[
https://issues.apache.org/jira/browse/SPARK-35316?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Yuming Wang updated SPARK-35316:
--------------------------------
Description:
It will not pushdown filters for In/InSet predicates:
{code:scala}
spark.range(50).selectExpr("cast(id as int) as
id").write.mode("overwrite").parquet("/tmp/parquet/t1")
spark.read.parquet("/tmp/parquet/t1").where("id in (1L, 2L, 4L)").explain
spark.read.parquet("/tmp/parquet/t1").where("id = 1L or id = 2L or id =
4L").explain
{code}
{noformat}
== Physical Plan ==
*(1) Filter cast(id#5 as bigint) IN (1,2,4)
+- *(1) ColumnarToRow
+- FileScan parquet [id#5] Batched: true, DataFilters: [cast(id#5 as bigint)
IN (1,2,4)], Format: Parquet, Location: InMemoryFileIndex(1
paths)[file:/tmp/parquet/t1], PartitionFilters: [], PushedFilters: [],
ReadSchema: struct<id:int>
== Physical Plan ==
*(1) Filter (((id#7 = 1) OR (id#7 = 2)) OR (id#7 = 4))
+- *(1) ColumnarToRow
+- FileScan parquet [id#7] Batched: true, DataFilters: [(((id#7 = 1) OR
(id#7 = 2)) OR (id#7 = 4))], Format: Parquet, Location: InMemoryFileIndex(1
paths)[file:/tmp/parquet/t1], PartitionFilters: [], PushedFilters:
[Or(Or(EqualTo(id,1),EqualTo(id,2)),EqualTo(id,4))], ReadSchema: struct<id:int>
{noformat}
was:
In/InSet missing this optimization:
{code:scala}
spark.range(50).selectExpr("cast(id as int) as
id").write.mode("overwrite").parquet("/tmp/parquet/t1")
spark.read.parquet("/tmp/parquet/t1").where("id in (1L, 2L, 4L)").explain
spark.read.parquet("/tmp/parquet/t1").where("id = 1L or id = 2L or id =
4L").explain
{code}
{noformat}
== Physical Plan ==
*(1) Filter cast(id#5 as bigint) IN (1,2,4)
+- *(1) ColumnarToRow
+- FileScan parquet [id#5] Batched: true, DataFilters: [cast(id#5 as bigint)
IN (1,2,4)], Format: Parquet, Location: InMemoryFileIndex(1
paths)[file:/tmp/parquet/t1], PartitionFilters: [], PushedFilters: [],
ReadSchema: struct<id:int>
== Physical Plan ==
*(1) Filter (((id#7 = 1) OR (id#7 = 2)) OR (id#7 = 4))
+- *(1) ColumnarToRow
+- FileScan parquet [id#7] Batched: true, DataFilters: [(((id#7 = 1) OR
(id#7 = 2)) OR (id#7 = 4))], Format: Parquet, Location: InMemoryFileIndex(1
paths)[file:/tmp/parquet/t1], PartitionFilters: [], PushedFilters:
[Or(Or(EqualTo(id,1),EqualTo(id,2)),EqualTo(id,4))], ReadSchema: struct<id:int>
{noformat}
> UnwrapCastInBinaryComparison support In/InSet predicate
> -------------------------------------------------------
>
> Key: SPARK-35316
> URL: https://issues.apache.org/jira/browse/SPARK-35316
> Project: Spark
> Issue Type: Improvement
> Components: SQL
> Affects Versions: 3.2.0
> Reporter: Yuming Wang
> Priority: Major
>
> It will not pushdown filters for In/InSet predicates:
> {code:scala}
> spark.range(50).selectExpr("cast(id as int) as
> id").write.mode("overwrite").parquet("/tmp/parquet/t1")
> spark.read.parquet("/tmp/parquet/t1").where("id in (1L, 2L, 4L)").explain
> spark.read.parquet("/tmp/parquet/t1").where("id = 1L or id = 2L or id =
> 4L").explain
> {code}
> {noformat}
> == Physical Plan ==
> *(1) Filter cast(id#5 as bigint) IN (1,2,4)
> +- *(1) ColumnarToRow
> +- FileScan parquet [id#5] Batched: true, DataFilters: [cast(id#5 as
> bigint) IN (1,2,4)], Format: Parquet, Location: InMemoryFileIndex(1
> paths)[file:/tmp/parquet/t1], PartitionFilters: [], PushedFilters: [],
> ReadSchema: struct<id:int>
> == Physical Plan ==
> *(1) Filter (((id#7 = 1) OR (id#7 = 2)) OR (id#7 = 4))
> +- *(1) ColumnarToRow
> +- FileScan parquet [id#7] Batched: true, DataFilters: [(((id#7 = 1) OR
> (id#7 = 2)) OR (id#7 = 4))], Format: Parquet, Location: InMemoryFileIndex(1
> paths)[file:/tmp/parquet/t1], PartitionFilters: [], PushedFilters:
> [Or(Or(EqualTo(id,1),EqualTo(id,2)),EqualTo(id,4))], ReadSchema:
> struct<id:int>
> {noformat}
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
