zhedoubushishi commented on a change in pull request #2049:
URL: https://github.com/apache/hudi/pull/2049#discussion_r543587976



##########
File path: 
hudi-client/hudi-spark-client/src/main/java/org/apache/hudi/execution/bulkinsert/GlobalSortPartitionerWithRows.java
##########
@@ -0,0 +1,45 @@
+/*
+ * 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.sql.Dataset;
+import org.apache.spark.sql.Row;
+import org.apache.spark.sql.functions;
+
+/**
+ * A built-in partitioner that does global sorting for the input Rows across 
partitions after repartition for bulk insert operation, corresponding to the 
{@code BulkInsertSortMode.GLOBAL_SORT} mode.
+ */
+public class GlobalSortPartitionerWithRows implements 
BulkInsertPartitioner<Dataset<Row>> {
+
+  @Override
+  public Dataset<Row> repartitionRecords(Dataset<Row> rows, int 
outputSparkPartitions) {
+    // Now, sort the records and line them up nicely for loading.
+    // Let's use "partitionPath + key" as the sort key.
+    return 
rows.sort(functions.col(HoodieRecord.PARTITION_PATH_METADATA_FIELD), 
functions.col(HoodieRecord.RECORD_KEY_METADATA_FIELD))

Review comment:
       Looks like coalesce action is not based on the sorting result.
   
   Here is an example, if I set ```outputSparkPartitions``` to 2, the partition 
column is ```event_type```:
   
   ```
   val df = Seq(
     (100, "event_name_16", "2015-01-01T13:51:39.340396Z", "type1"),
     (101, "event_name_546", "2015-01-01T12:14:58.597216Z", "type2"),
     (104, "event_name_123", "2015-01-01T12:15:00.512679Z", "type1"),
     (108, "event_name_18", "2015-01-01T11:51:33.340396Z", "type1"),
     (109, "event_name_19", "2014-01-01T11:51:33.340396Z", "type3"),
     (110, "event_name_20", "2014-02-01T11:51:33.340396Z", "type3"),
     (105, "event_name_678", "2015-01-01T13:51:42.248818Z", "type2")
     ).toDF("event_id", "event_name", "event_ts", "event_type")
   ```
   (Here I added a new column partitionID for better understanding) Based on 
the current logic, after sorting and coalesce, the df would become:
   ```
   val df2 = df.sort(functions.col("event_type"), 
functions.col("event_id")).coalesce(2)
   df2.withColumn("partitionID", spark_partition_id).show(false)
   
   +--------+--------------+---------------------------+----------+-----------+
   |event_id|event_name    |event_ts                   |event_type|partitionID|
   +--------+--------------+---------------------------+----------+-----------+
   |100     |event_name_16 |2015-01-01T13:51:39.340396Z|type1     |0          |
   |108     |event_name_18 |2015-01-01T11:51:33.340396Z|type1     |0          |
   |105     |event_name_678|2015-01-01T13:51:42.248818Z|type2     |0          |
   |110     |event_name_20 |2014-02-01T11:51:33.340396Z|type3     |0          |
   |104     |event_name_123|2015-01-01T12:15:00.512679Z|type1     |1          |
   |101     |event_name_546|2015-01-01T12:14:58.597216Z|type2     |1          |
   |109     |event_name_19 |2014-01-01T11:51:33.340396Z|type3     |1          |
   +--------+--------------+---------------------------+----------+-----------+
   ```
   You can see the coalescing result actually does not depend on the sorting 
result. Each spark partition id contains 3 types of Hudi partitions.




----------------------------------------------------------------
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.

For queries about this service, please contact Infrastructure at:
[email protected]


Reply via email to