Versions Compared

Key

  • This line was added.
  • This line was removed.
  • Formatting was changed.

...

HIVE-5119 has been created to track this issuefeature improvement.

Proposed Solution

Figure out the set of values from all small tables for each join column from big table (that is partition key). Using these set of values figure out the partitions from big table that should be scanned using metadata. Change the partitions to be scanned for big table before Map-Join starts streaming big table. This feature would be turned on only through an explicit configuration (name of that configuration is TBD).

...