Add support to fixed block size recording with pulseaudio

[mpariente]

To enable the mic to return a fixed number of frames at each call I introduce _record_chunk which will perform what record used to perform without reshaping the data at the output, record_chunk which will perform what record used to perform and record which will call _record_chunk and perform the buffering internally so that every call, the same number of frames is returned. I also implemented flush to empty the last _pending_frame.

[mpariente]

I agree with your comments and I'm working on them. Just one quick question, can we add a .gitignore to the repo? I'm recompiling soundcard at every change to make tests and it would be nice if we don't have to delete dist, build and .egg-info before every commit.

[bastibe]

Just one quick question, can we add a .gitignore to the repo?

Absolutely! Although your particular problem might be solved with pip install -e ., which will link the soundcard into your packages instead of copying, thus applying every change immediately.

[mpariente]

• I started by suppressing completely the while loop in _record_chunk, but it turned out more practical and more efficient to turn it into a while 1 so we don't have to check if a chunk was returned of not in record

• record now returns whatever pulseaudio is returning if numframes=None, which I set as the default. Should None be default? I think it's not a bad idea

• record_chunk doesn't exist anymore

Btw : Sorry for the three commits, I made dumb mistakes on the code.

[mpariente]

It's a pleasure !

[bastibe]

Thank you very much for this contribution!

I'll try to implement the same change in the other backends, and then publish a new version on PyPi. (unless you want to try on another platform.)