Github user andrewor14 commented on a diff in the pull request:
https://github.com/apache/spark/pull/7774#discussion_r36045507
--- Diff:
sql/core/src/main/scala/org/apache/spark/sql/execution/SparkSQLExecution.scala
---
@@ -0,0 +1,84 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one or more
+ * contributor license agreements. See the NOTICE file distributed with
+ * this work for additional information regarding copyright ownership.
+ * The ASF licenses this file to You 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 org.apache.spark.sql.execution
+
+import java.util.concurrent.atomic.AtomicLong
+
+import org.apache.spark.SparkContext
+import org.apache.spark.sql.{DataFrame, SQLContext}
+import org.apache.spark.util.Utils
+
+private[sql] object SparkSQLExecution {
+
+ val EXECUTION_ID_KEY = "spark.sql.execution.id"
+
+ private val _nextExecutionId = new AtomicLong(0)
+
+ private def nextExecutionId: Long = _nextExecutionId.getAndIncrement
+
+ /**
+ * Wrap a DataFrame action to track all Spark jobs in the body so that
we can connect them with
+ * an execution.
+ */
+ def withNewExecution[T](sqlContext: SQLContext, df: DataFrame)(body: =>
T): T = {
+ val sc = sqlContext.sparkContext
+ val oldExecutionId = sc.getLocalProperty(EXECUTION_ID_KEY)
+ try {
+ if (oldExecutionId == null) {
+ val executionId = SparkSQLExecution.nextExecutionId
+ sc.setLocalProperty(EXECUTION_ID_KEY, executionId.toString)
+ val callSite = Utils.getCallSite()
+ sqlContext.listener.onExecutionStart(
+ executionId, callSite.shortForm, callSite.longForm, df,
System.currentTimeMillis())
+ val r = body
+ sqlContext.listener.onExecutionEnd(executionId,
System.currentTimeMillis())
+ r
+ } else {
+ // Don't support nested `withNewExecution`. This is an example of
the nested
+ // `withNewExecution`:
+ //
+ // class DataFrame {
+ // def foo: T = withNewExecution {
something.createNewDataFrame().collect() }
+ // }
+ //
+ // Note: `collect` will call withNewExecution
+ // In this case, only the "executedPlan" for "collect" will be
executed. The "executedPlan"
+ // for the outer DataFrame won't be executed. So it's meaningless
to create a new Execution
+ // for the outer DataFrame. Even if we track it, since its
"executedPlan" doesn't run,
+ // all accumulator metrics will be 0. It will confuse people if we
show them in Web UI.
+ //
+ // A real case is the `DataFrame.count` method.
+ throw new IllegalArgumentException(s"$EXECUTION_ID_KEY is already
set")
--- End diff --
Ah, I get it now. It's because a DF operation might call other DF
operations internally (e.g. count calls collect) but the user only cares about
the top level operation (i.e. count, not collect). I think the proper semantics
should be to inherit the existing execution ID if it already exists. I'm not
sure if I see why we should throw an exception though.
By the way, I'm confused about one thing, since you didn't wrap `count` in
`withNewExecution` how did your screenshot display `count` in the call site
instead of `collect`?
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
