Oh yes, having newly been using C++ at work, I realized they were a "big something" [1] that enabled you, as it were, to do whatever unstructured unholy type trickery you want, and yes, even making classes A<B> and A<C> completely different things. (BUT! We could argue over this fact: Isn't it also one of the purposes of... Type Families? Where the TF Foo :: * -> *, can yield to datatypes Foo String and Foo Int being completely different and unrelated?)
I was more saying that you could roughly "emulate" Haskell classes in C++ with templates (minus a good type security).

[1] Vernacular, isn't it?


2011/6/10 Richard O'Keefe <ok@cs.otago.ac.nz>

On 9/06/2011, at 8:02 PM, Yves Parès wrote:

> Were templates an original feature of C++ or did they appear in a revision of the langage ?

The latter.  "C with classes" did not have multiple inheritance, exceptions, or templates.

Note that C++ templates are *not* the same kind of animal as Eiffel generics or Java generics
or Ada generics or Haskell parametric polymorphism.  The C++ template language lets you do
type-level functional programming, and different instances of a common "type constructor" may
in fact have quite different internal structures.  C++ templates are NOT 'merely keywords
around .. parametric polymorphism', they are a far more dangerous thing.