_read_or_exit() in faircode/cli.py is used for both the main dataset argument and --reference, so the reference baseline file already goes through the same loaders_extra.read_table() and supports .tsv/.xlsx/.json/.parquet - but the argparse help text still reads:
--reference PATH score against a reference baseline CSV (columns: column,group,share)
Small, low-risk doc fix: drop "CSV" (or say "file") so the help text matches actual behavior.
_read_or_exit()infaircode/cli.pyis used for both the main dataset argument and--reference, so the reference baseline file already goes through the sameloaders_extra.read_table()and supports.tsv/.xlsx/.json/.parquet- but the argparse help text still reads:Small, low-risk doc fix: drop "CSV" (or say "file") so the help text matches actual behavior.