voonhous opened a new issue, #20003:
URL: https://github.com/apache/hudi/issues/20003

   Follow-up to #18780 (RFC-105 hudi-trino migration), raised in review of 
trinodb/trino#30682.
   
   ### What
   
   `hudi-trino` adds a `hudi.parquet.use-column-index` config property and a 
`parquet_use_column_index` session property, both defaulting to `false`. The 
property first appeared with the migration of the connector into this repo 
(#13493).
   
   Trino's in-tree Hudi connector has no such property. It takes the setting 
from `ParquetReaderConfig`, where `parquet.use-column-index` defaults to 
`true`, and Trino's Hive, Iceberg and Delta Lake connectors also default 
`parquet_use_column_index` to `true`. So once trinodb/trino#30682 switches 
Trino to `hudi-trino`, standalone Hudi catalogs stop using Parquet column 
indexes by default, while Hudi tables read through Trino's lakehouse catalog 
keep using them.
   
   ### Question
   
   Is the `false` default a correctness guard (for example, a known problem 
with column indexes in Hudi-written files) or just a conservative choice?
   
   - If it is conservative, align the default with Trino's `true`.
   - If it guards against a known problem, document the reason so Trino can 
document the difference between Hudi and lakehouse catalogs.
   
   ### Why here
   
   Under RFC-105, `HudiConfig` and `HudiSessionProperties` live in 
`org.apache.hudi:hudi-trino`, so the default can only change here.
   


-- 
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