hmc-cs-mdrissi opened a new pull request #16466: URL: https://github.com/apache/beam/pull/16466
Follow up to this [pr](https://github.com/apache/beam/pull/15233). A couple months ago we loosen typing extensions to any typing extensions <4. Well, typing extensions release a version 4 that is backwards compatible with 3 as they no longer want to follow python versions. The old versioning of typing extensions was based on python version (3.10 meant python 3.10 features). So we should support version 4 at least. Typing extensions is about to release a version 5 too. The only backwards incompatible change is typing extensions version 5 drops support for 3.6 now that it is end of life. That should be safe as someone using 3.6 will automatically pick version 4 at most due to [python_requires](https://github.com/python/typing/blob/dea3e3eefb92748559fb8e2202f3e2d70a8d8517/typing_extensions/pyproject.toml#L25) line in typing extensions. I think backwards incompatible typing-extensions change except for EOL python is very unlikely as it would break typeshed/typing ecosystem. So I do not think beam needs to upper bound typing extensions. beam is only package out of 100s I depend on that upper bounds typing-extensions. The upper bound is currently blocking some usage of newer type features. I expect tensorflow/pytorch/numpy to all want to use TypeVarTuple in the near future because it allows proper tensor typing. That will be a feature of typing extensions 5. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
