Github user davies commented on a diff in the pull request:
https://github.com/apache/spark/pull/8512#discussion_r38594881
--- Diff:
sql/hive/src/main/scala/org/apache/spark/sql/hive/TableReader.scala ---
@@ -249,6 +256,68 @@ class HadoopTableReader(
}
/**
+ * If `spark.sql.hive.parallelFileListing` is true, then pre-calculate
input splits for all the
+ * partitions together that can be cached in HadoopRDDs.
+ */
+ private def populateInputSplitsCache(
+ hivePartitions: Map[HivePartition, Class[_ <: Deserializer]],
+ filterOpt: Option[PathFilter]): Map[String, Array[InputSplit]] = {
+ if (hivePartitions.isEmpty) {
+ Map[String, Array[InputSplit]]()
+ } else {
+ val inputSplitsCache = collection.mutable.Map[String,
ArrayBuffer[InputSplit]]()
+ // Compute input splits for all the partitions together if they have
the same input format.
+ // This is faster than computing them individually because listing
multiple input dirs can be
+ // done in parallel using
`mapreduce.input.fileinputformat.list-status.num-threads`.
+ if (sc.conf.parallelFileListing) {
+ val inputFormatClass = relation.hiveQlTable.getInputFormatClass
+ val homogeneousInputFormat = hivePartitions.forall {
+ case (part, _) => part.getInputFormatClass == inputFormatClass
+ }
+
+ if (homogeneousInputFormat) {
+ val jobConf = new JobConf(hiveExtraConf)
+ val minPartitions = _minSplitsPerRDD * hivePartitions.size
+ val combinedPaths = hivePartitions.map { case (part, _) =>
+ applyFilterIfNeeded(part.getDataLocation, filterOpt)
+ }.mkString(",")
+
+ HadoopTableReader.initializeLocalJobConfFunc(combinedPaths,
relation.tableDesc)(jobConf)
+
+ val inputSplits =
+ // If Hive table is stored on S3, we can use S3 bulk listing to
speed up listing
+ // even further. This is particularly faster when listing a
large number of files
+ // on S3.
+ if (sc.conf.s3BulkListing) {
--- End diff --
Should we also test that the paths is coming from S3? Or once we turn this
one, it will fail to load any table from HDFS.
---
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.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
