271 feature implementation of columntransformer support

[fernandonbarros]

Summary

• Updates the PLSRegression transformer to add compatibility with sklearn's ColumnTransformer.
• When setting transform_out="pandas", the PLSRegression transformer will output scores as a dataframe with columns LV1 to LVn.
• This PR also addresses issues with the test_optional_dependencies test failing.

Why is this change needed?

• This change will allow the passthrough of metadata when creating a pipeline to process pandas dataframes.

Closes

• Closes feature: Implement ColumnTransformer support in PLSRegression method #271

Related issues

Type of change

• Bug fix
• New feature
• Refactor
• Performance improvement
• Documentation update
• Tests only
• CI / DevOps / tooling
• Dependency update
• Breaking change

What changed?

• Updated chemotools/regression/_pls_regression.py:
  • Defined get_feature_names_out method to return an array ["LV1" ... "LVn"] to serve as column names for scores dataframe.
  • Updated the transform method to add a flag to define if y_scores should be returned by the method.
• Updated chemotools/tests/regression/test_pls_regression.py:
  • Added test_works_with_column_transformer test to ensure that the PLSRegression transformer is compatible with the ColumnTransformer.
• Updated chemotools/_optional.py and chemotools/tests/utils/test_optional_dependencies.py to fix issues with the test_import_optional_dependency_real_optional test.

What did NOT change?

• No other transformers other than PLSRegression were modified.

API and compatibility impact

• No public API changes
• Public API added
• Public API changed
• Deprecation introduced
• Breaking change introduced

Notes

• scikit-learn API compatibility: PLSRegression should still be fully compliant with SKLearn
• Affected modules/classes/functions:
  • PLSRegression class in chemotools/regression/_pls_regression.py was updated.
  • import_optional_dependency function in chemotools/_optional.py was updated.
• Backward compatibility considerations:
  • Instances where PLSRegression transformer is used and expect y_scores to be returned should be modified to include the return_y=True flag.

Validation

• task format-check
• task lint
• task spelling
• task type-check
• task test
• task coverage
• task test:matrix
• task docs:check
• task build

Validation notes

• test_import_optional_dependency_real_optional was originally failing, but issue was addressed by updating the test and the import_optional_dependency function.
• Failure observed when running task docs:check

Tests

• Added new tests
• Updated existing tests
• No tests added because this PR only changes docs / CI / metadata

Test coverage details

• Relevant test files:
  • chemotools/tests/utils/test_optional_dependencies.py
  • chemotools/tests/regression/test_pls_regression.py
• Edge cases covered:
  • Use of PLSRegression in a pipeline with ColumnTransformer where the transform_output="pandas"
• Numerical / estimator behavior checked: N/A

Documentation

• Docs updated
• Docstrings updated
• No docs update needed

Documentation notes

• Updates to methods and functions were reflected in the previously existing docstring.

Dependency / build / CI impact

• No dependency changes
• Runtime dependency change
• Dev dependency change
• GitHub Actions / CI changed
• Build/release process changed

Notes

Reviewer focus

Please focus on:

• The updated transform and the new get_feature_names_out methods for PLSRegression.
• The new/updated tests for utils and PLSRegression.

Screenshots / logs / benchmark output

Details

Checklist

• I linked the relevant issue(s) or explained why none exists
• I kept this PR scoped to a single purpose
• I added or updated tests where appropriate
• I updated documentation where appropriate
• I verified the change locally using the relevant tasks above
• I considered backward compatibility and public API impact
• I am ready for review

[Copilot]

Copilot encountered an error and was unable to review this pull request. You can try again by re-requesting a review.

[fernandonbarros]

It looks like the update to the PLSRegression API introduced a bug in the PLSRegressionInspector that was not caught be any of the existing tests. I created an additional commit to address the bug that was introduced.