Thread overview
[Issue 5601] A template forward reference error
Nov 22, 2013
Kenji Hara
Nov 22, 2013
Kenji Hara
November 22, 2013
https://d.puremagic.com/issues/show_bug.cgi?id=5601


Kenji Hara <k.hara.pg@gmail.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
            Version|D1 & D2                     |D2
         Resolution|                            |INVALID


--- Comment #3 from Kenji Hara <k.hara.pg@gmail.com> 2013-11-22 01:47:37 PST ---
(In reply to comment #0)
> A case of forward template forward reference (I am not sure this is actually a
> bug):
> 
> 
> import std.traits;
> auto foo() {
>     alias ReturnType!foo T;
>     return 0;
> }
> void main() {}
> 
> 
> DMD 2.052 shows (it's not a regression):
> 
> ...\dmd\src\phobos\std\traits.d(122): Error: template instance forward
> reference of foo
> ...\dmd\src\phobos\std\traits.d(105): Error: forward reference to foo
> ...\dmd\src\phobos\std\traits.d(122): Error: template instance
> std.traits.staticLength!(foo) error instantiating

This is correct behavior. Before function body semantic finished, getting type of auto function will cause forward reference error since its return type is not determined.

-- 
Configure issuemail: https://d.puremagic.com/issues/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
November 22, 2013
https://d.puremagic.com/issues/show_bug.cgi?id=5601



--- Comment #4 from bearophile_hugs@eml.cc 2013-11-22 02:52:35 PST ---
(In reply to comment #3)

> Before function body semantic finished, getting type
> of auto function will cause forward reference error since its return type is
> not determined.

Can't the type inferencer become a little smarter and see that the return type of foo is something defined in terms of itself, but it's also an int, and so decide it's an int?

-- 
Configure issuemail: https://d.puremagic.com/issues/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
November 22, 2013
https://d.puremagic.com/issues/show_bug.cgi?id=5601



--- Comment #5 from Kenji Hara <k.hara.pg@gmail.com> 2013-11-22 03:31:55 PST ---
(In reply to comment #4)
> (In reply to comment #3)
> 
> > Before function body semantic finished, getting type
> > of auto function will cause forward reference error since its return type is
> > not determined.
> 
> Can't the type inferencer become a little smarter and see that the return type of foo is something defined in terms of itself, but it's also an int, and so decide it's an int?

No. In statement scope, all of statements are ordered from top to the bottom. We can regard as the return type of auto function is declared at the end of the function body. So seeing return type inside auto-function is exactly same as forward reference issue.

Consider the following invalid code. If the return type is visible inside function, compiler cannot determine the return type of foo.

auto foo(int n)
{
    if (n == 1)
        return 1;

    // forward reference to return type
    static if (is(ReturnType!foo == int))
        return "";

    assert(0);
}

Therefore it is necessary limitation to avoid indeterminate case.

-- 
Configure issuemail: https://d.puremagic.com/issues/userprefs.cgi?tab=email
------- You are receiving this mail because: -------