Support for 3.0?

There’s a new graphene alpha release targeting graphql-core-next. Just curious: are plans yet to support this, or will it be delayed until the graphene interface reaches a beta/rc/stable version?

[DoctorJohn]

Just to let you folks know, I started working on an PR. Got it down to 8 failing tests, 5 of which are related to promises/dataloaders.

[zsiciarz]

We at Emplocity are willing to collaborate on compatibility with graphene 3. We haven’t contributed to the graphene ecosystem before, but we’re heavy users of the 2.x versions of libraries and are quite familiar with them. I took a look at @DoctorJohn’s PR and got the tests to run (not pass, that would be too much to ask 😉 ). There appears to be a few points that need a decision:

• Python 2 support. I think we can safely drop it, following upstream graphene.
• Batching. That’s the part I’m not very familiar with (we’re using eventlet, which unfortunately causes a ton of issues with promises used by graphene). The tests related to batching don’t error out on missing APIs, they just return wrong results. @DoctorJohn pointed out graphene dropped (?) promises in favor of asyncio and yet the DataLoader documented here uses promises. However, there’s a new asyncio-based approach, see: https://github.com/graphql-python/graphene/pull/1190. I guess we should follow that approach.
• Dependency versions. I needed to pin SQLAlchemy to < 1.4 to get the tests to run. I’ll have a look at what changed there and how graphene-sqlalchemy is using the removed/changed APIs.

Our WIP branch based on DoctorJohn’s work is here, for anyone interested: https://github.com/Emplocity/graphene-sqlalchemy/commits/graphene-v3