[GitHub] [hudi] nsivabalan commented on a diff in pull request #5328: [WIP][HUDI-3883] Fix Bulk Insert to repartition the dataset based on Partition Path

[GitBox <gi...@apache.org>]

nsivabalan commented on code in PR #5328:
URL: https://github.com/apache/hudi/pull/5328#discussion_r923159642


##########
hudi-client/hudi-spark-client/src/main/java/org/apache/hudi/execution/bulkinsert/PartitionSortPartitionerWithRows.java:
##########
@@ -19,19 +19,39 @@
 package org.apache.hudi.execution.bulkinsert;
 
 import org.apache.hudi.common.model.HoodieRecord;
+import org.apache.hudi.common.table.HoodieTableConfig;
 import org.apache.hudi.table.BulkInsertPartitioner;
 
+import org.apache.spark.sql.Column;
 import org.apache.spark.sql.Dataset;
 import org.apache.spark.sql.Row;
 
 /**
- * A built-in partitioner that does local sorting for each spark partitions after coalesce for bulk insert operation, corresponding to the {@code BulkInsertSortMode.PARTITION_SORT} mode.
+ * A built-in partitioner that does local sorting w/in the Spark partition,
+ * corresponding to the {@code BulkInsertSortMode.PARTITION_SORT} mode.
  */
-public class PartitionSortPartitionerWithRows implements BulkInsertPartitioner<Dataset<Row>> {
+public class PartitionSortPartitionerWithRows extends RepartitioningBulkInsertPartitionerBase<Dataset<Row>> {
+
+  public PartitionSortPartitionerWithRows(HoodieTableConfig tableConfig) {
+    super(tableConfig);
+  }
 
   @Override
-  public Dataset<Row> repartitionRecords(Dataset<Row> rows, int outputSparkPartitions) {
-    return rows.coalesce(outputSparkPartitions).sortWithinPartitions(HoodieRecord.PARTITION_PATH_METADATA_FIELD, HoodieRecord.RECORD_KEY_METADATA_FIELD);
+  public Dataset<Row> repartitionRecords(Dataset<Row> dataset, int outputSparkPartitions) {
+    Dataset<Row> repartitionedDataset;
+
+    // NOTE: Datasets being ingested into partitioned tables are additionally re-partitioned to better
+    //       align dataset's logical partitioning with expected table's physical partitioning to
+    //       provide for appropriate file-sizing and better control of the number of files created.
+    //
+    //       Please check out {@code GlobalSortPartitioner} java-doc for more details
+    if (isPartitionedTable) {
+      repartitionedDataset = dataset.repartition(outputSparkPartitions, new Column(HoodieRecord.PARTITION_PATH_METADATA_FIELD));

Review Comment:
   Did you have any discussion w/ @vinothchandar  around this. If I am not wrong, decision was to not touch of the existing sort modes and introduce new ones. If not, let me know if you had brainstormed this w/ anyone else already. 



-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: commits-unsubscribe@hudi.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org