Github user cloud-fan commented on a diff in the pull request:
https://github.com/apache/spark/pull/10604#discussion_r49824982
--- Diff:
sql/core/src/main/scala/org/apache/spark/sql/sources/interfaces.scala ---
@@ -683,7 +714,20 @@ abstract class HadoopFsRelation private[sql](
}
}
- buildInternalScan(requiredColumns, filters, inputStatuses,
broadcastedConf)
+ groupBucketFiles(inputStatuses).map { groupedBucketFiles =>
+ // For each bucket id, firstly we get all files belong to this
bucket, by detecting bucket
+ // id from file name. Then read these files into a RDD(use
one-partition empty RDD for empty
+ // bucket), and coalesce it to one partition. Finally union all
bucket RDDs to one result.
+ val perBucketRows = (0 until maybeBucketSpec.get.numBuckets).map {
bucketId =>
+ groupedBucketFiles.get(bucketId).map { inputStatuses =>
+ buildInternalScan(requiredColumns, filters, inputStatuses,
broadcastedConf).coalesce(1)
--- End diff --
This is a good point, but it requires more changes to the data source API.
Here we pass the bucket files to the underlying relation, and let it to scan
the files to a RDD. So we can't simply create a new hadooprdd to handle it, but
somehow enable the underlying relation to scan files to a RDD partition. How
about we do it in follow-up PR after think more about it?
---
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]
