I noticed that if try to build a DatetimeIndex doesn't play nice with matplotlib's date module. Consider the following:
In [1]: import pandas
In [2]: import datetime
In [3]: import matplotlib.dates as mdates
In [4]: start = mdates.num2date(mdates.datestr2num('2012-1-1'))
In [5]: end = mdates.num2date(mdates.datestr2num('2012-6-1'))
In [6]: pandas.DatetimeIndex(start=start, end=end, freq='MS')
---------------------------------------------------------------------------
AttributeError: '_UTC' object has no attribute 'normalize'
In [7]: pandas.version.version
Out[7]: '0.9.1'
In [8]: start
Out[8]: datetime.datetime(2012, 1, 1, 0, 0, tzinfo=<matplotlib.dates._UTC object at 0x04C92130>)
Setting matplotib.rcParams['timezone'] to *anything* that's not UTC fixes this problem
My guess is that isn't a big deal since pandas can parse the datestrings as well as matplotlib, making lines 4 and 5 unnecessary. Perhaps an note in the docs is in order? If so, that's something I'd be willing to take on myself and submit a PR. Should I also check in with the MPL devs about this?
Thanks,
-paul