fix: big number to handle NULL as it did in the past

[mistercrunch]

context: #9313

[etr2460]

can we add a unit test to this PR as this was a regression that was recently introduced?

[mistercrunch]

I can write a unit test covering get_data later tonight

[etr2460]

ok. I think i'd feel comfortable accepting then once you add the manual test plan and some screenshots

however, a unit test should fast follow

[willbarrett]

I'm working on a unit test for this PR now.

[willbarrett]

@etr2460 unfortunately I was unable to come up with a comprehensive unit test by the end of my day. This will have to wait for @mistercrunch to free up again this evening. Apologies.

[mistercrunch]

Definitely was worth taking the time to write the unit test. Pandas is doing some funky things here in pivot_table. While clearly np.isnan(np.sum(np.nan)) == True df.pivot_table({...}, aggfunc=np.sum) of a column that has only None returns 0.

Anyhow. This here works, I picked np.min as no aggregates apply (already at the right grain) and with np.min a None leads to nan, which works well in the end (same as before the original PR). I tested this same unit test on before this PR and it fails (as expected, with the 0 value in place of the None/nan/NULL), and tested on before the a commit prior to the PR that created the problem 753aeb482991f0dd3f36373f6d248ce277c67949 and this unit test passes. This gives pretty high confidence that None / NULL will be treated the same as before.

[codecov-io]

Codecov Report

Merging #9314 into master will not change coverage by %.
The diff coverage is n/a.

@@           Coverage Diff           @@
##           master    #9314   +/-   ##
=======================================
  Coverage   59.08%   59.08%           
=======================================
  Files         374      374           
  Lines       12205    12205           
  Branches     2989     2989           
=======================================
  Hits         7211     7211           
  Misses       4815     4815           
  Partials      179      179           

---

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 f1370c5...06eb8ea. Read the comment docs.

[etr2460]

makes sense to me, thanks for adding the unit test!

I appreciate the quick response to the regression called out in the revert PR

[villebro]

Makes sense, LGTM. It seems one really has to be careful when applying Pandas transformations.

[etr2460]

@mistercrunch is this good to merge now?