Re: [glob2-devel] Embedded scripting language (again)

[Bo Lorentsen]

Stéphane Magnenat wrote:

> We have been thinking about this for a long time indeed. Don't forget that the 
> determinism is not the only constraint, but another one is the ability to 
> checkpoint the VM in order to save and reload a game. Another one is the 
> ability to seamlessly run several threads (stories in SGSL) with co-routine 
> operations (wait in SGSL) and preemption on buggy thread (stories are 
> preempted after a thousand of continuous instructions in SGSL).
>  
I  find both Lua and Squirrel (that seems to be Lua based) wery neat, 
and both have co-rutines, but I don't know anything about persisting 
data and checkpoints (cross platform).

> Marv and I have looked into several languages, the bests candidates are indeed 
> Squirrel (http://squirrel.sourceforge.net/), but the multi-threading and 
> checkpoint stuff is not trivial, and tea (http://lue.dk/prj/tea), whose 
> author was willing to help, but who seems to have disappeared lately.
>  
Nope, he is right here, but working hard on TEA and some other projects 
(have to make a living) :-) I have just updated the link you gave to my 
trac/svn repos, in case anyone like to know what is happening with TEA, 
at the moment :-)

> Lisp syntax are a big joke, lua is squirrel--, python is probably, for our 
> use, slightly bloated, but this can be discussed.
>  
Python is a nice language, but for threads and corutines, I am not sure 
its the propper choice at the moment.

> I think that there is two main constraints for language choice:
> * the capabilities of the language
>  
And this is not trivial nor common :-)

> * the language from the user perspective
>
> I don't see how you want to fix missing capabilities afterwards, hacking on a 
> big language is not trivial. For hacking, tea is probably the best candidate.
>  
Yeps, that why I made it :-)

TEA need some work regarding persisting of both runtime state and data, 
but the framework is there :-)

/BL