phobos commit, revision 2123


user: kyllingstad

msg:
Bug 5163: meaningless error message with front() applied to void[]

It's not possible to index into void[] arrays, which means that they
cannot be ranges.  Therefore, I've added template constraints to
front(), back(), popFront(), and popBack() in std.array that prevents
them from being instantiated with void[].

http://www.dsource.org/projects/phobos/changeset/2123

I noticed something funny when fixing this.  In the template constraint
for back(), is(typeof(A.init[0])), is used to check whether the type A
supports indexing.  This makes sense.  In the constraint for front(),
however, is(typeof(A[0])) is used.  Does anyone know why this last case
works at all?  Outside of typeof(), A[0] is completely meaningless
unless A is a tuple, so why allow this special case?

-Lars



On Fri, 2010-11-05 at 08:12 -0400, dsource.org wrote:
> phobos commit, revision 2123
> 
> 
> user: kyllingstad
> 
> msg:
> Bug 5163: meaningless error message with front() applied to void[]
> 
> It's not possible to index into void[] arrays, which means that they
> cannot be ranges.  Therefore, I've added template constraints to
> front(), back(), popFront(), and popBack() in std.array that prevents
> them from being instantiated with void[].
> 
> http://www.dsource.org/projects/phobos/changeset/2123
> 
> _______________________________________________
> phobos mailing list
> phobos at puremagic.com
> http://lists.puremagic.com/mailman/listinfo/phobos

A[0] is a static array of A's of size 0.  However, I'm not sure why it ever passes, because typeof(X) where X is already a type should fail, I think?

-Steve



----- Original Message ----
> From: Lars Tandle Kyllingstad <lars at kyllingen.net>
> 
> I noticed something funny when fixing this.  In the template  constraint
> for back(), is(typeof(A.init[0])), is used to check whether the  type A
> supports indexing.  This makes sense.  In the constraint for  front(),
> however, is(typeof(A[0])) is used.  Does anyone know why this  last case
> works at all?  Outside of typeof(), A[0] is completely  meaningless
> unless A is a tuple, so why allow this special  case?
> 
> -Lars

So it's not only weird, it's actually wrong, since A[0] does in fact have a sensible meaning...

Anyway, this shows that the compiler does not interpret it as a static array type inside the typeof():

        template Elem(T)
        {
                alias typeof(T[0]) Elem;
        }

        static assert (is(Elem!(int[]) == int));

-Lars



On Fri, 2010-11-05 at 06:01 -0700, Steve Schveighoffer wrote:
> A[0] is a static array of A's of size 0.  However, I'm not sure why it ever passes, because typeof(X) where X is already a type should fail, I think?
> 
> -Steve
> 
> 
> 
> ----- Original Message ----
> > From: Lars Tandle Kyllingstad <lars at kyllingen.net>
> > 
> > I noticed something funny when fixing this.  In the template  constraint
> > for back(), is(typeof(A.init[0])), is used to check whether the  type A
> > supports indexing.  This makes sense.  In the constraint for  front(),
> > however, is(typeof(A[0])) is used.  Does anyone know why this  last case
> > works at all?  Outside of typeof(), A[0] is completely  meaningless
> > unless A is a tuple, so why allow this special  case?
> > 
> > -Lars
> 
> 
> 
> 
> _______________________________________________
> phobos mailing list
> phobos at puremagic.com
> http://lists.puremagic.com/mailman/listinfo/phobos