[GitHub] [iceberg] rdblue commented on a change in pull request #2141: Spark: Add DistributionAndOrderingUtils

[GitBox <gi...@apache.org>]

rdblue commented on a change in pull request #2141:
URL: https://github.com/apache/iceberg/pull/2141#discussion_r563949252



##########
File path: spark3-extensions/src/main/scala/org/apache/spark/sql/catalyst/optimizer/RewriteMergeInto.scala
##########
@@ -233,47 +215,23 @@ case class RewriteMergeInto(spark: SparkSession) extends Rule[LogicalPlan] with
     !(actions.size == 1 && hasUnconditionalDelete(actions.headOption))
   }
 
-  def buildWritePlan(
-     childPlan: LogicalPlan,
-     table: Table): LogicalPlan = {
-    val defaultDistributionMode = table match {
-      case iceberg: SparkTable if !iceberg.table.sortOrder.isUnsorted =>
-        TableProperties.WRITE_DISTRIBUTION_MODE_RANGE
-      case _ =>
-        TableProperties.WRITE_DISTRIBUTION_MODE_DEFAULT
-    }
-
+  private def buildWritePlan(childPlan: LogicalPlan, table: Table): LogicalPlan = {
     table match {
-      case iceTable: SparkTable =>
-        val numShufflePartitions = spark.sessionState.conf.numShufflePartitions
-        val table = iceTable.table()
-        val distributionMode: String = table.properties
-          .getOrDefault(TableProperties.WRITE_DISTRIBUTION_MODE, defaultDistributionMode)
-        val order = toCatalyst(toOrderedDistribution(table.spec(), table.sortOrder(), true), childPlan)
-        DistributionMode.fromName(distributionMode) match {
-          case DistributionMode.NONE =>
-            Sort(buildSortOrder(order), global = false, childPlan)
-          case DistributionMode.HASH =>
-            val clustering = toCatalyst(toClusteredDistribution(table.spec()), childPlan)
-            val hashPartitioned = RepartitionByExpression(clustering, childPlan, numShufflePartitions)
-            Sort(buildSortOrder(order), global = false, hashPartitioned)
-          case DistributionMode.RANGE =>
-            val roundRobin = Repartition(numShufflePartitions, shuffle = true, childPlan)
-            Sort(buildSortOrder(order), global = true, roundRobin)
+      case iceberg: SparkTable =>
+        val distribution = Spark3Util.buildRequiredDistribution(iceberg.table)

Review comment:
       Can we expose the same interface as in Spark 3.2 from `Table` and use that here instead? Then `Spark3Util` calls remain in `SparkTable`.




----------------------------------------------------------------
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:
users@infra.apache.org



---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscribe@iceberg.apache.org
For additional commands, e-mail: issues-help@iceberg.apache.org