Problem with YearBegin frequency?

[jseabold]

Bug or user/expectation error? Been a while since I looked at this. Shouldn't this work and return days at the start of April?

pandas.date_range("1972-04-02", "2007-04-02", freq="AS-APR")

#<class 'pandas.tseries.index.DatetimeIndex'>
#[1973-01-01 00:00:00, ..., 2007-01-01 00:00:00]
#Length: 35, Freq: AS-APR, Timezone: None

pandas.datetools.to_offset("AS-APR")
#<1 YearBegin: kwds={'month': 4}, month=4>

This does.

pandas.date_range("1972-04-02", "2007-04-02", freq="BAS-APR")

#<class 'pandas.tseries.index.DatetimeIndex'>
#[1972-04-03 00:00:00, ..., 2007-04-02 00:00:00]
#Length: 36, Freq: BAS-APR, Timezone: None

pandas.date_range("1972-04-02", "2007-04-02", freq="A-APR")

#<class 'pandas.tseries.index.DatetimeIndex'>
#[1972-04-30 00:00:00, ..., 2006-04-30 00:00:00]
#Length: 35, Freq: A-APR, Timezone: None

pandas.version.version
# '0.10.1'

[jseabold]

Some more information. It looks like YearBegin is broken to me unless I'm completely misunderstanding. It has never actually worked correctly going back to 0.8.0. I'll see if I can't figure out more since I need to see how this works for our time series models/forecasting.

pandas.date_range("1972-04-02", "2007-04-02", freq="AS-APR") + pandas.datetools.YearEnd(month=4)
#<class 'pandas.tseries.index.DatetimeIndex'>
#[1973-04-30 00:00:00, ..., 2007-04-30 00:00:00]
#Length: 35, Freq: A-APR, Timezone: None

pandas.date_range("1972-04-02", "2007-04-02", freq="AS-APR") + pandas.datetools.YearBegin(month=4)
#<class 'pandas.tseries.index.DatetimeIndex'>
#[1974-01-01 00:00:00, ..., 2008-01-01 00:00:00]
#Length: 35, Freq: AS-JAN, Timezone: None