Codecov Report

Merging #20481 into master will decrease coverage by <.01%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##           master   #20481      +/-   ##
==========================================
- Coverage   91.85%   91.84%   -0.01%     
==========================================
  Files         152      152              
  Lines       49231    49230       -1     
==========================================
- Hits        45220    45215       -5     
- Misses       4011     4015       +4

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 4fb963b...d13ae37. Read the comment docs.

should also be on nansum. how did you come to change this? e.g. what broke?

oh, now I don't see it so clearly and I'm probably wrong. Nothing has been broken, I was checking the code and I noticed that bottleneck_switch seems to do nothing when the function is nansum or nanprod according to _bn_ok_dtype, I thought it would be an unnecessary call.

@mmngreco no it IS unecessary since 0.22 (as we compute sum/prod inside pandas only). was just wondering how you detected it.

happy to have you remove the one on nansum as well. I don't think there is a test for this that would break, but if you have something would be great.

I have not found any test for bottleneck_switch, I understand that this would be necessary for this case. I haven't written anything yet. I will try in the next few days as I am not sure how to do it.

Thanks @mmngreco