you have a local repo clone of your fork. Unless you cloned it
   from the upstream, this would be your origin remote.
2. Make sure you have an upstream repo added as a remote too. In these
   instructions you'll refer to it by the name upstream. If you don't
   have it, here's how you can add it:

   $ git remote add upstream https://github.com/aio-libs/aiohttp.git

3. Ensure you have the latest copy of upstream and prepare a branch
   that will hold the backported code:

   $ git fetch upstream
   $ git checkout -b patchback/backports/3.11/1b88af2c3f5e5f992d0015f90927fd9c3e00bef0/pr-8736 upstream/3.11

4. Now, cherry-pick PR Improve performance of WebSocketReader #8736 contents into that branch:

   $ git cherry-pick -x 1b88af2c3f5e5f992d0015f90927fd9c3e00bef0

   If it'll yell at you with something like fatal: Commit 1b88af2c3f5e5f992d0015f90927fd9c3e00bef0 is a merge but no -m option was given., add -m 1 as follows instead:

   $ git cherry-pick -m1 -x 1b88af2c3f5e5f992d0015f90927fd9c3e00bef0

5. At this point, you'll probably encounter some merge conflicts. You must
   resolve them in to preserve the patch from PR Improve performance of WebSocketReader #8736 as close to the
   original as possible.
6. Push this branch to your fork on GitHub:

   $ git push origin patchback/backports/3.11/1b88af2c3f5e5f992d0015f90927fd9c3e00bef0/pr-8736

7. Create a PR, ensure that the CI is green. If it's not — update it so that
   the tests and any other checks pass. This is it!
   Now relax and wait for the maintainers to process your pull request
   when they have some cycles to do reviews. Don't worry — they'll tell you if
   any improvements are necessary when the time comes!

🤖 @patchback
I'm built with octomachinery and
my source is open — https://github.com/sanitizers/patchback-github-app.