jorisvandenbossche commented on issue #40153:
URL: https://github.com/apache/arrow/issues/40153#issuecomment-1956747712

   > Among the remaining failures, `test_dictionary_to_numpy` should probably 
be fixed in the test itself, though I'm not sure.
   
   Looking at the traceback, it seems this is actually failing in the code 
itself, here:
   
   
https://github.com/apache/arrow/blob/b51c318122ff5db1c6c4c70a69b0c804f9e31704/python/pyarrow/array.pxi#L1588-L1589
   
   We do a take with int64 indices, and for some reason that seems to fail 
inside numpy's internals when trying to cast the int64 indices to int32 
(because it's not "safe" casting).  
   The numpy `take` docs don't say anything about a limitation of the indices 
array to int32 on 32bit platforms, so not sure if that's something we should 
report to them (or if we need to cast to int32 up front when needed).


-- 
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]

Reply via email to