"David Abrahams" wrote

"Andy Little" writes:

...

Hi,

I am trying to transform from a fusion vector of values to a fusion vector of non-const references to those values, the ultimate aim being to assign the original sequence from another sequence.

The code below works ok with make_vector, but I want to be able to automate it for any fusion sequence. How can I do that?

You can find an example of exactly that on the zip_iterator_fusion branch of boost/iterator/zip_iterator.hpp

Yikes. I don't really understand what I am meant to do with that. I will have to study it in more detail. I reckon that a get_ref_vector(fusion::vector) would be quite useful. I guess there isnt much difference to a view though?. Anyway I have continued hacking away and so here is my latest efforts, using fusion to do a dot product of two vectors. Maybe it would be a useful example in the fusion docs?. Anyway here it is FWIW: regards Andy Little -------------------------- // Copyright Andrew Little 2006 // // Distributed under the Boost Software License, Version 1.0. // (See accompanying file LICENSE_1_0.txt or copy at // http://www.boost.org/LICENSE_1_0.txt) // // See QUAN_ROOT/quan_matters/index.html for documentation. /* dot product of two vectors using fusion. The input vectors are just raw fusion sequences */ #include #include #include #include #include #include #include #include #include #include #include <cassert> // Couldnt resist this... :-) namespace quan{ namespace fusion{ // This functor needed as fusion::fold sends // its args the opposite way to quan::operator_XX_equals // Lhs in invoked function is assumed to be // a non const boost::reference_wrapper template <typename F> struct reverse_assign_f_args{ template struct result{ BOOST_MPL_ASSERT((boost::is_reference_wrapper<Lhs>)); BOOST_MPL_ASSERT((boost::mpl::not_< boost::is_const<typename Lhs::type> >)); typedef Lhs type; }; template typename result::type operator()( Rhs const & rhs, Lhs lhs)const { F f; return boost::ref(f(lhs.get(),rhs)); } }; // more arg reversing template <typename F> struct reverse_f_args{ template struct result{ typedef typename F:: template result< Rhs,Lhs >::type type; }; template typename result::type operator()( Rhs const & rhs, Lhs const & lhs )const { F f; f(lhs,rhs); } }; }}//quan::fusion int main() { // The types of the input 'vectors' typedef boost::fusion::vector4< quan::length::mm,int,int,int > seq1_type; typedef boost::fusion::vector4< quan::reciprocal_length::per_m,int,int,int > seq2_type; // Use fusion to work out the type of the result of // the dot product... // First the result sequence of // multiplying the input sequences. typedef boost::fusion::result_of::as_vector< boost::fusion::result_of::transform< seq1_type,seq2_type,quan::operator_times >::type >::type multiplied_seq; // Now the add and accumulate result type deduction part. // Note that need to switch the args for use with my own functors, // (I have been here with boost::is_convertible, so I'll say no more;-)). // I am using + as the 2nd function applied, not +=, as used at runtime // as I am also trying to 'accumulate' the result_type at this stage. typedef boost::fusion::result_of::fold< multiplied_seq, boost::fusion::result_of::at_c::type , quan::fusion::reverse_f_argsquan::operator_plus >::type result_type; // There was a bit of fun and games to get here so ... BOOST_STATIC_ASSERT(( boost::is_same::value !=0)); // set up the input 'vectors' seq1_type seq1(quan::length::mm(1000),2,3,4); seq2_type seq2(quan::reciprocal_length::per_m(1),2,3,4); // do the dot_product result_type result = boost::fusion::fold( boost::fusion::transform(seq1,seq2,quan::operator_times()), boost::ref(result), quan::fusion::reverse_assign_f_argsquan::operator_plus_equals() ); assert(quan::compare(result,30,1e-12) ==0); std::cout << result <<'\n'; } -----------------------------

Andy Little wrote:

"Andy Little" wrote

Also I found I had to comment out the specialisation of as_vector<0> in < boost/fusion/sequence/conversion/detail/as_vector.hpp> in the Boost Review version to get it to compile OK in GCC4.0.1

I havent tried the CVS version of the file but it looks unchanged relative to my version.

Hmmm... Could you provide a minimal code that exhibits the problem? Cheers, -- Joel de Guzman http://www.boost-consulting.com http://spirit.sf.net

Andy Little wrote:

"Joel de Guzman" wrote in message news:edu1li$1sf$2@sea.gmane.org...

...

Andy Little wrote:

...

"Andy Little" wrote Also I found I had to comment out the specialisation of as_vector<0> in < boost/fusion/sequence/conversion/detail/as_vector.hpp> in the Boost Review version to get it to compile OK in GCC4.0.1

I havent tried the CVS version of the file but it looks unchanged relative to my version. Hmmm... Could you provide a minimal code that exhibits the problem?

Sure. Just include the as_vector header. Note this is the review version, not CVS

#include

Sorry I can't reproduce it. I don't have g++ 4.0.1 at the moment. I tried it with VC7.1, VC8.0, G++ 3.4 and Comeau 4.3.3. All the compilers compile it just fine. Cheers, -- Joel de Guzman http://www.boost-consulting.com http://spirit.sf.net

"Joel de Guzman" wrote

Andy Little wrote:

...

Maybe it would be a useful example in the fusion docs?. Anyway here it is FWIW:

Yes, definitely! I'll have to ask for documentation though. A simple readme file is ok, plus, lots of comments.

OK. Bear in mind that its reliant 3 functors, multiply, add, and add-and-assign. In order to be agnostic regarding type deduction schemes ( my own, phoenix, lambda etc), my thought is to present particular versions of the functors implemented using Boost.Typeof. For the assignment one it needs to be implemented in terms of boost::reference_wrapper anyway and with the Lvalue as the right hand argument, which is AFAICS particular to the use of fold in fusion. for example : struct example_multiply{ template struct result{ typedef BOOST_TYPEOF_TPL(A() * B()) type; }; template typename result::type operator(A const & a, B const & b) { return a * b; } }; Boost.Typeof will after all be in the next release alongside Boost.Fusion Is that acceptable? regards Andy Little

"Joel de Guzman" wrote in message news:edv1e0$n73$1@sea.gmane.org...

Sounds good. Okidoki.

Great. I will see if I can get on with it, rather than spend more time mucking about with fusion ;-) regards Andy Little

Andy Little wrote:

"Joel de Guzman" wrote in message news:edv1e0$n73$1@sea.gmane.org...

...

Sounds good. Okidoki.

OK. Its done. Its called fusion_example.zip and in The Boost vault in generic programming directory:

http://tinyurl.com/hw7js

There may be some typo errors and I could write in more detail, but there should be eneough there to start with anyawy.

Cool! I'll have a look at it as soon as I get some time; possibly over the weekend. Cheers! -- Joel de Guzman http://www.boost-consulting.com http://spirit.sf.net

"Andy Little" writes:

"David Abrahams" wrote

...

"Andy Little" writes:

...

Hi,

I am trying to transform from a fusion vector of values to a fusion vector of non-const references to those values, the ultimate aim being to assign the original sequence from another sequence.

The code below works ok with make_vector, but I want to be able to automate it for any fusion sequence. How can I do that?

You can find an example of exactly that on the zip_iterator_fusion branch of boost/iterator/zip_iterator.hpp

Yikes. I don't really understand what I am meant to do with that. I will have to study it in more detail.

mpl::transform > ref_tuple(( value_tuple )); Does that help? -- Dave Abrahams Boost Consulting www.boost-consulting.com