Re: Can you use a function reference passed in a template argument?

[Jonathan Wakely <jwakely.gcc@gmail.com>]

On Thu, 9 Nov 2023 at 21:24, Arthur Schwarz <home@slipbits.com> wrote:
>
>
> On 11/9/2023 1:11 PM, Jonathan Wakely via Gcc-help wrote:
> > On Thu, 9 Nov 2023 at 20:46, Arthur Schwarz <home@slipbits.com> wrote:
> >>
> >> Is there any way to use a function passed as an argument to a template
> >> (example below)? Couldn't the existence of the referenced function be
> >> established durint instantiation (Stack<some class> obj)? I realize that
> >> just doing analysis of the template that the existence of a referenced
> >> function can't be determined, but during instantiation it can be validated.
> > I have no idea what that code is trying to do. You're trying to call a
> > member function on a _type_ T.
> As if I knew what I was trying to do.
> >
> > And you said you want to call a function passed as an argument ... but
> > the template argument you pass is a type, not a function.
> >
> At the time of object instantiation the 'type' is a class and as a
> member of this
> class there is a function. It is resolvable that when a class is used
> that to satisfy
> the requirement that the template object is correct, the class must
> contain the
> indicated function. If a passed class does not contain the indicated
> function,
> then an error can be generated.

No, you have a category error. You call a non-static member function
on an object, not on a type.

You could write T().toString() which would create a temporary object
of type T, and call the function on _that_. But you can't call a
non-static member function on a type. You need an object.

>
> It looks like the determination of template instantiability is made when
> the
> template is 'compiled', and at this time it is not possible to determine
> that
> the referenced function, T.fun(), is available.

T.fun() is not even valid C++ syntax, it's just nonsense. So the
problem has nothing to do with when the template is instantiated or
compiled. You're just writing something that isn't C++.

> This validation only
> becomes
> possible during template instantiation, that is, when
> template_name<class_name>
> is instantiated. To me the question then becomes does the standard require
> that checking is not deferred until instantiation.
>
> >> As a nit, the repeated instances of "../header/" in the error message is
> >> an annoyance.
> > That looks like a problem with symlinks or your build system, not gcc's fault.
> >
> >> Diagnostic message and code given below.
> >>
> >> thanks
> >> art
> >>
> >> ../header/../header/../header/../header/Stack.h: In member function
> >> ‘std::string Stack<T>::toString()’:
> >> ../header/../header/../header/../header/Stack.h:184:46: error: expected
> >> primary-expression before ‘.’ token
> >>     184 |       str << setw(3) << s.size() << ": " << T.toString();
> >>         |                                              ^
> >>
> >> # include <iomanip>
> >> # include <sstream>
> >>
> >> using namespace std;
> >> template <class T>
> >> class Stack {
> >>      string toString() {
> >>         stringstream str;
> >>         str << setw(3) << s.size() << ": " << T.toString();
> >>         return str.str();
> >>      };
> >> };
> >>