Hi Jie, Let's suppose we have ((dimension_table Join fact_table1) join fact_table2). In the case where (dimension_table JOIN fact_table1) is small enough, the result ideally can be treated as another dimension table and thus used to prune the fact_table2. I don't find an easy way to implement it though.
On Fri, Nov 11, 2022 at 4:32 PM Jie Han <tunyu...@gmail.com> wrote: > FYI, > https://medium.com/@prabhakaran.electric/spark-3-0-feature-dynamic-partition-pruning-dpp-to-avoid-scanning-irrelevant-data-1a7bbd006a89 > > This blog may be helpful. Dynamic pruning often works for star schema > queries. So, your fact table is big_table which is used to join the others. > So there’s only one subqueryboradcast dynamicpruning plan before > big_table’s scan while there’s none for the others. > > I’m not sure that I’m correct. Hope it’s helpful to you. > > 2022年11月11日 21:43,hajyoussef amine <hajyoussef.am...@gmail.com> 写道: > > SubqueryBroadcast > > >