[SPARK-31620][SQL] Fix reference binding failure in case of an final …

…agg contains subquery

### What changes were proposed in this pull request?

Instead of using `child.output` directly, we should use `inputAggBufferAttributes` from the current agg expression  for `Final` and `PartialMerge` aggregates to bind references for their `mergeExpression`.

### Why are the changes needed?

When planning aggregates, the partial aggregate uses agg fucs' `inputAggBufferAttributes` as its output, see https://github.com/apache/spark/blob/v3.0.0-rc1/sql/core/src/main/scala/org/apache/spark/sql/execution/aggregate/AggUtils.scala#L105

For final `HashAggregateExec`, we need to bind the `DeclarativeAggregate.mergeExpressions` with the output of the partial aggregate operator, see https://github.com/apache/spark/blob/v3.0.0-rc1/sql/core/src/main/scala/org/apache/spark/sql/execution/aggregate/HashAggregateExec.scala#L348

This is usually fine. However, if we copy the agg func somehow after agg planning, like `PlanSubqueries`, the `DeclarativeAggregate` will be replaced by a new instance with new `inputAggBufferAttributes` and `mergeExpressions`. Then we can't bind the `mergeExpressions` with the output of the partial aggregate operator, as it uses the `inputAggBufferAttributes` of the original `DeclarativeAggregate` before copy.

Note that, `ImperativeAggregate` doesn't have this problem, as we don't need to bind its `mergeExpressions`. It has a different mechanism to access buffer values, via `mutableAggBufferOffset` and `inputAggBufferOffset`.

### Does this PR introduce _any_ user-facing change?

Yes, user hit error previously but run query successfully after this change.

### How was this patch tested?

Added a regression test.

Closes #28496 from Ngone51/spark-31620.

Authored-by: yi.wu <yi.wu@databricks.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>

sql/core/src/main/scala/org/apache/spark/sql/execution/aggregate/BaseAggregateExec.scala

@@ -18,7 +18,7 @@
 package org.apache.spark.sql.execution.aggregate
 
 import org.apache.spark.sql.catalyst.expressions.{Attribute, NamedExpression}
-import org.apache.spark.sql.catalyst.expressions.aggregate.AggregateExpression
+import org.apache.spark.sql.catalyst.expressions.aggregate.{AggregateExpression, Final, PartialMerge}
 import org.apache.spark.sql.execution.{ExplainUtils, UnaryExecNode}
 
 /**
@@ -40,4 +40,28 @@ trait BaseAggregateExec extends UnaryExecNode {
        |${ExplainUtils.generateFieldString("Results", resultExpressions)}
        |""".stripMargin
   }
+
+  protected def inputAttributes: Seq[Attribute] = {
+    val modes = aggregateExpressions.map(_.mode).distinct
+    if (modes.contains(Final) || modes.contains(PartialMerge)) {
+      // SPARK-31620: when planning aggregates, the partial aggregate uses aggregate function's
+      // `inputAggBufferAttributes` as its output. And Final and PartialMerge aggregate rely on the
+      // output to bind references for `DeclarativeAggregate.mergeExpressions`. But if we copy the
+      // aggregate function somehow after aggregate planning, like `PlanSubqueries`, the
+      // `DeclarativeAggregate` will be replaced by a new instance with new
+      // `inputAggBufferAttributes` and `mergeExpressions`. Then Final and PartialMerge aggregate
+      // can't bind the `mergeExpressions` with the output of the partial aggregate, as they use
+      // the `inputAggBufferAttributes` of the original `DeclarativeAggregate` before copy. Instead,
+      // we shall use `inputAggBufferAttributes` after copy to match the new `mergeExpressions`.
+      val aggAttrs = aggregateExpressions
+        // there're exactly four cases needs `inputAggBufferAttributes` from child according to the
+        // agg planning in `AggUtils`: Partial -> Final, PartialMerge -> Final,
+        // Partial -> PartialMerge, PartialMerge -> PartialMerge.
+        .filter(a => a.mode == Final || a.mode == PartialMerge).map(_.aggregateFunction)
+        .flatMap(_.inputAggBufferAttributes)
+      child.output.dropRight(aggAttrs.length) ++ aggAttrs
+    } else {
+      child.output
+    }
+  }
 }

sql/core/src/main/scala/org/apache/spark/sql/execution/aggregate/HashAggregateExec.scala

