HyukjinKwon commented on a change in pull request #23559: [SPARK-26630][SQL]
Fix ClassCastException in TableReader while creating HadoopRDD
URL: https://github.com/apache/spark/pull/23559#discussion_r249276853
##########
File path: sql/hive/src/main/scala/org/apache/spark/sql/hive/TableReader.scala
##########
@@ -288,29 +286,65 @@ class HadoopTableReader(
}
}
+ /**
+ * The entry of creating a RDD.
+ */
+ private def createHadoopRDD(
+ inputClassName: String, localTableDesc: TableDesc, inputPathStr:
String): RDD[Writable] = {
+ if (classOf[org.apache.hadoop.mapreduce.InputFormat[_, _]]
+ .isAssignableFrom(Utils.classForName(inputClassName))) {
+ createNewHadoopRdd(localTableDesc, inputPathStr, inputClassName)
+ } else {
+ createOldHadoopRdd(localTableDesc, inputPathStr, inputClassName)
+ }
+ }
+
/**
* Creates a HadoopRDD based on the broadcasted HiveConf and other job
properties that will be
* applied locally on each slave.
*/
- private def createHadoopRdd(
- tableDesc: TableDesc,
- path: String,
- inputFormatClass: Class[InputFormat[Writable, Writable]]): RDD[Writable] =
{
+ private def createOldHadoopRdd(
+ tableDesc: TableDesc, path: String, inputClassName: String):
RDD[Writable] = {
val initializeJobConfFunc =
HadoopTableReader.initializeLocalJobConfFunc(path, tableDesc) _
val rdd = new HadoopRDD(
sparkSession.sparkContext,
_broadcastedHadoopConf.asInstanceOf[Broadcast[SerializableConfiguration]],
Some(initializeJobConfFunc),
- inputFormatClass,
+ Utils.classForName(inputClassName)
+
.asInstanceOf[java.lang.Class[org.apache.hadoop.mapred.InputFormat[Writable,
Writable]]],
classOf[Writable],
classOf[Writable],
_minSplitsPerRDD)
// Only take the value (skip the key) because Hive works only with values.
rdd.map(_._2)
}
+
+ /**
+ * Creates a HadoopRDD based on the broadcasted HiveConf and other job
properties that will be
+ * applied locally on each slave.
+ */
+ private def createNewHadoopRdd(
+ tableDesc: TableDesc, path: String, inputClassName: String):
RDD[Writable] = {
+
+ val newJobConf = new JobConf(hadoopConf)
+ (HadoopTableReader.initializeLocalJobConfFunc(path, tableDesc)
_).apply(newJobConf)
Review comment:
`HadoopTableReader.initializeLocalJobConfFunc(path, tableDesc)(newJobConf)`
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
With regards,
Apache Git Services
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
