BUG: groupby().any() returns true for groups with timedelta all NaT

[Petroncini]

PR Summary:

This PR addresses an issue where groupby().any() returns True for groups where all Timedelta values are NaT, which is inconsistent with other dtypes (like integers and strings). For non-Timedelta dtypes, groupby().any() correctly returns False when all values in a group are null.
Reproduction:

python

import pandas as pd

Current behavior: incorrect output for all NaT groups

df = pd.DataFrame([pd.Timedelta(1), pd.NaT]).groupby([0, 1]).any()

For other dtypes (integers and strings):

python

df = pd.DataFrame([1, None]).groupby([0, 1]).any() # Returns False
df = pd.DataFrame(["a", None]).groupby([0, 1]).any() # Returns False

Expected behavior for Timedelta values: groupby().any() should return False for groups where all values are NaT.
Fix Details:

Modified the _call_cython_op function in the groupby aggregation to correctly handle Timedelta types. Now, when all values in a group are NaT, the any() method will return False.
This ensures consistent behavior across all dtypes, including Timedelta.

Tests Added:

A new test has been added to check that groupby().any() returns False for groups with only NaT Timedelta values and True for groups with at least one non-NaT Timedelta.

Closing:

This PR fixes the inconsistent behavior of groupby().any() for Timedelta values and improves the overall correctness of the method. Let me know if there are any additional changes needed.

Resolves: #59712

Note: I am a complete novice to open source and this is my first ever PR. I'm taking an open source systems class is college and we're suppposed to attempt to contribute to a existing project we use and I chose pandas as it is a tool that has helped me alot. So i ask for any guidance anyone can give me about any potential mistakes or improvements i can make here. Appreciate it!

[mroeschke]

Thanks for the pull request but this is already being worked on in #59750. Closing to let that contributor finish the effort, but we can reopen this if it goes stale

[Petroncini]

Hi @mroeschke , could this be reopened?

[rhshadrach]

The linked issue suggests moving up the determination of mask below to be prior to viewing the values as int64 on L375. Is there a reason this doesn't work?

[Petroncini]

Sorry, it appears i went in the wrong direction with my implementation. Just did a commit with this better solution. Could you give me some pointers on all these check failures?

[rhshadrach]

Merging main should fix the CI issues. However, I do not believe the changes here will fix the issue and will break other tests. We need to move the determination of mask = isna(values) to happen before values = values.view("int64") but not change the order of any other logic.

[rhshadrach]

Looking good! This needs a line in the whatsnew. Can you add a line to docs/source/whatsnew/v3.0.0.rst in the groupby subsection of the bug fixes.

[rhshadrach]

The comments in this test repeat the code verbatim. Can you remove them.

In addition, can you add a reference to the issue as the first line of this test.

def test_groupby_any_with_timedelta():
    # GH#59712

[Petroncini]

I added these changes.

[Petroncini]

Hi @rhshadrach, how's this?

[rhshadrach]

lgtm

[rhshadrach]

Thanks @Petroncini!

[Petroncini]

Thanks for all the help and guidance, @rhshadrach!