wangyum commented on pull request #29243: URL: https://github.com/apache/spark/pull/29243#issuecomment-676111794
Before this pr. We can only prune on `fact_stats.store_id` column: ``` == Physical Plan == *(3) Project [store_id#2705, code#2706, product_id#2708L] +- *(3) BroadcastHashJoin [cast(store_id#2705 as bigint)], [store_id#2709L], Inner, BuildRight, false :- *(3) Project [store_id#2705, code#2706] : +- *(3) BroadcastHashJoin [store_id#2705], [store_id#2707], Inner, BuildRight, false : :- *(3) ColumnarToRow : : +- FileScan parquet default.fact_stats[store_id#2705] Batched: true, DataFilters: [], Format: Parquet, Location: InMemoryFileIndex[file:/Users/yumwang/spark/SPARK-27227/sql/core/spark-warehouse/org.apache.spark..., PartitionFilters: [isnotnull(store_id#2705), dynamicpruningexpression(cast(store_id#2705 as bigint) IN subquery#2716)], PushedFilters: [], ReadSchema: struct<> : : +- Subquery subquery#2716, [id=#245] : : +- *(2) HashAggregate(keys=[store_id#2709L#2715L], functions=[]) : : +- Exchange hashpartitioning(store_id#2709L#2715L, 5), true, [id=#241] : : +- *(1) HashAggregate(keys=[store_id#2709L AS store_id#2709L#2715L], functions=[]) : : +- *(1) Filter ((isnotnull(product_id#2708L) AND (product_id#2708L < 3)) AND isnotnull(store_id#2709L)) : : +- *(1) ColumnarToRow : : +- FileScan parquet default.product[product_id#2708L,store_id#2709L] Batched: true, DataFilters: [isnotnull(product_id#2708L), (product_id#2708L < 3), isnotnull(store_id#2709L)], Format: Parquet, Location: InMemoryFileIndex[file:/Users/yumwang/spark/SPARK-27227/sql/core/spark-warehouse/org.apache.spark..., PartitionFilters: [], PushedFilters: [IsNotNull(product_id), LessThan(product_id,3), IsNotNull(store_id)], ReadSchema: struct<product_id:bigint,store_id:bigint> : +- BroadcastExchange HashedRelationBroadcastMode(List(cast(input[1, int, true] as bigint)),false), [id=#275] : +- *(1) ColumnarToRow : +- FileScan parquet default.code_stats[code#2706,store_id#2707] Batched: true, DataFilters: [], Format: Parquet, Location: InMemoryFileIndex[file:/Users/yumwang/spark/SPARK-27227/sql/core/spark-warehouse/org.apache.spark..., PartitionFilters: [isnotnull(store_id#2707)], PushedFilters: [], ReadSchema: struct<code:int> +- BroadcastExchange HashedRelationBroadcastMode(List(input[1, bigint, false]),false), [id=#283] +- *(2) Filter ((isnotnull(product_id#2708L) AND (product_id#2708L < 3)) AND isnotnull(store_id#2709L)) +- *(2) ColumnarToRow +- FileScan parquet default.product[product_id#2708L,store_id#2709L] Batched: true, DataFilters: [isnotnull(product_id#2708L), (product_id#2708L < 3), isnotnull(store_id#2709L)], Format: Parquet, Location: InMemoryFileIndex[file:/Users/yumwang/spark/SPARK-27227/sql/core/spark-warehouse/org.apache.spark..., PartitionFilters: [], PushedFilters: [IsNotNull(product_id), LessThan(product_id,3), IsNotNull(store_id)], ReadSchema: struct<product_id:bigint,store_id:bigint> ``` After this pr. We can also prune on `code_stats.store_id` column: ``` == Physical Plan == *(3) Project [store_id#2705, code#2706, product_id#2708L] +- *(3) BroadcastHashJoin [cast(store_id#2705 as bigint)], [store_id#2709L], Inner, BuildRight, false :- *(3) Project [store_id#2705, code#2706] : +- *(3) BroadcastHashJoin [store_id#2705], [store_id#2707], Inner, BuildRight, false : :- *(3) ColumnarToRow : : +- FileScan parquet default.fact_stats[store_id#2705] Batched: true, DataFilters: [], Format: Parquet, Location: