BiteTheDDDDt opened a new pull request, #64413:
URL: https://github.com/apache/doris/pull/64413

   ### What problem does this PR solve?
   
   Issue Number: N/A
   
   Related PR: N/A
   
   Problem Summary:
   
   This is an alternative to keeping separate storage and execution schemas in 
the BE segment iterator for storage expression pushdown.
   
   For AGG key tables and non-MOW unique key tables, the planner now expands 
the OLAP scan tuple so the scan schema starts with the storage key columns in 
storage order. Missing key slots are marked in thrift as FE-filled key slots. 
The scan node projects back to the original execution output after storage 
filters and runtime filters are translated.
   
   On the BE side, the marked key slots are mapped to tablet column ids. Direct 
reader paths can fill those storage-only key columns without reading their data 
pages, while merge and aggregation reader paths still keep real key columns for 
storage semantics. Since FE now provides the storage-aligned key prefix, the 
non-direct scanner no longer rebuilds the return column list as "all keys plus 
value outputs"; it only checks the key prefix and keeps the existing 
sequence-column expansion.
   
   ### Release note
   
   None
   
   ### Check List (For Author)
   
   - Test: Manual test
       - `build-support/clang-format.sh`
       - `git diff --check`
       - `./build.sh --fe`
       - `./build.sh --be`
       - `build-support/run-clang-tidy.sh --build-dir be/build_Release --base 
23e21f44f0080e25deff5dcdda2e61af3efc9480` was attempted, but this checkout 
reports existing file-level clang-tidy diagnostics in the touched storage files 
before producing a clean changed-line gate.
   - Behavior changed: No user-visible behavior change. Internal OLAP scan 
tuple planning and storage-read column handling are changed.
   - Does this need documentation: No
   


-- 
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