[GitHub] [hudi] Zouxxyy commented on a diff in pull request #7872: [HUDI-5716] Cleaning up `Partitioner`s hierarchy

["Zouxxyy (via GitHub)" <gi...@apache.org>]

Zouxxyy commented on code in PR #7872:
URL: https://github.com/apache/hudi/pull/7872#discussion_r1105531548


##########
hudi-client/hudi-spark-client/src/main/java/org/apache/hudi/execution/bulkinsert/RowCustomColumnsSortPartitioner.java:
##########
@@ -19,43 +19,70 @@
 package org.apache.hudi.execution.bulkinsert;
 
 import org.apache.hudi.common.model.HoodieRecord;
+import org.apache.hudi.common.table.HoodieTableConfig;
 import org.apache.hudi.config.HoodieWriteConfig;
-import org.apache.hudi.table.BulkInsertPartitioner;
-
+import org.apache.spark.sql.Column;
 import org.apache.spark.sql.Dataset;
 import org.apache.spark.sql.Row;
+import scala.collection.JavaConverters;
 
 import java.util.Arrays;
+import java.util.stream.Collectors;
+
+import static org.apache.hudi.common.util.ValidationUtils.checkState;
+import static org.apache.hudi.execution.bulkinsert.RDDCustomColumnsSortPartitioner.getOrderByColumnNames;
 
 /**
- * A partitioner that does sorting based on specified column values for each spark partitions.
+ * A partitioner that does local sorting for each RDD partition based on the tuple of
+ * values of the columns configured for ordering.
  */
-public class RowCustomColumnsSortPartitioner implements BulkInsertPartitioner<Dataset<Row>> {
+public class RowCustomColumnsSortPartitioner extends RepartitioningBulkInsertPartitionerBase<Dataset<Row>> {
 
-  private final String[] sortColumnNames;
+  private final String[] orderByColumnNames;
 
-  public RowCustomColumnsSortPartitioner(HoodieWriteConfig config) {
-    this.sortColumnNames = getSortColumnName(config);
+  public RowCustomColumnsSortPartitioner(HoodieWriteConfig config, HoodieTableConfig tableConfig) {
+    super(tableConfig);
+    this.orderByColumnNames = getOrderByColumnNames(config);
+
+    checkState(orderByColumnNames.length > 0);
   }
 
-  public RowCustomColumnsSortPartitioner(String[] columnNames) {
-    this.sortColumnNames = columnNames;
+  public RowCustomColumnsSortPartitioner(String[] columnNames, HoodieTableConfig tableConfig) {
+    super(tableConfig);
+    this.orderByColumnNames = columnNames;
+
+    checkState(orderByColumnNames.length > 0);
   }
 
   @Override
-  public Dataset<Row> repartitionRecords(Dataset<Row> records, int outputSparkPartitions) {
-    final String[] sortColumns = this.sortColumnNames;
-    return records.sort(HoodieRecord.PARTITION_PATH_METADATA_FIELD, sortColumns)
-        .coalesce(outputSparkPartitions);
+  public Dataset<Row> repartitionRecords(Dataset<Row> dataset, int targetPartitionNumHint) {
+    Dataset<Row> repartitionedDataset;
+
+    // NOTE: In case of partitioned table even "global" ordering (across all RDD partitions) could
+    //       not change table's partitioning and therefore there's no point in doing global sorting
+    //       across "physical" partitions, and instead we can reduce total amount of data being
+    //       shuffled by doing do "local" sorting:

Review Comment:
   As far as I know, RowCustomColumnsSortPartitioner will only be used in `cluster`. At this time, the files in the same FG should already be in one physical partition.



##########
hudi-client/hudi-spark-client/src/main/java/org/apache/hudi/execution/bulkinsert/NonSortPartitioner.java:
##########
@@ -31,38 +29,15 @@
  * <p>

Review Comment:
   `enforceNumOutputPartitions` has been deleted, maybe the expression here can be modified



##########
hudi-client/hudi-spark-client/src/main/java/org/apache/hudi/execution/bulkinsert/SparkBulkInsertPartitionerBase.java:
##########
@@ -0,0 +1,62 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements.  See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership.  The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License.  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package org.apache.hudi.execution.bulkinsert;
+
+import org.apache.hudi.common.model.HoodieRecord;
+import org.apache.hudi.table.BulkInsertPartitioner;
+import org.apache.spark.api.java.JavaRDD;
+import org.apache.spark.sql.Dataset;
+import org.apache.spark.sql.HoodieUnsafeUtils$;
+import org.apache.spark.sql.Row;
+import org.apache.spark.sql.SparkSession$;
+
+public abstract class SparkBulkInsertPartitionerBase<T> implements BulkInsertPartitioner<T> {
+
+  protected static Dataset<Row> tryCoalesce(Dataset<Row> dataset, int targetPartitionNumHint) {
+    // NOTE: In case incoming [[Dataset]]'s partition count matches the target one,
+    //       we short-circuit coalescing altogether (since this isn't done by Spark itself)
+    if (targetPartitionNumHint > 0 && targetPartitionNumHint != HoodieUnsafeUtils$.MODULE$.getNumPartitions(dataset)) {
+      return dataset.coalesce(targetPartitionNumHint);
+    }
+
+    return dataset;
+  }
+
+  protected static <T> JavaRDD<HoodieRecord<T>> tryCoalesce(JavaRDD<HoodieRecord<T>> records,
+                                                            int targetPartitionNumHint) {
+    // NOTE: In case incoming [[RDD]]'s partition count matches the target one,
+    //       we short-circuit coalescing altogether (since this isn't done by Spark itself)
+    if (targetPartitionNumHint > 0 && targetPartitionNumHint != records.getNumPartitions()) {

Review Comment:
   When `targetPartitionNumHint` > `records.getNumPartitions` coalesce is meaningless, maybe we can use `targetPartitionNumHint !< records.getNumPartitions()` as the condition
   
   In addition, I actually have a question, if `targetPartitionNumHint > records.getNumPartitions`, should we use `repartition` instead of `coalesce`?



-- 
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