[Rd] Baffled with as.matrix

[Brian Diggs]

On 11/30/2012 4:53 AM, Terry Therneau wrote:
> I'm puzzled by as.matrix.  It appears to work differently for Surv objects.
> Here is a session from my computer:
>
> tmt% R --vanilla
>  > library(survival)
> Loading required package: splines
>  > ytest <- Surv(1:3, c(1,0,1))
>  > is.matrix(ytest)
>  >[1] TRUE

It seems that survival:::as.matrix.Surv is not recognized as a 
dispatch-able method for as.matrix:

 > methods("as.matrix")
[1] as.matrix.data.frame as.matrix.default    as.matrix.dist*
[4] as.matrix.noquote    as.matrix.POSIXlt    as.matrix.raster*
[7] as.matrix.ratetable*

I'm guessing it is not declared as such in the NAMESPACE file.

>  > attr(ytest, 'type')
> [1] "right"
>  > attr(as.matrix(ytest), 'type')
> [1] "right"
>  >
>  > y2 <- ytest
>  > class(y2) <- "charlie"
>  > as.matrix.charlie <- survival:::as.matrix.Surv

This new function does show up:

 > methods("as.matrix")
[1] as.matrix.charlie    as.matrix.data.frame as.matrix.default
[4] as.matrix.dist*      as.matrix.noquote    as.matrix.POSIXlt
[7] as.matrix.raster*    as.matrix.ratetable*

and is being called as expected.

>  > attr(y2, 'type')
> [1] "right"
>  > attr(as.matrix(y2), 'type')
> NULL
>
>  > survival:::as.matrix.Surv
> function (x)
> {
>      y <- unclass(x)
>      attr(y, "type") <- NULL
>      y
> }
> <bytecode: 0x91c1610>
> <environment: namespace:survival>

What is being called by as.matrix(ytest) is as.matrix.default

 > as.matrix.default
function (x, ...)
{
     if (is.matrix(x))
         x
     else array(x, c(length(x), 1L), if (!is.null(names(x)))
         list(names(x), NULL)
     else NULL)
}
<bytecode: 0x000000000631a0b8>
<environment: namespace:base>

which will just return ytest is is.matrix(ytest) is TRUE (as you show).

> --------
> It appears that Surv objects are being processed by as.matrix.default,
> but "charlie" objects by the
> actual method.  One more verification:
>
>  > attr(survival:::as.matrix.Surv(ytest), 'type')
> NULL
>  > attr(as.matrix.default(y2), 'type')
> [1] "right"

Correct.

> Context: In testing the next survival release (2.37), it has lost this
> "special" behavior.  One package that depends on survival expects this
> behavior and thus fails.  I'm at a loss to figure out how my package got
> this attribute in the first place, or how it lost it.  Can anyone shed
> light?

The behavior I showed above was with 2.36-14, even.

In your NAMESPACE file, there are two lines that say

S3method(as.matrix, ratetable)

I bet one of them was meant to be

S3method(as.matrix, Surv)


> Terry Therneau
>
> PS I'm on vacation for the next few days so will be intermittent with
> email.  (Off to see my first grandchild!)
>
> ---------------------
>
>  > sessionInfo()
> R version 2.15.2 (2012-10-26)
> Platform: i686-pc-linux-gnu (32-bit)
>
> locale:
>   [1] LC_CTYPE=en_US.UTF-8       LC_NUMERIC=C
>   [3] LC_TIME=en_US.UTF-8        LC_COLLATE=C
>   [5] LC_MONETARY=en_US.UTF-8    LC_MESSAGES=en_US.UTF-8
>   [7] LC_PAPER=C                 LC_NAME=C
>   [9] LC_ADDRESS=C               LC_TELEPHONE=C
> [11] LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C
>
> attached base packages:
> [1] splines   stats     graphics  grDevices utils     datasets  methods
> [8] base
>
> other attached packages:
> [1] survival_2.36-14
>


-- 
Brian S. Diggs, PhD
Senior Research Associate, Department of Surgery
Oregon Health & Science University