Re: [boost] [proto] transform for flattening an expression - how?

From: Larry Evans (cppljevans_at_[hidden])
Date: 2008-03-20 08:44:50

On 03/18/08 17:23, Markus Werle wrote:
[snip]
> 
> I am dreaming of:
> 
> scalar_field<0> rho;
> scalar_field<1> u;
> scalar_field<2> v;
> ....
> 
> Solver.AddPDESystem()
> [ddt(rho)     + ddx(rho * u)         + ddy(rho * v)  + sourceterm  == 0],
> [ddt(rho * u) + ddx(rho * pow<2, 1>(u)) + ddy(rho * u * v)      == 0],
> [.. some other 10-20 partial differential eqns. omitted        ];
> 
[snip]
> In chemistry there are systems with more than 100 equations of this form. 
> One day proto will bend before such usecases. You've been warned ;-)

Spirit has something similar except it's equations always have a
single variable (a.k.a. nonterminal in grammar parlance) on one
side of the equation.  For example, the calculator::definition
CTOR here:

http://www.boost.org/libs/spirit/example/fundamental/full_calc.cpp

Spirit uses pointer's to the rhs stored in spirit::rule's which
avoids, somewhat, the compiler limitations.  Maybe you could
do something similar.

This sequence of grammar equations (a.k.a. productions)
and a solution to these equations for non-terminal
lookaheads is prototyped using proto in cfg_lookahead_extends.zip
here:

http://www.boost-consulting.com/vault/index.php?&directory=Strings%20-%20Text%20Processing

However, instead of using '[]' to separate the equations, like spirit, 
this prototype uses ','.

Maybe the lookahead prototype could give you some ideas for
constructing the equations and solving them for your
PDE solver.