DOC: infer_dtypes document is inconsistent with behaviour regarding boolean

[tsinggggg]

Location of the documentation

pandas/pandas/_libs/lib.pyx

Line 1315 in 0711b7a

>>> infer_dtype([True, False, np.nan])

Documentation problem

In this example the return value should be 'boolean' as mentioned in this test case

pandas/pandas/tests/dtypes/test_inference.py

Line 679 in 1db3aa2

arr = np.array([True, np.nan, False], dtype="O")

Suggested fix for documentation

change 'mix' to 'boolean'

[arw2019]

Am I misunderstanding or this isn't a docs problem? infer_dtypes returns mixed in this case

[tsinggggg]

Am I misunderstanding or this isn't a docs problem? infer_dtypes returns mixed in this case

@arw2019 i think it actually returns boolean, as shown in the test case

[arw2019]

is the difference that one is a python list holding [True, False, np.nan] versus an ndarray np.array([True, np.nan, False], dtype="O") constructed from that list

[tsinggggg]

is the difference that one is a python list holding [True, False, np.nan] versus an ndarray np.array([True, np.nan, False], dtype="O") constructed from that list

it seems even with a list it's returning boolean
https://colab.research.google.com/drive/1s8kRGIkeQk2TXprc_Q2iYcNAT3u1QrPa?usp=sharing

[arw2019]

ok. Are you interested in submitting a PR with the correction?

(I know that in .py files we verify the docstrings but maybe in .pyx we don't...)

[tsinggggg]

ok. Are you interested in submitting a PR with the correction?

(I know that in .py files we verify the docstrings but maybe in .pyx we don't...)

yes sure!

[FullBeardDev]

Hello there, as I'm starting to work on some open source contributions, I'm happy to submit a PR with the correction.

[tsinggggg]

take