[BUG] Unlike its pandas counterpart, cudf.date_range doesn't include the end if it's specified

[amanlai]

If I create a DatetimeIndex object using cudf.date_range() by passing start, end and freq parameters:

cudf.date_range('2020-01-01', '2020-01-05', freq='D')

DatetimeIndex(['2020-01-01', '2020-01-02', '2020-01-03', '2020-01-04'], 
              dtype='datetime64[ns]', freq='D')

As you can see, the result is an object with length 4. The same code in pandas results in an object with length 5:

pd.date_range('2020-01-01', '2020-01-05', freq='D')

DatetimeIndex(['2020-01-01', '2020-01-02', '2020-01-03', '2020-01-04', '2020-01-05'],
              dtype='datetime64[ns]', freq='D')

Is it possible to make it consistent with pandas' date_range() in that the result includes both ends? Because, closed parameter is nonfunctional at the moment, to make cudf.date_range() produce the same output as its pandas counterpart, we need to add 1 freq to end, which creates a whole lot of work: convert the string to datetime, convert freq to timedelta and add them to define a new end.

Anyway, is it possible to make it clear in the documentation that this is different from its pandas counterpart?

^{I didn't know how to tag this since it's not really a bug. It's just different from pandas API that results in a subtle bug in my code.}

[shwina]

I think this is a legitimate bug. Thanks for reporting!

[wence-]

Also #12133

[Matt711]

Was this closed by #16277?

cc. @mroeschke

[mroeschke]

Was this closed by #16277?

cc. @mroeschke

Appears so! Closing