REF: smarter NaN handling in remove_unused_levels() #18438
Conversation
toobaz
force-pushed
the
ref_unused_nan
branch
from
9d3c436 to
d08fd29
Compare
Codecov Report
@@ Coverage Diff @@
## master #18438 +/- ##
==========================================
- Coverage 91.35% 91.33% -0.02%
==========================================
Files 163 163
Lines 49691 49688 -3
==========================================
- Hits 45397 45385 -12
- Misses 4294 4303 +9
Continue to review full report at Codecov.
|
Codecov Report
@@ Coverage Diff @@
## master #18438 +/- ##
==========================================
- Coverage 91.35% 91.33% -0.02%
==========================================
Files 163 163
Lines 49691 49688 -3
==========================================
- Hits 45397 45385 -12
- Misses 4294 4303 +9
Continue to review full report at Codecov.
|
jreback
added
Clean
Missing-data
|
thanks! @toobaz separately, it might be nice to see (on a bigger MI), how much time this actually takes, if its small then it might be theoretically feasible to actually do this (or at least check whether we need to do this). |
Do you mean to implicitly do this (e.g. when slicing)? |
|
yep - seeing if this is feasible |
|
But do we actually want this? E.g. a categorical does not loose categories when it is sliced. |
|
it’s actuaiky confusing to users |
git diff upstream/master -u -- "*.py" | flake8 --diffSorry for the bad timing @jreback , only after you merged #18426 this simpler way to proceed came to my mind.