Re: [Boost-users] [fusion - I think!] Multiple nested range traversal.

Subject: Re: [Boost-users] [fusion - I think!] Multiple nested range traversal.
From: Steven Watanabe (watanabesj_at_[hidden])
Date: 2009-01-12 13:02:24

AMDG

Robert Jones wrote:
> I recently asked a question on here about traversal of multiple 
> indices, which
> produced a luke-warm reaction, mainly because I wasn't really asking
> the right question!
>
> So here's another try. Consider a bit of code like this
>
> for ( unsigned i=0; i!=3; ++i )
>   for ( unsigned j=2; j!=5;++j )
>     for ( unsigned k=4;k!=10;++k )
>       f( i, j, k );
>
> I'd guess this comes up pretty often, and by generalising it to
>
> * an arbitrary number of ranges, of arbitrary types
> * f() implemented as a lambda or bind function
>
> would have quite wide applicability.
>
> I've considered writing an implemetation, which has dragged me
> into examining fusion and mpl in some detail, but doing so does
> not seem to me be trivial at all.
>
> Is there already something in Boost which does this?

I don't think so.  Something like this would work:

make_product_range(
  make_counting_range(0, 3),
  make_counting_range(2, 5),
  make_counting_range(4, 10)
)

should return a range whose elements are (0, 2, 4), (0, 2, 5), etc.

Then as I recall, fusion has function object adapters for splitting
these up into separate arguments to a function.

I've attached a solution that handles the two argument case.

In Christ,
Steven Watanabe