linliu-code commented on code in PR #12597:
URL: https://github.com/apache/hudi/pull/12597#discussion_r1934908471
##########
hudi-client/hudi-spark-client/src/main/java/org/apache/hudi/DefaultSparkRecordMerger.java:
##########
@@ -44,40 +47,46 @@ public String getMergingStrategy() {
}
@Override
- public Option<Pair<HoodieRecord, Schema>> merge(HoodieRecord older, Schema
oldSchema, HoodieRecord newer, Schema newSchema, TypedProperties props) throws
IOException {
+ public Option<Pair<HoodieRecord, Schema>> merge(HoodieRecord older,
+ Schema oldSchema,
+ HoodieRecord newer,
+ Schema newSchema,
+ TypedProperties props)
throws IOException {
ValidationUtils.checkArgument(older.getRecordType() ==
HoodieRecordType.SPARK);
ValidationUtils.checkArgument(newer.getRecordType() ==
HoodieRecordType.SPARK);
- if (newer instanceof HoodieSparkRecord) {
- HoodieSparkRecord newSparkRecord = (HoodieSparkRecord) newer;
- if (newSparkRecord.isDelete(newSchema, props)) {
- // Delete record
- return Option.empty();
- }
- } else {
- if (newer.getData() == null) {
- // Delete record
- return Option.empty();
- }
+ if (newer instanceof HoodieEmptyRecord) {
+ return Option.empty();
Review Comment:
Good point. For Cow table, it works since empty record will not be write.
For MOR table, a delete record should be new; therefore it should already be in
the buffer; returning empty will not change the buffer. The only case we don't
handle is: for event time ordering, if there are more updates records after the
delete; we can fix that in a separate PR if needed.
--
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]
