Inconsistent comparison between datetime and series of datetime

[hemflit]

similar to #3416 (also failing in 0.11)

Using pandas 0.10.1 and numpy 1.6.1:

>>> import pandas; import numpy
>>> s = pandas.Series(pandas.date_range('2013-01-01', periods=3))
>>> d = numpy.datetime64('2013-01-02')
>>> s > d
0    False
1    False
2     True
>>> d < s
0    True
1    True
2    True
>>> s.dtype
dtype('datetime64[ns]')
>>> d.dtype
dtype('datetime64[us]')

All other comparison operators also give expected results with operands (s,d), and seem to compare them as unscaled integers when the order is (d,s).

[jreback]

thanks for the report, this fails in 0.11 as well, for now use use a datetime object instead

[hemflit]

Thanks.

For completeness, at least in 0.10.1, a datetime.datetime as the left operand triggers a "TypeError: can't compare datetime.datetime to Series". (Still definitely preferable to the numpy.datetime64 treatment.)

I guess the workaround for the moment is to convert to datetime.datetime and reorder operands.

[jreback]

This will have to wait till 0.12 (I just tried to fix this), because of the way python tried to do reverse comparison operators.....

[jreback]

@hemflit

This actually does work in 0.12 and numpy 1.7.

but the semantics are weird here anyhow

you are comparing a scalar (like) on the lhs with a Series on the rhs

and that's why I can't intercept it

so just reverse the comparison

closing as unfixable/not-a-bug