@@ -129,7 +129,7 @@ case class HashAggregateExec(
             resultExpressions,
             (expressions, inputSchema) =>
               MutableProjection.create(expressions, inputSchema),
-            child.output,
+            inputAttributes,
             iter,
             testFallbackStartsAt,
             numOutputRows,
@@ -334,7 +334,7 @@ case class HashAggregateExec(
   private def doConsumeWithoutKeys(ctx: CodegenContext, input: Seq[ExprCode]): String = {
     // only have DeclarativeAggregate
     val functions = aggregateExpressions.map(_.aggregateFunction.asInstanceOf[DeclarativeAggregate])
-    val inputAttrs = functions.flatMap(_.aggBufferAttributes) ++ child.output
+    val inputAttrs = functions.flatMap(_.aggBufferAttributes) ++ inputAttributes
     // To individually generate code for each aggregate function, an element in `updateExprs` holds
     // all the expressions for the buffer of an aggregation function.
     val updateExprs = aggregateExpressions.map { e =>
@@ -931,7 +931,7 @@ case class HashAggregateExec(
       }
     }
 
-    val inputAttr = aggregateBufferAttributes ++ child.output
+    val inputAttr = aggregateBufferAttributes ++ inputAttributes
     // Here we set `currentVars(0)` to `currentVars(numBufferSlots)` to null, so that when
     // generating code for buffer columns, we use `INPUT_ROW`(will be the buffer row), while
     // generating input columns, we use `currentVars`.

sql/core/src/main/scala/org/apache/spark/sql/execution/aggregate/ObjectHashAggregateExec.scala

@@ -123,7 +123,7 @@ case class ObjectHashAggregateExec(
             resultExpressions,
             (expressions, inputSchema) =>
               MutableProjection.create(expressions, inputSchema),
-            child.output,
+            inputAttributes,
             iter,
             fallbackCountThreshold,
             numOutputRows)

sql/core/src/main/scala/org/apache/spark/sql/execution/aggregate/SortAggregateExec.scala

@@ -88,7 +88,7 @@ case class SortAggregateExec(
         val outputIter = new SortBasedAggregationIterator(
           partIndex,
           groupingExpressions,
-          child.output,
+          inputAttributes,
           iter,
           aggregateExpressions,
           aggregateAttributes,

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

@@ -973,4 +973,43 @@ class DataFrameAggregateSuite extends QueryTest
       assert(error.message.contains("function count_if requires boolean type"))
     }
   }
+
+  Seq(true, false).foreach { value =>
+    test(s"SPARK-31620: agg with subquery (whole-stage-codegen = $value)") {
+      withSQLConf(SQLConf.WHOLESTAGE_CODEGEN_ENABLED.key -> value.toString) {
+        withTempView("t1", "t2") {
+          sql("create temporary view t1 as select * from values (1, 2) as t1(a, b)")
+          sql("create temporary view t2 as select * from values (3, 4) as t2(c, d)")
+
+          // test without grouping keys
+          checkAnswer(sql("select sum(if(c > (select a from t1), d, 0)) as csum from t2"),
+            Row(4) :: Nil)
+
+          // test with grouping keys
+          checkAnswer(sql("select c, sum(if(c > (select a from t1), d, 0)) as csum from " +
+            "t2 group by c"), Row(3, 4) :: Nil)
+
+          // test with distinct
+          checkAnswer(sql("select avg(distinct(d)), sum(distinct(if(c > (select a from t1)," +
+            " d, 0))) as csum from t2 group by c"), Row(4, 4) :: Nil)
+
+          // test subquery with agg
+          checkAnswer(sql("select sum(distinct(if(c > (select sum(distinct(a)) from t1)," +
+            " d, 0))) as csum from t2 group by c"), Row(4) :: Nil)
+
+          // test SortAggregateExec
+          var df = sql("select max(if(c > (select a from t1), 'str1', 'str2')) as csum from t2")
+          assert(df.queryExecution.executedPlan
+            .find { case _: SortAggregateExec => true }.isDefined)
+          checkAnswer(df, Row("str1") :: Nil)
+
+          // test ObjectHashAggregateExec
+          df = sql("select collect_list(d), sum(if(c > (select a from t1), d, 0)) as csum from t2")
+          assert(df.queryExecution.executedPlan
+            .find { case _: ObjectHashAggregateExec => true }.isDefined)
+          checkAnswer(df, Row(Array(4), 4) :: Nil)
+        }
+      }
+    }
+  }
 }