sbernauer edited a comment on issue #1845: URL: https://github.com/apache/hudi/issues/1845#issuecomment-663065111
Yes, to be correct you have to specify both reader and writer schema to GenericDatumReader. Hudi currenty only passes one schema here https://github.com/apache/hudi/blob/c3279cd5989805946267b046007ea23ba4b615c2/hudi-common/src/main/java/org/apache/hudi/avro/HoodieAvroUtils.java#L107 and here https://github.com/apache/hudi/blob/c3279cd5989805946267b046007ea23ba4b615c2/hudi-common/src/main/java/org/apache/hudi/avro/HoodieAvroUtils.java#L117 I think that's a problem I would have used GenericDatumReader with reader and writer schema, but first I wanted to append new evoluted events to an existing file. Looking at "Each datum in a file is of the same schema" this seems expected behavior to be impossible. "Each datum in a file is of the same schema" - How is the strategy of Hudi when there are Upserts for an evoluted event? The new events can't simply be written to an existing file. Every time a file must be updated, read it correctly with the **reader and writer** schema and than create a new file with the new schema? ---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org
