Re: ObjC dynamic parser

[glen e. p. ropella]

Marcus G. Daniels wrote:

> However, you then suggested using such a parser for genetic
> programming, and I think that GP is one of a large class of
> applications where (in my experience) it is it is better to stick with
> a general purpose language implementation.  It's just as easy, more
> reliable, better defined, and more powerful.

> Absolutely -- assuming that a mini-language is really the best way to go.
> Unix and X11 are full of funny file formats and mini-languages.
> It seems like lots of useless minor variation to me.
>

I think some pretty good arguments can be made for using mini-languages
(good word, BTW) in some cases.  For instance, in GP, if one is really
trying to grow the kind of programs we write (i.e. that solve the kinds
of problems human-written programs try to solve), then you're absolutely
right that the genetic substrate should be coded in a language capable of
expression as complex as the programs we write.  And that implies using
a rich language.

But, if the problem to be solved is much simpler, like writing programs
for the behavior of cells in a CA or writing programs where no floating
point arithmetic is necessary, then one might be able to prune the set of
possible programs quite quickly by using a special mini-language.  And
in many cases, this *seems* to be a good idea.  One of the things that
struck me about Koza's work is that he stresses the front-end work
of designing your language so that the GA algorithm is almost certain
to find a good solution.  Kinda backhanded, in a way. [grin]  But,
I can also imagine cases where using a mini-language in which to
encode your gene might prune out all the good (or efficient) solutions
in advance, keeping the GA from ever finding a good solution.

Anyway, the gist of all this is that I think it would be a good thing for
the user to be able to install her own dynamically loadable parser,
whether that parser is for a full- or mini- language.  Heck, I can
even imagine people wanting to load a Scheme parser. [grin]
glen
--
Glen E. Ropella                  (505) 424-0448
The Swarm Corporation            address@hidden




                  ==================================
   Swarm-Modelling is for discussion of Simulation and Modelling techniques
   esp. using Swarm.  For list administration needs (esp. [un]subscribing),
   please send a message to <address@hidden> with "help" in the
   body of the message.
                  ==================================