manojpec commented on a change in pull request #4352:
URL: https://github.com/apache/hudi/pull/4352#discussion_r786377970
##########
File path:
hudi-client/hudi-spark-client/src/main/java/org/apache/hudi/index/bloom/SparkHoodieBloomIndexHelper.java
##########
@@ -70,26 +69,44 @@ public static SparkHoodieBloomIndexHelper getInstance() {
JavaRDD<Tuple2<String, HoodieKey>> fileComparisonsRDD =
HoodieJavaRDD.getJavaRDD(fileComparisonPairs)
.map(pair -> new Tuple2<>(pair.getLeft(), pair.getRight()));
- Map<String, Long> comparisonsPerFileGroup = computeComparisonsPerFileGroup(
- config, recordsPerPartition, partitionToFileInfo, fileComparisonsRDD,
context);
- int inputParallelism =
-
HoodieJavaPairRDD.getJavaPairRDD(partitionRecordKeyPairs).partitions().size();
+
+ int inputParallelism =
HoodieJavaPairRDD.getJavaPairRDD(partitionRecordKeyPairs).partitions().size();
int joinParallelism = Math.max(inputParallelism,
config.getBloomIndexParallelism());
LOG.info("InputParallelism: ${" + inputParallelism + "}, IndexParallelism:
${"
+ config.getBloomIndexParallelism() + "}");
- if (config.useBloomIndexBucketizedChecking()) {
+ JavaRDD<List<HoodieKeyLookupResult>> keyLookupResultRDD;
+ if (config.isMetaIndexBloomFilterEnabled()) {
+ // Step 1: Sort by file id
+ JavaRDD<Tuple2<String, HoodieKey>> sortedFileIdAndKeyPairs =
+ fileComparisonsRDD.sortBy(entry -> entry._1, true, joinParallelism);
+
+ // Step 2: Use bloom filter to filter and the actual log file to get the
record location
+ final boolean isBloomFiltersBatchLoadEnabled =
config.isMetaIndexBloomFilterBatchLoadEnabled();
+ if (isBloomFiltersBatchLoadEnabled) {
Review comment:
Right, batch is performant for small to medium dataset. For a very large
dataset, lazy mode is ok.
--
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]
