typedef & template

[Dmytro]

Hi!

This code is ok:

template<class T>
struct Container
{
T x;
typedef int size_type;
size_type size() { return sizeof T; }
};

Why the compiler (VC7) does not compile the following code?

template<class T>
struct Container
{
T x;
typedef int size_type;
size_type size() const;
};

template<class T>
Container<T>::size_type Container<T>::size() { return sizeof T; }

 

[Ian Collins]

Hi!

This code is ok:

template<class T>
struct Container
{
T x;
typedef int size_type;
size_type size() { return sizeof T; }
};

Why the compiler (VC7) does not compile the following code?

template<class T>
struct Container
{
T x;
typedef int size_type;

Why use int for a size_type? Does a negative size make sense?

size_type size() const;
};

template<class T>
Container<T>::size_type Container<T>::size() { return sizeof T; }

You left out two important things:

typename Container<T>::size_type Container<T>::size() const {
return sizeof T; }

The typename is required to tell the compiler that
Container<T>::size_type is a type and the const because the prototype
function is const.

 

[Jakob Bieling]

Why the compiler (VC7) does not compile the following code?

template<class T>
struct Container
{
T x;
typedef int size_type;
size_type size() const;
};

template<class T>
Container<T>::size_type Container<T>::size() { return sizeof T; }

Where (which line in the above snippet) and which error do you get?
My guess is you're missing a 'typename' in the last line.

hth

 

[Rolf Magnus]

Why use int for a size_type? Does a negative size make sense?

Actually, it does. If you want to be able to get an offset between two
arbitrary elements, that offset must be a signed type and will only be able
to span half of the range of the unsigned size type. So the size couldn't
be more anyway. And then, it's always a bad idea to mix signed and
unsigned, so actually, I would prefer a signed size type.

 

[Frederick Gotham]

Rolf Magnus posted:

And then, it's always a bad idea to mix
signed and unsigned, so actually, I would prefer a signed size type.

Some people simply used "signed int" all the time. The situation is very
similar to how some people will always use "i++" and have no intention of
changing their style.

I myself only use signed integer types when I have to. Elsewhere, I use
unsigned integer types.