oliversheridanmethven commented on issue #53954: URL: https://github.com/apache/arrow/issues/53954#issuecomment-6572214416
@mohamd764 - I see the restrictions surrounding the expressiveness in the package metadata for such a conditional constraint, and while some local build systems support this sort of conditional package requirement, the standard shipping mechanism might not be as expressive. Similarly, I see the point raised by @jorisvandenbossche, that much of the code does not require numpy. > various parts of pyarrow can be used without having numpy installed This is a very interesting point. From my perspective, it seems that in lieu of some sort of optional dependency and a `pyarrow[numpy]`, then whether 1% or 99% of the code requires numpy or not in order to be used is almost immaterial. If some part of it requires numpy, then that is the part that gets noticed, and numpy should be a dependency. For managing numerous repos which in turn have numerous dependencies, it becomes unscalable to read all the docs for optional/conditional dependencies for each package. The only scalable solutions are the standard resolution mechanisms in a package manager, (uv, poetry, etc). If the desired expressiveness is not possible, I would prefer to have a package "over specify" and for me to pick up the package bloat and it work in all scenarios, rather than underspecify and for me to not have an entirely working solution. (I recognise though that others might be more adverse to bloat). -- 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]
