Where: faircode/loaders_extra.py's JSON reading path (same area as the split/index-orient handling).
The gap: pandas.DataFrame.to_json(orient="table") - a real, valid, non-obscure pandas JSON export format (includes a "schema" key alongside "data") - isn't recognized at all. Verified directly:
>>> df.to_json('table.json', orient='table')
>>> from faircode.loaders_extra import read_table
>>> read_table('table.json')
ValueError: Mixing dicts with non-Series may lead to ambiguous ordering.That message comes from deep inside pandas/core/internals/construction.py - exactly the kind of "internal, version-specific message" this module's own comment (describing why it special-cases split-orient) says it wants to avoid, but here for a completely valid JSON file pandas itself produces.
Why it matters: a user who exported their data with to_json(orient="table") (a reasonable choice - it's the only orientation that round-trips dtypes) gets a confusing, implementation-detail error instead of a clear "unsupported/unrecognized JSON shape" message.
Suggested fix: detect the {"schema": ..., "data": ...} shape the same way split-orient is already detected, and either dispatch to pd.read_json(path, orient="table") or raise the module's own clear ValueError for it.
Where:
faircode/loaders_extra.py's JSON reading path (same area as the split/index-orient handling).The gap:
pandas.DataFrame.to_json(orient="table")- a real, valid, non-obscure pandas JSON export format (includes a"schema"key alongside"data") - isn't recognized at all. Verified directly:That message comes from deep inside
pandas/core/internals/construction.py- exactly the kind of "internal, version-specific message" this module's own comment (describing why it special-cases split-orient) says it wants to avoid, but here for a completely valid JSON file pandas itself produces.Why it matters: a user who exported their data with
to_json(orient="table")(a reasonable choice - it's the only orientation that round-trips dtypes) gets a confusing, implementation-detail error instead of a clear "unsupported/unrecognized JSON shape" message.Suggested fix: detect the
{"schema": ..., "data": ...}shape the same way split-orient is already detected, and either dispatch topd.read_json(path, orient="table")or raise the module's own clearValueErrorfor it.