BUG: Groupby median on timedelta column with NaT returns odd value (#… #57957
Conversation
JJLLWW
force-pushed
the
groupby-median-nat
branch
from
a132083 to
f9a6d3d
Compare
…andas-dev#57926) Handle NaT correctly in group_median_float64
JJLLWW
force-pushed
the
groupby-median-nat
branch
from
f9a6d3d to
44d56aa
Compare
…_datetimelike is true
| cdef float64_t median_linear(float64_t* a, int n) noexcept nogil: | ||
| cdef float64_t median_linear( | ||
| float64_t* a, | ||
| int n, |
There was a problem hiding this comment.
FYI these should be Py_ssize_t not int. int is limited to 2 ** 31 - 1 on most platforms, so this would cause undefined behavior if ever called with an n that exceeded that (probably unlikely within groupby...but still possible)
Happy to have that in a follow up PR
|
Thanks @JJLLWW |
|
@WillAyd - why the backport? I didn't think this was a regression. |
|
Ah OK - I thought we were doing bugfixes on that branch now. Didn't realize it was limited to regressions. Can just keep this on 3.x then |
|
This might be just me, but I like to keep patch versions as small as possible since bugfixes can induce new bugs and would like patch upgrade to be as safe as possible for users. But in the past, I know others have been more inclined to backport more than I prefer - cc @phofl for any thoughts. |
|
Agree with @rhshadrach we backported fixes for new features that were introduced in the version before but generally not much else |
Ah, I think that's where I remember a disagreement - thanks. And yea, I've come around to doing bug fixes for new features. |
…57926)
Handle NaT correctly in group_median_float64
doc/source/whatsnew/vX.X.X.rstfile if fixing a bug or adding a new feature.