Note: migrated from original JIRA: https://issues.apache.org/jira/browse/ARROW-11991
Currently it returns Partitioning::UnknownPartitioning(num_partitions) based on the sum of partition counts (if available) of the underlying inputs. In case of another partition scheme - such as hash partitioning, it would be better to keep that information available, so other optimizations could be used.
Note: migrated from original JIRA: https://issues.apache.org/jira/browse/ARROW-11991
Currently it returns Partitioning::UnknownPartitioning(num_partitions) based on the sum of partition counts (if available) of the underlying inputs. In case of another partition scheme - such as hash partitioning, it would be better to keep that information available, so other optimizations could be used.