zhangyue19921010 opened a new issue, #11115:
URL: https://github.com/apache/paimon/issues/11115

   ### Search before asking
   
   - [x] I searched the issues and found no report of this statistics-dispatch 
failure.
   
   ### Paimon version
   
   master, `3821db197` (2.2-SNAPSHOT).
   
   ### Compute Engine
   
   Java API, JDK 11.
   
   ### Minimal reproduce step
   
   Evaluate a NULL comparison against complete min/max bounds with unavailable 
null counts:
   
   ```java
   Predicate predicate = new 
PredicateBuilder(RowType.of(DataTypes.INT())).equal(0, null);
   predicate.test(
           3, GenericRow.of(1), GenericRow.of(3),
           BinaryArray.fromLongArray(new Long[0]));
   ```
   
   This throws from `CompareUtils.compareLiteral`. A null entry in a nonempty 
null-count array also triggers it. BETWEEN with a null bound follows the same 
failing path.
   
   ### What doesn't meet your expectations?
   
   Statistics filtering should reject these comparisons without throwing, as 
row-level evaluation already does. In `LeafBinaryFunction` and 
`LeafTernaryFunction`, the literal check currently runs only when the null 
count is known.
   
   ### Anything else?
   
   The fix checks null literals independently of the null count. Both 
regression methods fail on unchanged production code and pass with the fix. A 
full Maven `verify` run passed 106 related tests and Checkstyle, Spotless, 
Enforcer and RAT on JDK 11.
   
   Related #10852 handles omitted null-count arrays; this is the subsequent 
comparison dispatch. #9647 fixes a different path in range-predicate 
optimization.
   
   This report and patch were generated with OpenAI Codex and reviewed by an 
independent AI agent; they have not received human review.
   
   ### Are you willing to submit a PR?
   
   - [x] I'm willing to submit a PR!
   
   Prepared fix: 
https://github.com/zhangyue19921010/flink-table-store/commit/aac1edc9ee593d72b3835397054671a40184b5d7
   
   Please assign this issue to me so I can submit the tested fix.
   


-- 
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