fix: StaticCovariatesTransformer IndexError with OneHotEncoder drop param

[JKDasondee]

Summary

Fixes #2880.

StaticCovariatesTransformer raises IndexError when using OneHotEncoder with drop='first' (or drop='if_binary', etc.). The transformed output has fewer columns than categories, but the column mapping was built from transformer_cat.categories_ which includes all categories regardless of the drop setting.

Root cause

In _create_category_mappings(), the column name mapping iterates over transformer_cat.categories_:

for col, categories in zip(cols_cat, transformer_cat.categories_):
    for cat in categories:
        col_map_cat_i.append(str(col) + "_" + str(cat))

With drop='first' and 3 categories per column, categories_ has 3 entries but the encoder only outputs 2 columns. Later in _add_back_static_covs(), the mapping tries to index position 4 in an array of size 4 (0-indexed), causing the IndexError.

Fix

Use get_feature_names_out() instead of iterating over categories_. This method returns only the actual output column names, correctly reflecting any dropped categories.

feature_names = transformer_cat.get_feature_names_out(cols_cat)
feat_idx = 0
for col in cols_cat:
    col_map_cat_i = []
    prefix = str(col) + "_"
    while feat_idx < len(feature_names) and feature_names[feat_idx].startswith(prefix):
        name = feature_names[feat_idx]
        col_map_cat_i.append(name)
        inv_col_map_cat[name] = [col]
        feat_idx += 1
    col_map_cat[col] = col_map_cat_i

Tests

Adds test_one_hot_encoder_with_drop parametrized on ['first', 'if_binary']. Each variant tests:

• Single-series fit_transform + inverse_transform round-trip
• Multi-series fit_transform + inverse_transform round-trip
• Asserts static_covariates DataFrames match exactly after round-trip

TDD verification

Without the fix, the new test fails with:

IndexError: index 4 is out of bounds for axis 1 with size 4

at static_covariates_transformer.py:456 in _add_back_static_covs().

With the fix, all 10 tests pass (8 existing + 2 new):

$ pytest darts/tests/dataprocessing/transformers/test_static_covariates_transformer.py -v
10 passed in 0.20s

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 95.73%. Comparing base (447c7d3) to head (19571c3).

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #3065      +/-   ##
==========================================
- Coverage   95.80%   95.73%   -0.07%     
==========================================
  Files         158      158              
  Lines       17315    17320       +5     
==========================================
- Hits        16588    16582       -6     
- Misses        727      738      +11     

☔ 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.