[BUG] No direct UFCS syntax for (captured function object, dependent base member, function object prvalue) #748

Activity

1. JohelEGP

   added

   bugSomething isn't workingSomething isn't working

   on Oct 12, 2023
2. JohelEGP commented on Nov 11, 2023

   

   JohelEGP

   on Nov 11, 2023

   ContributorAuthor

   More actions

   From #741 (comment):

   . :(x) zip_transform(std::logical_or(), x, scan_left(x, true, std::not_equal_to())) ()

   That's not valid grammar (https://cpp2.godbolt.org/z/9avjYh6sb):

   Right, I like the call helper well enough that I'm waiting to see if there's really a need to write a new function expression in the middle of a postfix-expression... it's doable but is it needed?

   You could say it's needed for #748,
   so we could write w: (f) :(x) x.(f$)();
   and have x.(f$)() call (f$)(x).
3. JohelEGP commented on Nov 25, 2023

   

   JohelEGP

   on Nov 25, 2023

   ContributorAuthor

   More actions

   Also, can't do UFCS on a dependent base member (#741 (comment)):

   . :(x) zip_transform(std::logical_or(), x, scan_left(x, true, std::not_equal_to())) ()

   That's not valid grammar (https://cpp2.godbolt.org/z/9avjYh6sb):

   Right, I like the call helper well enough that I'm waiting to see if there's really a need to write a new function expression in the middle of a postfix-expression... it's doable but is it needed?

   You could say it's needed for #748,
   so we could write w: (f) :(x) x.(f$)();
   and have x.(f$)() call (f$)(x).

   There is another use case.
   In a member function of a type template,
   UFCS on a dependent base member requires explicit qualification (https://cpp2.godbolt.org/z/Mff1G9331):

   u: @struct <T> type = {
     f: (_) true;
   }
   
   t: @struct <T> type = {
     this: u<T>;
   
     operator(): (this, r: std::span<const int>) -> _ = {
       for r do (e) {
         // OK.
         return this.f(e);
   
         // error: explicit qualification required to use member 'f' from dependent base class
         //   16 |       return CPP2_UFCS_0(f, e);
         //      |                          ^
         return e.f();
   
         // error: expected unqualified-id
         //   21 |       return CPP2_UFCS_0(f, e.(*this));
         //      |                               ^
         return e.this.f();
   
         // OK.
         return e.u<T>::f();
   
         // What I would like to write:
         // return e.(this.f)();
         //  That translates to this:
         return (this.f)(e);
       }
     }
   }
   
   main: () = {
     r: std::vector = (0, 1, 2);
     assert(t<int>()(r));
   }
   
4. JohelEGP

   changed the title [-][BUG] Can't capture function object for UFCS[/-] [+][BUG] Can't use UFCS on captured function object or dependent base member[/+] on Nov 25, 2023
5. JohelEGP commented on Nov 25, 2023

   

   JohelEGP

   on Nov 25, 2023

   ContributorAuthor

   More actions

   Also, can't do UFCS on a dependent base member (#741 (comment)):

   For now, I can workaround this with using u<T>::f;, which makes return e.f(); well-formed: https://cpp2.godbolt.org/z/T5b5sxbYr.
6. JohelEGP commented on Nov 26, 2023

   

   JohelEGP

   on Nov 26, 2023

   ContributorAuthor

   More actions

   Function object prvalue

   I forgot to mention this other case where you first need to construct a callable.

   Consider the call in the assert in the main above: t<int>()(r).
   t<int>() is the function object we want to call.

   Let's rewrite it in UFCS order: r.t<int>()().
   The function to call becomes t<int> instead.
   The second () applies to the result of the UFCS.

   The call helper does help:
   r.call(t<int>()) (https://cpp2.godbolt.org/z/Mfc5TP8xz).
7. JohelEGP

   changed the title [-][BUG] Can't use UFCS on captured function object or dependent base member[/-] [+][BUG] No direct UFCS syntax for (captured function object, dependent base member, function object prvalue)[/+] on Nov 26, 2023