Qualification tool: Parsing Execs to get the ExecInfo #2 (NVIDIA#5426)

* Qualification tool: Parsing Execs to get the ExecInfo #2

Signed-off-by: Niranjan Artal <[email protected]>

tools/src/main/scala/com/nvidia/spark/rapids/tool/planparser/CartesianProductExecParser.scala

@@ -0,0 +1,42 @@
+/*
+ * Copyright (c) 2022, NVIDIA CORPORATION.
+ *
+ * Licensed 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 com.nvidia.spark.rapids.tool.planparser
+
+import com.nvidia.spark.rapids.tool.qualification.PluginTypeChecker
+
+import org.apache.spark.sql.execution.ui.SparkPlanGraphNode
+
+case class CartesianProductExecParser(
+    node: SparkPlanGraphNode,
+    checker: PluginTypeChecker,
+    sqlID: Long) extends ExecParser {
+
+  val fullExecName = node.name + "Exec"
+
+  override def parse: Seq[ExecInfo] = {
+    // CartesianProduct doesn't have duration
+    val duration = None
+    val (speedupFactor, isSupported) = if (checker.isExecSupported(fullExecName)) {
+      (checker.getSpeedupFactor(fullExecName), true)
+    } else {
+      (1, false)
+    }
+    // TODO - add in parsing expressions - average speedup across?
+    Seq(ExecInfo(sqlID, node.name, "", speedupFactor,
+      duration, node.id, isSupported, None))
+  }
+}

tools/src/main/scala/com/nvidia/spark/rapids/tool/planparser/GenerateExecParser.scala

@@ -0,0 +1,42 @@
+/*
+ * Copyright (c) 2022, NVIDIA CORPORATION.
+ *
+ * Licensed 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 com.nvidia.spark.rapids.tool.planparser
+
+import com.nvidia.spark.rapids.tool.qualification.PluginTypeChecker
+
+import org.apache.spark.sql.execution.ui.SparkPlanGraphNode
+
+case class GenerateExecParser(
+    node: SparkPlanGraphNode,
+    checker: PluginTypeChecker,
+    sqlID: Long) extends ExecParser {
+
+  val fullExecName = node.name + "Exec"
+
+  override def parse: Seq[ExecInfo] = {
+    // Generate doesn't have duration
+    val duration = None
+    val (speedupFactor, isSupported) = if (checker.isExecSupported(fullExecName)) {
+      (checker.getSpeedupFactor(fullExecName), true)
+    } else {
+      (1, false)
+    }
+    // TODO - add in parsing expressions - average speedup across?
+    Seq(ExecInfo(sqlID, node.name, "", speedupFactor,
+      duration, node.id, isSupported, None))
+  }
+}

tools/src/main/scala/com/nvidia/spark/rapids/tool/planparser/GlobalLimitExecParser.scala

@@ -0,0 +1,42 @@
+/*
+ * Copyright (c) 2022, NVIDIA CORPORATION.
+ *
+ * Licensed 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 com.nvidia.spark.rapids.tool.planparser
+
+import com.nvidia.spark.rapids.tool.qualification.PluginTypeChecker
+
+import org.apache.spark.sql.execution.ui.SparkPlanGraphNode
+
+case class GlobalLimitExecParser(
+    node: SparkPlanGraphNode,
+    checker: PluginTypeChecker,
+    sqlID: Long) extends ExecParser {
+
+  val fullExecName = node.name + "Exec"
+
+  override def parse: Seq[ExecInfo] = {
+    // GlobalLimit doesn't have duration
+    val duration = None
+    val (speedupFactor, isSupported) = if (checker.isExecSupported(fullExecName)) {
+      (checker.getSpeedupFactor(fullExecName), true)
+    } else {
+      (1, false)
+    }
+    // TODO - add in parsing expressions - average speedup across?
+    Seq(ExecInfo(sqlID, node.name, "", speedupFactor,
+      duration, node.id, isSupported, None))
+  }
+}

tools/src/main/scala/com/nvidia/spark/rapids/tool/planparser/LocalLimitExecParser.scala

