Series .equals does not ignore NaNs

[rockg]

This is a datatype thing I believe. When I construct a DataFrame and then compare rows of that, NaNs are not ignored contrary to the documentation.

import pandas as pd
import numpy as np
a = pd.DataFrame([{'A': 1, 'B': 'b', 'C': np.NAN}, {'A': 1, 'B': 'b', 'C': 1}])
b = pd.DataFrame([{'A': 1, 'B': 'b', 'C': np.NAN}, {'A': 1, 'B': 'b', 'C': 1}])

a.loc[0]
A      1
B      b
C    NaN
Name: 0, dtype: object
aa = a.loc[0]
bb = b.loc[0]
aa.equals(bb)
False

# which essentially does this in Block
aa.values == bb.values
Out[35]: array([ True,  True, False], dtype=bool)

Constructing the series outright works as advertised.

a = pd.Series({'A': 1, 'B': 'b', 'C': np.NAN})
b = pd.Series({'A': 1, 'B': 'b', 'C': np.NAN})
a.equals(b)
True
a.values == b.values
Out[30]: array([ True,  True,  True], dtype=bool)

[jorisvandenbossche]

What version of pandas? Works for me in 0.15.1

[rockg]

0.14.0. Glad to hear it's fixed. I'll close.

[jorisvandenbossche]

Possibly fixed by #8443