caicancai commented on code in PR #3911:
URL: https://github.com/apache/calcite/pull/3911#discussion_r1713013287
##########
core/src/main/java/org/apache/calcite/sql2rel/StandardConvertletTable.java:
##########
@@ -2135,11 +2137,34 @@ private static class TimestampAddConvertlet implements
SqlRexConvertlet {
final RexNode op2;
switch (call.operandCount()) {
case 2:
- // BigQuery-style 'TIMESTAMP_ADD(timestamp, interval)'
- final SqlBasicCall operandCall = call.operand(1);
- qualifier = operandCall.operand(1);
- op1 = cx.convertExpression(operandCall.operand(0));
- op2 = cx.convertExpression(call.operand(0));
+ // Oracle-style 'ADD_MONTHS(date, integer months)'
+ if (call.getOperator() == SqlLibraryOperators.ADD_MONTHS) {
+ qualifier = new SqlIntervalQualifier(TimeUnit.MONTH, null,
SqlParserPos.ZERO);
+ RexNode op = cx.convertExpression(call.operand(1));
+ RelDataTypeFactory typeFactory = cx.getTypeFactory();
+
+ // In Calcite, cast('1.2' as integer) is invalid.
+ // For details, see
https://issues.apache.org/jira/browse/CALCITE-1439
+ // When trying to cast a string value to an integer type, Calcite may
+ // encounter errors if the string value cannot be successfully
converted.
+ // To handle this, the string needs to be first converted to a
double type,
+ // and then the double value can be converted to an integer type.
+ // Since the final target type is integer, converting the string to
double first
+ // will not lose precision for the add_months operation.
+ if (op.getType().getSqlTypeName() == SqlTypeName.CHAR) {
Review Comment:
> The Oracle Functions documentation describes it as: "The date argument can
be a datetime value or any value that can be implicitly converted to DATE".
Although the Spark documentation doesn't explicitly mention this, such implicit
conversions exist. I assume we can use makeCast to let the user determine
compatibility between types? WDYT
thanks, you are right
--
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]
