gortiz commented on code in PR #12265:
URL: https://github.com/apache/pinot/pull/12265#discussion_r1498970724
##########
pinot-segment-local/src/main/java/org/apache/pinot/segment/local/segment/index/loader/IndexLoadingConfig.java:
##########
@@ -303,7 +303,7 @@ private <C extends IndexConfig> ColumnConfigDeserializer<C>
getDeserializer(Inde
deserializer = IndexConfigDeserializer.fromMap(table ->
fromIndexLoadingConfig);
} else if (_segmentTier == null) {
deserializer =
- IndexConfigDeserializer.fromMap(table ->
fromIndexLoadingConfig).withFallbackAlternative(stdDeserializer);
+
stdDeserializer.withFallbackAlternative(IndexConfigDeserializer.fromMap(table
-> fromIndexLoadingConfig));
Review Comment:
I think this is the change @Jackie-Jiang was worried about. And I think it
makes sense to apply the change. Before this PR, our index config test verified
the behavior in the ideal scenario where `IndexingLoadingConfig` was not used.
But this PR adds a new check to verify that semantics are conserved when
`IndexLoadingConfig` is used... and it proves there are issues right now in
this case.
I'm not 100% sure if this is the correct solution, but in indexes like
ForwardIndex that are enabled by default, if we prioritizes
`IndexConfigDeserializer.fromMap(table -> fromIndexLoadingConfig)` instead of
`stdDeserializer`, the `stdDeserializer` is always ignored which is clearly
incorrect.
IMHO the correct solution would be to remove `IndexingLoadingConfig`, as we
planned during `index-spi`.
--
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]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
