[Rd] Bug in copying of S4 objects (PR#8112)

[Luke Tierney]

I've poked in the source a bit.  Here are some notes in case someone
has time to look into this.

The main internal difference between <<- and <- for complex
assignments is that <- calls EnsureLocal which calls duplicate on the
value of the left hand side value if NAMED == 2.  In principle I don't
tink this should be necessary, but it means that assignment functions
called for <- assignments will not see NAMED == 2.  This protects
agains internal assignment functions that destructively modify without
looking at NAMED, which is what @<- and slot<- do.  But <<- does not
do this defensive duplicating, hence the problem with <<-.  Similar
problems occur if structures contain environments used to implement
reference behaviour since the copying stops at the environment.

The appropriate fix is to insure that @<- and slot<- respect NAMED and
duplicate when NAMED == 2, as attr<- does.  This will require either
making @<- and slot<- into SPECIALSXP's or some sort of
underhandedness to allow them to remain closures.  One possibility
might be to define a SPECIALSXP that looks at the NAMED value of the
object argument (by inspecting the promise before it is forced).  This
can then be passed into the internal R_set_slot function via .Call.

luke

On Thu, 1 Sep 2005, Luke Tierney wrote:

> Another variant of what is probably the same issue:
>
>    > setClass("foo", representation(a = "numeric"))
>    [1] "foo"
>    > f <- function() x at a <<- 2
>    > x <- new("foo",a=1)
>    > y <- x
>    > f()
>    > x
>    An object of class “foo”
>    Slot "a":
>    [1] 2
>
>    > y
>    An object of class “foo”
>    Slot "a":
>    [1] 2
>
> luke
>
> On Wed, 31 Aug 2005, murdoch at stats.uwo.ca wrote:
>
>> If I have an S4 object, and I make a copy, changes to the original
>> aren't reflected in the copy:
>> 
>> > setClass("foo", representation(slot="numeric"))
>> > x <- new("foo", slot=1)
>> > y <- x
>> > x at slot <- 2
>> > y
>> An object of class "foo"
>> Slot "slot":
>> [1] 1
>> 
>> This is as it should be.  However, if I call the slot assignment
>> function in a funny way, y *does* receive the changes:
>> 
>> > x <- new("foo", slot=1)
>> > y <- x
>> > assign("x", "@<-"(x, "slot", 2))
>> > y
>> An object of class "foo"
>> Slot "slot":
>> [1] 2
>> 
>> This happens in the current R-devel in Windows, and R-patched too.
>> 
>> > version
>>          _
>> platform i386-pc-mingw32
>> arch     i386
>> os       mingw32
>> system   i386, mingw32
>> status   Under development (unstable)
>> major    2
>> minor    2.0
>> year     2005
>> month    08
>> day      31
>> svn rev  35467
>> language R
>> 
>> Duncan Murdoch
>> 
>> ______________________________________________
>> R-devel at r-project.org mailing list
>> https://stat.ethz.ch/mailman/listinfo/r-devel
>> 
>
>

-- 
Luke Tierney
Chair, Statistics and Actuarial Science
Ralph E. Wareham Professor of Mathematical Sciences
University of Iowa                  Phone:             319-335-3386
Department of Statistics and        Fax:               319-335-3017
    Actuarial Science
241 Schaeffer Hall                  email:      luke at stat.uiowa.edu
Iowa City, IA 52242                 WWW:  http://www.stat.uiowa.edu