[R] Can an object reference itself?

[Janko Thyson]

Hadley,

thanks for your comment. What you're saying is true, of course, and possibly
I did not really chose the best header to describe the actual issue I was
addressing. However, in the help archive I found something that gets me what
I want (calling a certain "slot function" without having to specify
arguments at calling time). I adapted it to my situation and maybe it's of
interest for someone facing a similar problem.

setClass(
	Class="Testclass",
	representation=representation(
		env="environment",
		data="function"
	)
)

data.fun <- function(env) get("actual.data", env=env)

setMethod(
	f="initialize", 
	signature=signature("Testclass"), 
	definition=function(
		.Object, 
		actual.data=NULL, 
		...
	) {
	     env 			<- new.env(parent=emptyenv())
	     env$actual.data 	<- actual.data
		 
		data.fun.wrapper <- function() data.fun(env)
		 
	     callNextMethod(.Object, env=env, data=data.fun.wrapper, ...)
 	}
)

obj <- new("Testclass", actual.data=1:10)
obj at data()
obj at env$actual.data	<- 1:5
obj at data()

obj.mod <- new("Testclass")
obj.mod at env$actual.data	<- "horst"
obj.mod at data()

Loading an object from harddrive could be implemented in a similar way by
putting the path-filename-combination in "obj at env" and adapting the function
"data.fun()" accordingly.

Putting "data.fun()" into the wrapper "data.fun.wrapper()" allows for
changes in "data.fun()" without having to re-source the method for
"initialize()" and to reassign "obj" each time.

Regards,
Janko

> -----Ursprüngliche Nachricht-----
> Von: hadley wickham [mailto:h.wickham at gmail.com]
> Gesendet: Freitag, 15. Januar 2010 19:35
> An: Janko Thyson
> Cc: r-help at r-project.org
> Betreff: Re: [R] Can an object reference itself?
> 
> On Fri, Jan 15, 2010 at 5:30 PM, Janko Thyson
> <janko.thyson at ku-eichstaett.de> wrote:
> > Dear List,
> >
> > I am not really familiar with any other language than R, but I’ve
> heard that
> > in other languages there is something called “self referencing”.
> >
> > Here’s what I’m trying to get an answer for:
> > Suppose there is a function that takes as its input a value of a slot
> of an
> > S4 object. The function itself is stored in another slot of the SAME
> S4
> > object. Is it then possible to have the function automatically
> “recognize”
> > the name of the object in which slot it is placed via some sort of
> “self
> > referencing” of the S4 object (to avoid having to explicitly state
> the
> > required function argument for different S4 object instances with
> different
> > names)?
> >
> > I hope the following code snippets will give you an idea what I’m
> trying to
> > do:
> >
> > obj.for.slot <- data.frame(a=1); obj.for.slot <- data.frame(a=100)
> > save(obj.for.slot, file="C:/obj.1.for.slot.Rdata");
> save(obj.for.slot,
> > file="C:/obj.2.for.slot.Rdata")
> >
> > slotfun <- function(obj.name)
> > {
> >        file.fqn.char <- paste("file.fqn <- ", obj.name, "@file.fqn",
> > sep="")
> >        eval(parse(text=file.fqn.char))
> >        load(file=file.fqn)
> >        return(obj.for.slot)
> > }
> >
> > setClass(
> >        Class="Testclass",
> >        representation=representation(
> >                file.fqn="character",
> >                data="function"
> >        ),
> >        prototype=prototype(
> >                file.fqn="blabla",
> >                data=slotfun
> >        )
> > )
> >
> > test            <- new("Testclass")
> > test.mod        <- new("Testclass")
> >
> >
> > test at file.fqn           <- "C:/obj.1.for.slot.Rdata"
> > test.mod at file.fqn       <- "C:/obj.2.for.slot.Rdata"
> >
> > test at data(obj.name="test")
> > test.mod at data(obj.name="test.mod")
> 
> An object does not have a unique name.  What would happen in the
> following situations?
> 
> t3 <- test.mod
> tests <- list(
>   test,
>   test.mod
> )
> 
> Hadley
> 
> --
> http://had.co.nz/