Hi,

Currently the ORC reader can read ORC metadata of a table. However it
doesn’t actually write the obtained metadata into the schema of the
resulting Arrow Table. The ORC writer can not write either field-specific
or schema-level metadata from Arrow Tables into ORC.

I wonder whether we should change the current behavior and read ORC
metadata as schema-level metadata in Arrow.

As for writing I’m not sure whether we should begin to write schema-level
metadata as ORC metadata. Maybe we can make that optional? It is even less
clear what we should do about the field-level metadata. Shall we add the
field name & an underscore as a prefix and write it to ORC? That doesn’t
look very elegant.

Thanks,
Ian

Reply via email to