Fixed bug in reading nested vectors of vectors

[iguinn]

Recent merge broke nested VoV. This is a fix for that
We should add some tests for this feature; I think including some evt tier data in the test data would help with this (legend-exp/legend-testdata#21)

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 78.34%. Comparing base (782f566) to head (7a6e180).
Report is 11 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #115      +/-   ##
==========================================
+ Coverage   78.05%   78.34%   +0.28%     
==========================================
  Files          46       46              
  Lines        3368     3362       -6     
==========================================
+ Hits         2629     2634       +5     
+ Misses        739      728      -11     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[gipert]

Mmmh I'm not sure I'm completely comfortable with this. I think we should try to figure out why that reference is not unique...

[iguinn]

This was popping up while reading in the VoV, so there wasn't some sort of external reference causing this. Somehow another reference must be popping up in the process of reading out the evt tier.

I will add that this is likely to flare up during iterative use. In an iterator loop, it is going to be quite reasonable to view_as a table, and the reference in that view will exist when reading in the next chunk of entries.

[iguinn]

Had a look at this. The reason the reference was not unique is that I was accessing it with the iterator, and view_asing it. This is an old, known problem that doesn't have an easy solution. One thing that works is to just del the pandas/awkward view at the end of the loop block, but I think people will have trouble with that. Otherwise, the nested VoV thing was fixed by the other line that was changed.