dwsmith1983 opened a new issue, #6039:
URL: https://github.com/apache/datafusion-comet/issues/6039

   ### What is the problem the feature request solves?
   
   The native Iceberg scan skips Parquet pages: `IcebergScanExec` enables 
iceberg-rust's row selection and passes the residual predicate, so iceberg-rust 
loads the page index and builds a page-level row selection. Measured locally on 
one data file with one row group and 158 `id` pages, `WHERE id BETWEEN 1000 AND 
1100` reads about 550 KB of a 21 MB file.
   
   Nothing in the test suite would notice if that stopped. The behaviour hinges 
on one builder flag in the scan operator and on iceberg-rust internals that 
change with every pin bump, and `CometIcebergNativeScanExec` reports no 
page-index counters (iceberg-rust's scan metrics expose only bytes read), so a 
regression would show up only as slower queries.
   
   ### Describe the potential solution
   
   A regression test in `CometIcebergNativeSuite` that writes a 
single-row-group, many-page table, verifies that layout from the footer so the 
fixture cannot degrade into a shape where file-level or row-group pruning 
explains the savings, and asserts that a narrow range reads a small fraction of 
the full read's `bytes_scanned`.
   
   ### Additional context
   
   Page-index counters on the Iceberg node, matching the Parquet scan's 
`page_index_pages_pruned`, need iceberg-rust to expose them first. Related: 
#6026.


-- 
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]

Reply via email to