seawinde commented on code in PR #62607:
URL: https://github.com/apache/doris/pull/62607#discussion_r3263215457
##########
fe/fe-core/src/main/java/org/apache/doris/nereids/rules/exploration/mv/AbstractMaterializedViewWindowRule.java:
##########
@@ -104,20 +117,141 @@ protected Plan rewriteQueryByView(MatchMode matchMode,
StructInfo queryStructInf
viewToQuerySlotMapping,
ImmutableMap.of(), cascadesContext
);
- // Can not rewrite, bail out
+ // If generic rewrite fails, try roll up from query expressions.
if (expressionsRewritten.isEmpty()) {
- materializationContext.recordFailReason(queryStructInfo,
- "Rewrite expressions by view in window scan fail",
- () -> String.format("expressionToRewritten is %s,\n
mvExprToMvScanExprMapping is %s,\n"
- + "targetToSourceMapping = %s",
queryStructInfo.getExpressions(),
-
materializationContext.getShuttledExprToScanExprMapping(),
- viewToQuerySlotMapping));
- return null;
+ expressionsRewritten =
rollupWindowAggregateFunctions(queryStructInfo.getExpressions(),
Review Comment:
Can this fallback actually be reached for the documented case? Before
rewriteQueryByView(), the MV rewrite flow calls
StructInfo.isGraphLogicalEquals(), and window nodes appear to be compared as
full
WindowExpression objects. A query window using xx_merge(...) and an MV
window using xx_union(...) may be rejected before this fallback runs. Please
add a regression test proving the PR example is
rewritten, or handle this equivalence during graph comparison.
--
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]
