Make get_loc with nan for FloatIndex consistent with other index types #39382
Conversation
phofl
added
the
Indexing
|
i think test failures should be fixed by merging master |
phofl
changed the title
| @@ -57,7 +57,10 @@ cdef class {{name}}Engine(IndexEngine): | |||
| with warnings.catch_warnings(): | |||
| # e.g. if values is float64 and `val` is a str, suppress warning | |||
| warnings.filterwarnings("ignore", category=FutureWarning) | |||
| indexer = values == val | |||
| if cmath.isnan(val): | |||
There was a problem hiding this comment.
probably should use tempita to only do this check for FloatEngine
might be able to use something in the cnp namespace for the isnan check
There was a problem hiding this comment.
If we do this only for float, test_slice_locs_na raises from /indexes/test_base.py
Actually we don't need imports I think. Changed it
| @@ -57,7 +57,10 @@ cdef class {{name}}Engine(IndexEngine): | |||
| with warnings.catch_warnings(): | |||
| # e.g. if values is float64 and `val` is a str, suppress warning | |||
| warnings.filterwarnings("ignore", category=FutureWarning) | |||
| indexer = values == val | |||
| if val != val: | |||
| indexer = values != values | |||
There was a problem hiding this comment.
something like
{{ if dtype in ["float64", "float32"]}}
if val != val:
indexer = np.isnan(values)
else:
indexer = values == val
{{else}}
indexer = values == val
{{endif}}
There was a problem hiding this comment.
I did the same thing yesterday, this caused the test to fail. Weird...
|
@phofl can you merge master. ok with fixing this here or in followup to raise. |
|
This is raising now if nan is not in Index |
|
very nice! this may close other issues (hard to search for but looking for indexing and nan) |
Have not added a whatsnew because I am not sure if 2 is expected here. You could also make a point for retruning nan. But to handle these cases consistently, we have to check for method here