[SPARK-26865] DataSourceV2Strategy should push normalized filters

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 3.0.0
• Fix Version/s: 3.0.0
• Component/s: SQL
• Assignee: Dongjoon Hyun
• Reporter: Wenchen Fan
• Created: 13/Feb/19 03:03
• Updated: 14/Feb/19 00:08
• Resolved: 14/Feb/19 00:06

Description

Although we designed `SupportsPushDownFilters` in the same way by using `Filter`. DSv1 and DSv2 passes different filters.

  /**
   * Pushes down filters, and returns filters that need to be evaluated after scanning.
   */
  Filter[] pushFilters(Filter[] filters);

Specifically, DSv2 doesn't guarantee that filter expressions match the underlying schema in terms of case-sensitivity.

buildReaderWithPartitionValues(..., filters: Seq[Filter], ...)
- IsNotNull(ID)

DataSourceV2Strategy.pushFilters
- IsNotNull(id)

steps to reproduce:

spark.range(10).write.orc("/tmp/o1")
spark.read.schema("ID long").orc("/tmp/o1").filter("id > 5").show

java.util.NoSuchElementException: key not found: id
  at scala.collection.immutable.Map$Map1.apply(Map.scala:114)
  at org.apache.spark.sql.execution.datasources.orc.OrcFilters$.createBuilder(OrcFilters.scala:263)
  at org.apache.spark.sql.execution.datasources.orc.OrcFilters$.buildSearchArgument(OrcFilters.scala:153)
  at org.apache.spark.sql.execution.datasources.orc.OrcFilters$.$anonfun$convertibleFilters$1(OrcFilters.scala:99)
  at scala.collection.TraversableLike.$anonfun$flatMap$1(TraversableLike.scala:244)
  at scala.collection.IndexedSeqOptimized.foreach(IndexedSeqOptimized.scala:36)
  at scala.collection.IndexedSeqOptimized.foreach$(IndexedSeqOptimized.scala:33)
  at scala.collection.mutable.WrappedArray.foreach(WrappedArray.scala:39)
  at scala.collection.TraversableLike.flatMap(TraversableLike.scala:244)
  at scala.collection.TraversableLike.flatMap$(TraversableLike.scala:241)
  at scala.collection.AbstractTraversable.flatMap(Traversable.scala:108)
  at org.apache.spark.sql.execution.datasources.orc.OrcFilters$.convertibleFilters(OrcFilters.scala:98)
  at org.apache.spark.sql.execution.datasources.orc.OrcFilters$.createFilter(OrcFilters.scala:87)
  at org.apache.spark.sql.execution.datasources.v2.orc.OrcScanBuilder.pushFilters(OrcScanBuilder.scala:50)

Issue Links

• is caused by: SPARK-23817 Create file source V2 framework and migrate ORC read path
• links to: GitHub Pull Request #23770