Gabriel39 commented on PR #66498:
URL: https://github.com/apache/doris/pull/66498#issuecomment-5306921317

   Thanks for tracing this against Paimon 1.4.2. I agree with the technical 
conclusion that the existing commit options do not provide a row-level 
concurrency fence for ordinary primary-key updates. However, that confirms the 
original P1 rather than resolving it.\n\nThis PR exposes standard SQL  /  and 
can report success after either overwriting columns from a concurrent writer or 
having its stale row discarded by . Following the native upsert conflict model 
does not make those outcomes safe UPDATE semantics. Since a correct fence 
requires an upstream atomic primitive and a full statement retry, the safe 
scope for this PR is to defer/reject UPDATE and MERGE UPDATE until that 
primitive exists. A latest-snapshot pre-check, commit-only retry, or Doris-only 
lock would indeed be insufficient.\n\nThe short-circuit implementation change 
itself looks correct: rebuilding through  preserves . There are still two test 
issues to address:\n\n1. The inactive-assignment assertion is currently folde
 d away.  is simplified by  to  because both result branches are equal. Please 
use distinct branches, for example , so evaluating the inactive assignment 
would actually fail.\n2. The  regression exercises  through , but it does not 
exercise . Please add focused FE tests for both rewrite paths and assert that 
the rewritten expression remains a .\n\nThe later failing predicate is still 
useful coverage for the  path, but the current test does not substantiate the 
inactive-assignment or second-rewriter claims.


-- 
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