feat: support polars dataframe in CachedFoundryClient.save_dataset

[Waschenbacher]

Summary

Allow polars dataframe in CachedFoundryClient.save_dataset

Checklist

• [x ] You agree with our CLA
• [x ] Included tests (or is not applicable).
• [x ] Updated documentation (or is not applicable).
• [x ] Used pre-commit hooks to format and lint the code.

[Copilot]

Pull Request Overview

This PR adds support for polars DataFrame in CachedFoundryClient.save_dataset, allowing users to pass polars DataFrame in addition to pandas and pyspark DataFrames.

• Updated type hints and docstrings to incorporate polars support.
• Modified test cases to validate functionality with both pandas and polars DataFrames.

Reviewed Changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated no comments.

File	Description
tests/unit/test_cached_foundry_client.py	Added parameterized tests for both pandas and polars DataFrames.
libs/foundry-dev-tools/src/foundry_dev_tools/cached_foundry_client.py	Extended type hints, error messages, and functionality to support polars DataFrame in save_dataset.

Comments suppressed due to low confidence (2)

tests/unit/test_cached_foundry_client.py:38

• [nitpick] Consider renaming 'test_save_pandas' to 'test_save_dataframe' to reflect that the test now covers both pandas and polars DataFrames.

@pytest.mark.parametrize(

libs/foundry-dev-tools/src/foundry_dev_tools/cached_foundry_client.py:222

• Update the error message to include polars support, for example: 'Please provide a spark, pandas, or polars dataframe object with parameter "df"'.

        if df is None:

[nicornk]

Thanks for the contribution.

can you double check we don’t run into this issue again when polars is not available (because it’s an optional dependency)

#83

[Waschenbacher]

Sure. In an environment without polars, the current code works fine. The if-clause for polars is the same as pandas:

if not pd.__fake__ and isinstance(df, pd.DataFrame):
    ...
elif not pl.__fake__ and isinstance(df, pl.DataFrame):
    ...

When polars is no available in the env, not pl.__fake__ gives False and isinstance(df, pl.DataFrame) will be skipped, hence no issue.