zzwqqq commented on code in PR #5385:
URL: https://github.com/apache/calcite/pull/5385#discussion_r4454219290
##########
core/src/main/java/org/apache/calcite/sql2rel/RelDecorrelator.java:
##########
@@ -464,9 +497,48 @@ public RelNode removeCorrelationViaRule(RelNode root) {
* {@link org.apache.calcite.rel.core.Correlate}s might be removable in such
way).
*/
public RelNode removeCorrelationViaRule(RelNode root, RuleSet ruleSet) {
+ root = prepareCorrelates(root);
return applyHepProgram(root, ruleSetToHepProgram(ruleSet));
}
+ /** Removes unused RHS computations before matching rules or building frames.
+ * The checked input must also be the input used by the rewrite.
+ */
+ private RelNode prepareCorrelates(RelNode root) {
+ final RelNode prepared = root.accept(new RelHomogeneousShuttle() {
Review Comment:
Thanks. Following your suggestion below to trim globally,
`prepareForDecorrelation` now uses the existing `RelFieldTrimmer`. I've removed
shuttle from `prepareCorrelates`, so there's no new shuttle class.
##########
core/src/main/java/org/apache/calcite/sql2rel/RelDecorrelator.java:
##########
@@ -464,9 +497,48 @@ public RelNode removeCorrelationViaRule(RelNode root) {
* {@link org.apache.calcite.rel.core.Correlate}s might be removable in such
way).
*/
public RelNode removeCorrelationViaRule(RelNode root, RuleSet ruleSet) {
+ root = prepareCorrelates(root);
return applyHepProgram(root, ruleSetToHepProgram(ruleSet));
}
+ /** Removes unused RHS computations before matching rules or building frames.
+ * The checked input must also be the input used by the rewrite.
+ */
+ private RelNode prepareCorrelates(RelNode root) {
+ final RelNode prepared = root.accept(new RelHomogeneousShuttle() {
+ @Override public RelNode visit(RelNode other) {
+ final RelNode rel = super.visit(other);
+ if (!(rel instanceof Correlate)) {
+ return rel;
+ }
+ final Correlate correlate = (Correlate) rel;
+ RelNode right = correlate.getRight();
+ if (!isDeterministic(right)) {
+ final RelBuilder trimBuilder =
relBuilderFactory().create(right.getCluster(), null);
+ final RelNode trimmed = new RelFieldTrimmer(null,
trimBuilder).trim(right);
Review Comment:
Thanks, `prepareForDecorrelation` now calls the existing
`RelFieldTrimmer.trim(root)` on the whole expanded plan, replacing the
anonymous shuttle in `prepareCorrelates`. It runs after the scalar rules and
project extraction, before recursive decorrelation. I kept the separate trim
for the correlation-key projection because it may need fewer columns.
--
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]
