cloud-fan commented on a change in pull request #26231: [SPARK-29572][SQL] add
v1 read fallback API in DS v2
URL: https://github.com/apache/spark/pull/26231#discussion_r367283434
##########
File path:
sql/core/src/main/scala/org/apache/spark/sql/execution/datasources/v2/DataSourceV2Strategy.scala
##########
@@ -19,40 +19,72 @@ package org.apache.spark.sql.execution.datasources.v2
import scala.collection.JavaConverters._
-import org.apache.spark.sql.{AnalysisException, Strategy}
+import org.apache.spark.sql.{AnalysisException, SparkSession, Strategy}
import org.apache.spark.sql.catalyst.analysis.{ResolvedNamespace,
ResolvedTable}
-import org.apache.spark.sql.catalyst.expressions.{And, PredicateHelper,
SubqueryExpression}
+import org.apache.spark.sql.catalyst.expressions.{And, Expression,
NamedExpression, PredicateHelper, SubqueryExpression}
import org.apache.spark.sql.catalyst.planning.PhysicalOperation
import org.apache.spark.sql.catalyst.plans.logical.{AlterNamespaceSetLocation,
AlterNamespaceSetOwner, AlterNamespaceSetProperties, AlterTable, AppendData,
CommentOnNamespace, CommentOnTable, CreateNamespace, CreateTableAsSelect,
CreateV2Table, DeleteFromTable, DescribeNamespace, DescribeTable,
DropNamespace, DropTable, LogicalPlan, OverwriteByExpression,
OverwritePartitionsDynamic, RefreshTable, RenameTable, Repartition,
ReplaceTable, ReplaceTableAsSelect, SetCatalogAndNamespace,
ShowCurrentNamespace, ShowNamespaces, ShowTableProperties, ShowTables}
-import org.apache.spark.sql.connector.catalog.{Identifier,
StagingTableCatalog, SupportsNamespaces, TableCapability, TableCatalog,
TableChange}
+import org.apache.spark.sql.connector.catalog.{StagingTableCatalog,
SupportsNamespaces, TableCapability, TableCatalog, TableChange}
+import org.apache.spark.sql.connector.read.V1Scan
import org.apache.spark.sql.connector.read.streaming.{ContinuousStream,
MicroBatchStream}
-import org.apache.spark.sql.execution.{FilterExec, ProjectExec, SparkPlan}
+import org.apache.spark.sql.execution.{FilterExec, LeafExecNode, ProjectExec,
RowDataSourceScanExec, SparkPlan}
import org.apache.spark.sql.execution.datasources.DataSourceStrategy
import
org.apache.spark.sql.execution.streaming.continuous.{ContinuousCoalesceExec,
WriteToContinuousDataSource, WriteToContinuousDataSourceExec}
+import org.apache.spark.sql.sources.{BaseRelation, TableScan}
import org.apache.spark.sql.util.CaseInsensitiveStringMap
-object DataSourceV2Strategy extends Strategy with PredicateHelper {
+class DataSourceV2Strategy(session: SparkSession) extends Strategy with
PredicateHelper {
import DataSourceV2Implicits._
import org.apache.spark.sql.connector.catalog.CatalogV2Implicits._
+ private def withProjectAndFilter(
+ project: Seq[NamedExpression],
+ filters: Seq[Expression],
+ scan: LeafExecNode,
+ needsUnsafeConversion: Boolean): SparkPlan = {
+ val filterCondition = filters.reduceLeftOption(And)
+ val withFilter = filterCondition.map(FilterExec(_, scan)).getOrElse(scan)
+
+ if (withFilter.output != project || needsUnsafeConversion) {
+ ProjectExec(project, withFilter)
+ } else {
+ withFilter
+ }
+ }
+
override def apply(plan: LogicalPlan): Seq[SparkPlan] = plan match {
+ case PhysicalOperation(project, filters,
+ relation @ DataSourceV2ScanRelation(_, v1Scan: V1Scan, output)) =>
+ val pushedFilters =
relation.getTagValue(V2ScanRelationPushDown.PUSHED_FILTERS_TAG)
+ .getOrElse(Seq.empty)
+ val v1Relation = v1Scan.toV1TableScan[BaseRelation with
TableScan](session.sqlContext)
+ if (v1Relation.schema != v1Scan.readSchema()) {
+ throw new IllegalArgumentException(
+ "The fallback v1 relation reports inconsistent schema:\n" +
+ "Schema of v2 scan: " + v1Scan.readSchema() + "\n" +
+ "Schema of v1 relation: " + v1Relation.schema)
+ }
+ val rdd = v1Relation.buildScan()
+ val unsafeRowRDD = DataSourceStrategy.toCatalystRDD(v1Relation, output,
rdd)
+ val originalOutputNames = relation.table.schema().map(_.name)
+ val requiredColumnsIndex =
output.map(_.name).map(originalOutputNames.indexOf)
+ val dsScan = RowDataSourceScanExec(
+ output,
+ requiredColumnsIndex,
+ pushedFilters.toSet,
Review comment:
After some more thoughts, I think we should only track the pushed filters.
In data source v1, what we track is all the filters Spark pushes down, and
the ones accepted by the source, which is calculated by `all filters --
post-scan filters`. This is actually imprecise. For example, parquet source can
leverage the data filters but still need Spark to evaluate the filter again, so
they are post-scan filters and won't be treated as "accepted filters".
In data source v2, we fix it by asking the source to report the pushed
filters (via `SupportsPushDownFiters.pushedFilters`), and we only report the
pushed filters in plan string. v1 fallback API is still part of DS v2 and we
should follow the v2 way to report filters.
----------------------------------------------------------------
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:
[email protected]
With regards,
Apache Git Services
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
