'new' template usage

[adrian.hawryluk@gmail.com]

Hi everyone,

I've been using templates for a while, but I'm not at full power yet
(knowledge=powe r) ;). Does anyone know where I can get information on
this 'new' template usage?

template<a (b, c, d)>

Does one define it like:

template<typena me A (typename B, typename C, typename D)>
{templated class/function definition here}

What is its use?

I also think that I've heard that ellipsis can also be used in the
template list. Is this true? If so, how is it used, and for what
purpose?

Thanks all,


Adrian

[Piyo]

Re: 'new' template usage

adrian.hawryluk @gmail.com wrote:

Hi everyone,
>
I've been using templates for a while, but I'm not at full power yet
(knowledge=powe r) ;). Does anyone know where I can get information on
this 'new' template usage?
>
template<a (b, c, d)>
>
Does one define it like:
>
template<typena me A (typename B, typename C, typename D)>
{templated class/function definition here}
>
What is its use?
>
I also think that I've heard that ellipsis can also be used in the
template list. Is this true? If so, how is it used, and for what
purpose?
>
Thanks all,
>
>
Adrian
>

Hi Adrian,

The new template usage that you are referring to is not part of
the template standards of C++ I think. You will probably need to
verify it for yourself in comp.std.c++. Having said that, that
template syntax is used by <boost/function.hpp>. The boost people
use boost::preproce ssor (a preprocessor meta-programming library)
that essentially uses very nifty preprocessor tricks to map

foo<int (float, double)to a proper looking template like
template<typena me T1, typename T2, typename T3>
class foo;

You can see it is action for yourself if you look inside
boost::function . I like the work they did on it and find it
pretty awesome for them to use macros to help create partial
template specializations on the fly. As to what is the use of
this type of programming, well, boost::function sums it up
pretty nicely. I hope that answers your first question.

For the ellipsis, again, I am not sure but once more I think that
the "..." syntax has not yet be standardized. You will need to
verify it for yourself. What is its use? Imagine this:

I would like to write a max function that takes in arbitrary
number of parameters like so

template<typena me T1, typename T2, ..., typename TN>
max( T1 &a, T2 &b, ..., TN &z );

this way, I can now call it like this

max( 1, 2, 3 ); or
max( 1.0, 2.0, 3.0, 4.5 );

Now having said that, looking again at boost function we notice that
it requires arbitrary number of arguments also. Yet using a combination
of nifty tricks, they are able to circumvent the problem that the
"..." syntax is unavailable. So it makes me wonder if the "..." syntax
is merely a convenience for the current ways to make it work.

You'll probably get more mileage if you post on the c++ standards
boards instead :)

But I hope I helped by giving you a better idea where templates can
go in the future. :)

[=?iso-8859-1?q?Erik_Wikstr=F6m?=]

Re: 'new' template usage

On 7 Mar, 01:11, adrian.hawry... @gmail.com wrote:

Hi everyone,
>
I've been using templates for a while, but I'm not at full power yet
(knowledge=powe r) ;). Does anyone know where I can get information on
this 'new' template usage?
>
template<a (b, c, d)>
>
Does one define it like:
>
template<typena me A (typename B, typename C, typename D)>
{templated class/function definition here}
>
What is its use?
>
I also think that I've heard that ellipsis can also be used in the
template list. Is this true? If so, how is it used, and for what
purpose?

I'm sorry to say (since I'd like to have them) that neither of those
are part of the current standard but may be part of the next (the
ellipsis thing I'm quite sure of, but the first I don't know). The
problem with this is that since they are not part of any standard yet,
which means that there probably are no compilers supporting it. And if
there are the functionality of those constructs might change until
they have been standardized.

As I understand things they hope to have the new standard out and
approved by 2009, which means that most things will have settles
sometime by 2008 and widespread compiler support should probably not
be expected until 2010 or later. Worst case scenario would be that
they can't agree about how something is supposed to work and throw it
out of the standard to be part of a separate TR or the next version.
So my advice is to not bother with these kinds of things for at least
a year in any serious project. You can of course play around to stay
up to date.

For more information about what will/might be in the next standard see

http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2142.html

--
Erik Wikström