David Abrahams wrote:

"Neil Hunt" writes:

...

I guess the main thrust of the question is "Does the parameters library support runtime enumeration?"

Not directly, but it seems like something we (sh/c)ould support, probably by making ArgumentPacks conforming Fusion sequences so you could use boost/spirit/fusion/algorithm/for_each.hpp to enumerate them. The problem is that I don't have expertise with Fusion-1 and AFAIK it's not fully documented. Specifically, I don't know how to make a conforming Fusion sequence. Maybe Joel can help. Joel?

Sure. I'll look into it. Should be a good example for the docs in the making. I'm a bit hesitant to write this for Fusion-1, however, as it will be replaced by Fusion-2 --which, BTW, has a "map" that might be a good replacement for boost parameter's internals. If boost.parameter uses fusion::map, then you automatically have iteration support plus all the algorithms etc. I bet fusion::map will simplify boost.parameter significantly too. Regards, -- Joel de Guzman http://www.boost-consulting.com http://spirit.sf.net

Thanks David (and Joel), Appreciate the time. That's helped me along the path. As an aside, I think the docs in http://www.boost.org/libs/parameter/doc/html/#applying-sfinae-to-the-overloa... reference a superceded example. I think "when the name argument is not convertible to const char*." should be "when the root_vertex argument is not convertible to int." as per the section above it. Thanks again, Neil "David Abrahams" wrote in message news:u8xutrqem.fsf@boost-consulting.com...

"Neil Hunt" writes:

...

I guess the main thrust of the question is "Does the parameters library support runtime enumeration?"

Not directly, but it seems like something we (sh/c)ould support, probably by making ArgumentPacks conforming Fusion sequences so you could use boost/spirit/fusion/algorithm/for_each.hpp to enumerate them. The problem is that I don't have expertise with Fusion-1 and AFAIK it's not fully documented. Specifically, I don't know how to make a conforming Fusion sequence. Maybe Joel can help. Joel?

...

or should I be looking at MPL?

Wouldn't help you, at least not directly. Well, here's one thing you could do (ALL_CAPS should be reserved for macros, so I'm going with lower_case):

// untested BOOST_PARAMETER_KEYWORD(tag, person_window) BOOST_PARAMETER_KEYWORD(tag, address_window) BOOST_PARAMETER_KEYWORD(tag, employer_window)

typedef mpl::vector< tag::person_window, tag::address_window, tag::employer_window

...

all_windows;

template struct process_arg { process_arg(ArgumentPack args, F f) : args(args), f(f) {}

template <class Keyword> void operator()(Keyword k) const { f(k, args[k]); }

private: ArgumentPack const& args; F const& f; };

template for_each_arg(ArgumentPack const& args, F const& f) { mpl::for_each<Keywords>( process_arg(args,f) ); }

Now you can for_each_arg(args, f) where f is some binary function you'd like to be applied to each argument. The first parameter is the keyword (just in case you need it) and the second one is the original argument passed.

This won't cope well with missing default arguments, so it seems to me making our ArgumentPacks conforming fusion sequences is really the way to go.

-- Dave Abrahams Boost Consulting www.boost-consulting.com

Hi, I'd like to ask how to create device and stream class using stream class from iostreams library so that read function in my device class can accept n bytes argument properly? I know that from my last discussion with Jonathan Turkanish i have to set stream's buffer first using stream constructor or open method say, in x bytes. Each times i call read function in my stream class to read n bytes, read function in my device class always accept argument in x bytes. I'd like to have flexible read method so that my device class can figure out how many bytes that my stream want read. Can anybody help me? Regards Hendra Hidayat __________________________________________________ Do You Yahoo!? Tired of spam? Yahoo! Mail has the best spam protection around http://mail.yahoo.com