Re: Class structures

[Jamie Webb <j@...>]

On Thursday 20 May 2004 22:37, Ivan Kolev wrote:
> I realize that
> many people have developed such tools before, and if we joined efforts,
> we might bring closer the idea of Lua as a powerful standalone RAD
> language.

You're right, LuaCheia does badly need more man-hours. Perhaps you should join 
and bring the software you've developed with you? It seems like the obvious 
focal point for the sort of work you're talking about, and it needs the 
critical mass so that people start developing bindings, etc. which work with 
LuaCheia rather than a bunch of separate upstream projects which have to be 
integrated. Then, hopefully it will become exactly the sort of environment 
you are after.

On the object system front, I once had great plans to add a sophisticated 
object system system to Lua. It never got past the drawing board, because I 
realised that an implementation would be about the same size as Lua itself, 
and hopelessly inefficient. I couldn't correct those problems without 
removing features which I might want, and I didn't want to write an object 
system that I would then be constantly fighting. The thing just didn't adhere 
to the 'Lua way'.

In practice, I've found ad-hoc object systems written in 10 lines of code to 
be quite adequate (and extensible with particular features as I need them). 
Lua provides just enough consistency (if you don't try to be too clever) for 
there to be no real compatibility issues between different object systems. On 
the other hand, I've likewise given up on waiting for a Lua GUI library to 
become useable and started on my own. That will feature a very simple and 
very thin dynamic object model which will permit subclassing between C and 
Lua. The external interface for the entire GUI library is (presently) 5 
functions, 2 macros, and 2 structs.

-- Jamie Webb