reimplemented the MemQueue class to avoid unordered inserts #295
Conversation
| quotient, _ = divmod(mem_queue_duration, queue_duration) | ||
| assert quotient < 30 |
There was a problem hiding this comment.
Arbitrary number. It's just set to make sure the class does not get slower over time when we change it
| item = "small stuff" | ||
| queue = MemQueue( | ||
| maxmemsize=get_size(item) * 2 + 1, refresh_interval=0.1, refresh_timeout=30 | ||
| ) | ||
| max_size = 0 |
There was a problem hiding this comment.
I think a comment would help to understand what's going on in the test. I vaguely understand it, but not 100% sure and not 100% sure what is the actual problem.
I can add comments, but hardly refactor because it's based on the |
I see. Maybe pointing out every change to the original implementation can help? |
this is what I did with the two comments (1/ 2/) in the cod, we have two changes -- the original method is here https://github.com/python/cpython/blob/3.10/Lib/asyncio/queues.py#L111 |
artem-shelkovnikov
force-pushed
the
tarekziade/doc
branch
from
996d30d
to
8c37a63
Compare
artem-shelkovnikov
force-pushed
the
tarekziade/doc
branch
from
8c37a63
to
3aa23c4
Compare
I closed #281 because I figured out that we did not have a race condition.
See #281 (comment)
But then I realized the implementation can lead to unordered inserts in the queue. So this new patch reintroduces the change which does respect the ordering.