Question

After implementing a custom frequency in pandas by subclassing DateOffset, is it possible to “register” an offset alias for that frequency so that the alias can be used in built-in pandas functions such as date_range and resample?

For example, suppose I implement a custom twice-monthly frequency:

from pandas.tseries.offsets import DateOffset, CacheableOffset

class TwiceMonthly(DateOffset, CacheableOffset):
    def apply(self, other):
        # Some date logic here

    @property
    def rule_code(self):
        return 'TM'

Now, instead of using TwiceMonthly() everywhere, I want to use the offset alias TM.

# Suppose s is a time series
s.resample('TM', how='sum')

Answer 1

That’s not possible right now unfortunately. Aliasing is static.
It would be a good feature to add though. Please check back on github (https://github.com/pydata/pandas/issues/2085). Additional feedback or a PR would be appreciated.