ENH: implement Timedelta._as_unit #47162
Conversation
jreback
added
Timedelta
| @@ -954,7 +954,7 @@ void pandas_timedelta_to_timedeltastruct(npy_timedelta td, | |||
| case NPY_FR_s: | |||
| // special case where we can simplify many expressions bc per_sec=1 | |||
|
|
|||
| per_day = 86400000LL; | |||
| per_day = 86400LL; | |||
There was a problem hiding this comment.
Did this fix a bug somewhere?
There was a problem hiding this comment.
turns out not to be relevant, could actually get rid of per_day here entirely
| assert rt.value == td.value | ||
| assert rt._reso == td._reso | ||
|
|
||
| def test_as_unit_overflows(self): |
There was a problem hiding this comment.
Just confirming aloud, based on the multiplication operation we can't get to a state where underflowing occurs (and therefore can't test that)?
There was a problem hiding this comment.
this is all in integers, so i dont think underflows are an issue
| td._as_unit("ms", round_ok=False) | ||
|
|
||
| def test_as_unit_non_nano(self): | ||
| # case where we are going neither to nor from nano |
There was a problem hiding this comment.
Not really sure what neither to nor means (if it's an idiom)
There was a problem hiding this comment.
"not to nano" and "not from nano"
Yes _as_unit is the one that will eventually be user-facing. Just didnt want to make it public yet in case we want to change the name |
doc/source/whatsnew/vX.X.X.rstfile if fixing a bug or adding a new feature.