[Python] Deprecate `pyarrow.gandiva`

[pitrou]

Describe the enhancement requested

I don't think pyarrow.gandiva is getting a lot of usage. The last non-trivial change was made 2 years ago by @niyue . It is not packaged on PyPI, but it's part of the pyarrow-all package on conda-forge which is not downloaded much (35K downloads for 6 months compared to 3M for pyarrow-core).

Deprecating and then dropping pyarrow.gandiva would simplify some CI build and remove some light maintenance burden from the team. It would also make it easier to later move Gandiva C++ to a separate repository.

Component(s)

Python

[pitrou]

Thoughts @AlenkaF @raulcd @kou ?

[raulcd]

I am in favor of this. Not being part of the provided wheels to PyPI helps on the decision.

As a concrete plan we could deprecate it for 24.0.0 and drop it for 25.0.0.

We should send an email to the dev mailing list to give visibility to this issue in case there are any concerns raised.

[rok]

Agreed with all said above.

[kou]

We should send an email to the dev mailing list to give visibility to this issue in case there are any concerns raised.

+1

[pitrou]

ML thread: https://lists.apache.org/thread/k83vy5p4hcz638z5bo09c4zdj9b5nd0b

[AlenkaF]

There are no responses/concerns raised via ML AFAICT. I can try to get a PR up today to get the deprecation part in for 24.0.0 so we can drop it in a later version.