voonhous opened a new pull request, #20758:
URL: https://github.com/apache/hudi/pull/20758

   ### Describe the issue this Pull Request addresses
   
   Part of #20630. Stacked on #20757: only the last two commits are new; review 
them 
[here](https://github.com/voonhous/hudi/compare/trino-mapper-column-types...trino-record-context).
 The diff against master shrinks to them once #20757 lands.
   
   ### Summary and Changelog
   
   With `hudi.schema-column-types.enabled` on, MOR merge reads plug a 
`HoodieReaderContext<TrinoRow>` into the file-group reader instead of the Avro 
one. A `TrinoRow` is a view of a page position, so a base row with no log 
update is never converted: it is copied into the output page. Avro is used only 
where hudi-common hands it over or asks for it: classic Avro log blocks, 
payload merges and spill.
   
   - `TrinoRecordContext` returns the value classes the Avro path orders by, so 
ordering values compare with Avro log rows and delete records as before.
   - Commit-time, event-time and payload merges take `TrinoRow`s. Tables with a 
custom `HoodieRecordMerger` stay on the Avro context, as does everything with 
the property off.
   - Merged splits report a `hudiTrinoRowMergeSplits` connector metric.
   - The archived timeline reader moves from `HudiAvroSerializer` to 
`HoodieSchemaAvroSerializer`.
   
   ### Impact
   
   None by default. With the property on, MOR merges skip the per-row Avro 
conversion. Not yet measured with the #20626 benchmarks.
   
   ### Risk Level
   
   low: gated, off by default. The merge-mode, payload and custom-merger suites 
run with the property off and on, and the metric checks which context each 
split took.
   
   ### Documentation Update
   
   none
   
   ### Contributor's checklist
   
   - [x] Read through [contributor's 
guide](https://hudi.apache.org/contribute/how-to-contribute)
   - [x] Enough context is provided in the sections above
   - [x] Adequate tests were added if applicable
   


-- 
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