set_index/reindex bugs with time indices

[jseabold]

[14]: pd.version.version
[14]: '0.13.1-426-ge19b2eb'

Noticed these. This one's minor. set_index seems to drop frequency information for DatetimeIndex. I'm not sure what other metadata is ignored.

import statsmodels.api as sm
import pandas as pd
index = pd.PeriodIndex(start='1959Q1', end='2009Q3')
dta = sm.datasets.macrodata.load_pandas().data

assert dta.set_index(index.to_timestamp()).index.freq == index.to_timestamp().freq

This one is less minor. set_index and reindex are broken for a PeriodIndex. Treats it as regular integer index I guess. Not sure if it ever worked, though I don't recall running into this before.

dta.set_index(index).head()

dta.reindex(index).head()

[jreback]

you can just do

dta.index = index

I see where the issue is in set_index, will mark as a bug

[jseabold]

Yeah, that's what I have in all the statsmodels examples, though I thought the setting of the attribute was a no-no (don't recall why though). Just trying to disseminate best practices.