On Wed, Oct 23, 2002 at 06:11:08PM +0200, Andre Poenitz wrote:
> You mean something like
>
> template <class Inset>
> struct inset_traits {
> bool shouldBeUsedInTocList = false;
> };
>
> template <>
> struct inset_traits<FooBarInset> {
> bool shouldBeUsedInTocList = true;
> };
> I don't see much that's worth the extra complexity, _especially_ if you
> want to access that with base class pointers/references as this basically
> ends up implementing something like vtables. The compiler is better at
> that.
So how /would/ this work with base class pointers ?
I find the current scheme not scalable and very ugly. Not to mention the
recompile of everything when you add a new one. I am trying to reduce
the horrendous size of the Inset/UpdatableInset classes.
They are too large to work with. I like the idea of centralising the
traits, it is far easier to understand.
> Moreover, not many people will understand that "pattern" whereas even the
> average Java programmer has a good grasp of virtual functions (even if
> "virtual" sounds unfamiliar). So from a purely practical POV: Why bother?
I don't buy this argument.
regards
john
--
"This is mindless pedantism up with which I will not put."
- Donald Knuth on Pascal's lack of default: case statement
