wait for transport buffer flush about aiozmq HOT 6 CLOSED

Sorry for delay.
I'll take a look this weekend.

from aiozmq.

Not a problem at all. You're really doing a great job with this lib!

from aiozmq.

Please explain me why do you need to wait for transport's write buffer flush?
Looks like you are seeking for pause_writing/resume_writing pair.
I'we implemented it in stream branch as part of creating ZmqStream class for giving more human interface that transports/protocols.
The code is almost ready for merging but I need to make several tests for 100% coverage.
I've postponed my work on streams for sake of other my projects but can force development if the stream branch is that you needed.

from aiozmq.

I run a task with run_until_complete, which issues some writes. The task is done after the last write, so the event loop stops running and and some messages don't get written, as the write callback can't happen when the loops stops. As the loop gets other tasks to run later, I don't want to call close() and wait for connection_lost(), but still I want all buffers flushed before the task is done and the loop suspended.

from aiozmq.

Aaah, understood.
Sorry, I don't want to do anything to fix your particular requirement.
The reason is: ZmqTransport has modeled after asyncio ones.
Neither of asyncio transports: TCP, UDP, SSL, Unix, pipe -- doesn't have any user callback to signal that data has been moved from python buffer to OS one. In the most cases (regular usage with run_forever or for client sockets) that doesn't make sense.
Twisted's protocols has no such callback also.

I'll add the callback only after asyncio will do it.

Good news for you: probably you don't need for callback at all.
ZeroMQ executes separate IO thread for writing and has reasonable big default buffer sizes.
If your ZMQ socket counterpart consumes ZMQ messages quickly enough you simply never get situation when ZmqTrasport will need to use own buffer.

from aiozmq.

ok. thank you anyway. perhaps I raise the issue for asyncio

from aiozmq.