rangareddy commented on issue #17064:
URL: https://github.com/apache/hudi/issues/17064#issuecomment-5391453358

   This issue was reviewed as part of the JIRA-migrated backlog triage 
(HUDI-9544).
   
   **Findings: still open.**
   
   The method named in the description, 
`HoodieBackedTableMetadata#readRecordIndexWithMapping`, still exists.
   
   The reported cause is structural rather than a single defect: metrics 
recorded inside a lazily-evaluated RDD transformation never fire if the caller 
only builds the RDD and never forces it. That means the fix is a design choice 
- record the metric in an action, use an accumulator, or document that the 
metric is only valid after materialisation - rather than a line to change.
   
   **This is one of four related tickets**, all the same underlying problem - 
the `HoodieData` abstraction is not honoured end to end, so callers cannot pass 
a distributed implementation where the API says they can:
   
   - #17052 (HUDI-9515) - index lookup requires the set to fit on one machine; 
needs an RDD-based implementation
   - #17063 (HUDI-9542) - `HoodieListData` / `HoodieListPairData` need a notion 
of partitions
   - #17064 (HUDI-9544) - the RLI metric does not fire for RDD because of lazy 
evaluation
   - #17101 (HUDI-9613) - `HoodieBackedTableMetadata` asserts `instanceof 
HoodieListData` at runtime
   
   Fixing them independently risks four partial answers. Recommend one design 
covering what `HoodieData` guarantees - partitioning, laziness, and which 
implementations each API accepts - with these as its consequences.
   
   Keeping this open.
   


-- 
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]

Reply via email to