Description
This task consists of a couple of sub-tasks.
1. Removing the map-join hint completely. Moving all processing to runtime. Currently, bucketed map joins and sort merge joins are completely
driven off hints.
2. Create a performance suite for performance for bucketed map-joins for different scenarios: the big table has many buckets, whereas the small table has fewer buckets etc.
3. Infer bucketing and sorting from the user query, and store that in the metastore. This can be used for follow-up queries on those outputs.