Merge remote-tracking branch 'upstream/master' into SPARK-22856

sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/codegen/CodeGenerator.scala

@@ -1006,7 +1006,7 @@ class CodegenContext {
       subexprFunctions += s"${addNewFunction(fnName, fn)}($INPUT_ROW);"
       val state = SubExprEliminationState(GlobalValue(isNull, JAVA_BOOLEAN),
         GlobalValue(value, javaType(expr.dataType)))
-      e.foreach(subExprEliminationExprs.put(_, state))
+      subExprEliminationExprs ++= e.map(_ -> state).toMap
     }
   }
 

sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/codegen/GenerateUnsafeProjection.scala

@@ -336,7 +336,9 @@ object GenerateUnsafeProjection extends CodeGenerator[Seq[Expression], UnsafePro
         $evalSubexpr
         $writeExpressions
       """
-    ExprCode(code, FalseLiteral, GlobalValue(result, "UnsafeRow"))
+    // `rowWriter` is declared as a class field, so we can access it directly in methods.
+    ExprCode(code, FalseLiteral, StatementValue(s"$rowWriter.getRow()", "UnsafeRow",
+      canDirectAccess = true))
   }
 
   protected def canonicalize(in: Seq[Expression]): Seq[Expression] =

sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/misc.scala

@@ -143,8 +143,14 @@ case class Uuid(randomSeed: Option[Long] = None) extends LeafExpression with Sta
     randomGenerator.getNextUUIDUTF8String()
 
   override def doGenCode(ctx: CodegenContext, ev: ExprCode): ExprCode = {
-    ev.copy(code = s"final UTF8String ${ev.value} = " +
-      s"UTF8String.fromString(java.util.UUID.randomUUID().toString());",
+    val randomGen = ctx.freshName("randomGen")
+    ctx.addMutableState("org.apache.spark.sql.catalyst.util.RandomUUIDGenerator", randomGen,
+      forceInline = true,
+      useFreshName = false)
+    ctx.addPartitionInitializationStatement(s"$randomGen = " +
+      "new org.apache.spark.sql.catalyst.util.RandomUUIDGenerator(" +
+      s"${randomSeed.get}L + partitionIndex);")
+    ev.copy(code = s"final UTF8String ${ev.value} = $randomGen.getNextUUIDUTF8String();",
       isNull = FalseLiteral)
   }
 

sql/catalyst/src/test/scala/org/apache/spark/sql/catalyst/expressions/CodeGenerationSuite.scala

@@ -448,6 +448,8 @@ class CodeGenerationSuite extends SparkFunSuite with ExpressionEvalHelper {
     val ref = BoundReference(0, IntegerType, true)
     val add1 = Add(ref, ref)
     val add2 = Add(add1, add1)
+    val dummy = SubExprEliminationState(VariableValue("dummy", "boolean"),
+      VariableValue("dummy", "boolean"))
 
     // raw testing of basic functionality
     {
@@ -457,7 +459,7 @@ class CodeGenerationSuite extends SparkFunSuite with ExpressionEvalHelper {
       ctx.subExprEliminationExprs += ref -> SubExprEliminationState(e.isNull, e.value)
       assert(ctx.subExprEliminationExprs.contains(ref))
       // call withSubExprEliminationExprs
-      ctx.withSubExprEliminationExprs(Map(add1 -> SubExprEliminationState("dummy", "dummy"))) {
+      ctx.withSubExprEliminationExprs(Map(add1 -> dummy)) {
         assert(ctx.subExprEliminationExprs.contains(add1))
         assert(!ctx.subExprEliminationExprs.contains(ref))
         Seq.empty
@@ -475,7 +477,7 @@ class CodeGenerationSuite extends SparkFunSuite with ExpressionEvalHelper {
       ctx.generateExpressions(Seq(add2, add1), doSubexpressionElimination = true) // trigger CSE
       assert(ctx.subExprEliminationExprs.contains(add1))
       // call withSubExprEliminationExprs
-      ctx.withSubExprEliminationExprs(Map(ref -> SubExprEliminationState("dummy", "dummy"))) {
+      ctx.withSubExprEliminationExprs(Map(ref -> dummy)) {
         assert(ctx.subExprEliminationExprs.contains(ref))
         assert(!ctx.subExprEliminationExprs.contains(add1))
         Seq.empty