From hourly DatetimeIndex to daily DatetimeIndex

Question

I am trying to convert an hourly DatetimeIndex to a daily DatetimeIndex with no repetitions.

I have two DatetimeIndex:

import pandas as pd

snapshots = pd.date_range("2050-01-01 00:00", "2050-12-31 23:00", freq="H")

days = pd.date_range("2050-01-01", "2050-12-31", freq="d")

However I would like to create days directly from snapshots. I tried to convert to PeriodIndex but then I struggle with unique dates:

days = snapshots.to_period().asfreq('d')

(I have 'YYYY-MM-DD' data, but 8760 for every hour of the year)

Answer 1

Simpliest is pass minimal and maximal value of snapshots to date_range:

days = pd.date_range(snapshots.min(), snapshots.max(), freq="d")
print (days)
DatetimeIndex(['2050-01-01', '2050-01-02', '2050-01-03', '2050-01-04',
               '2050-01-05', '2050-01-06', '2050-01-07', '2050-01-08',
               '2050-01-09', '2050-01-10',
               ...
               '2050-12-22', '2050-12-23', '2050-12-24', '2050-12-25',
               '2050-12-26', '2050-12-27', '2050-12-28', '2050-12-29',
               '2050-12-30', '2050-12-31'],
              dtype='datetime64[ns]', length=365, freq='D')

Or use DatetimeIndex.normalize for set times to 00:00:00 and remove duplicates - by Index.unique or Index.drop_duplicates:

days = snapshots.normalize().unique()
print (days)
DatetimeIndex(['2050-01-01', '2050-01-02', '2050-01-03', '2050-01-04',
               '2050-01-05', '2050-01-06', '2050-01-07', '2050-01-08',
               '2050-01-09', '2050-01-10',
               ...
               '2050-12-22', '2050-12-23', '2050-12-24', '2050-12-25',
               '2050-12-26', '2050-12-27', '2050-12-28', '2050-12-29',
               '2050-12-30', '2050-12-31'],
              dtype='datetime64[ns]', length=365, freq=None)

---

days = snapshots.normalize().drop_duplicates()
print (days)
DatetimeIndex(['2050-01-01', '2050-01-02', '2050-01-03', '2050-01-04',
               '2050-01-05', '2050-01-06', '2050-01-07', '2050-01-08',
               '2050-01-09', '2050-01-10',
               ...
               '2050-12-22', '2050-12-23', '2050-12-24', '2050-12-25',
               '2050-12-26', '2050-12-27', '2050-12-28', '2050-12-29',
               '2050-12-30', '2050-12-31'],
              dtype='datetime64[ns]', length=365, freq=None)