Hi ...
I have been looking into Neko quite a lot lately, as it seems to be a
project with some good momentum (thanks to HaXe i quess). I have tried
to make my own language once (http://trac.lue.dk/tea), and it fun to
make a language, but its hard maintain it outside its narrow domain.
Well, the VM in tea are stack based (not very optimal, but fun to
write), so the type system was the really special part (all values was
the same base type). In the domain i wrote TEA in needed to be able to
resolve complex types (objects) as late/lazy as possible, but type safe.
The reason for this was that I had a DB, and all elements (tables) in
the database needed to be visible to the language, but I only created
the full object if the script tried to access the object (this was at
compile time, not runtime). Also, I need to define the lazy object based
on meta data, and not just by loading data (that i did not have, at
compile time).
This method made it possible to validate if the scripts was accessing
the prober elements in the DB, and that the types was right (based on
meta data directly from DB), and when there were very few subjects left
to resolve besides bade controle statements :-)
Also, I used refcount for values, as I disliked the sweep delete problem
(strange delays), and I just love destructor's, for non memory resources :-)
So my question is, it it possible to in Neko to create types lazy, at
compile time like descried, or something like it ? Also is it possible
to have real refcount values, and destructor's ?
Kind regards ...
/BL
--
Neko : One VM to run them all
(http://nekovm.org)
