[GitHub] [arrow] jorisvandenbossche commented on pull request #11606: [Python] Experiment with pinning numpy using oldest-supported-numpy dependency

[GitBox <gi...@apache.org>]

jorisvandenbossche commented on pull request #11606:
URL: https://github.com/apache/arrow/pull/11606#issuecomment-984636005


   I suppose it is not specific to Python 3.7, but due to a combination of versions of other packages that get installed. One guess is that because of having an older pandas version (1.0.5) but the latest numpy (1.21.3), numpy raises some warnings that are not only fixed in newer pandas versions. 
   When using pandas 1.0.5 to test, we probably should also install the numpy version that was releases around the same time as pandas 1.0.5, to ensure compatibility between numpy and pandas. 
   
   For Python 3.6 this doesn't happen because an older version of numpy gets installed (I suppose there are no wheels for the latest numpy release for python 3.6)


-- 
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: github-unsubscribe@arrow.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org