faircode/loaders_extra.py's .json branch tries pd.read_json(path) then falls back to orient="split" on ValueError, and tests cover exactly those two orientations (records, split). Pandas' other orientations (index, columns, values, table) aren't tested, and it's unclear whether the current two-attempt fallback would silently produce a wrong-shaped DataFrame for one of them (a dict-of-arrays can parse "successfully" into the wrong shape without raising) rather than erroring out.
Request: either document the two supported orientations explicitly in the CLI help/README as the only guaranteed ones, or add test coverage (and if needed, explicit orient detection) for the remaining pandas orientations.
faircode/loaders_extra.py's.jsonbranch triespd.read_json(path)then falls back toorient="split"onValueError, and tests cover exactly those two orientations (records,split). Pandas' other orientations (index,columns,values,table) aren't tested, and it's unclear whether the current two-attempt fallback would silently produce a wrong-shaped DataFrame for one of them (a dict-of-arrays can parse "successfully" into the wrong shape without raising) rather than erroring out.Request: either document the two supported orientations explicitly in the CLI help/README as the only guaranteed ones, or add test coverage (and if needed, explicit orient detection) for the remaining pandas orientations.