RE: tables and the gc

["Nick Trout" <nick@...>]

    function create_resource() return "X" end
    function delete_resource(r) end
    
    class1 = {} -- simplification
        
    function class1:create()
        -- create hardware resource
        self.res = create_resource()
        return self
    end
    
    function class1:delete()
        -- delete the resource
        print("delete!")
        self.res = delete_resource(self.res)
        -- returning nothing returns nil
    end
    
    
    class2 = {}  -- simplification
        
    function class2:create()
        self.class1 = class1:create()
        print(self.class1.res)
    end
        
    function class2:free_hwr()
        -- delete resource and set self.class1 to nil
        self.class1 = class1:delete()
        print(self.class1)
    end
    
    class2:create()
    class2:free_hwr()
    
Gives you:

X
delete!
nil


I think this is what you want?

Using the GC to collect objects immediately will not be efficient.
Delete user objects explicitly and let the GC collect the Lua bits in
its own time.

You really don't want to go the way of forcing GC, especially as
incremental GC is on its way.

Mmmm. Friday. Sunshine. Beer. Have a good weekend and I hope I answered
your question.

Regards,
Nick


-----Original Message-----
From: Tom Spilman [mailto:tom@sickheadgames.com] 
Sent: Friday, April 25, 2003 3:23 PM
To: Multiple recipients of list
Subject: tables and the gc


I have the following code (i'm using luabind, but this is a lua 
issue )...


  class 'my_class1'

  function my_class1:__init()

    self._my_hardware_resource = create_hardware_resource()

  end


  class 'my_class2'

  function my_class2:__init()

    self._my_class1 = my_class1()

  end

  function my_class2:free_hardware_resources()

    self._my_class1 = nil

    -- note that the class above is still not
    -- released... must gc.
    collectgarbage()

    -- now 'my_class1' has been gc'ed, but
    -- the internal member '_my_hardware_resource'
    -- has not... must call gc again to do that.
    collectgarbage()

  end


  Looking at my_class2.free_hardware_resources() you can see what my 
problem is.  It seems that the first garbage collection pass will 
free a table ( the class members are in a table ), but will not free 
newly unreferenced members of that table.  A second garbage 
collection pass is needed to do that.  I assume that if i had a 
deeper hierarchy i'd need to call the garbage collection call once 
per level to ensure resources i have to have freed are.

  Ok... stop the madness... what is the right way for me to free 
something that must be destroyed immediately?  I'm thinking 
a "destructor" for 'my_class1' would be once solution as i could nil 
the member before the gc call.  Still calling collectgarbage() to 
force the destruction of one object seems like overkill to me.  
Should i just be using a different design pattern than what i've 
used in other languages for years?

  Tom