github-actions[bot] commented on code in PR #61813:
URL: https://github.com/apache/doris/pull/61813#discussion_r2999756345
##########
fe/fe-core/src/main/java/org/apache/doris/nereids/rules/analysis/ExpressionAnalyzer.java:
##########
@@ -875,6 +875,33 @@ public Expression
visitComparisonPredicate(ComparisonPredicate cp, ExpressionRew
@Override
public Expression visitCaseWhen(CaseWhen caseWhen,
ExpressionRewriteContext context) {
+ if (caseWhen.getValue().isPresent()) {
+ // Simple case: CASE value WHEN cond THEN result ...
+ // Analyze value once, then construct EqualTo(analyzedValue,
analyzedCond) per WhenClause
+ Expression analyzedValue = caseWhen.getValue().get().accept(this,
context);
+
+ List<WhenClause> newWhenClauses = new ArrayList<>();
+ for (WhenClause whenClause : caseWhen.getWhenClauses()) {
+ Expression operand = whenClause.getOperand().accept(this,
context);
+ Expression result = whenClause.getResult().accept(this,
context);
+ EqualTo rawEqualTo = new EqualTo(analyzedValue, operand);
+ registerPlaceholderIdToSlot(rawEqualTo, context,
analyzedValue, operand);
+ Expression equalTo =
TypeCoercionUtils.processComparisonPredicate(rawEqualTo);
Review Comment:
Nit: This `registerPlaceholderIdToSlot` call is effectively dead code in
this context. The method checks whether `cp.right()` or `cp.left()` is a
`Placeholder` instance. In `visitComparisonPredicate`, the original
pre-analysis `cp` is passed (so Placeholder children are still visible), but
here `rawEqualTo` is freshly constructed with already-analyzed children
(`analyzedValue` and `operand`), which have already had their Placeholders
resolved by `accept(this, context)`. So `rawEqualTo.right() instanceof
Placeholder` and `rawEqualTo.left() instanceof Placeholder` will always be
false.
This is harmless (just two instanceof checks that always fail), but if the
intent is to support prepared statement short-circuit optimization for simple
CASE expressions, the approach would need to pass the original unanalyzed
expressions as the `ComparisonPredicate` argument — similar to how
`visitComparisonPredicate` passes the original `cp`.
Consider either removing this call (since it's a no-op) or adding a comment
explaining it's kept for future correctness if the prepared statement path
needs to handle this case.
--
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]
