All occurences of freq should be a Union[str, BaseOffset]

[gandhis1]

The one bug I could find was the method bdate_range which is annotated as a str, but my suspicion is there are other occurrences as freq is a common parameter

[Dr-Irv]

Please provide an example that illustrates a type checker showing an error.

[gandhis1]

import pandas as pd
from pandas.tseries.offsets import Day, BusinessDay

x = pd.date_range("1/1/2022", "2/1/2022", freq="1D")
y = pd.date_range("1/1/2022", "2/1/2022", freq=Day())
z = pd.bdate_range("1/1/2022", "2/1/2022", freq=BusinessDay())

test_case.py:6: error: Argument "freq" to "bdate_range" has incompatible type "BusinessDay"; expected "str"
Found 1 error in 1 file (checked 1 source file)

[ojdo]

Sorry to piggy-back here, but this seems like the best place to ask: in a tiny script, mypy complains about

Argument "freq" to "date_range" has incompatible type "timedelta"; expected "Union[str, BaseOffset]"

As far as I understand, and offsets.pyi seems to agree, argument "freq" should also support timedelta. Is this a bug in mypy, or should one of the lines in this PR include another timedelta | somewhere?

If so, I would be happy to provide a PR. I am just very new to mypy with such heavy stubbing.

[gandhis1]

The Pandas documentation for some functions that accept an offset does not seem to include timedelta. This interface may vary.

https://pandas.pydata.org/docs/reference/api/pandas.date_range.html

[gandhis1]

I will note that this may be in error as the underlying code ultimately calls to_offset. So whatever that supports should be supported by the overlying function.

[Dr-Irv]

As far as I understand, and offsets.pyi seems to agree, argument "freq" should also support timedelta. Is this a bug in mypy, or should one of the lines in this PR include another timedelta | somewhere?

The main issue here is that the docs for pd.date_range() don't say that a timedelta (or pd.Timedelta()) is a valid argument for freq, but the implementation accepts it. With the stubs, we have been following what's documented.

So can you open an issue in the pandas repo saying that pd.date_range accepts timedelta for the freq argument, but it's not documented, and then a determination can be made by the pandas team whether the docs are wrong or the implementation is wrong.

If the docs are wrong, then create a new issue here to update the stubs. If the implementation is wrong, we leave things here as they are.

[gandhis1]

My PR to update the Pandas documentation to include timedelta was just accepted. I can open a PR here to match later, unless anyone else wants to go ahead and take care of it.

[Dr-Irv]

My PR to update the Pandas documentation to include timedelta was just accepted. I can open a PR here to match later, unless anyone else wants to go ahead and take care of it.

Please submit a PR here. Thanks!