ValueError on df.columns.isin(pd.Series())

[wikiped]

Code Sample, a copy-pastable example if possible

    df = pd.DataFrame(columns=list('ab'))
    s1 = pd.Series(['a'])
    s2 = pd.Series()
    df.columns.isin(s1)
    df.columns.isin(s2)

Problem description

The second call to df.columns.isin(s2) fails with

D:\Anaconda\envs\py3k\lib\site-packages\pandas\core\algorithms.py in <lambda>(x, y)
    402     # work-around for numpy < 1.8 and comparisions on py3
    403     # faster for larger cases to use np.in1d
--> 404     f = lambda x, y: htable.ismember_object(x, values)
    405     if (_np_version_under1p8 and compat.PY3) or len(comps) > 1000000:
    406         f = lambda x, y: np.in1d(x, y)

pandas\_libs\hashtable_func_helper.pxi in pandas._libs.hashtable.ismember_object (pandas\_libs\hashtable.c:30162)()

ValueError: Buffer dtype mismatch, expected 'Python object' but got 'double'

Expected Output

array([ False, False], dtype=bool)

Output of pd.show_versions()

INSTALLED VERSIONS
------------------
commit: None
python: 3.5.3.final.0
python-bits: 64
OS: Windows
OS-release: 10
machine: AMD64

pandas: 0.20.3
numpy: 1.13.1

Might be linked to #16394

[gfyoung]

Reproduced on master. That's a bug alright. PR is welcome!

[gfyoung]

@jreback : Tagging this for 0.21 because IMO we should be able to address this bug for the release. If no one picks it up, I can patch it before then.

[jreback]

i almost always tag with next release

otherwise u end up moving them later back if they r not done

just so many reports it's hard to judge what to tag unless there is an actual PR

[gfyoung]

otherwise u end up moving them later back if they r not done

That's true. I do have this one on my radar though, so when the time comes to release and this still hasn't been resolved, feel free to call me out on it 😄