[SPARK-43781][SQL] Fix IllegalStateException when cogrouping two datasets derived from the same source

sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/analysis/DeduplicateRelations.scala

@@ -19,7 +19,7 @@ package org.apache.spark.sql.catalyst.analysis
 
 import scala.collection.mutable
 
-import org.apache.spark.sql.catalyst.expressions.{Alias, AttributeMap, AttributeSet, NamedExpression, OuterReference, SubqueryExpression}
+import org.apache.spark.sql.catalyst.expressions.{Alias, Attribute, AttributeMap, AttributeReference, AttributeSet, Expression, NamedExpression, OuterReference, SubqueryExpression}
 import org.apache.spark.sql.catalyst.plans.logical._
 import org.apache.spark.sql.catalyst.rules.Rule
 import org.apache.spark.sql.catalyst.trees.TreePattern._
@@ -220,7 +220,46 @@ object DeduplicateRelations extends Rule[LogicalPlan] {
           if (attrMap.isEmpty) {
             planWithNewChildren
           } else {
-            planWithNewChildren.rewriteAttrs(attrMap)
+            def rewriteAttrsMatchWithSubPlan[T <: Expression](
+                attrs: Seq[T],
+                attrMap: AttributeMap[Attribute],
+                planOutput: Seq[Attribute]): Seq[T] = {
+              val canRewriteAttrs = attrMap.filter(a => planOutput.contains(a._2))
+              attrs.map(attr => {
+                attr.transformWithPruning(_.containsPattern(ATTRIBUTE_REFERENCE)) {
+                  case a: AttributeReference =>
+                    canRewriteAttrs.getOrElse(a, a)
+                }.asInstanceOf[T]
+              })
+            }
+
+            planWithNewChildren match {
+              case c: CoGroup =>
+                // SPARK-43781: CoGroup is a special case, `rewriteAttrs` will incorrectly update
+                // some fields that do not need to be updated. We need to update the output
+                // attributes of CoGroup manually.
+                val newLeftAttr = c.leftAttr.map(attr => attrMap.getOrElse(attr, attr))
+                val newRightAttr = rewriteAttrsMatchWithSubPlan(c.rightAttr, attrMap,
+                  c.right.output)
+                val newLeftGroup = rewriteAttrsMatchWithSubPlan(c.leftGroup, attrMap, c.left.output)
+                val newRightGroup = rewriteAttrsMatchWithSubPlan(c.rightGroup, attrMap,
+                  c.right.output)
+                val newLeftOrder = rewriteAttrsMatchWithSubPlan(c.leftOrder, attrMap,
+                  c.left.output)
+                val newRightOrder = rewriteAttrsMatchWithSubPlan(c.rightOrder, attrMap,
+                  c.right.output)
+                val newKeyDes = c.keyDeserializer.asInstanceOf[UnresolvedDeserializer]
+                  .copy(inputAttributes = newLeftGroup)
+                val newLeftDes = c.leftDeserializer.asInstanceOf[UnresolvedDeserializer]
+                  .copy(inputAttributes = newLeftAttr)
+                val newRightDes = c.rightDeserializer.asInstanceOf[UnresolvedDeserializer]
+                  .copy(inputAttributes = newRightAttr)
+                c.copy(keyDeserializer = newKeyDes, leftDeserializer = newLeftDes,
+                  rightDeserializer = newRightDes, leftGroup = newLeftGroup,
+                  rightGroup = newRightGroup, leftAttr = newLeftAttr, rightAttr = newRightAttr,
+                  leftOrder = newLeftOrder, rightOrder = newRightOrder)
+              case _ => planWithNewChildren.rewriteAttrs(attrMap)
+            }
           }
         } else {
           planWithNewSubquery.withNewChildren(newChildren.toSeq)

sql/core/src/test/scala/org/apache/spark/sql/DatasetSuite.scala

@@ -916,6 +916,32 @@ class DatasetSuite extends QueryTest
     }
   }
 
+  test("SPARK-43781: cogroup two datasets derived from the same source") {
+    val inputType = StructType(Array(StructField("id", LongType, false),
+      StructField("type", StringType, false)))
+    val keyType = StructType(Array(StructField("id", LongType, false)))
+
+    val inputRows = new java.util.ArrayList[Row]()
+    inputRows.add(Row(1L, "foo"))
+    inputRows.add(Row(1L, "bar"))
+    inputRows.add(Row(2L, "foo"))
+    val input = spark.createDataFrame(inputRows, inputType)
+    val fooGroups = input.filter("type = 'foo'").groupBy("id").as(ExpressionEncoder(keyType),
+      ExpressionEncoder(inputType))
+    val barGroups = input.filter("type = 'bar'").groupBy("id").as(ExpressionEncoder(keyType),
+      ExpressionEncoder(inputType))
+
+    val result = fooGroups.cogroup(barGroups) { case (row, iterator, iterator1) =>
+      iterator.toSeq ++ iterator1.toSeq
+    }(ExpressionEncoder(inputType)).collect()
+    assert(result.length == 3)
+
+    val result2 = fooGroups.cogroupSorted(barGroups)($"id")($"id") {
+      case (row, iterator, iterator1) => iterator.toSeq ++ iterator1.toSeq
+    }(ExpressionEncoder(inputType)).collect()
+    assert(result2.length == 3)
+  }
+
   test("SPARK-34806: observation on datasets") {
     val namedObservation = Observation("named")
     val unnamedObservation = Observation()