Split library into two parts: DB-API 2.0 compatible and Asyncio style

[runfalk]

I am very interested in using this library since type support in psycopg2 is awful to work with. However, some projects still use Python 2.7, and will for another 4 years. Would it be possible to split asyncpg into two parts, one with an asynchronous API, usable with Python 3.5 and above, and one synchronous API that can be compatible across both 2.7 and 3.5 and above?

My proposal is a DB-API 2.0 (PEP 249) compatible API that can be utilized by the asyncio part of the library. The benefits is that applications that do not need asynchronous calls can use the same underlying library as those who do. With this code that doesn't rely on co-routines can use the type support provided by asyncpg.

I haven't used this project yet, or dived into its internals, and don't know whether it is feasible at all to do this. I am just venting my ideas.

[1st1]

We're considering to do this. I can't promise that it will happen very soon, but maybe in a month or couple.

[necaris]

@1st1 count me in as another volunteer to try and make this happen! From reading your comments on HN it sounds like it's feasible if non-trivial, and I'd like to help!

[auvipy]

I have planned to work on django orm part based on asyncpg https://github.com/auvipy/django-asyncpg

[mumbleskates]

Looking forward to / hoping for progress on this!

[Skyross]

@1st1. Thanks for an awesome and performant library. Have any information about progress on this ticket?

[1st1]

We are going to do it soon, but have no concrete ETA.

[gwerbin]

I'd like to contribute here as well. Is there some kind of roadmap or list of tasks to be done? NB I am not a professional developer, so I don't want to commit to something I won't be able to handle.

[piero10]

thank for your work, I hope I could use this lib in my python 2.7 soon....

[auvipy]

@piero10 2.7 is legacy... py3k is present and future.

[runfalk]

@auvipy: Even if Python 3 is the current Python version there is still a ton of code out there that is not trivial to refactor. Maybe @piero10 is heavily invested in a particular library which hasn't been ported yet? Statements like that doesn't benefit anyone.

Python 2 is still officially supported until 2020. Dropbox, who currently employs Guido van Rossum, is still partially on Python 2 for crying out loud.

Also, please keep the discussion related to the feature request.

[1st1]

thank for your work, I hope I could use this lib in my python 2.7 soon....

Sorry, Python 2.7 isn't going to happen for asyncpg; I can tell this with almost 100% certainty. It's too much work for us for a relatively small group of potential users (psycopg2 still exists and works just fine for many).

Speaking on the subject: splitting asyncpg in two parts will happen eventually, when we need it for edgedb (sooner than later).

[piero10]

((( I'm so sorry....

[ipfans]

It's almost one year passed, so can we talking about how to reach the goal?