TypeError: cannot do label indexing on <class 'pandas.core.indexes.base.Index'> with these indexers [nan] of <class 'float'>

[kzhui125]

import pandas as pd
import numpy as np

df = pd.Series([1, '{1,2}', 1, None])
vc = df.value_counts(dropna=False)
print(vc.loc[np.nan])
print(vc[np.nan])

TypeError: cannot do label indexing on <class 'pandas.core.indexes.base.Index'> with these indexers [nan] of <class 'float'>

Hello, now I have trouble getting null value count, help.

[jbrockmendel]

This works on master, what pandas version are you using?

[kzhui125]

@jbrockmendel

1.0.1

[NicolasCwy]

Facing the same problem Pandas 1.0.5. What code should I be providing?

TypeError: cannot do label indexing on <class 'pandas.core.indexes.base.Index'> with these indexers [nan] of <class 'float'>

[jbrockmendel]

This works in 1.1.0

[simonjayhawkins]

@jbrockmendel i've labelled this as needs tests. Any idea which commit fixed this and whether additional tests are needed.

[alexhtn]

I think that commit d4293f0 fixed this issue. But I'm not sure about how I find it. I took commits from git log --pretty=oneline and in automated mode go from newest to oldest until code from first post started raising errors. On commit d4293f0 all works fine and on the next commit 58df0ac starts errors.

[simonjayhawkins]

Thanks @alexhtn I think #31962 may also have contributed to fixing the issue. (commit before this KeyError: nan was raised)

d206479 is the first fixed commit
commit d206479
Author: jbrockmendel [email protected]
Date: Tue Mar 3 07:27:17 2020 -0800

REF: remove _convert_scalar_indexer (#31962)

@jbrockmendel i've labelled this as needs tests. Any idea which commit fixed this and whether additional tests are needed.

no tests were added in #31962 and the test added in #31866 are for error checking so safe to say should add a test for the code sample in OP to close issue

But I'm not sure about how I find it.

~/test.py

import numpy as np
import pandas as pd
print(pd.__version__)
s = pd.Series([1, "{1,2}", 1, None])
vc = s.value_counts(dropna=False)
print(vc)

try:
    res = vc.loc[np.nan]
    print(res)
except:
    pass
else:
    sys.exit(1)

and in shell

git bisect start HEAD v1.0.5 --term-new fixed --term-old broken
git bisect run /bin/bash -c "python setup.py build_ext -i;python ~/test.py"

see #35685

[wesleyboelrijk]

take

[callmekofi]

facing the same proble

m

[jbrockmendel]

@callmekofi what version of pandas are you using? for me the example in the OP currently raises KeyError, which I think is correct