DOC: DateOffset.is_anchored track down intention, fix docstring · Issue #55388 · pandas-dev/pandas (original) (raw)

def is_anchored(self) -> bool: # TODO: Does this make sense for the general case? It would help # if there were a canonical docstring for what is_anchored means.

The docstring on the base class is_anchored pretty clearly doesn't make sense in the non-base class methods, so I suspect it is wrong. It has never been clear to me exactly what the intention of is_anchored is, but my best guess is something like offsets that are "anchored" on a particular day of the week or month of the year or etc.

The task here is to dig into the GH history to determine if the intent and correct behavior was ever clear. If so, to fix the implementation and/or docstring (and then remove the comment). The method isn't used outside the tests and I doubt many users use it, so OK to deprecate if it isn't useful or doesn't make sense.

I suspect something like it can make sense and be useful for DatetimeIndex set operations to determine when we can use RangeIndex fastpaths, xref #44025.

For historical digging/grepping it may be useful to know that this used to be isAnchored.