The `endswith(ts)` check in `get_column_groups` matched columns
incorrectly when time shifts shared a numeric suffix — e.g.
"metric__22 weeks ago".endswith("2 weeks ago") returned True,
causing columns to be misclassified. Use `endswith(TIME_COMPARISON + ts)`
to include the "__" separator in the match.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* Bump pandas 1.4 and pyarrow 6
* Use engine="pyarrow" for pd.read_csv()
* Refactoring
* Refactoring
* Refactoring
* Use bytes in pd.read_json()
* Fix test_contribution
* Fix pandas issue when 'arrays' are empty but 'names' contain values
* fix: ValueError: For argument "ascending" expected type bool, received type NoneType.
* Remove engine="pyarrow" and convert bytes to string
* make copy of selected df to fix regression
* Simplify pd.read_json() and pd.read_csv() for example data
Co-authored-by: Ville Brofeldt <ville.brofeldt@apple.com>
