yadavay-amzn commented on code in PR #56101:
URL: https://github.com/apache/spark/pull/56101#discussion_r3338528603


##########
sql/core/src/main/scala/org/apache/spark/sql/execution/joins/BroadcastNearestByJoinExec.scala:
##########
@@ -0,0 +1,149 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one or more
+ * contributor license agreements.  See the NOTICE file distributed with
+ * this work for additional information regarding copyright ownership.
+ * The ASF licenses this file to You under the Apache License, Version 2.0
+ * (the "License"); you may not use this file except in compliance with
+ * the License.  You may obtain a copy of the License at
+ *
+ *    http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package org.apache.spark.sql.execution.joins
+
+import java.util.{PriorityQueue => JPriorityQueue}
+
+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.{InnerLike, JoinType, LeftOuter, 
NearestByDirection, NearestByDistance}
+import org.apache.spark.sql.catalyst.plans.physical._
+import org.apache.spark.sql.execution.{BinaryExecNode, SparkPlan}
+import org.apache.spark.sql.execution.metric.SQLMetrics
+import org.apache.spark.sql.types.DoubleType
+
+/**
+ * Physical operator for NearestByJoin that avoids materializing the full 
cross product.
+ * For each left row, iterates all broadcast right rows maintaining a bounded 
priority
+ * queue of size k, then emits the top-k matches directly.
+ *
+ * The right side is fully broadcast to all partitions. This operator only 
fires when
+ * the right side fits within [[SQLConf.AUTO_BROADCAST_JOIN_THRESHOLD]]. For 
right tables
+ * exceeding this threshold, the existing cross-product + aggregate rewrite is 
used as
+ * fallback. Tie-breaking among equal ranking values is non-deterministic 
(matches the
+ * existing rewrite behavior).
+ */
+case class BroadcastNearestByJoinExec(
+    left: SparkPlan,
+    right: SparkPlan,
+    joinType: JoinType,
+    numResults: Int,
+    rankingExpression: Expression,
+    direction: NearestByDirection) extends BinaryExecNode {
+
+  override def output: Seq[Attribute] = joinType match {
+    case LeftOuter =>
+      left.output ++ right.output.map(_.withNullability(true))
+    case _: InnerLike =>
+      left.output ++ right.output
+    case other =>
+      throw SparkException.internalError(
+        s"$nodeName does not support join type: $other")
+  }
+
+  override lazy val metrics = Map(
+    "numOutputRows" -> SQLMetrics.createMetric(sparkContext, "number of output 
rows"),
+    "streamedRows" -> SQLMetrics.createMetric(sparkContext, "number of left 
rows processed"))
+
+  override def requiredChildDistribution: Seq[Distribution] =
+    UnspecifiedDistribution :: BroadcastDistribution(IdentityBroadcastMode) :: 
Nil
+
+  override def outputPartitioning: Partitioning = left.outputPartitioning
+
+  override def outputOrdering: Seq[SortOrder] = Nil
+
+  protected override def doExecute(): RDD[InternalRow] = {
+    val broadcastedRight = right.executeBroadcast[Array[InternalRow]]()
+    val numOutput = longMetric("numOutputRows")
+    val k = numResults
+    val isDistance = direction == NearestByDistance
+    val leftOutput = left.output
+    val rightOutput = right.output
+    val rankExpr = rankingExpression
+    val allOutput = output
+
+    left.execute().mapPartitionsInternal { leftIter =>
+      val rightRows = broadcastedRight.value
+      if (rightRows.isEmpty && joinType != LeftOuter) {
+        Iterator.empty
+      } else {
+        val joinedRow = new JoinedRow
+        val rankingProj = UnsafeProjection.create(
+          Seq(Cast(rankExpr, DoubleType)), leftOutput ++ rightOutput)
+        val resultProj = UnsafeProjection.create(allOutput, allOutput)
+        val streamedRowsMetric = longMetric("streamedRows")
+
+        // Hoist heap outside flatMap to reduce GC pressure

Review Comment:
   Replaced `(Int, Any)` tuples with `private[joins] case class 
HeapEntry(index: Int, rankingValue: Any)` — the `Int` field is stored unboxed. 
Also switched from SAM lambda comparators to explicit `Comparator[HeapEntry]` 
instances.



-- 
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: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]


---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to