STYLE: dont' use pd.api.types anywhere in tests

[jreback]

example here: https://github.com/pandas-dev/pandas/pull/37367/files#diff-69c698518ca64a05f8b137ff9bb5445ce9fd21139b396e9b0d4ce7338b2a3984R132

we should just ban this with a precommit style rule (and import things directly). the doc-strings also should be updated to do the import once.

grep -r pd.api.types ~/pandas-dev/pandas
/home/jreback/pandas-dev/pandas/_libs/lib.pyx:    >>> pd.api.types.is_scalar(dt)
/home/jreback/pandas-dev/pandas/_libs/lib.pyx:    >>> pd.api.types.is_scalar([2, 3])
/home/jreback/pandas-dev/pandas/_libs/lib.pyx:    >>> pd.api.types.is_scalar({0: 1, 2: 3})
/home/jreback/pandas-dev/pandas/_libs/lib.pyx:    >>> pd.api.types.is_scalar((0, 2))
/home/jreback/pandas-dev/pandas/_libs/lib.pyx:    >>> pd.api.types.is_scalar(Fraction(3, 5))
/home/jreback/pandas-dev/pandas/tests/frame/test_ufunc.py:    if pd.api.types.is_extension_array_dtype(dtype) or isinstance(dtype, dict):
/home/jreback/pandas-dev/pandas/tests/frame/test_ufunc.py:        pd.api.types.is_extension_array_dtype(dtype_a)
/home/jreback/pandas-dev/pandas/tests/frame/test_ufunc.py:        or pd.api.types.is_extension_array_dtype(dtype_b)
/home/jreback/pandas-dev/pandas/tests/frame/test_ufunc.py:    if pd.api.types.is_extension_array_dtype(dtype) or isinstance(dtype, dict):
/home/jreback/pandas-dev/pandas/tests/extension/base/dtype.py:        return not pd.api.types.is_string_dtype(dtype)
/home/jreback/pandas-dev/pandas/tests/extension/base/dtype.py:        return not pd.api.types.is_object_dtype(dtype)
/home/jreback/pandas-dev/pandas/tests/extension/arrow/arrays.py:        if pd.api.types.is_scalar(item):
/home/jreback/pandas-dev/pandas/tests/extension/arrow/test_bool.py:    assert pd.api.types.is_bool_dtype(data)
/home/jreback/pandas-dev/pandas/tests/extension/json/array.py:            if pd.api.types.is_bool_dtype(item.dtype):
/home/jreback/pandas-dev/pandas/tests/extension/decimal/array.py:        if pd.api.types.is_list_like(value):
/home/jreback/pandas-dev/pandas/tests/extension/decimal/array.py:            if pd.api.types.is_scalar(key):
/home/jreback/pandas-dev/pandas/tests/extension/list/array.py:        elif pd.api.types.is_string_dtype(dtype) and not pd.api.types.is_object_dtype(
/home/jreback/pandas-dev/pandas/tests/extension/test_integer.py:                and pd.api.types.is_integer_dtype(other.dtype)
/home/jreback/pandas-dev/pandas/tests/extension/test_floating.py:                and pd.api.types.is_float_dtype(other.dtype)
/home/jreback/pandas-dev/pandas/core/dtypes/inference.py:    >>> pd.api.types.is_number(1)
/home/jreback/pandas-dev/pandas/core/dtypes/inference.py:    >>> pd.api.types.is_number(7.15)
/home/jreback/pandas-dev/pandas/core/dtypes/inference.py:    >>> pd.api.types.is_number(False)
/home/jreback/pandas-dev/pandas/core/dtypes/inference.py:    >>> pd.api.types.is_number("foo")
/home/jreback/pandas-dev/pandas/core/dtypes/inference.py:    >>> pd.api.types.is_number("5")
/home/jreback/pandas-dev/pandas/core/generic.py:        >>> cat_dtype = pd.api.types.CategoricalDtype(

[nofarm3]

Hi @jreback!

To make sure I get this right:

1. Change it to be:
   from pd.api.types import is_number
   and then just use it in the test.
2. add a precommit rule

[nofarm3]

take

[MarcoGorelli]

Hi @nofarm3 - I haven't tried this yet but I think you'll need from pandas.api.types import ... (rather than from pd.api.types import ...) - feel free to reach out if you want/need help with anything!

[nofarm3]

Thanks @MarcoGorelli !

[jorisvandenbossche]

the doc-strings also should be updated to do the import once.

How do you want the docstrings changed?
Note that those are public functions, and thus need to be documented using their public import (which is pd.api.types)

In addition, IMO we should keep all occurrences in tests/extension as is (or at least the ones in the array classes). Those tests are meant for external projects to use and adapt, and thus should use public APIs to the extent possible.

[jreback]

the request is to simply import infer_dtype rather than use the long form

[jreback]

having a one off is -1

[nofarm3]

@jreback
I want to make sure I fix it the way you expect (this is my first time):

For example:
if pd.api.types.is_integer_dtype(other.dtype):

I'll change it to be:

from pandas._libs import lib
if lib.infer_dtype(other.dtype) == "integer":

Or

from pandas.api.types import is_integer_dtype
if is_integer_dtype(other.dtype):

I guess the first option is correct, am I right?
Thanks

[MarcoGorelli]

@nofarm3 the fastest way to get feedback is usually to just open a pull request, even if you're not sure it's correct, reviewers can make suggestions there

[jreback]

pls use the public api (2nd way)