fix: treat pandas Boolean as numeric in clinical_kernel

[55Kamiryo]

Checklist

• closes Bug: clinical_kernel silently drops pandas Boolean columns; ClinicalKernelTransform.fit raises TypeError on the same input #589
• pytest passes
• tests are included
• code is well formatted
• documentation renders correctly

What does this implement/fix? Explain your changes

Fixes the two divergent Boolean-handling bugs in sksurv/kernels/clinical.py described in #589:

1. clinical_kernel silent drop. _get_continuous_and_ordinal_array used x.select_dtypes(include=[np.number]), which excludes np.bool_ because numpy treats it as a sibling of np.number rather than a subclass. Boolean columns matched neither the numeric nor the object/category filter, were silently dropped, and the resulting matrix was biased by (n_features - n_bool_cols) / n_features because normalization still used x.shape[1].

   Fix: select_dtypes(include=[np.number, "bool"]).

2. ClinicalKernelTransform.fit TypeError. _prepare_by_column_dtype uses pandas.api.types.is_numeric_dtype, which returns True for Boolean, but col.max() - col.min() then failed with TypeError: numpy boolean subtract on numpy ≥ 1.25.

   Fix: cast Boolean columns to np.uint8 before computing the range.

Both fixes align the pandas path with the policy already established in this codebase: pandas.api.types.is_numeric_dtype(bool) is True. After this change, clinical_kernel(df) and clinical_kernel(df.astype({col: 'uint8'})) produce identical kernel matrices, and ClinicalKernelTransform().fit(df) no longer crashes.

Tests

Added TestClinicalKernel.test_bool_column_treated_as_numeric asserting:

• clinical_kernel returns the same matrix for a Boolean column and its uint8 equivalent.
• ClinicalKernelTransform().fit produces matching _numeric_ranges and X_fit_ for both dtypes.
• The Boolean column is classified into _numeric_columns, not _nominal_columns.

Local verification:

• pytest tests/ (including slow tests): 966 passed, 48 skipped, 0 failed.
• pytest --doctest-modules --pyargs sksurv.kernels: 1 passed.
• ruff check on the whole repo: clean.
• pre-commit run on changed files: all hooks pass.

Behavior change note

Users whose pipelines previously called clinical_kernel on a pandas frame containing Boolean columns will see different (corrected) kernel values after this change. The previous values were silently biased, so the change is a bug fix rather than an API change, but it may be worth flagging in the changelog.

[codacy-production]

Up to standards ✅

🟢 Issues 0 issues

Results:
0 new issues

View in Codacy

🟢 Metrics 0 complexity · 0 duplication

Metric	Results
Complexity	0
Duplication	0

View in Codacy

_{NEW Get contextual insights on your PRs based on Codacy's metrics, along with PR and Jira context, without leaving GitHub. Enable AI reviewer
TIP This summary will be updated as you push new changes.}

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 98.46%. Comparing base (6a1df8a) to head (fd54167).

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #590   +/-   ##
=======================================
  Coverage   98.46%   98.46%           
=======================================
  Files          38       38           
  Lines        3713     3715    +2     
  Branches      480      481    +1     
=======================================
+ Hits         3656     3658    +2     
  Misses         27       27           
  Partials       30       30           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[sebp]

Thanks a lot for this!