C / C++ / MFC

TheGreatAndPowerfulOz12-Sep-11 12:23

12-Sep-11 12:23

The reason was it was ambiguous.

They could have provided a different dereferencing operator for member function pointers, something like ::* maybe, but alas they didn't.

If your actions inspire others to dream more, learn more, do more and become more, you are a leader." - John Quincy Adams
You must accept one of two basic premises: Either we are alone in the universe, or we are not alone in the universe. And either way, the implications are staggering” - Wernher von Braun

image processing toolkits | batch image processing

C^hris L^osinger12-Sep-11 15:20

C^hris L^osinger

12-Sep-11 15:20

ahmed zahmed wrote:
The reason was it was ambiguous.

no, it isn't. or, it wouldn't be. (it is now, since the C++ rules don't allow for this kind of thing, but that's not my point)

the compiler already knows how to do the implicit 'this' with member functions. it could do the same thing with member fn ptrs. assume unqualified member references are to 'this'. if a member function pointer for the class is used inside a non-static member function? assume 'this'.

TheGreatAndPowerfulOz13-Sep-11 12:54

13-Sep-11 12:54

yes, it could do that, but then calling a c-style function pointer would need a new syntax then.

image processing toolkits | batch image processing

C^hris L^osinger13-Sep-11 15:51

C^hris L^osinger

13-Sep-11 15:51

why? one syntax works for function calls now:

strcpy(foo, bar);
memberFnOfThisClass(foo, bar);
staticMemberFnOfThisClass(foo, bar);
std::sort(blah1, blah2);

Erudite_Eric13-Sep-11 22:50

13-Sep-11 22:50

When you use a pointer data member the syntax is *pVar = 10; or some such. No need to write
this->*pVar = 10;

So why not with member func pointers?

In fact func pointers are very useful, once you get to have used them a fair bit, they are indispensible, I just never realised how crappy the syntax for them is in C++. Smile | :)

==============================

Nothing to say.

jschell12-Sep-11 8:30

jschell

12-Sep-11 8:30

Fat__Eric wrote:
So I use a lot of function pointers in C

You don't however use member function pointers in C. Because they do not exist.

That said you shouldn't be using a "lot" of function pointers in C++ because there are probably better ways to do what you want in C++ without using them.

Perhaps what you want is a static member function pointer.

Or even just a function pointer - which would be exactly like it is in C.

TheGreatAndPowerfulOz12-Sep-11 12:24

12-Sep-11 12:24

jschell wrote:
they do not exist

True, unless you think of function pointers in a struct as member functions. Then the dereferencing syntax is very similar.

Erudite_Eric12-Sep-11 20:39

12-Sep-11 20:39

You know, if they are going to put the feature in the language, then saying "you shouldnt use that feature because there are better ways of doing it" is total bollocks. Smile | :)

The fact is the feature exists, yet the syntax is inconsistent with non pointer functions.

OK, take pointer member variables. Do you have to reference them as 'this->*pInt = 1;'

No, you just write '*pInt = 1;' in any function belonging to that class.

Yet you cant for a function pointer. You cant do (*pProc)(blah blah blah) you have to do (this->*pProc)(blah blah blah).

It is wrong! Smile | :)

==============================

Nothing to say.

jschell13-Sep-11 11:38

jschell

13-Sep-11 11:38

Fat__Eric wrote:
Yet you cant for a function pointer. You cant do (*pProc)(blah blah blah) you
have to do (this->*pProc)(blah blah blah).

I don't see that.

At best you are suggesting a shortcut that would apply to limited situations. It would require all of the following.
1. One needs a member function pointer to method X in class A.
2. Class A needs a variable of that type.
3. Class A needs to use tht variable to call itself.

I have very rarely seen a need for 1. Perhaps less that 10 times.
I have never seen a case for 2.

If I saw 3 I would suspect that there is a design flaw. I wouldn't be surprised if 2 represented a design flaw as well.

If there is any valid case for 3 then the there are so few situations that complicating the syntax just to provide a shortcut isn't worth it.

Erudite_Eric13-Sep-11 22:43

13-Sep-11 22:43

jschell wrote:
At best you are suggesting a shortcut that would apply to limited situations

As stated, when you use a pointer data member you use *pInt = 10; or some such. You do not need to use this->*pInt = 10;

So IMO it should be the same with pointer function members. Smile | :)

==============================

Nothing to say.

jschell14-Sep-11 13:16

jschell

14-Sep-11 13:16

Fat__Eric wrote:
As stated, when you use a pointer data member you use *pInt = 10; or some such. You do not need to use this->*pInt = 10;

Which actually would seem to clarify exactly why your proposition is not reasonable.

The scope of the your example is the referenced int. Whether it is in the current class or another class or even global the scope is still the referenced int. One single scope. No other scope is possible.

A member function pointer belongs to a single class instances. Thus there can be zero, one or many (hundreds) of scopes.

The scope must be specified when the member function pointer is called. There is no other possibility.

TheGreatAndPowerfulOz12-Sep-11 12:18

12-Sep-11 12:18

If you think of a class as a struct with function pointers, then the syntax makes sense. It would actually be similar syntax in straight C:

That is if I have:

typedef ULONG (*SomeFuncPtr)(PULONG, PVOID)

struct foo
{
SomeFuncPtr func;
};

Then code in some other function:

ULONG SomeFunc(PULONG pVal, foo* bar, PVOID pOpt)
{
...
(*bar->func)(&x, NULL);
...
}

the * is just moved is all...

Erudite_Eric12-Sep-11 20:34

12-Sep-11 20:34

Yes, it is easier with an instance of the object used externally to the object. My gripe was when used internally; why calling a func pointer specifically typedefed as belonging to the class should require an explicit this-> but calling a func doesnt.

Its almost racist. Smile | :)

(Well, pointerist perhaps)

==============================

Nothing to say.

TheGreatAndPowerfulOz13-Sep-11 12:55

Re: Member function pionters in C++ [modified]

13-Sep-11 12:55

see my replies to Chris Loslinger above.

Stefan_Lang13-Sep-11 23:01

Stefan_Lang

13-Sep-11 23:01

The problem is that operator*(MyFunc) is ambiguous, whereas operator->*(MyFunc) is not. You may have missed the point that this->*pProc is not equivalent to this->(*pProc) or (*this).(*pProc). neither of the latter versions would compile, because they are just as ambigous as (*pProc)!

A member function pointer isn't simply an address, it is a small struct that contains various bits of information that are needed to determine the actual function address, depending on whether the function is static, inline, virtual, or ordinary, whether the function type was defined in a base class or not, and maybe other factors as well. The compiler therefore wouldn't be able to tell if by invoking operator* you want to access these struct members, or rather call a function whose address is encoded by that struct!

At least that is my (incomplete) understanding of the matter. I may not be entirely correct, but you're free to google for more details. I've found 'Pointers to member functions' a very insightful article on the matter, although I couldn't pinpoint a concise answer to your question in there.

P.S.: I've just thought to search for this topic on CP and found this article. Just like the above link it explains a lot about member function pointers and their (lack of) use, but you might not like the article as much because it makes a point of how utterly useless member function pointers really are, except for the one thing this article is about: fast delegates. Wink | ;)

modified on Wednesday, September 14, 2011 5:25 AM

Erudite_Eric14-Sep-11 20:17