I don't like the way default parameters of templates are handled when
the compiler is matching a template to a formal argument in a template
parameter list. It would be ideal if the following were true.

template<
 class T
 class Policy_1 = P_1
 //...
 class Policy_n = P_n
 >
 class I_take_N_parameters_N_minus_1_of_which_have_default_types{...}

template<
 class T
 template <class> class TTP_taking_one_type
 >
 class I_will_compile{

  TTP_taking_one_type<T> var;

 //....
 };

 I_will_compile<int,
  I_take_N_parameters_N_minus_1_of_which_have_default_types> OK;
///yes!!


I think the current restriction on template template parameters (TTP)
is too draconian and removes much of the contruct's potential power. I
somewhat understand the need to specify the number of arguments the
TTP expects to be instatiated with; the philosophy being that it is
better to catch errors at the template definition rather than at
instatiation. However, if a template can legally be instatiated with
one type (like std::vector) and a TTP expects to be instatiated with
one type why disallow that?

Here is a more realistic example:

template<
 class T,
 template<class U, class A = std::allocator<U> > class C = std::deque,
 template<class> class Ptr = boost::shared_ptr
 //...
 >
 class Manager{
  C<Ptr<T> > SharedObjects;

  //...
 };

Instiantiating it like:

Manager<Sprite, std::list, Loki::SmartPtr> SpriteManager;

is asking for trouble from the compiler. I think that is unfortunate.
I'd have to use a suboptimal and somewhat unsafe workaround such as:

template<class T> class MyLokiSmartPtr : public
Loki::SmartPtr<T>{/*...*/}

//^^Yes, inheritance. There's no way I'm writing all those forwarding
functions

Manager<Sprite, std::list, MyLokiSmartPtr> SpriteManager;

BTW, I found it interesting that Bjarne Stroustrup, in appendix C:13.3
of TC++PL, uses TTPs in exactly the same way the current
interpretation disallows.

I think it would be in the best interest of C++ generic programming to
change the way default parameters of templates and TTPs interact.
IMHO, the current definition is quite unnatural. Am I missing some
terrible ramification of allowing the above?

---
[ comp.std.c++ is moderated.  To submit articles, try just posting with ]
[ your news-reader.  If that fails, use mailto:std-c++@ncar.ucar.edu    ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://www.jamesd.demon.co.uk/csc/faq.html                       ]