@@ -0,0 +1,42 @@
+/*
+ * Copyright (c) 2022, NVIDIA CORPORATION.
+ *
+ * Licensed 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 com.nvidia.spark.rapids.tool.planparser
+
+import com.nvidia.spark.rapids.tool.qualification.PluginTypeChecker
+
+import org.apache.spark.sql.execution.ui.SparkPlanGraphNode
+
+case class LocalLimitExecParser(
+    node: SparkPlanGraphNode,
+    checker: PluginTypeChecker,
+    sqlID: Long) extends ExecParser {
+
+  val fullExecName = node.name + "Exec"
+
+  override def parse: Seq[ExecInfo] = {
+    // LocalLimit doesn't have duration
+    val duration = None
+    val (speedupFactor, isSupported) = if (checker.isExecSupported(fullExecName)) {
+      (checker.getSpeedupFactor(fullExecName), true)
+    } else {
+      (1, false)
+    }
+    // TODO - add in parsing expressions - average speedup across?
+    Seq(ExecInfo(sqlID, node.name, "", speedupFactor,
+      duration, node.id, isSupported, None))
+  }
+}

tools/src/main/scala/com/nvidia/spark/rapids/tool/planparser/SQLPlanParser.scala

@@ -80,6 +80,8 @@ object SQLPlanParser extends Logging {
       app: AppBase
   ): Seq[ExecInfo] = {
     node match {
+      case c if (c.name == "CartesianProduct") =>
+        CartesianProductExecParser(c, checker, sqlID).parse
       case c if (c.name == "Coalesce") =>
         CoalesceExecParser(c, checker, sqlID).parse
       case c if (c.name == "CollectLimit") =>
@@ -88,12 +90,20 @@ object SQLPlanParser extends Logging {
         ExpandExecParser(e, checker, sqlID).parse
       case f if (f.name == "Filter") =>
         FilterExecParser(f, checker, sqlID).parse
+      case g if (g.name == "Generate") =>
+        GenerateExecParser(g, checker, sqlID).parse
+      case g if (g.name == "GlobalLimit") =>
+        GlobalLimitExecParser(g, checker, sqlID).parse
+      case l if (l.name == "LocalLimit") =>
+        LocalLimitExecParser(l, checker, sqlID).parse
       case p if (p.name == "Project") =>
         ProjectExecParser(p, checker, sqlID).parse
       case r if (r.name == "Range") =>
         RangeExecParser(r, checker, sqlID).parse
       case s if (s.name == "Sample") =>
         SampleExecParser(s, checker, sqlID).parse
+      case s if (s.name == "Sort") =>
+        SortExecParser(s, checker, sqlID).parse
       case t if (t.name == "TakeOrderedAndProject") =>
         TakeOrderedAndProjectExecParser(t, checker, sqlID).parse
       case u if (u.name == "Union") =>

tools/src/main/scala/com/nvidia/spark/rapids/tool/planparser/SortExecParser.scala

@@ -0,0 +1,42 @@
+/*
+ * Copyright (c) 2022, NVIDIA CORPORATION.
+ *
+ * Licensed 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 com.nvidia.spark.rapids.tool.planparser
+
+import com.nvidia.spark.rapids.tool.qualification.PluginTypeChecker
+
+import org.apache.spark.sql.execution.ui.SparkPlanGraphNode
+
+case class SortExecParser(
+    node: SparkPlanGraphNode,
+    checker: PluginTypeChecker,
+    sqlID: Long) extends ExecParser {
+
+  val fullExecName = node.name + "Exec"
+
+  override def parse: Seq[ExecInfo] = {
+    // Sort doesn't have duration
+    val duration = None
+    val (speedupFactor, isSupported) = if (checker.isExecSupported(fullExecName)) {
+      (checker.getSpeedupFactor(fullExecName), true)
+    } else {
+      (1, false)
+    }
+    // TODO - add in parsing expressions - average speedup across?
+    Seq(ExecInfo(sqlID, node.name, "", speedupFactor,
+      duration, node.id, isSupported, None))
+  }
+}

tools/src/test/scala/com/nvidia/spark/rapids/tool/planparser/SqlPlanParserSuite.scala

