REF: stop allowing iNaT in TimedeltaBlock methods #27411
Conversation
|
|
||
| # Note: we have to do do `astype` and not view because in general we | ||
| # have float result at this point, not i8 | ||
| return result.astype(values.dtype) |
There was a problem hiding this comment.
why are you not letting this then fall thru, e.g. if there is masking here?
There was a problem hiding this comment.
If we just return result here, at this point it is in general float64. In the calling function we're going to do a result.view(values.dtype), which will be incorrect
There was a problem hiding this comment.
right my point is to do this later in the function after masking and checking for 0-len.
There was a problem hiding this comment.
just tried that, breaks because it ends up double-masking
There was a problem hiding this comment.
i wouldnt hang up on this too much, since we're going to have to move this to earlier in the call stack in #27428
There was a problem hiding this comment.
ok, if you can clean this up later would be great.
jreback
added
Compat
|
|
||
| # Note: we have to do do `astype` and not view because in general we | ||
| # have float result at this point, not i8 | ||
| return result.astype(values.dtype) |
There was a problem hiding this comment.
right my point is to do this later in the function after masking and checking for 0-len.
|
gentle ping; this is a blocker for some pretty nice internals simplification |
|
Seems reasonable to me. |
Working towards minimizing the amount of special-casing we do within internals