kou commented on code in PR #47297:
URL: https://github.com/apache/arrow/pull/47297#discussion_r2268403021
##########
cpp/src/arrow/compute/kernel.cc:
##########
@@ -475,23 +475,78 @@ std::string OutputType::ToString() const {
return "computed";
}
+// ----------------------------------------------------------------------
+// MatchConstraint
+
+std::shared_ptr<MatchConstraint> DecimalsHaveSameScale() {
+ class DecimalsHaveSameScaleConstraint : public MatchConstraint {
+ public:
+ bool Matches(const std::vector<TypeHolder>& types) const override {
+ DCHECK_GE(types.size(), 2);
+ DCHECK(std::all_of(types.begin(), types.end(),
+ [](const TypeHolder& type) { return
is_decimal(type.id()); }));
+ auto ty0 = checked_cast<const DecimalType*>(types[0].type);
+ DCHECK_NE(ty0, nullptr);
Review Comment:
> Sorry I'm not following. Those two `static_assert`s only assert the from
and to types are classes. There can be cases that both are classes but without
inheritance, that can still result in `nullptr`?
Oh, sorry. I haven't looked at these `static_assert`s carefully... You're
right.
> This `is_decimal` merely checks the internal `type_id`. I'm not sure if we
can assume `ty.type_id == Type::DECIMAL` implies `typeof(ty) == class
DecimalType`.
I think that we can assume it based on the current implementation but I'm OK
with keeping this `DCHECK_NE()` now.
https://github.com/apache/arrow/blob/2b8db0094baed09875d44ebe33b78dd7ce859ab5/cpp/src/arrow/type_traits.h#L1109-L1120
https://github.com/apache/arrow/blob/2b8db0094baed09875d44ebe33b78dd7ce859ab5/cpp/src/arrow/type.h#L1047-L1049
https://github.com/apache/arrow/blob/2b8db0094baed09875d44ebe33b78dd7ce859ab5/cpp/src/arrow/type.h#L1082-L1084
https://github.com/apache/arrow/blob/2b8db0094baed09875d44ebe33b78dd7ce859ab5/cpp/src/arrow/type.h#L1116-L1118
https://github.com/apache/arrow/blob/2b8db0094baed09875d44ebe33b78dd7ce859ab5/cpp/src/arrow/type.h#L1149-L1151
--
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]
