• Trass3r (4/4) Jul 16 2011 I guess this is the cause for
• Timon Gehr (10/14) Jul 16 2011 mixin template foo(){
• Trass3r (1/16) Jul 16 2011
• Trass3r (3/3) Jul 16 2011 I know but what is so bad about issuing an error message and you rename ...
• Timon Gehr (11/14) Jul 17 2011 First of all, you would break some existing code for no benefit, includi...
• Trass3r (2/3) Jul 17 2011 It isn't limited to constructors. Overload resolution for any function
• Jacob Carlborg (8/12) Jul 17 2011 I've been knowing that template mixins behave like this for a long time....

Trass3r <un known.com> writes:

I guess this is the cause for  
http://d.puremagic.com/issues/show_bug.cgi?id=3332
It makes it practically impossible to mixin constructors because overload  
resolution doesn't work.

Timon Gehr <timon.gehr gmx.ch> writes:

Trass3r wrote:
 I guess this is the cause for
 http://d.puremagic.com/issues/show_bug.cgi?id=3332
 It makes it practically impossible to mixin constructors because overload
 resolution doesn't work.


mixin template foo(){
    immutable x = 1;
    int bar(){return x;}
}

immutable x = 2;

mixin foo!();

void main(){assert(bar == 1);}

Cheers,
-Timon

Trass3r <un known.com> writes:

Am 16.07.2011, 20:41 Uhr, schrieb Timon Gehr <timon.gehr gmx.ch>:

 Trass3r wrote:
 I guess this is the cause for
 http://d.puremagic.com/issues/show_bug.cgi?id=3332
 It makes it practically impossible to mixin constructors because  
 overload
 resolution doesn't work.


 mixin template foo(){
     immutable x = 1;
     int bar(){return x;}
 }

 immutable x = 2;

 mixin foo!();

 void main(){assert(bar == 1);}

 Cheers,
 -Timon

Trass3r <un known.com> writes:

I know but what is so bad about issuing an error message and you rename  
one of the two.
As I said it doesn't work out as soon as functions are involved.

Timon Gehr <timon.gehr gmx.ch> writes:

Trass3r wrote:
 I know but what is so bad about issuing an error message and you rename
 one of the two.
 As I said it doesn't work out as soon as functions are involved.

First of all, you would break some existing code for no benefit, including code
that mixes in two templates with clashing members.

Furthermore, it would be more difficult to implement, because templates already
come with their own scope.

Mixin templates are there to allow mixing in declarations without coping with
the
problems of string mixins. An own scope is pretty much required for that.

Therefore, I think mixin templates having their own scope is sensible.

The bug fix is to just search for constructors in mixin scopes too.

Cheers,
-Timon

Trass3r <un known.com> writes:

 The bug fix is to just search for constructors in mixin scopes too.

It isn't limited to constructors. Overload resolution for any function  
needs to be fixed.

Jacob Carlborg <doob me.com> writes:

On 2011-07-16 20:32, Trass3r wrote:
 I guess this is the cause for
 http://d.puremagic.com/issues/show_bug.cgi?id=3332
 It makes it practically impossible to mixin constructors because
 overload resolution doesn't work.

I've been knowing that template mixins behave like this for a long time. 
Don't know why I created that ticket, must have forgotten how mixins 
work. What is confusing is that the documentation says that a template 
mixin is just like copying the code and pasting it in the mixin place. 
Which it's not, shown by this ticket.

-- 
/Jacob Carlborg