RE: To "colon" or not to "colon"

["Erik Cassel" <erik@...>]

> In order for bar1 and bar2 to be different functions, you'd have to
> create each one using lua_pushcclosure().

Yes, I use the "lazy" approach.  I certainly don't want to create 500
closures!


> foo1 = newObj()
> foo2 = newObj()
> assert( getmetatable(foo1) == getmetatable(foo2), "huh?" )
> bar1 = foo1.bar
> bar2 = foo2.bar

Hey, that brings up another interesting point:

	foo1 = newObj()
	foo2 = newObj()
	assert( foo1.bar == foo2.bar, "huh?"  )

That assertion fails, too.  To make matters worse, the lazy pushclosure
approach leads to the following even more bizarre assertion failure:

	foo1 = newObj()
	assert( foo1.bar == foo1.bar, "huh????!!!!!!!!"  )


I see more and more reasons to not bend the rules.  Thanks to everybody for
their input!

-Erik





> -----Original Message-----
> From: lua-bounces@bazar2.conectiva.com.br [mailto:lua-
> bounces@bazar2.conectiva.com.br] On Behalf Of Mark Edgar
> Sent: Wednesday, July 12, 2006 4:32 PM
> To: Lua list
> Subject: Re: To "colon" or not to "colon"
> 
> Erik Cassel wrote:
> [snip]
> > However, it occurs to me that this isn't strictly necessary.  When my
> C++
> > object handles the __index request, it already knows what "self" is,
> because
> > it is passed in on the stack, along with the function name.  It is
> redundant
> > to require the object again as the first function argument.  My Lua code
> > could look like this:
> >
> > 	foo = newFoo()
> > 	foo.prop = 2
> > 	foo.bar()
> 
> Note that the colon syntax applies to function calls ("methods") and not
> assignments or index operations ("properties").  This is illegal syntax:
> 
> foo:prop = 2
> 
> That aside, I believe that you'll find you are mistaken the redundany of
> the self parameter as the __index metamethod only retrieves the member,
> but does not call it.
> 
> foo1 = newObj()
> foo2 = newObj()
> assert( getmetatable(foo1) == getmetatable(foo2), "huh?" )
> bar1 = foo1.bar
> bar2 = foo2.bar
> 
> In order for bar1 and bar2 to be different functions, you'd have to
> create each one using lua_pushcclosure().  In other words, each object
> would have to have its own set of functions instead of all objects of a
> class sharing a single set of functions.  In fact, the typical
> implementation has the metatable as the __index instead of an __index
> function.
> 
> You _can_ do this, of course, but do you really want to?  If you have a
> class with 5 methods, and you create 100 instances of this class, you
> would have to create 500 closures.  You could lazily create these
> closures in the __index metamethod, but it's probably a bit of overhead
> for not much syntactical gain.
> 
> 					-Mark