zhangyue19921010 commented on code in PR #11004:
URL: https://github.com/apache/paimon/pull/11004#discussion_r4545850780


##########
paimon-common/src/main/java/org/apache/paimon/predicate/LeafPredicateExtractor.java:
##########
@@ -46,4 +49,25 @@ public Map<String, LeafPredicate> visit(CompoundPredicate 
predicate) {
         }
         return Collections.emptyMap();
     }
+
+    /**
+     * Returns the conjuncts that contain a transform leaf, which this 
extractor skips and callers
+     * must evaluate on rows, or null if there are none.
+     */
+    @Nullable
+    public static Predicate transformConjuncts(@Nullable Predicate predicate) {
+        List<Predicate> conjuncts =
+                PredicateBuilder.splitAnd(predicate).stream()
+                        .filter(LeafPredicateExtractor::containsTransform)

Review Comment:
   [P2] Preserve unconsumed sibling conjuncts with the transform residual
   
   On `07a3f3d`, a real `$snapshots` SDK read with `LOWER(commit_kind) = 
'append' AND schema_id = 999`, including `newRead().executeFilter()`, returns 
both append snapshots with schema ID 0 instead of no rows. The reader only 
pushes `snapshot_id`; this filter drops the plain `schema_id` conjunct and 
evaluates only `LOWER(commit_kind)`.
   
   The compound query previously failed on its transform, so accepting it while 
silently losing a sibling changes the complete-filter behavior. Could these 
readers retain all conjuncts their ID optimization does not actually consume, 
while preserving the existing Java String IN path? The same structure exists in 
the consumers, schemas and manifests readers. I added 
`SnapshotsTableTest#testTransformWithUnconsumedConjunctExecuteFilter`: 71 
existing tests passed and this additional regression failed.



-- 
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