[R] behavior of [<-.foo

[Prof Brian Ripley]

I've not seen an actual answer to this, which is that this is a 
misunderstanding as to how NextMethod works.

First,

> +     x <- unclass(x)

looks wrong.  NextMethod uses the next method at the call to the generic, 
and subsequent changes to the object 'x' do not alter the class that would 
be dispatched on.  Given that the next method might not be the default 
method, unclassing here seems potentially damaging.

Second, the matched call is

Called from: `[<-.foo`(`*tmp*`, , value = 100)

and it is that call which is going to be passed on to the next method. 
As the help page says:

      'NextMethod' works by creating a special call frame for the next
      method.  If no new arguments are supplied, the arguments will be
      the same in number, order and name as those to the current method
      but their values will be promises to evaluate their name in the
      current method and environment.

Since 'j' was not an argument of the original call, it is ignored.

Now, [<- is an internal generic without a visible default method, but it 
is as if you have called `[<-.default`(`*tmp*`, 1:5, value = 100), which 
explains the result you got.

(That NextMethod invokes the generic as a possible default method is not 
documented anywhere that I can see, and the description in 2.3.1 is all 
about methods invoked via UseMethod.  There are issues with the above 
description when the method invoked is a primitive such as [<-, as that 
uses positional matching.  I would not be confident that this works as 
intended for multi-argument primitives.)

x[,] <- 100 is perhaps what you intended for a matrix-like class, and that 
does not work (it seems because the argument matching does indeed not work 
as intended).  You need something like

`[<-.foo` <- function(x, i, j, value) {
      if(missing(i)) i <- 1:nrow(x)
      if(missing(j)) j <- 1:ncol(x)
      cl <- class(x)
      cll <- length(cl)
      m <- match("foo", cl, cll)
      oldClass(x) <- if(m == cll) NULL else cl[(m+1):cll]
      x[i,j] <- value
      class(x) <- cl
      x
}


On Fri, 22 Sep 2006, Armstrong, Whit wrote:

> Can someone help me understand the following behavior of "[<-" ?
>
> If I define a simple class based on a matrix, the [<- operation only
> inserts into the first column:
>
>
>> x <- matrix(rnorm(10),nrow=5,ncol=2)
>>  class(x) <- "foo"
>> "[<-.foo" <- function(x, i, j, value) {
> +     if(missing(i)) i <- 1:nrow(x)
> +     if(missing(j)) j <- 1:ncol(x)
> +
> +     x <- unclass(x)
> +     x <- NextMethod(.Generic)
> +     class(x) <- "foo"
> +     x
> + }
>>
>> x[] <- 100.0
>> x
>     [,1]       [,2]
> [1,]  100 -0.1465296
> [2,]  100 -0.2615796
> [3,]  100 -0.8882629
> [4,]  100 -0.2886357
> [5,]  100 -0.9565273
> attr(,"class")
> [1] "foo"
>
> Based on the behavior of [<- for a matrix, I would have thought that the
> data for the whole object would be replaced.
>
> for instance:
>
>> y <- matrix(rnorm(10),nrow=5,ncol=2)
>> y
>            [,1]       [,2]
> [1,] -0.55297049 -1.1896488
> [2,]  0.06157438 -0.6628254
> [3,] -0.28184208 -2.5260177
> [4,]  0.61204398 -0.3492488
> [5,]  0.43971216  1.8990789
>> y[] <- 100
>> y
>     [,1] [,2]
> [1,]  100  100
> [2,]  100  100
> [3,]  100  100
> [4,]  100  100
> [5,]  100  100

[...]

-- 
Brian D. Ripley,                  ripley at stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595