Re: [Boost-users] [Review] Phoenix V3: mini-review starts February 19th

On Saturday, February 26, 2011 11:28:58 PM alfC wrote:

On Feb 25, 11:57 pm, Thomas Heller wrote:

...

On Saturday, February 26, 2011 08:49:25 AM alfC wrote:

...

On Jan 31, 5:43 pm, "Hartmut Kaiser" wrote:

...

Hi List,

PhoenixV3 is nearing its completion.

With the port to Boost.Proto we tried to address the outstanding issues of the review and would like to announce that themini-review will start on February 19th.

Thank you for the new version of Phoenix-3. I wonder if things like phoenix-expression matching is now possible as "promised" in this

thread: http://lists.boost.org/boost-users/2010/05/59491.php

Yes, stuff like this could be indeed made possible. Though, I don't see where phoenix comes into the game in the examples in this thread.

I was doing things like this with Phoenix2 by hacking the internals:

in a function call like this

integral( 'arbitrary phoenix expression' , interval(2., 4.));

I was matching, first that the passed expresion is of the form #1/ ( phoenix::_1 - #2), if so it checked that #2 is a 'double-type' value, then extracted that value and then passed the whole thing to a numerical intergrator that handle functions with sigularities at the value of #2. If the expression didn't match then a generic integrator is called.

Nothing easier than that, no need to hack (more or less) undocumented internals of the library. Every phoenix expression can be matched with proto's grammar matching facilities: struct singularity : proto::when< proto::divides< proto::_ , proto::minus< phoenix::expression::argument<1> , proto::_ > > , proto::_value(proto::_right(proto::_right)) > {}; Match arbitratry expressions with proto::matches ... call the transform to get the singularity, etc.

Thanks, Alfredo