Should I declare method "const" if it is "shallow const" only, not "deep const"?

Question

I have class Foo and its member bar_ is a pointer to some data. Method modify modifies the data, but not the pointer itself. Therefore I can declare the method as const:

class Foo {
public:
    Foo() : bar_(new double) {};
    void modify() const {*bar_ += 1;};
private:
    double *bar_;
};

If I declare the method as const, it will accessible from other const methods, which is more flexible. At the same time I can drop const as a hint for other developers and users that method modifies the data indirectly (and let think that the data is owned by the class). So I have a choice here: declare modify as const or drop const: void modify() const or void modify().

What are pros and cons of each approach? What do guidelines say? What should I do?

Answer 1

const after a method declaration is a statement of intent - const methods are meant to be idempotent; they do not modify the state of the object and can be called on const instances.

If the value pointed-to by bar_ is part of the object's state, then a const method should not modify it.

In addition, even the name modify() sounds like it modifies the object in some way, and thus should not be declared const.

But also the other way is important - if a method does not modify the state, it is recommended to declare it const (see C++ Core Guideline Con.2).