News for hackers. Stuff that matters.
Re:Clay: A new language for generic programming (based on LLVM) by Chickencha on Monday July 26, 2010 @09:31AM (#1547954)
It looks pretty nice. I'm curious about how the code generation works and whether Clay has the potential to produce the kind of code bloat you sometimes see with C++ templates. Say I have a function

  min(a, b) {
      if(a < b)
          return a;
      return b;
  }
From what I understand, this function can accept two Int32s, or a Uint8 and a Float32, or basically any other combination of numeric types. Is a new function for each type combination generated? I don't think there's any way around this to some degree, but I'm curious if you've been able to do anything to ease the pain.
Re:Clay: A new language for generic programming (based on LLVM) by kssreeram on Monday July 26, 2010 @09:47AM (#1547987)
You can control type specialization to a certain extent with overloading. For instance, if a procedure takes two arguments, and if both these types can vary independently, then you can use overloading to ensure that both arguments are converted to a common type.

    procedure foo;
    
    [T1,T2]
    overload foo(a:T1, b:T2) {
        // by default, convert both arguments
        // to the type of first argument.
        foo(T1(a), T1(b));
    }
    
    // the following overload specializes for the case
    // when both arguments are of the same type.
    [T]
    overload foo(a:T, b:T) {
        ...
        implement the logic here
        ...
    }