Cannot Add Dates that are the 1st of a Month and on a Sunday to a DatetimeIndex

[jeffshek]

Running into a problem when I attempt to add the first date of a month that is a Sunday.

import datetime
from pandas.tseries.index import DatetimeIndex

d1 = datetime.date(2002, 9, 1)
d2 = datetime.date(2013, 10, 27)

d3 = datetime.date(2012, 9, 30)

i1 = DatetimeIndex([d1, d2])
i2 = DatetimeIndex([d3])

i1 + i2

ValueError: Could not evaluate WOM-1SUN

Returns a ValueError "Could not evaluate WOM-1SUN". However, if I change d1 and then follow the same steps

d1 = datetime.date(2002, 9, 2)
i1 = DatetimeIndex([d1, d2])
i1 + i2

Out[2]: <class 'pandas.tseries.index.DatetimeIndex'>
[2002-09-02 00:00:00, ..., 2013-10-27 00:00:00]
Length: 3, Freq: None, Timezone: None

This create a DatetimeIndex adding the the index i1 and i2.

[cancan101]

This is caused by two issues: #5115 and and #5152

[jeffshek]

Since some of the underlying has fixed for the example above (while understanding that the full fix is for a future release). Providing an additional scenario that still showcases this in 15.2 (if helpful)

d1 = datetime.date(2014, 3, 31)
d2 = datetime.date(2014, 6, 30)
d3 = datetime.date(2014, 12, 29)

i1 = DatetimeIndex([d1, d2])
i2 = DatetimeIndex([d3])

i1 + i2
ValueError: Could not evaluate WOM-5MON

[jreback]

hmm, that is a weird error. I'll mark as a bug. pull-requests welcome!