InMemoryFileIndex[file:/Users/yumwang/spark/SPARK-27227/sql/core/spark-warehouse/org.apache.spark..., PartitionFilters: [isnotnull(store_id#2705), dynamicpruningexpression(cast(store_id#2705 as bigint) IN subquery#2716)], PushedFilters: [], ReadSchema: struct<> : : +- Subquery subquery#2716, [id=#250] : : +- *(2) HashAggregate(keys=[store_id#2709L#2715L], functions=[]) : : +- Exchange hashpartitioning(store_id#2709L#2715L, 5), true, [id=#246] : : +- *(1) HashAggregate(keys=[store_id#2709L AS store_id#2709L#2715L], functions=[]) : : +- *(1) Filter ((isnotnull(product_id#2708L) AND (product_id#2708L < 3)) AND isnotnull(store_id#2709L)) : : +- *(1) ColumnarToRow : : +- FileScan parquet default.product[product_id#2708L,store_id#2709L] Batched: true, DataFilters: [isnotnull(product_id#2708L), (product_id#2708L < 3), isnotnull(store_id#2709L)], Format: Parquet, Location: InMemoryFileIndex[file:/Users/yumwang/spark/SPARK-27227/sql/core/spark-warehouse/org.apache.spark..., PartitionFilters: [], PushedFilters: [IsNotNull(product_id), LessThan(product_id,3), IsNotNull(store_id)], ReadSchema: struct<product_id:bigint,store_id:bigint> : +- BroadcastExchange HashedRelationBroadcastMode(List(cast(input[1, int, true] as bigint)),false), [id=#309] : +- *(1) ColumnarToRow : +- FileScan parquet default.code_stats[code#2706,store_id#2707] Batched: true, DataFilters: [], Format: Parquet, Location: InMemoryFileIndex[file:/Users/yumwang/spark/SPARK-27227/sql/core/spark-warehouse/org.apache.spark..., PartitionFilters: [isnotnull(store_id#2707), dynamicpruningexpression(cast(store_id#2707 as bigint) IN subquery#2718)], PushedFilters: [], ReadSchema: struct<code:int> : +- Subquery subquery#2718, [id=#279] : +- *(2) HashAggregate(keys=[store_id#2709L#2717L], functions=[]) : +- Exchange hashpartitioning(store_id#2709L#2717L, 5), true, [id=#275] : +- *(1) HashAggregate(keys=[store_id#2709L AS store_id#2709L#2717L], functions=[]) : +- *(1) Filter ((isnotnull(product_id#2708L) AND (product_id#2708L < 3)) AND isnotnull(store_id#2709L)) : +- *(1) ColumnarToRow : +- FileScan parquet default.product[product_id#2708L,store_id#2709L] Batched: true, DataFilters: [isnotnull(product_id#2708L), (product_id#2708L < 3), isnotnull(store_id#2709L)], Format: Parquet, Location: InMemoryFileIndex[file:/Users/yumwang/spark/SPARK-27227/sql/core/spark-warehouse/org.apache.spark..., PartitionFilters: [], PushedFilters: [IsNotNull(product_id), LessThan(product_id,3), IsNotNull(store_id)], ReadSchema: struct<product_id:bigint,store_id:bigint> +- BroadcastExchange HashedRelationBroadcastMode(List(input[1, bigint, false]),false), [id=#317] +- *(2) Filter ((isnotnull(product_id#2708L) AND (product_id#2708L < 3)) AND isnotnull(store_id#2709L)) +- *(2) ColumnarToRow +- FileScan parquet default.product[product_id#2708L,store_id#2709L] Batched: true, DataFilters: [isnotnull(product_id#2708L), (product_id#2708L < 3), isnotnull(store_id#2709L)], Format: Parquet, Location: InMemoryFileIndex[file:/Users/yumwang/spark/SPARK-27227/sql/core/spark-warehouse/org.apache.spark..., PartitionFilters: [], PushedFilters: [IsNotNull(product_id), LessThan(product_id,3), IsNotNull(store_id)], ReadSchema: struct<product_id:bigint,store_id:bigint> ``` ---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org --------------------------------------------------------------------- To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org For additional commands, e-mail: reviews-h...@spark.apache.org