codope commented on code in PR #11153:
URL: https://github.com/apache/hudi/pull/11153#discussion_r1590512117
##########
hudi-client/hudi-client-common/src/main/java/org/apache/hudi/metadata/HoodieBackedTableMetadataWriter.java:
##########
@@ -575,6 +598,40 @@ private Pair<Integer, HoodieData<HoodieRecord>>
initializeRecordIndexPartition()
return Pair.of(fileGroupCount, records);
}
+ private static HoodieData<HoodieRecord>
readRecordKeysFromFileSlices(HoodieEngineContext engineContext,
Review Comment:
There is already `HoodieTableMetadataUtil#readRecordKeysFromFileSlices`.
Let's reuse that if possible, or maybe consolidate the two (i guess
`HoodieMergedReadHandle` is also using `HoodieMergedLogRecordScanner`)?
##########
hudi-spark-datasource/hudi-spark/src/test/scala/org/apache/hudi/functional/TestRecordLevelIndex.scala:
##########
@@ -55,6 +59,76 @@ class TestRecordLevelIndex extends RecordLevelIndexTestBase {
saveMode = SaveMode.Overwrite)
}
+ @ParameterizedTest
+ @EnumSource(classOf[HoodieTableType])
+ def testRLIInitializationForMorGlobalIndex(tableType: HoodieTableType): Unit
= {
Review Comment:
Is this test useful for COW? Shall we run it only for MOR? I am trying to
cut down extra test time. I think for COW, the test should pass even with
current code.
##########
hudi-client/hudi-client-common/src/main/java/org/apache/hudi/metadata/HoodieBackedTableMetadataWriter.java:
##########
@@ -575,6 +598,40 @@ private Pair<Integer, HoodieData<HoodieRecord>>
initializeRecordIndexPartition()
return Pair.of(fileGroupCount, records);
}
+ private static HoodieData<HoodieRecord>
readRecordKeysFromFileSlices(HoodieEngineContext engineContext,
+
List<Pair<String, FileSlice>> partitionFileSlicePairs,
Review Comment:
should work.. For COW, file slice won't have any log files for filegroup id,
which is handled by the HoodieMergedLogRecordScanner.
##########
hudi-spark-datasource/hudi-spark/src/test/scala/org/apache/hudi/functional/TestRecordLevelIndex.scala:
##########
@@ -55,6 +59,76 @@ class TestRecordLevelIndex extends RecordLevelIndexTestBase {
saveMode = SaveMode.Overwrite)
}
+ @ParameterizedTest
+ @EnumSource(classOf[HoodieTableType])
+ def testRLIInitializationForMorGlobalIndex(tableType: HoodieTableType): Unit
= {
+ val hudiOpts = commonOpts + (DataSourceWriteOptions.TABLE_TYPE.key ->
tableType.name()) +
+ (HoodieMetadataConfig.RECORD_INDEX_MIN_FILE_GROUP_COUNT_PROP.key -> "1")
+
+ (HoodieMetadataConfig.RECORD_INDEX_MAX_FILE_GROUP_COUNT_PROP.key -> "1")
+
+ (HoodieIndexConfig.INDEX_TYPE.key -> "RECORD_INDEX") +
+ (HoodieIndexConfig.RECORD_INDEX_UPDATE_PARTITION_PATH_ENABLE.key ->
"true") -
+ HoodieMetadataConfig.RECORD_INDEX_ENABLE_PROP.key
+
+ val dataGen1 = HoodieTestDataGenerator.createTestGeneratorFirstPartition()
+ val dataGen2 = HoodieTestDataGenerator.createTestGeneratorSecondPartition()
+
+ // batch1 inserts
+ val instantTime1 = getInstantTime()
+ val latestBatch = recordsToStrings(dataGen1.generateInserts(instantTime1,
5)).asScala
+ var operation = INSERT_OPERATION_OPT_VAL
+ val latestBatchDf =
spark.read.json(spark.sparkContext.parallelize(latestBatch, 1))
+ latestBatchDf.cache()
+ latestBatchDf.write.format("org.apache.hudi")
+ .options(hudiOpts)
+ .mode(SaveMode.Overwrite)
+ .save(basePath)
+ val deletedDf1 = calculateMergedDf(latestBatchDf, operation, true)
+ deletedDf1.cache()
+
+ // batch2. upsert. update few records to 2nd partition from partition1 and
insert a few to partition2.
+ val instantTime2 = getInstantTime()
+
+ val latestBatch2_1 =
recordsToStrings(dataGen1.generateUniqueUpdates(instantTime2, 3)).asScala
+ val latestBatchDf2_1 =
spark.read.json(spark.sparkContext.parallelize(latestBatch2_1, 1))
+ val latestBatchDf2_2 = latestBatchDf2_1.withColumn("partition",
lit(HoodieTestDataGenerator.DEFAULT_SECOND_PARTITION_PATH))
+ .withColumn("partition_path",
lit(HoodieTestDataGenerator.DEFAULT_SECOND_PARTITION_PATH))
+ val latestBatch2_3 =
recordsToStrings(dataGen2.generateInserts(instantTime2, 2)).asScala
+ val latestBatchDf2_3 =
spark.read.json(spark.sparkContext.parallelize(latestBatch2_3, 1))
+ val latestBatchDf2Final = latestBatchDf2_3.union(latestBatchDf2_2)
+ latestBatchDf2Final.cache()
+ latestBatchDf2Final.write.format("org.apache.hudi")
+ .options(hudiOpts)
+ .mode(SaveMode.Append)
+ .save(basePath)
+ operation = UPSERT_OPERATION_OPT_VAL
+ val deletedDf2 = calculateMergedDf(latestBatchDf2Final, operation, true)
+ deletedDf2.cache()
+
+ val hudiOpts2 = commonOpts + (DataSourceWriteOptions.TABLE_TYPE.key ->
tableType.name()) +
+ (HoodieMetadataConfig.RECORD_INDEX_MIN_FILE_GROUP_COUNT_PROP.key -> "1")
+
+ (HoodieMetadataConfig.RECORD_INDEX_MAX_FILE_GROUP_COUNT_PROP.key -> "1")
+
+ (HoodieIndexConfig.INDEX_TYPE.key -> "RECORD_INDEX") +
+ (HoodieIndexConfig.RECORD_INDEX_UPDATE_PARTITION_PATH_ENABLE.key ->
"true") +
+ (HoodieMetadataConfig.RECORD_INDEX_ENABLE_PROP.key -> "true")
+
+ val instantTime3 = getInstantTime()
+ // batch3. updates to partition2
+ val latestBatch3 =
recordsToStrings(dataGen2.generateUniqueUpdates(instantTime3, 2)).asScala
+ val latestBatchDf3 =
spark.read.json(spark.sparkContext.parallelize(latestBatch3, 1))
+ latestBatchDf3.cache()
+ latestBatchDf.write.format("org.apache.hudi")
+ .options(hudiOpts2)
+ .mode(SaveMode.Append)
+ .save(basePath)
+ val deletedDf3 = calculateMergedDf(latestBatchDf, operation, true)
+ deletedDf3.cache()
+ validateDataAndRecordIndices(hudiOpts, deletedDf3)
+ }
+
+ private def getInstantTime(): String = {
Review Comment:
Can we reuse `HoodieInstantTimeGenerator`? For the test, we have something
in `InProcessTimeGenerator`.
--
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: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
