Module#=== should call #kind_of? on the object rather than rb_obj_is_kind_of which only searches the ancestor heirarchy
In trying to implement a Decorator pattern which wraps Hashes and/or Arrays, I would like case equality to work such that the decorator would be handled by case statements on Array or Hash:
case when 'Array' # handle Arrays when 'Hash' # handle Hashes end
My decorators delegate #is_a? and #kind_of? to the decorated object so that #is_a?(Hash) or #kind_of?(Hash) return true even though the decorator does not inherit from Array or Hash.
When trying to use the decorator in case statements, however, the case-equality is handled by Array#=== or Hash#===. This is a method inherited from Module#=== which calls rb_obj_is_kind_of:
The implementation of rb_obj_is_kind_of does not call arg#kind_of?(mod) as its name might suggest, but instead searches the ancestor hierarchy:
The result of this is that even though my decorator implements the Hash contract, even though it responds to #is_a?(Hash) and #kind_of?(Hash) with true, it cannot be considered a Hash in a case statement short of monkeypatching the core Hash#=== operator.
A proposed fix would be to simply replace the rb_obj_is_kind_of call with:
rb_funcall(arg, rb_intern("kind_of?"), mod)
A more aggressive fix might be to modify the implementation of rb_obj_is_kind_of but that is called at multiple other places in the source code.
This is likely only going to be acceptable for ruby-3.0 and be considered a potentially breaking change I would imagine.
Updated by matz (Yukihiro Matsumoto) 8 months ago
- Status changed from Open to Rejected
I understand your needs but the receiver of
=== is a class/module, not the decorator. So if you want to override the behavior of the
case statement, you need some complex mechanism like coercing. I am not positive about adding that kind of complex system since it could slow down every