@@ -23,7 +23,7 @@ import org.scalatest.{BeforeAndAfterEach, FunSuite}
 
 import org.apache.spark.internal.Logging
 import org.apache.spark.sql.{SparkSession, TrampolineUtil}
-import org.apache.spark.sql.functions.col
+import org.apache.spark.sql.functions.{col, explode}
 import org.apache.spark.sql.rapids.tool.qualification.QualificationAppInfo
 
 class SQLPlanParserSuite extends FunSuite with BeforeAndAfterEach with Logging {
@@ -39,7 +39,9 @@ class SQLPlanParserSuite extends FunSuite with BeforeAndAfterEach with Logging {
       .getOrCreate()
   }
 
-  test("WholeStage with Filter and Project") {
+  private val logDir = ToolTestUtils.getTestResourcePath("spark-events-qualification")
+
+  test("WholeStage with Filter, Project and Sort") {
     TrampolineUtil.withTempDir { eventLogDir =>
       val (eventLog, _) = ToolTestUtils.generateEventLog(eventLogDir, "sqlmetric") { spark =>
         import spark.implicits._
@@ -48,6 +50,7 @@ class SQLPlanParserSuite extends FunSuite with BeforeAndAfterEach with Logging {
         df.select( $"value" as "a")
           .join(df2.select($"value" as "b"), $"a" === $"b")
           .filter($"b" < 100)
+          .sort($"b")
       }
 
