typed-ast end of life?

[srittau]

I suggest we officially end of life typed-ast:

• Officially stop support after Python 3.10 (i.e. close Compilation fails on Python 3.11 #167).
• Add recommendation to use ast on Python 3.8+ in the package's long_description and more prominently in the README.
• Add < 3.11 to python_requires.

We can keep maintaining it until 3.7 (or 3.10) reaches end of life, and improved ast is generally available.

[hauntsaninja]

Bullet point 2 sounds good to me.

I'm opposed to point 1... I think typed_ast will continue to have value until 3.7 reaches EOL, and I'm happy to accept PRs from people fixing real issues like #167 / I'm happy to make releases.

Open source is always best effort, so if in a year no effort has surfaced to get things working on 3.11, we can do bullet point 3.

[srittau]

Adding python_requires is about setting expectations. Sure, if we receive a PR to support 3.11, we can lift that restriction, but without it, people expect that typed_ast will work on 3.11. By setting it, we communicate clearly that it doesn't.

My amended suggestion:

• Keep Compilation fails on Python 3.11 #167 open for now.
• Add recommendation to use ast on Python 3.8+ in the package's long_description and more prominently in the README.
• Add < 3.11 to python_requires, until we receive a PR that lifts that restriction.
• Archive this project after 3.7 reaches end of life.

[JukkaL]

This seems fine to me, if we can wait until 3.7 EoL (and are open to making a 3.11 compatible release, in case somebody makes a PR).

[vstinner]

I hit #167 issue on Python 3.11 while trying to port mypyc to Python 3.11 (to test my PR): python/mypy#11652

I proposed a simple fix to port typed_ast to Python 3.11: #176

Right now, mypyc still requires typed_ast, so making it compatible with Python 3.11 solves a practical issue ;-)

[srittau]

Since @vstinner fixed #167 and we now suggest to use ast, my proposal is now basically down to "archive this project after 3.7 reaches end of life".

[vstinner]

Well, you can still plan the death of the project if you want, but in that case, please help users (like mypy) to migrate to something else ;-) typed_ast seems to be quite popular: https://libraries.io/pypi/typed-ast

• Dependent packages: 277
• Dependent repositories: 5.18K

[srittau]

We will most likely keep typed_ast around for a bit even after 3.7 reaches end of life, but upgrading to ast (Python 3.8+) should be fairly trivial. Current users often still support Python 3.6 or 3.7.

[JelleZijlstra]

I don't know what the packages depending on typed-ast are using it for, but typed-ast can't parse Python 3.8+, so they probably should look for something else.

mypy and black need typed-ast only for parsing under old versions of Python, and should be able to drop usage when they drop those old Python versions.

[srittau]

(Also, archiving the project does not mean that typed-ast is gone from pypi, just that it won't receive any updates.)

[hauntsaninja]

my proposal is now basically down to "archive this project after 3.7 reaches end of life".

To avoid user confusion, I summarised this in #179

[JelleZijlstra]

I'm going to keep posting here to avoid bringing confusion to the other issue :)

I realized that tools that want to consume stub files and support only 3.6+ already don't need to use typed-ast, because ast.parse in 3.6 should support everything stub files support. This affects my typeshed-client project: JelleZijlstra/typeshed_client#30.

[CAM-Gerlach]

@hauntsaninja @JelleZijlstra Just as a quick suggestion, to further avoid user confusion :) , I'd suggest locking #179 as resolved so only collaborators can post, and bolding/unparenthesizing your redirect at the bottom. In my experience with similar issues, unfortunately all too many users miss such things, so its best to be proactive.