I am not sure what is the correct thing to do here. I think that we should prevent users from explicitly adding duplicate fields (as far as I remember BQ SQL disallows duplicate fields, don't know about standard). But I don't know if this can be a valid thing to do internally. For example, when implementing a join there potentially can exist an intermediary joined schema that combines left and right side of the join. Ambiguities should be dealt with by projection which is applied after that. Does this sound like something that can happen?
[ Full content available at: https://github.com/apache/beam/pull/6316 ] This message was relayed via gitbox.apache.org for [email protected]
