add test

Author: cloud-fan

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

@@ -968,7 +968,22 @@ class Analyzer(
    * }}}
    */
   object RemoveEvaluationFromSort extends Rule[LogicalPlan] {
+    private def hasAlias(expr: Expression) = {
+      expr.find {
+        case a: Alias => true
+        case _ => false
+      }.isDefined
+    }
+
     override def apply(plan: LogicalPlan): LogicalPlan = plan transform {
+      // The ordering expressions have no effect to the output schema of `Sort`,
+      // so `Alias`s in ordering expressions are unnecessary and we should remove them.
+      case s @ Sort(ordering, _, _) if ordering.exists(hasAlias) =>
+        val newOrdering = ordering.map(_.transformUp {
+          case Alias(child, _) => child
+        }.asInstanceOf[SortOrder])
+        s.copy(order = newOrdering)
+
       case s @ Sort(ordering, global, child)
         if s.expressions.forall(_.resolved) && s.childrenResolved && !s.hasNoEvaluation =>
 

sql/catalyst/src/test/scala/org/apache/spark/sql/catalyst/analysis/AnalysisSuite.scala

@@ -165,11 +165,39 @@ class AnalysisSuite extends AnalysisTest {
 
   test("pull out nondeterministic expressions from Sort") {
     val plan = Sort(Seq(SortOrder(Rand(33), Ascending)), false, testRelation)
-    val projected = Alias(Rand(33), "_nondeterministic")()
+    val analyzed = caseSensitiveAnalyzer.execute(plan)
+    analyzed.transform {
+      case s: Sort if s.expressions.exists(!_.deterministic) =>
+        fail("nondeterministic expressions are not allowed in Sort")
+    }
+  }
+
+  test("remove still-need-evaluate ordering expressions from sort") {
+    val a = testRelation2.output(0)
+    val b = testRelation2.output(1)
+
+    def makeOrder(e: Expression): SortOrder = SortOrder(e, Ascending)
+
+    val noEvalOrdering = makeOrder(a)
+    val noEvalOrderingWithAlias = makeOrder(Alias(Alias(b, "name1")(), "name2")())
+
+    val needEvalExpr = Coalesce(Seq(a, Literal("1")))
+    val needEvalExpr2 = Coalesce(Seq(a, b))
+    val needEvalOrdering = makeOrder(needEvalExpr)
+    val needEvalOrdering2 = makeOrder(needEvalExpr2)
+
+    val plan = Sort(
+      Seq(noEvalOrdering, noEvalOrderingWithAlias, needEvalOrdering, needEvalOrdering2),
+      false, testRelation2)
+
+    val evaluatedOrdering = makeOrder(AttributeReference("_sortCondition", StringType)())
+    val materializedExprs = Seq(needEvalExpr, needEvalExpr2).map(e => Alias(e, "_sortCondition")())
+
     val expected =
-      Project(testRelation.output,
-        Sort(Seq(SortOrder(projected.toAttribute, Ascending)), false,
-          Project(testRelation.output :+ projected, testRelation)))
+      Project(testRelation2.output,
+        Sort(Seq(makeOrder(a), makeOrder(b), evaluatedOrdering, evaluatedOrdering), false,
+          Project(testRelation2.output ++ materializedExprs, testRelation2)))
+
     checkAnalysis(plan, expected)
   }
 }

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

@@ -841,7 +841,6 @@ class DataFrameSuite extends QueryTest with SQLTestUtils {
     }
   }
 
-<<<<<<< HEAD
   test("SPARK-8608: call `show` on local DataFrame with random columns should return same value") {
     // Make sure we can pass this test for both codegen mode and interpreted mode.
     withSQLConf(SQLConf.CODEGEN_ENABLED.key -> "true") {