Re: deprecating 'for x in t'

[Rici Lake <lua@...>]

On 1-Nov-04, at 12:43 PM, Lucas Ackerman wrote:

> I would like to see this resolved as well.  Either officially kill it,
> give a compiler option to dissallow it (probably not good to fragment 
> the
> language base), or make it syntactic sugar for 'for x in pairs(t)', so
> that it always goes through the pairs function.

It is not a compiler issue, as I see it. The syntax:

  for k in iterator, object, current do
  end

is all the compiler knows about.

>   The latter option is my
> preference, as it allows us to override the default table iteration in 
> a
> clean way, short of the metatable __pairs ideas (which of course don't
> address the basic 'for' issue for regular meta-free tables).

The proposal I posted a few hours ago does cover regular meta-free 
tables
(by doing the same thing as the current standard pairs function). A 
simpler
implementation, which ignored the possibility of tables with __call
metamethods, would have simply used the standard pairs function.

R.