Add bench_async_func() #124
Conversation
pyperf/_runner.py
Outdated
| # use fast local variables | ||
| local_timer = time.perf_counter | ||
| local_func = func | ||
| loop = asyncio.get_event_loop() |
There was a problem hiding this comment.
Would it make sense to attempt reimplementing asyncio.run() here, but only measure run_until_complete() timing? I'm talking about creating a fresh event loop each time, and shutdown asynchronous generators and the default executor. Well, same things than asyncio.run().
There was a problem hiding this comment.
I use asyncio.run() instead of reimplement it.
|
I tested the example on my Fedora 35 with Python 3.10. I get: That's interesting :-) Using |
|
Oh, sadly pyperf currently still supports Python 3.6: https://pyperf.readthedocs.io/en/latest/run_benchmark.html#install-pyperf |
pyperf/_runner.py
Outdated
| return dt | ||
|
|
||
| import asyncio | ||
| try: |
There was a problem hiding this comment.
I would prefer to testing hasattr(asyncio, 'main'), since main() can raise AttributeError.
doc/api.rst
Outdated
|
|
||
| .. note:: | ||
| If the ``func()`` leaves any tasks, it affects to next iteration. | ||
| So you should care about task leak for reliable benchmark. |
There was a problem hiding this comment.
This note is now outdated, since a new event loop is created at each iteration, no?
|
With the proposed implementation (measure time inside asyncio.run and not outside), I agree that the implementation increases the accuracy and so it's worth it to add a new method ;-) |
|
I completed the changelog in commit 0814a4a. |
|
This change is now part of the just released pyperf 2.3.1. |
Ref #121.