Async support

[cdce8p]

Currently HAP-python uses threads which, at least on a larger scale, does impact performance quite a bit. Are there plans to change to an async framework, like asyncio in the future?

Unfortunately this would also require a more recent min python version, to take full advantage of the new async syntax introduced with py3.5.

[ikalchev]

This is actually a very good idea, which I think we might pursue. I was recently thinking on the threads starting to be an issue and it looks like asyncio can fit the bill nicely.

Dropping 3.4 will not be a huge issue I believe. I had similar requests before, but no good reasons to do it.

I am currently busier at work than what is usual, but I expect things to be back to normal from this weekend onwards. I will definitely spend some time on this soon and try to make some first steps. Of course, you are welcome to do further suggestions and PRs if you want.

On a side note, @jslay88 started a chat for the project on discord I think recently which I hadn't had the time to "take over" yet. I will try to kick this off as well.

Best,
Ivan

[jslay88]

https://discord.gg/BhxZ4Dy

So lonely :)

[cdce8p]

Good to here that. I'm happy to help, although I just started watching talks and tutorials about this two weeks ago 😅 I didn't wanted to start before knowing if this is something you want. It's quite some work and will require major changes to the code base.

Maybe we can coordinate efforts on the discord channel than

[ikalchev]

FYI started transforming the Accessory/Bridge run methods to use asyncio (seemed easiest starting point)

[cdce8p]

👍 Could you open a [WIP] PR for it?

[ikalchev]

Yes, good point, I am doing this now since I moved to dev and master branches. I hadn't had anything significant to push recently due to lack of time.

[ikalchev]

@cdce8p have you used aiohttp https://aiohttp.readthedocs.io/en/stable/

I am considering it as the next step (after the Accessories run method).

[cdce8p]

@ikalchev Not yet, since I didn't have/had any experience with networking prior to working with HAP-python. aiohttp seems to be the right choice though.

Did you open a WIP PR? That would really help. Especially to know what will be changed and to give suggestions.

[ikalchev]

I will, I just need to try something before the PR, I am learning asyncio as well

[ikalchev]

By the way what is WIP?

[cdce8p]

Work in progress

[cdce8p]

@ikalchev @thomaspurchas What do you guys think about creating a separate branch, maybe dev-async to support the async conversion. For now all async related PR's would be based against this branch, new features from dev merged via PR to dev-async and when we are ready we merge it into dev (later master) with the complete async support.

That way we don't need to guaranty that every version of the conversion works. Ultimately this would allow us to work more freely than if we were bound to just dev.

@thomaspurchas I don't think to merge #76 into a feature branch (ultimately a PR) is a wise idea here.

@ikalchev If you want, I could support this from an administrative standpoint as well.

[thomaspurchas]

Seems like a reasonable thing to do. #76 is just doing some work that is related to #74, porting the rest of the accessories. I don’t really care where it ends up.

…

On 2 Apr 2018, 01:38 +0100, cdce8p ***@***.***>, wrote: @ikalchev @thomaspurchas What do you guys think about creating a separate branch, maybe dev-async to support the async conversion. For now all async related PR's would be based against this branch, new features from dev merged via PR to dev-async and when we are ready we merge it into dev (later master) with the complete async support. That way we don't need to guaranty that every version of the conversion works. Ultimately this would allow us to work more freely than if we were bound to just dev. @thomaspurchas I don't think to merge #76 into a feature branch (ultimately a PR) is a wise idea here. @ikalchev If you want, I could support this from an administrative standpoint as well. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or mute the thread.