Is newproxy (still) a hack?

[Asko Kauppi <askok@...>]

I remember hearing newproxy mentioned a number of times on the list.

However, it still has not been added to the documentation, and a short  
glimpse on list archive shows Roberto calling it a "hack".

Has that status been changed since 2006, or is it going to?

-asko

http://lua-users.org/lists/lua-l/2006-06/msg00436.html


David Manura kirjoitti 23.1.2009 kello 6:53:

> On Wed, Jan 21, 2009 at 4:20 AM, Asko Kauppi wrote:
> > Using traditional GC mechanism for these scopes is deterministic,  
> > since the
> > whole Lua state is cleared at return/error. However, I could survey  
> > this a
> > little further and maybe add some ease of use way to add  
> > "finalizers" to the
> > lanes.
>
> In that case, either the caller or callee could have that
> responsibility.  For example, the callee could probably do
>
>  local function on_exit(f)
>    local o = newproxy(true)
>    getmetatable(o).__gc = f
>    _G[{}] = o
>  end
>
>  on_exit(function() print 'deleting file..' end)

Yes, this seems like the way to go.

Thanks for the sample on using 'newproxy' - I haven't used it anywhere  
until now.

> Similarly, in the more general case, one "could" call collectgarbage()
> after every pcall to deterministically cleanup most objects, though
> relying on this could be more error prone compared to an explicit
> mechanism that disposes all objects out of scope, regardless if they
> are finalizable.

With Lanes, the Lua state is actually closed after the pcall.  That  
should be very deterministic. :)

-asko