Rewritten library using asyncio

[albaintor]

Hi,

I rewrote the library using asyncio instead of threads. I had issues with the old behaviour and it was not suitable to use with an async program.

I tested it successfully on my side. Of course there are some breaking changes :
Instead of calling wait method, we use the connect method directly like this :

chromecast=pychromecast.get_chromecast_from_host((host, 8009, None, None, None))
await chromecast.connect(timeout=5)

As you can see the code is much more simpler as a lot of work is performed directly in standard wrappers of asyncio.

[elupus]

Sorry pressed wrong button. I like the effort, but i wonder how we can handle this in a non breaking way for current users of the library.

[albaintor]

There is indeed some breaking changes but not so much, only those 2 :

• the initialization of the connection should be replaced with an await prefix in an async method
• all the calls to commands should be prefixed with await in an async method

The rest remains unchanged, including all the registrations to the listeners.

This is a great benefit I think especially for HA which works only with a single event loop in a single thread.
I experienced random deadlocks with the current implementation with events / threading.

Also the new implementation is more efficient in performances / CPU usage.

Maybe you could create a forked branch (with dedicated releases, eg beta 15.0) in your repository that could live in parallel, awaiting for users to migrate to the new implementation.

[elupus]

I just wanted this to get better support for happyeyeballs connection of sockets. Are you still interested in trying to get this further?

[elupus]

There is no guarantee this data is the full payload. It could haven been split in chunks. So this need to accumulate as much as it need.

I think a standard asyncio task with just recv calls would be easier. It would also reduce the rather huge diff on this file.