       TrampolineUtil.withTempDir { outpath =>
@@ -63,25 +66,33 @@ class SQLPlanParserSuite extends FunSuite with BeforeAndAfterEach with Logging {
         assert(app.sqlPlans.size == 1)
         app.sqlPlans.foreach { case(sqlID, plan) =>
           val planInfo = SQLPlanParser.parseSQLPlan(plan, sqlID, pluginTypeChecker, app)
-          assert(planInfo.execInfo.size == 9)
+          assert(planInfo.execInfo.size == 11)
           val wholeStages = planInfo.execInfo.filter(_.exec.contains("WholeStageCodegen"))
-          assert(wholeStages.size == 5)
-          // only 2 in the above example have projects and filters
+          assert(wholeStages.size == 6)
+          // only 2 in the above example have projects and filters and the other 3 have sort
           val numSupported = wholeStages.filter(_.isSupported).size
-          assert(numSupported == 2)
+          assert(numSupported == 5)
           assert(wholeStages.forall(_.duration.nonEmpty))
           val allChildren = wholeStages.flatMap(_.children).flatten
-          assert(allChildren.size == 9)
-          val filters = allChildren.filter(_.exec == "FilterExec")
+          assert(allChildren.size == 10)
+          val filters = allChildren.filter(_.exec == "Filter")
+          assert(filters.size == 2)
           assert(filters.forall(_.speedupFactor == 2))
           assert(filters.forall(_.isSupported == true))
           assert(filters.forall(_.children.isEmpty))
           assert(filters.forall(_.duration.isEmpty))
-          val projects = allChildren.filter(_.exec == "ProjectExec")
+          val projects = allChildren.filter(_.exec == "Project")
+          assert(projects.size == 2)
           assert(projects.forall(_.speedupFactor == 2))
           assert(projects.forall(_.isSupported == true))
           assert(projects.forall(_.children.isEmpty))
           assert(projects.forall(_.duration.isEmpty))
+          val sorts = allChildren.filter(_.exec == "Sort")
+          assert(sorts.size == 3)
+          assert(sorts.forall(_.speedupFactor == 2))
+          assert(sorts.forall(_.isSupported == true))
+          assert(sorts.forall(_.children.isEmpty))
+          assert(sorts.forall(_.duration.isEmpty))
         }
       }
     }
@@ -136,4 +147,79 @@ class SQLPlanParserSuite extends FunSuite with BeforeAndAfterEach with Logging {
       }
     }
   }
+
+  test("Parse Execs - CartesianProduct and Generate") {
+    TrampolineUtil.withTempDir { eventLogDir =>
+      val (eventLog, _) = ToolTestUtils.generateEventLog(eventLogDir, "sqlmetric") { spark =>
+        import spark.implicits._
+        val genDf = spark.sparkContext.parallelize(List(List(1, 2, 3), List(4, 5, 6)), 4).toDF
+        val joinDf1 = spark.sparkContext.makeRDD(1 to 10, 4).toDF
+        val joinDf2 = spark.sparkContext.makeRDD(1 to 10, 4).toDF
+        genDf.select(explode($"value")).collect
+        joinDf1.crossJoin(joinDf2)
+      }
+      TrampolineUtil.withTempDir { outpath =>
+        val hadoopConf = new Configuration()
+        val (_, allEventLogs) = EventLogPathProcessor.processAllPaths(
+          None, None, List(eventLog), hadoopConf)
+        val pluginTypeChecker = new PluginTypeChecker()
+        assert(allEventLogs.size == 1)
+        val appOption = QualificationAppInfo.createApp(allEventLogs.head, hadoopConf,
+          pluginTypeChecker, 20)
+        assert(appOption.nonEmpty)
+        val app = appOption.get
+        assert(app.sqlPlans.size == 2)
+        val supportedExecs = Array("CartesianProduct", "Generate")
+        app.sqlPlans.foreach { case (sqlID, plan) =>
+          val planInfo = SQLPlanParser.parseSQLPlan(plan, sqlID, pluginTypeChecker, app)
+          for (execName <- supportedExecs) {
+            val supportedExec = planInfo.execInfo.filter(_.exec == execName)
+            if (supportedExec.nonEmpty) {
+              assert(supportedExec.size == 1)
+              assert(supportedExec.forall(_.children.isEmpty))
+              assert(supportedExec.forall(_.duration.isEmpty))
+              assert(supportedExec.forall(_.speedupFactor == 2), execName)
+              assert(supportedExec.forall(_.isSupported == true))
+            }
+          }
+        }
+      }
+    }
+  }
+
+  // GlobalLimit and LocalLimit is not in physical plan when collect is called on the dataframe.
+  // We are reading from static eventlogs to test these execs.
+  test("Parse execs - LocalLimit and GlobalLimit") {
+    val logFile = s"$logDir/global_local_limit_eventlog.zstd"
+    TrampolineUtil.withTempDir { outpath =>
+      val hadoopConf = new Configuration()
+      val (_, allEventLogs) = EventLogPathProcessor.processAllPaths(
+        None, None, List(logFile), hadoopConf)
+      val pluginTypeChecker = new PluginTypeChecker()
+      assert(allEventLogs.size == 1)
+      val appOption = QualificationAppInfo.createApp(allEventLogs.head, hadoopConf,
+        pluginTypeChecker, 20)
+      assert(appOption.nonEmpty)
+      val app = appOption.get
+      assert(app.sqlPlans.size == 1)
+      val supportedExecs = Array("GlobalLimit", "LocalLimit")
+      app.sqlPlans.foreach { case (sqlID, plan) =>
+        val planInfo = SQLPlanParser.parseSQLPlan(plan, sqlID, pluginTypeChecker, app)
+        // GlobalLimit and LocalLimit are inside WholeStageCodegen. So getting the children of
+        // WholeStageCodegenExec
+        val wholeStages = planInfo.execInfo.filter(_.exec.contains("WholeStageCodegen"))
+        val allChildren = wholeStages.flatMap(_.children).flatten
+        for (execName <- supportedExecs) {
+          val supportedExec = allChildren.filter(_.exec == execName)
+          if (supportedExec.nonEmpty) {
+            assert(supportedExec.size == 1)
+            assert(supportedExec.forall(_.children.isEmpty))
+            assert(supportedExec.forall(_.duration.isEmpty))
+            assert(supportedExec.forall(_.speedupFactor == 2), execName)
+            assert(supportedExec.forall(_.isSupported == true))
+          }
+        }
+      }
+    }
+  }
 }