repost: template function specialization & basic question...

[NKOBAYE027]

FIRST POST

Hi All: I'm trying to write a simple specialization before moving on to
something a bit more complex - always a good idea in my case, at least. )

I'm trying to adapt the example from Stroustrup, 3rd ed., The C++
Programming Language p. 344

I'm using MSDev 6.0 in case that's an issue. Here's the source...

/* begin snippet */
/* specializations.hpp */

#include <functional>
using namespace std;

class foo
{
public:
int m_foo;
foo():m_foo(0){}
};

template<> bool less<foo>(const foo& afoo, const foo& bfoo)
{
return afoo.m_foo < bfoo.m_foo;
}

/* specializationstest.cpp */
#include "specializations.hpp"

int main()
{
foo f,g;
g.m_foo = 8;

bool done = less<foo>(f,g);

return done;
}

Compiling...
specializationstest.cpp

error C2935: 'less<class foo>' : template-class-id redefined as a global
function

error C2912: explicit specialization; 'bool __cdecl less<foo>(const class
foo &,const class foo &)' is not a function template
see declaration of 'less<class foo>'

error C2661: 'less<class foo>::less<class foo>' : no overloaded function
takes 2 parameters
Error executing cl.exe.

specializations.exe - 3 error(s), 0 warning(s)


/* end snippet */
Any help would be appreciated.

I'm trying to do this with the intent of specializing an iterator operator*
and & in the stuff I'm working on right now.

i.e.
something like

modified by OP:
had been glibly using foo as the template args for the lists
in the example code below - was 3 am, gimme a break )
this makes more sense

template<typename T>
class foo<T>
{
public:
list<T> foolist;
list<T*>p_foolist;
list<int> somestupidlist;
typedef list<T*>::iterator iterator;
}

template<> T& foo::iterator:perator*(){/*...*/}

where dereferencing the iterator will return an object reference, not a
pointer to the object which is what the list contains.
so, any suggestions here would be welcome as well...

And thanks to all those who replied to my prior questions...

regards,
L.

SECOND POST

Hi folks: Can anyone tell me what the difference is between the two
functions below? In particular why does the function that returns a
reference give me warnings about returning the address of a temporary
variable, whereas the prior function compiles without any errors or warnings

template<typename foo>
class fooset<foo>
{
...
list<foo> fooList;
typedef list<foo>::iterator iterator;
...

public:

iterator begin()
{
return fooList.begin();
}

/*
iterator& begin()
{
return fooList.begin();
}
*/
iterator
};

thanks,
and regards,
L.

 

[bartek]

Hi All: I'm trying to write a simple specialization before moving on
to something a bit more complex - always a good idea in my case, at
least. )

I'm trying to adapt the example from Stroustrup, 3rd ed., The C++
Programming Language p. 344

I'm using MSDev 6.0 in case that's an issue. Here's the source...

(...)

template<> bool less<foo>(const foo& afoo, const foo& bfoo)
{
return afoo.m_foo < bfoo.m_foo;
}

1. std::less is not a function, but a functor. It provides an overloaded
operator(), so it can be 'invoked' like a function.

2. Function templates cannot be specialized.

3. Why don't you write an operator < overload for your foo class?

Cheers,
bartek

 

[Adam H. Peterson]

1. std::less is not a function, but a functor. It provides an overloaded

operator(), so it can be 'invoked' like a function.

2. Function templates cannot be specialized.

Function templates can be specialized. They cannot be
_partially_specialized_. (OTOH, MSVC6 doesn't support partial template
specialization anyway.)

 

[bartek]

Function templates can be specialized. They cannot be
_partially_specialized_. (OTOH, MSVC6 doesn't support partial
template specialization anyway.)

Oops, of course right. Sorry for the confusion.

Cheers,
bartek

 

[NKOBAYE027]

Thanks bartek.

The reason I don't overload the < is because I'm not aiming to do that. This
was a first attempt at specializing a function - and I failed miserably. )
I see now that I was missing the original definition of the template
function 'less' provided in the text (on the same page, no less...npi). So,
I guess, what I should do is simply ask if specializing a list<class
T*>::iterator is possible such that dereferencing the iterator will return
the object pointed to by T* instead of T*. If so how?

Sorry for the confusion - it's all mine!

I've since tried the specialized version and it works when I appropriately
define it and the original, but would still greatly appreciate any input on
the iterator issue mentioned above.

regards,
L.

 

[Adam H. Peterson]

Thanks bartek.

The reason I don't overload the < is because I'm not aiming to do that. This
was a first attempt at specializing a function - and I failed miserably. )
I see now that I was missing the original definition of the template
function 'less' provided in the text (on the same page, no less...npi). So,
I guess, what I should do is simply ask if specializing a list<class
T*>::iterator is possible such that dereferencing the iterator will return
the object pointed to by T* instead of T*. If so how?

I don't think you can do this with template specialization. In my
experience (and I think in all cases) when you specialize a function or
class, you can't alter the signatures of the functions -- what
parameters they take, their constness, their return type.

It looks like you're trying to create something akin to a polymorphic
container. You can probably search on that -- there's been lots of
discussion, research, etc. on that topic.

Personally, I would start to solve such a problem by using composition
with a standard list<>. Something like:


#include <list>

template<typename T>
class ptr_list {
public:
class iterator;
// Yada yada
private:
std::list<T*> data;
};

template<typename T>
class ptr_list<T>::iterator {
public:
T &operator*() const {return **it;}
// Yada yada
private:
std::list<T*>::iterator it;
};


It would take some work to get the full iterator interface working, but
you may not need all of that depending on your usage.

Adam H. Peterson

 

[NKOBAYE027]

Thanks, Adam - that was exactly what I was looking for. Your pointer about
specializations not being able to change function signatures is what I
needed. So, I do need to reinvent the wheel, so-to-speak, for the iterator
class. I can see no other way than the one you mentioned - and I'd already
considered that but was hoping for an easier way out...Ugh...c'est la vie,
non?

regards,
L.