It could be done, but again: lexical scoping. Using "nonlocal" would make _ENV less elegant and it would make closures more verbose, for no practical benefit.
can you clarify with an example how _ENV becomes less elegant?
same for closures: why should they be more verbose?
A closure example:
function make_counter()
local x = 1
return function() x = x + 1; return x; end
end
This is very trivial, but this would require a "nonlocal" definition even in a simple example. _ENV is more or less a different way of approaching closures. A more complex closure just ends up needing more nonlocal declarations. As I said, this doesn't impact the correctness of it, but this is a case where you absolutely HAVE to have a declared local inside the outer scope, regardless of the syntax, so "nonlocal" can only make this code more verbose.
I am not sure I fully understand what you mean by "symmetric" behavior.
writing: "x = 1" can be an assignment to local, upvalue, or global in both cases; it depends if/how x was declared before
reading: " ... = x" means referencing local, upvalue, or global in both cases; it depends if/how x was declared before
function read()
return x
end
function write()
x = 1
end
Since neither of these have any declarations, I would hope that they would behave the same way with respect to the x variable -- either both local (which makes read() an error) or both global, but in a language with "nonlocal" then read() is global while write() is local.
/s/ Adam
