Re: a new proposal for environments

[Enrico Colombini <erix@...>]

Tomas Guisasola Gorham wrote:
> > _G = { a = 4 }
> > print(a) --> 3  [*]
> > print(_G.a) --> 4
> >
> > [*] with _ENV instead of _G, this would be 4 (if I'm not mistaken).
>
>     In fact, it will throw an error, because print (or _ENV.print)
> does not exist :-) 

Right :-) I should have used { a = 4, print = print }

> > In short, _G is just a reference to the globals table; changing _G 
> > will not change the global table.
> > On the other hand, _ENV is an upvalue containing the current 'globals' 
> > table; changing _ENV will change the 'globals'.
>     Not exactly:
>
> _G.print = nil -- equivalent to _ENV.print = nil

Ah, yes: I meant assigning _G itself, not its fields. My point was that 
_G references the globals table but the VM keeps its own reference to it 
(whatever happens to _G), while assigning _ENV actually changes the 
globals table used by the VM just like setfenv does.

  Enrico