Feature request: writing es scripts inside

[Neon22]

with reference to this flickr conversation and image
http://flickr.com/photos/30278668@N04/3294456952/?addedcomment=1#comment72157614186656057

Is it possible to use the Javascript in Structure synth to write es scripts. Thereby allowing generation of 'random' or other kinds of es scripts ?

[Mikael Hvidtfeldt Christensen]

The JavaScript interface is far from done, but yes, when its done it should be possible to create random scripts.
The JavaScript interface was mostly intended for automating the workflow when producing animations, though.

I would really like to build a more complex random script generator explorer, based on the idea of a '9-up' interface, where the user interactively could 'mutate' the generated script (as Groovelock suggested <a href="http://flickr.com/photos/30278668@N04/3243441274/">here</a>).

It will probably be a while before I get around to implement this :-)

[Neon22]

Here is a nice simple(?) opensource genetic algorithm controller. its in Python but moving it to Javascript looks quite possible.
I'm trying to get it to make SS strings but no luck yet.
http://www.freenet.org.nz/python/pygene/
probably have to use "DiscreteGene" type

[Mikael Hvidtfeldt Christensen]

I'm not sure that Genetic Algorithms are the right choice if the user has to manually score the structures - you probably do not want to rate all of the structures in a population?

Also finding a good representation of the EisenScript programs is certainly not trivial - you would want something that is both syntactically correct - and interesting. So using symbol sequences as a representation may be difficult.

[Neon22]

Actually I was thinking the user would be the approver of fitness and to just use the genetic part to make the mutations.
So make nine - then let user choose which ones to mutate further...

as for the symbol representations. I like their example of deriving quadratic solution.
But I think the symbols for SS would have to be more complex
E.g.
   the symbols might be all the core operators e.g. "ry"
   but they would need to be assembled in syntactic patterns with numbers and in {}
      (hence thinking about using the "DiscreteGene" methods)
e.g. {ry 20 rx 2.0 hue 20} rule/object

But I expect you are right in that as only mutation and permutation are required maybe just direct coding it is easier :-)