ENH: Creating a now Period for a frequency of a multiple unit

[eumiro]

Feature Type

• Adding new functionality to pandas

• Changing existing functionality in pandas

• Removing existing functionality in pandas

Problem Description

I wish there was a straightforward way to create the now Period for the frequency that is a multiple of a unit (e.g. "15T")

Feature Description

>>> pd.Period.now("15T")
Period('2023-06-29 12:34', '15T')

>>> pd.Period.now("15T", floor=True)
Period('2023-06-29 12:30', '15T')

The first case is the status quo: the Period starts in the current minute. The second is my idea (and someone will certainly have a better name replacement for the floor= parameter) and it starts at the previous 15T stop (:00, :15, :30, :45).

Alternative Solutions

The most compact way I found to get the right result is:

>>> pd.Timestamp.now().floor("15T").to_period("15T")
Period('2023-06-29 12:30', '15T')

with 15T appearing twice in the line, which is what I'd like to avoid.

My proposal should work also for frequencies that do not fit within one hour:

>>> pd.Period.now("17T", floor=True)
>>> pd.Timestamp.now().floor("17T").to_period("17T")
Period('2023-06-29 12:31', '17T')

Additional Context

No response

[jesse-sealand]

I follow your logic here. Can you share a use-case for this?

[eumiro]

@jesse-sealand When I need to fetch some data, I prefer to pass a PeriodIndex to a method. It is more bullet-proof than start/end/freq/whatever, because the returned DataFrame has to have exactly that PeriodIndex as index.

When I need the previous day, I use:

>>> pd.period_range(end=pd.Period.now("D") - 1, periods=1)
PeriodIndex(['2023-08-19'], dtype='period[D]')

When I need more than one day, either I say periods=20 for the previous 20 days or I replace periods with start=pd.Period("2023-01-01") for an index since the start of the year.

II can do the same for the previous hours (D replaced with H). However if my frequency is not the whole unit (S/T/H/D/…), but e.g. 15T or 3H, then I have to go the floor/to_period day.