Within a member function of a class in C++, does it make a difference, if I use this->dataMember
or just dataMember
? What is considered better style? Is there any performance difference?
(I am not talking about the case where a local variable has the same name as the data member, in which case you must, to my knowledge, use this->
to distinguish between them.)
As a general rule, it's a question of local conventions. Most of the
places I've seen do not use this->
except when necessary, and that's
the convention I prefer as well, but I've heard of people who prefer to
use it systematically.
There are two cases when it is necessary. The first is if you've hidden
the name with the same name in local scope; if e.g. you have a member
named toto
, and you also named your function argument toto
. Many
coding conventions mark either the member or argments to avoid this
case, e.g. all member names start with my
or m_
, or a parameter name
will start with the
.
The other case is that this->
can be used in a template to make a name
dependent. This is relevant if a template class inherits from a
dependent type, and you want to access a member of the base, e.g.:
template <typename T>
class Toto : public T
{
public:
int f()
{
return this->g();
}
};
Without the this->
here, g()
would be a non-dependent name, and the
compiler would look it up in the context of the template definition,
without taking the base class into consideration.
This is a matter of style. Some people like the extra this->
to make it more obvious that you are accessing a class member. But if you feel it's obvious enough without it, there will be no difference in the generated code or performance.
(Besides the case you mentioned with overlapping scopes, this->
can also be mandatory in a template when trying to name a member of a type-dependent base class.)
it's simply redundant to use this->
to call members, unless you want to semantically distinguish between locals and members quickly.
a lot of people use the m_
prefix for class members, to avoid writing this->
all the time.
I always use this
when calling member functions.
- It turns the function name into a dependent name so that base class member functions are found within a class template.
- It suppresses argument-dependent lookup. ADL has its advantages, but it can lead to surprising behavior, and I like it if it's not getting in my way.
- It has no real disadvantages, and so I use it for all member function calls for consistency reasons.
- I program in Python a lot where an explicit
self
is mandatory, so it's not a real burden for me.
But for data members I use it only when necessary because there is no ADL taking place. To answer your specific questions:
Within a member function of a class in C++, does it make a difference, if I use this->dataMember or just dataMember?
Yes, if this is within a class template. Then dataMember
is considered a non-dependent name, which can lead to semantic differences. For example:
#include <iostream>
int i = 1;
struct R {
int i;
R(): i(2) { }
};
template<typename T>
struct S: T {
void f() {
std::cout << i << ' ' // selects ::i
<< this->i // selects R::i
<< std::endl;
}
};
int main() {
S<R>().f();
}
What is considered better style?
I don't think that there is a strong opinion within the community about this. Use either style, but be consistent.
Is there any performance difference?
I'm pretty sure there isn't.
use this when you have a hidden/private member =) in any other case it does not make a difference =)
from the IBM information center i quote the following
Unless a class member name is hidden, using the class member name is equivalent to using
the class member name with the this pointer and the class member access operator (->).
using "this->" is better (you are sure it's the members) but it's doesn't make a difference
If a template function makes a call to a member function such that the call does not depend on any template parameters, this->
can be used to help the compiler as an alternative to MyUtopicClass<int, double, double>::vin()
.