[SPARK-19168][STRUCTURED STREAMING] StateStore should be aborted upon error

## What changes were proposed in this pull request?

We should call `StateStore.abort()` when there should be any error before the store is committed.

## How was this patch tested?

Manually.

Author: Liwei Lin <lwlin7@gmail.com>

Closes apache#16547 from lw-lin/append-filter.

Author: lw-lin

sql/core/src/main/scala/org/apache/spark/sql/execution/streaming/StatefulAggregate.scala

@@ -31,6 +31,7 @@ import org.apache.spark.sql.execution.streaming.state._
 import org.apache.spark.sql.execution.SparkPlan
 import org.apache.spark.sql.streaming.OutputMode
 import org.apache.spark.sql.types.StructType
+import org.apache.spark.TaskContext
 
 
 /** Used to identify the state store for a given operator. */
@@ -150,6 +151,13 @@ case class StateStoreSaveExec(
         val numTotalStateRows = longMetric("numTotalStateRows")
         val numUpdatedStateRows = longMetric("numUpdatedStateRows")
 
+        // Abort the state store in case of error
+        TaskContext.get().addTaskCompletionListener(_ => {
+          if (!store.hasCommitted) {
+            store.abort()
+          }
+        })
+
         outputMode match {
           // Update and output all rows in the StateStore.
           case Some(Complete) =>

sql/core/src/main/scala/org/apache/spark/sql/execution/streaming/state/HDFSBackedStateStoreProvider.scala

@@ -203,7 +203,7 @@ private[state] class HDFSBackedStateStoreProvider(
     /**
      * Whether all updates have been committed
      */
-    override private[state] def hasCommitted: Boolean = {
+    override private[streaming] def hasCommitted: Boolean = {
       state == COMMITTED
     }
 

sql/core/src/main/scala/org/apache/spark/sql/execution/streaming/state/StateStore.scala

@@ -83,7 +83,7 @@ trait StateStore {
   /**
    * Whether all updates have been committed
    */
-  private[state] def hasCommitted: Boolean
+  private[streaming] def hasCommitted: Boolean
 }