Inconsistencies in index comparisons

[csiz]

Comparing indexes returns two different data types. This is problematic since one of the types is a bool while the other is a Series which explicitly can't be treated in a bool context. Not to mention that the next section says == returns a Series object.

import pandas as pd

print pd.__version__
#0.14.0

index = pd.Int64Index([1, 2, 3])

print index == pd.Int64Index([])
# False
print index == pd.Int64Index([1])
# [True, False, False]
print index == pd.Int64Index([1, 2])
# False
print index == pd.Int64Index([4, 5, 6])
# [False, False, False]

series = pd.Series([1, 2, 3])

print series == pd.Series([1, 2])
# ValueError: Series lengths must match to compare

I'd suggest that comparing indexes with different lengths raise a ValueError like Series do. Possibly keeping the 1 element vs many as a special case

[jreback]

You are using a pretty old version of pandas, IIRC this was corrected in 0.16.1 or so. Current is 0.17.1, and 0.18.0 is coming out shortly

In [81]: index = pd.Int64Index([1, 2, 3])

In [82]: index == pd.Int64Index([])
ValueError: Lengths must match to compare

In [83]: index == pd.Int64Index([1])
ValueError: Lengths must match to compare

In [85]: index == pd.Int64Index([1, 2])
ValueError: Lengths must match to compare

In [86]: index == pd.Int64Index([4, 5, 6])
Out[86]: array([False, False, False], dtype=bool)