Github user rxin commented on a diff in the pull request:
https://github.com/apache/spark/pull/1385#discussion_r14867102
--- Diff:
sql/hive/src/main/scala/org/apache/spark/sql/hive/TableReader.scala ---
@@ -206,17 +202,10 @@ class HadoopTableReader(@transient _tableDesc:
TableDesc, @transient sc: HiveCon
tableDesc: TableDesc,
path: String,
inputFormatClass: Class[InputFormat[Writable, Writable]]):
RDD[Writable] = {
-
- val initializeJobConfFunc =
HadoopTableReader.initializeLocalJobConfFunc(path, tableDesc) _
-
- val rdd = new HadoopRDD(
- sc.sparkContext,
-
_broadcastedHiveConf.asInstanceOf[Broadcast[SerializableWritable[Configuration]]],
- Some(initializeJobConfFunc),
- inputFormatClass,
- classOf[Writable],
- classOf[Writable],
- _minSplitsPerRDD)
+ val jobConf = new
JobConf(_broadcastedHiveConf.value.value.asInstanceOf[Configuration])
--- End diff --
Yes. With our current implementation, each Hive partition in Spark SQL
creates one HadoopRDD. We absolutely cannot afford broadcasting the conf for
each HadoopRDD/Hive partition.
---
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 [email protected] or file a JIRA ticket
with INFRA.
---
