Adding Period and Offset not implemented

[jseabold]

This works

from pandas.tseries.frequencies import to_offset
pd.Timestamp('1990-9-30') + 2*to_offset('Q')

This doesn't. Shouldn't it? Workaround is to convert with to_timestamp.

pd.Period('1990q1') + 2*to_offset('Q')

[jreback]

You want this, but also ability to do it with an offset. What if the offset is incompatible to the inferred_freq, raise?

In [17]: pd.Period('1990q1')
Out[17]: Period('1990Q1', 'Q-DEC')

In [18]: pd.Period('1990q1') + 2
Out[18]: Period('1990Q3', 'Q-DEC')

[jseabold]

Hmm, maybe that is sufficient. I guess I was expecting to work in both cases with a conformable offset, so I can treat period and timestamp the same and not special case everything.

I'm starting to think this is a design problem on our end though and we should special case for PeriodIndex vs DatetimeIndex with a freq. I'm not really sure. I'd like to be able to treat them both the same, but I don't know if that's a reasonable expectation.

[jreback]

hmm. they should have the same API, but right now they inherit from the base class separately, so this is kind of tricky

This will fix it (and its not really that hard), just need to find time for this: #6469

But I agree they should be pretty comparable.

That said, I think we can leave this issue open for now (i'll link to the other one), for reference

[jorisvandenbossche]

This seems to work now:

In [57]: pd.Period('1990q1')
Out[57]: Period('1990Q1', 'Q-DEC')

In [58]: pd.Period('1990q1') + 2*to_offset('Q')
Out[58]: Period('1990Q3', 'Q-DEC')

and was closed by #7966 (which already included tests for this)