CLN: various related to numeric indexes #41615
Conversation
| else: | ||
| raise NotImplementedError(f"wrong dtype {dtype}") | ||
|
|
||
| return Index(values, dtype=dtype, name=name) |
There was a problem hiding this comment.
In #41153 we'll only need to change this to return NumericIndex(values, dtype=dtype, name=name) after this.
| @@ -292,21 +296,40 @@ def makeBoolIndex(k=10, name=None): | |||
| return Index([False, True] + [False] * (k - 2), name=name) | |||
|
|
|||
|
|
|||
| def makeNumericIndex(k=10, name=None, *, dtype): | |||
| dtype = pandas_dtype(dtype) | |||
There was a problem hiding this comment.
do we expect non-numpy dtypes here?
There was a problem hiding this comment.
I’ve added an assert for that here.
There was a problem hiding this comment.
my idea was to just use dtype = np.dtype(dtype)
There was a problem hiding this comment.
>>> np.dtype("Int8") # capital Int
DeprecationWarning: Numeric-style type codes are deprecated and will result in an error in the future.
np.dtype("Int8")
dtype('int8')So I think we should hold on this until the above raises IMO.
There was a problem hiding this comment.
this will not ever raise as these are pandas dtypes
There was a problem hiding this comment.
@jbrockmendel raises a valid point; if u r expecting only numpy dtypes then why are pandas dtypes passed?
There was a problem hiding this comment.
this will not ever raise as these are pandas dtypes
assert isinstance(dtype, np.dtype) on line 304 ensures that it will raise if it is a pandas dtype.
if u r expecting only numpy dtypes then why are pandas dtypes passed?
I'm not intending to pass pandas dtypes here, but I'm using pandas_dtype to avoid an ambiguity: The string "Int64" in Pandas means Int64Dtype, but in numpy it means np.int64. So by using dtype = pandas_dtype(dtype), then an assert(dtype, np.dtype), we ensure the correct dtype..
pandas/_testing/__init__.py
Outdated
|
|
||
| if dtype.kind == "i": | ||
| values = list(range(k)) | ||
| elif dtype.kind == "u": |
There was a problem hiding this comment.
can u use np.arrange in these i and u cases
or make the i and u cases the same structure ; they are completely different now
There was a problem hiding this comment.
I've made a better version.
topper-123
force-pushed
the
clean_various_numeric_index
branch
from
6d672ba to
c5dbfa4
Compare
topper-123
force-pushed
the
clean_various_numeric_index
branch
from
c5dbfa4 to
02f54b2
Compare
|
Ping? Np.dtype("Int64") doesn’t raise, so I do think using pandas_dtype + asserting a numpy dtype is better than np.dtype. |
|
thanks @topper-123 |
A few bits to make #41153 smaller.
Also fixes an issue with the current implementation of NumericIndex._engine_type:
The implementation currently uses
self.dtype.type, which is a problem when we are given platform-specific dtypes, e.g.np.intc.For example:
So the _engine_type dict should use np.dtype(np.int64) etc., so avoid such problems.