akashrn5 commented on a change in pull request #3846:
URL:
https://github.com/apache/carbondata/pull/3846#discussion_r457857926##########
File path: integration/spark/src/main/scala/org/apache/spark/sql/execution/command/mutation/merge/CarbonMergeDataSetCommand.scala
##########
@@ -175,6 +178,21 @@ case class CarbonMergeDataSetCommand(
LOGGER.error("writing of update status file failed")
throw new CarbonMergeDataSetException("writing of update status file failed")
}
+ if (carbonTable.isHivePartitionTable) {
+ // If load count is 0 and if merge action contains delete operation, update
+ // tableUpdateStatus file name in loadMeta entry
+ if (count == 0 && hasDelAction && !tuple._1.isEmpty) {
+ val loadMetaDataDetails = SegmentStatusManager.readTableStatusFile(CarbonTablePath
+ .getTableStatusFilePath(carbonTable.getTablePath))
+ CarbonUpdateUtil.updateTableMetadataStatus(loadMetaDataDetails.map(l =>
+ new Segment(l.getMergedLoadName,
Review comment:
rename variable `l` to `loadMetadataDetails`
##########
File path: integration/spark/src/test/scala/org/apache/carbondata/spark/testsuite/merge/MergeTestCase.scala
##########
@@ -725,6 +725,45 @@ class MergeTestCase extends QueryTest with BeforeAndAfterAll {
checkAnswer(sql("select * from target order by key"), Seq(Row("c", "200"), Row("d", "3"), Row("e", "100")))
}
+ test("check the ccd delete with partition") {
+ sql("drop table if exists target")
+
+ val initframe = sqlContext.sparkSession.createDataFrame(Seq(
+ Row("a", "0"),
+ Row("b", "1"),
+ Row("c", "2"),
+ Row("d", "3")
+ ).asJava, StructType(Seq(StructField("key", StringType), StructField("value", StringType))))
+
+ initframe.write
+ .format("carbondata")
+ .option("tableName", "target")
+ .option("partitionColumns", "value")
+ .mode(SaveMode.Overwrite)
+ .save()
+ val target = sqlContext.read.format("carbondata").option("tableName", "target").load()
+ var ccd =
+ sqlContext.sparkSession.createDataFrame(Seq(
+ Row("a", null, true, 1),
+ Row("b", null, true, 2),
+ Row("c", null, true, 3),
+ Row("e", "100", false, 6)
+ ).asJava,
+ StructType(Seq(StructField("key", StringType),
+ StructField("newValue", StringType),
+ StructField("deleted", BooleanType), StructField("time", IntegerType))))
+ ccd.createOrReplaceTempView("changes")
+
+ ccd = sql("SELECT key, latest.newValue as newValue, latest.deleted as deleted FROM ( SELECT key, max(struct(time, newValue, deleted)) as latest FROM changes GROUP BY key)")
+
+ target.as("A").merge(ccd.as("B"), "A.key=B.key").
+ whenMatched("B.deleted=true").delete().execute()
+
+ assert(getDeleteDeltaFileCount("target", "0") == 0)
Review comment:
if delete operation is successful, then it should contain delete delta file, please check and add proper assert for delete delta file count
----------------------------------------------------------------
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:
[hidden email]
Locked
