[GitHub] spark pull request #15704: [SPARK-17732][SQL] ALTER TABLE DROP PARTITION sho...

[DazhuangSu <gi...@git.apache.org>]

Github user DazhuangSu commented on a diff in the pull request:

    https://github.com/apache/spark/pull/15704#discussion_r132893104
  
    --- Diff: sql/core/src/main/scala/org/apache/spark/sql/execution/command/ddl.scala ---
    @@ -418,27 +419,55 @@ case class AlterTableRenamePartitionCommand(
      */
     case class AlterTableDropPartitionCommand(
         tableName: TableIdentifier,
    -    specs: Seq[TablePartitionSpec],
    +    specs: Seq[Expression],
         ifExists: Boolean,
         purge: Boolean)
    -  extends RunnableCommand {
    +  extends RunnableCommand with PredicateHelper {
    +
    +  private def isRangeComparison(expr: Expression): Boolean = {
    +    expr.find(e => e.isInstanceOf[BinaryComparison] && !e.isInstanceOf[EqualTo]).isDefined
    +  }
     
       override def run(sparkSession: SparkSession): Seq[Row] = {
         val catalog = sparkSession.sessionState.catalog
         val table = catalog.getTableMetadata(tableName)
    +    val resolver = sparkSession.sessionState.conf.resolver
         DDLUtils.verifyAlterTableType(catalog, table, isView = false)
         DDLUtils.verifyPartitionProviderIsHive(sparkSession, table, "ALTER TABLE DROP PARTITION")
     
    -    val normalizedSpecs = specs.map { spec =>
    -      PartitioningUtils.normalizePartitionSpec(
    -        spec,
    -        table.partitionColumnNames,
    -        table.identifier.quotedString,
    -        sparkSession.sessionState.conf.resolver)
    +    specs.foreach { expr =>
    +      expr.references.foreach { attr =>
    +        if (!table.partitionColumnNames.exists(resolver(_, attr.name))) {
    +          throw new AnalysisException(s"${attr.name} is not a valid partition column " +
    +            s"in table ${table.identifier.quotedString}.")
    +        }
    +      }
         }
     
    -    catalog.dropPartitions(
    -      table.identifier, normalizedSpecs, ignoreIfNotExists = ifExists, purge = purge)
    +    if (specs.exists(isRangeComparison)) {
    +      val partitionSet = specs.flatMap { spec =>
    +        val partitions = catalog.listPartitionsByFilter(table.identifier, Seq(spec)).map(_.spec)
    +        if (partitions.isEmpty && !ifExists) {
    +          throw new AnalysisException(s"There is no partition for ${spec.sql}")
    +        }
    +        partitions
    +      }.distinct
    +      catalog.dropPartitions(
    +        table.identifier, partitionSet, ignoreIfNotExists = ifExists, purge = purge)
    +    } else {
    +      val normalizedSpecs = specs.map { expr =>
    +        val spec = splitConjunctivePredicates(expr).map {
    +          case BinaryComparison(AttributeReference(name, _, _, _), right) => name -> right.toString
    +        }.toMap
    +        PartitioningUtils.normalizePartitionSpec(
    +          spec,
    +          table.partitionColumnNames,
    +          table.identifier.quotedString,
    +          resolver)
    +      }
    +      catalog.dropPartitions(
    +        table.identifier, normalizedSpecs, ignoreIfNotExists = ifExists, purge = purge)
    +    }
    --- End diff --
    
    The function named isRangeComparison treats BinaryComparison and EqualTo as 2 different scenarios. For BinaryComparison,  it lists partitions by filter which are defined by expressions. But for EqualTo, it gets TablePartitionSpec through PartitioningUtils.normalizePartitionSpec using spec(Map[String, String]).
    Why don’t we just treat EqualTo as a special case for BinaryComparison and let catalog.listPartitionsByFilter do the work to get TablePartitionSpec. 
    In my test, this can solve the problem when using int literal after patching this pr.


---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastructure@apache.org or file a JIRA ticket
with INFRA.
---

---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org