LadyForest commented on code in PR #22478:
URL: https://github.com/apache/flink/pull/22478#discussion_r1179903642
##########
flink-table/flink-table-planner/src/main/java/org/apache/flink/table/planner/plan/nodes/exec/serde/RexNodeJsonSerializer.java:
##########
@@ -230,12 +226,10 @@ private static void serializeLiteralValue(
case BIGINT:
gen.writeNumberField(FIELD_NAME_VALUE, ((BigDecimal)
value).longValue());
break;
- case FLOAT:
Review Comment:
In `RexNodeJsonSerdeTest`, we have a test spec, see
https://github.com/apache/flink/blob/3664609c7622ccae80e36e85099a1b79b5935fe9/flink-table/flink-table-planner/src/test/java/org/apache/flink/table/planner/plan/nodes/exec/serde/RexNodeJsonSerdeTest.java#L578
```java
rexBuilder.makeApproxLiteral(
BigDecimal.valueOf(Float.MAX_VALUE),
FACTORY.createSqlType(SqlTypeName.FLOAT))
```
`makeApproxLiteral` will create the literal with `FLOAT` as the type of the
literal, and `DOUBLE` as the SQL type of literal, see
```java
/**
* Creates an approximate numeric literal (double or float).
*
* @param bd literal value
* @param type approximate numeric type
* @return new literal
*/
public RexLiteral makeApproxLiteral(@Nullable BigDecimal bd, RelDataType
type) {
assert SqlTypeFamily.APPROXIMATE_NUMERIC.getTypeNames().contains(
type.getSqlTypeName());
return makeLiteral(bd, type, SqlTypeName.DOUBLE);
}
```
So now we're passing the type of the literal (i.e. the `FLOAT`) to the
`serializeLiteral` method
```java
serializeLiteralValue(value, literal.getType().getSqlTypeName(), gen)
```
If we still serialize the `FLOAT` type using `FLOAT`, the serialized data
may lose precision.
The reason why the previous version does not have this issue is that the
parameter passed to the method is`DOUBLE`(the SQL type of literal)

--
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]