Re: [boost] Re: Hello MPL world!

From: David Abrahams (dave_at_[hidden])
Date: 2003-01-06 09:20:11

Alisdair Meredith <alisdair.meredith_at_[hidden]> writes:

> Terje Slettebø wrote:
>
>> The results don't necessarily have to be printed out at
>> compile-time (Erwin Unruh printed the results using compiler-warnings, but
>> that is of course highly implementation dependent).
>
>> "Hello, world" in compile-time programming doesn't necessarily have to be
>> the same kind of program as in run-time programming, since the way it works
>> is different.
>
> I suspect a metaprogram that does nothing but issue a few diagnostics
> would be dismissed as a curiosity, maybe a cool toy, but hardy seen as
> an interesting tool by someone new to the idea.
>
> Creating a static array of size 5! shows a little more potential, but
> what to do with the array?

And that's still primarily numerical computation.

> However, it is far too easy to be negative when I have nothing better to
> offer in return (although I will be quite keen to see the final example,
> MPL is one of those items on my 'to-do' list that never quite hits the
> top)

My current working idea is really simple:

  struct impl1 { ... };
  struct impl2 { ... };

  template <class T>
  struct my_container : if_<is_pointer<T>, impl1, impl2>::type
  {
     ...
  };

This is something I want to do all the time; it lets you swap out a
completely different implementation for the class (well, except for
the constructors) based on type properties. It also shows the
convenient interaction of type_traits metafunctions with MPL
algorithms.

Can we do better?

> What are people actively doing with MPL anyway?  Perhaps and answer
> to that question will indicate what 'Hello world' should look like.
> It should be more than a neat exercise in syntax, it should
> demonstrate the idea of real-world use.
>
> Currently I mainly use compile-time techniques for static assertions.  

It's a good thing to use MPL for, but I suspect the idea of causing
errors isn't going to entice most people to dig deeper ;-)

> I have written some toy curiousities such as the classic factorial
> generator and recursive array-like containers.  I am yet to make the
> leap from playing with the syntax to applying it to real world
> problems, and if we can come up with a simple example that makes
> that leap, we'll have hit gold [where 'we' here clearly means 'you'
> <g>]

That's what we're interested in.  My favorite examples are much too
involved to be used for "hello, world".

-- 
                       David Abrahams
   dave_at_[hidden] * http://www.boost-consulting.com
Boost support, enhancements, training, and commercial distribution