What is the best practice for a C++ getter method which is supposed to return a non trivial type, but a member which is of type class, or struct.
MyType MyClass::getMyType() { return mMyType; }
const MyType& MyClass::getMyType() { return mMyType; }
MyType* MyClass::getMyType() { return &mMyType; }
where
class MyType { /* ... */ };
class MyClass
{
private:
MyType mMyType;
}
I specifically worry about the following usages of this method. Can you please elaborate in details how this might affect copying the object, and the danger of dangling references and wild gone pointers if function()
wants to save it for further usage.
MyType* savedPointer;
SomeType function(MyType* pointer) { savedPointer = pointer; };
a. valid for 1. and 2.
{
MyType t = myClass.getMyType();
function(&t);
}
// is savedPointer still valid here?
b. valid for 1. and 2.
{
const MyType& t = myClass.getMyType();
function(&t);
}
// is savedPointer still valid here?
c. valid for 1. and 2.
{
MyType& t = myClass.getMyType();
function(&t);
}
// is savedPointer still valid here?
d. valid for 3.
{
MyType* t = myClass.getMyType();
function(t);
}
// is savedPointer still valid here?
where myClass
is an object of type MyClass
.
You can provide both const and non-const versions:
MyType & MyClass::getMyType() { return mMyType; }
MyType const & MyClass::getMyType() const { return mMyType; }
I wouldn't provide a pointer version, since that implies that the return value might be the null pointer, which it can never be in this instance.
The real point, however, is that you are basically giving the caller direct access to the internal object. If this is your intent, then you may as well make the data member public. If it isn't, then you will need to work harder to hide the object.
One option is to retain the MyType const &
accessor, but provide more indirect means to modify the internal object (setMyType(…)
or something more tailored to the semantics that you are trying to express at the level of the containing class).