From db7a46a169d1789ff221d7f84315edc3df04511a Mon Sep 17 00:00:00 2001
From: Cheng Lian <lian@databricks.com>
Date: Mon, 22 Jun 2015 00:45:39 -0700
Subject: [PATCH] More comments

---
 .../apache/spark/sql/sources/commands.scala   | 24 +++++++++++++++++++
 1 file changed, 24 insertions(+)

diff --git a/sql/core/src/main/scala/org/apache/spark/sql/sources/commands.scala b/sql/core/src/main/scala/org/apache/spark/sql/sources/commands.scala
index 3a4d04ed793b6..215e53c020849 100644
--- a/sql/core/src/main/scala/org/apache/spark/sql/sources/commands.scala
+++ b/sql/core/src/main/scala/org/apache/spark/sql/sources/commands.scala
@@ -58,6 +58,28 @@ private[sql] case class InsertIntoDataSource(
   }
 }
 
+/**
+ * A command for writing data to a [[HadoopFsRelation]].  Supports both overwriting and appending.
+ * Writing to dynamic partitions is also supported.  Each [[InsertIntoHadoopFsRelation]] issues a
+ * single write job, and owns a UUID that identifies this job.  Each concrete implementation of
+ * [[HadoopFsRelation]] should use this UUID together with task id to generate unique file path for
+ * each task output file.  This UUID is passed to executor side via a property named
+ * `spark.sql.sources.writeJobUUID`.
+ *
+ * Different writer containers, [[DefaultWriterContainer]] and [[DynamicPartitionWriterContainer]]
+ * are used to write to normal tables and tables with dynamic partitions.
+ *
+ * Basic work flow of this command is:
+ *
+ *   1. Driver side setup, including output committer initialization and data source specific
+ *      preparation work for the write job to be issued.
+ *   2. Issues a write job consists of one or more executor side tasks, each of which writes all
+ *      rows within an RDD partition.
+ *   3. If no exception is thrown in a task, commits that task, otherwise aborts that task;  If any
+ *      exception is thrown during task commitment, also aborts that task.
+ *   4. If all tasks are committed, commit the job, otherwise aborts the job;  If any exception is
+ *      thrown during job commitment, also aborts the job.
+ */
 private[sql] case class InsertIntoHadoopFsRelation(
     @transient relation: HadoopFsRelation,
     @transient query: LogicalPlan,
@@ -429,6 +451,8 @@ private[sql] class DefaultWriterContainer(
       writer.close()
       super.commitTask()
     } catch { case cause: Throwable =>
+      // This exception will be handled in `InsertIntoHadoopFsRelation.insert$writeRows`, and will
+      // cause `abortTask()` to be invoked.
       throw new RuntimeException("Failed to commit task", cause)
     }
   }