github-actions[bot] commented on code in PR #67737:
URL: https://github.com/apache/doris/pull/67737#discussion_r4079180539
##########
fe/fe-core/src/main/java/org/apache/doris/nereids/trees/expressions/Cast.java:
##########
@@ -316,26 +319,46 @@ public static boolean mayFailOnNonNullInput(DataType
sourceType, DataType target
return false;
}
- // BE casts to a character type through to_string_batch for these
source types.
- // JSON and VARIANT take separate paths and are intentionally left
conservative.
- boolean concreteNumber = (sourceType.isIntegralType() &&
sourceType.width() > 0)
- || sourceType.isFloatLikeType() ||
sourceType.isDecimalLikeType();
+ // Every valid typed value has a string representation. The generic
serde path and the
+ // dedicated JSON/VARIANT paths do not report data-dependent
conversion failures.
if (targetType.isStringLikeType()) {
- return !(sourceType.isStringLikeType() ||
sourceType.isBooleanType() || concreteNumber
- || sourceType.isDateLikeType() || sourceType.isTimeType()
- || sourceType.isArrayType() || sourceType.isMapType() ||
sourceType.isStructType());
+ return false;
Review Comment:
**[P1] Keep JSON-to-string out of the failure-free COUNT path**
This makes the following plan eligible:
```text
LogicalAggregate(count(CAST(j AS STRING)))
PhysicalStorageLayerAggregate(COUNT)
PhysicalOlapScan(j JSON NOT NULL)
```
For a real row containing a valid JSON object, the cast is non-NULL and the
unpushed count is 1. The pushed OLAP path instead has `VStatisticsIterator`
insert a source-typed default; JSON uses `ColumnString`, whose default is zero
bytes, and both JSON-to-string implementations turn that invalid empty JSONB
payload into SQL NULL. The retained upper count can therefore return 0 for a
non-empty segment. Keep JSON conservative here:
```suggestion
return sourceType.isJsonType();
```
Please also update the JSON classifier assertion and add an OLAP runtime
case with a valid `JSON NOT NULL` object so this storage-default boundary is
exercised.
--
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]
