Please consider using DeprecationWarning rather than FutureWarning for API changes

[mwaskom]

Pandas version checks

• I have checked that this issue has not already been reported.

• I have confirmed this bug exists on the latest version of pandas.

• I have confirmed this bug exists on the main branch of pandas.

Reproducible Example

import pandas as pd
df = pd.DataFrame({"x": pd.Categorical(["a", "b"]), "y": [1, 2]})
df.groupby("x")

Issue Description

As of #51811 this fires a FutureWarning because the default value for observed is going to change.

It seems like this is one of several pending deprecations in 2.1.0 to simplify / sand off the API.

Expected Behavior

(Sorry this is not exactly a bug but the other issue categories didn't work either)

Please consider introducing such deprecations with a DeprecationWarning rather than a FutureWarning so that downstream packages have a chance to adjust their internal calls to the Pandas API. Otherwise, users who update their pandas will immediate see confusing and non-actionable FutureWarning messages coming out of other libraries.

Pandas is often used interactively and so I agree that it's important to promote this to a more-visible FutureWarning at some point, but this could happen after a few release cycles.

Installed Versions

2.1.0

[mwaskom]

Bumping this as I am getting nearly daily false issue reports about these warnings (and doubt that my package is the only one affected).

[lithomas1]

Sure, delaying the FutureWarning by 1 release seems reasonable to me.

Assuming this is for seaborn, I noticed you aren't testing against pandas nightly.
(Ideally, it's best to have these kinds of things tracked before the release.)
Do you want me to open a PR for this?

[mwaskom]

Having nightly tests could be useful but not sufficient for this issue — I don't have resources to drop what I'm doing and cut a new release just because pandas has some breaking changes coming down the pike. DeprecationWarnings would be surfaced in tests but not to end users.

[mwaskom]

(To be clear, pandas devs were very helpful about proactively sending PRs which was much appreciated; but seaborn wasn't able to get a release out before pandas did, and so now users are affected).

[lithomas1]

Shoot, this somehow slipped off of my radar for 2.1.1. Sorry about that!

I will tag this as a blocker and put up a PR today so this doesn't get lost again.