On 2/19/2011 10:46 PM, Lutger Blijdestijn wrote:
Nick wrote:

I know I can parametrize template mixin types, but how about mixin
identifiers? This is the code:

----------------------------------------------------------------

mixin template Box(T) {
    T val;
}

class Set(A, B) {
    mixin Box!A a;
    mixin Box!B b;
}

alias Set!(int, int) IntSet;

int main(string[] argv) {
    scope auto s = new IntSet();
    s.a.val = 3;
}

----------------------------------------------------------------

As you can see, the A and B types can be changed and with them the boxed
values. But they need disambiguation, since they have identically named
members. I know only to hard-code that disambiguation.

I need this because I intend to change (A, B) with a type list
eventually and I need to make operations on the boxed types independent
on the template mixin  identifiers (a, b).

This pattern comes from C++ code, where Box(T) would be classes, Set (A,
B, ...) would recursively apply multiple inheritance on the type list.
Then each Box members could be addressed by casting to the correct Box(T).

Any idea if this is possible in D?

Thanks!

Possible yes, but probably not so pretty. There must be some places in
phobos where something like this is done (better), perhaps in the code for
tuples in std.typecons. Here is my unpolished attempt:

mixin template Box(T)
{
     T val;
}

mixin template MixinFields(P...)
{
     /* P[0] is the template to use as mixin
      * P[1] is the type to be use as parameter for P[0]
      * P[2] is the symbol name the mixin will be aliased to
      */
     mixin("mixin " ~ __traits(identifier, P[0]) ~ "!(" ~ P[1].stringof ~ ")
           " ~ P[2] ~ ";");
     static if(P.length>  3)
         mixin MixinFields!(P[0], P[3..$]);
}


class Set(T...)
{
     mixin MixinFields!(Box, T);
}

alias Set!(int, "a", int, "b") IntSet;

void main()
{
     scope auto s = new IntSet();
     s.a.val = 3;
}

Thanks!

I was staying away from string mixins as they seem (although powerful) rather difficult to grok. But if I must, I must...

Reply via email to