aokolnychyi commented on a change in pull request #32921:
URL: https://github.com/apache/spark/pull/32921#discussion_r662433500
##########
File path:
sql/core/src/main/scala/org/apache/spark/sql/execution/datasources/v2/BatchScanExec.scala
##########
@@ -17,38 +17,96 @@
package org.apache.spark.sql.execution.datasources.v2
+import com.google.common.base.Objects
+
+import org.apache.spark.SparkException
import org.apache.spark.rdd.RDD
import org.apache.spark.sql.catalyst.InternalRow
import org.apache.spark.sql.catalyst.expressions._
import org.apache.spark.sql.catalyst.plans.QueryPlan
-import org.apache.spark.sql.connector.read.{InputPartition,
PartitionReaderFactory, Scan}
+import org.apache.spark.sql.catalyst.plans.physical.SinglePartition
+import org.apache.spark.sql.catalyst.util.truncatedString
+import org.apache.spark.sql.connector.read.{InputPartition,
PartitionReaderFactory, Scan, SupportsRuntimeFiltering}
+import org.apache.spark.sql.execution.datasources.DataSourceStrategy
/**
* Physical plan node for scanning a batch of data from a data source v2.
*/
case class BatchScanExec(
output: Seq[AttributeReference],
- @transient scan: Scan) extends DataSourceV2ScanExecBase {
+ @transient scan: Scan,
+ runtimeFilters: Seq[Expression]) extends DataSourceV2ScanExecBase {
@transient lazy val batch = scan.toBatch
// TODO: unify the equal/hashCode implementation for all data source v2
query plans.
override def equals(other: Any): Boolean = other match {
- case other: BatchScanExec => this.batch == other.batch
- case _ => false
+ case other: BatchScanExec =>
+ this.batch == other.batch && this.runtimeFilters == other.runtimeFilters
+ case _ =>
+ false
}
- override def hashCode(): Int = batch.hashCode()
+ override def hashCode(): Int = Objects.hashCode(batch, runtimeFilters)
@transient override lazy val partitions: Seq[InputPartition] =
batch.planInputPartitions()
+ @transient private lazy val filteredPartitions: Seq[InputPartition] = {
+ val dataSourceFilters = runtimeFilters.flatMap {
+ case DynamicPruningExpression(e) =>
DataSourceStrategy.translateRuntimeFilter(e)
+ case _ => None
+ }
+
+ if (dataSourceFilters.nonEmpty) {
+ val originalPartitioning = outputPartitioning
+
+ // the cast is safe as runtime filters are only assigned if the scan can
be filtered
+ val filterableScan = scan.asInstanceOf[SupportsRuntimeFiltering]
+ filterableScan.filter(dataSourceFilters.toArray)
+
+ // call toBatch again to get filtered partitions
+ val newPartitions = scan.toBatch.planInputPartitions()
+
+ originalPartitioning match {
+ case p: DataSourcePartitioning if p.numPartitions !=
newPartitions.size =>
+ throw new SparkException(
+ "Data source must have preserved the original partitioning during
runtime filtering; " +
+ s"reported num partitions: ${p.numPartitions}, " +
+ s"num partitions after runtime filtering: ${newPartitions.size}")
+ case _ =>
+ // no validation is needed as the data source did not report any
specific partitioning
+ }
+
+ newPartitions
+ } else {
+ partitions
+ }
+ }
+
override lazy val readerFactory: PartitionReaderFactory =
batch.createReaderFactory()
override lazy val inputRDD: RDD[InternalRow] = {
- new DataSourceRDD(sparkContext, partitions, readerFactory,
supportsColumnar, customMetrics)
+ if (filteredPartitions.isEmpty && outputPartitioning == SinglePartition) {
Review comment:
We check the number of partitions match only if the source reported a
specific partitioning through `SupportsReportPartitioning`. Only in that case
we have `DataSourcePartitioning`. This situation, on the other hand, can happen
if we inferred `SinglePartition` but the source did not report anything.
--
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.
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
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
