Details
-
Improvement
-
Status: Open
-
Major
-
Resolution: Unresolved
-
1.16.0
-
None
-
None
Description
When using different Apache Parquet files in the same request, if the Mode(optional/required) is not the same for certain columns, it can produce some errors.
Example:
Error: UNSUPPORTED_OPERATION ERROR: Hash aggregate does not support
schema change
It's very annoying. So why not propose an option that allow to read NOT NULL columns as NULLABLE columns.
I imagine that certain optimizations can't be realized in this case but the queries will work without having to take again the datasets