Pandas invalid type comparison error

Question

For some reason that I can't find in the Pandas summary for 0.17.1, comparing the datetime series with an int (Unix epoch) value no longer works. Can someone explain this or point me to the right section in the change log?

Work in 0.16.2

>>> import pandas as pd
>>> import datetime
>>> d = pd.Series([datetime.datetime(2016, 1, 1), datetime.datetime(2016, 1, 1)])
>>> d
0   2016-01-01
1   2016-01-01
dtype: datetime64[ns]
>>> d.dtype
dtype('<M8[ns]')
>>> d > 10
0    True
1    True
dtype: bool

Error in 0.17.1

>>> import pandas as pd
>>> import datetime
>>> d = pd.Series([datetime.datetime(2016, 1, 1), datetime.datetime(2016, 1, 1)])
>>> d > 10
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/Users/sven/tmp/pandastest/pandas-0.17.1/lib/python2.7/site-packages/pandas/core/ops.py", line 726, in wrapper
    res = na_op(values, other)
  File "/Users/sven/tmp/pandastest/pandas-0.17.1/lib/python2.7/site-packages/pandas/core/ops.py", line 657, in na_op
    raise TypeError("invalid type comparison")
TypeError: invalid type comparison

Answer 1

You can still use an explicit conversion:

u_time_ns = d.apply(lambda x: x.to_datetime64().view('int64'))
u_time_ns

0    1451606400000000000
1    1451606400000000000
dtype: int64

u_time_ns > 10

0    True
1    True
dtype: bool

Or, if you want to rely on pandas timestamps saved as datetime64[ns]:

u_time_ns = d.view('int64')

Sorry, I don’t know why this has changed.