danny0405 commented on code in PR #9936:
URL: https://github.com/apache/hudi/pull/9936#discussion_r1375610014
##########
hudi-flink-datasource/hudi-flink/src/main/java/org/apache/hudi/sink/bulk/RowDataKeyGen.java:
##########
@@ -99,7 +99,7 @@ protected RowDataKeyGen(
this.recordKeyProjection = null;
} else {
this.recordKeyFields = recordKeys.get().split(",");
- if (this.recordKeyFields.length == 1) {
+ if (this.recordKeyFields.length == 1 && this.partitionPathFields.length
== 1) {
Review Comment:
Nice findings ~
I kind of think the handling of `RowDataKeyGen` is correct, what we need to
fix is the `ComplexAvroKeyGenerator`, it composite the record key string as
`key_name:key_val` even if the record key has only one field.
If we take a look at the `ComplexKeyGenerator.getRecordKey(InternalRow
internalRow, StructType schema)`, it also eliminate the `key_name` from the
record key string even if there are multiple partiton fields. And for
compatible interface `getRecordKey(GenericRecord record)`, it uses the delegate
generator `ComplexAvroKeyGenerator`, which causes the inconsistency of the two
interfaces.
In summary, I think `key_name:key_val` should be a mistake when the record
key fields is single, it should be simplified as `key_val`.
--
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]
