[R] Cacheing computationally expensive getter methods for S4 objects

[Martin Morgan]

Steve Lianoglou wrote:
> Very clever, that looks to do the trick!

I think though that all Square's then share the same environment
(created in the prototype) and hence area.

> area(new("Square", length=50, width=100))
Accessing
[1] 50


Which is quite efficient at doing the calculation, but maybe not what is
expected...

The solution is to create the area environment in an 'initialize' method.

A different approach, but along similar lines, might make 'area' a
function that exploits lexical scope. Here's an area 'factory'

areaf <- function() {
    area <- NULL
    function(x) {
        if (is.null(area)) {
            message("expensive")
            area <<- x at length * x at width
        }
        area
    }
}

that we use in the initialize method

setMethod(initialize, "Rect",
          function(.Object, ..., length=.Object at length,
                   width=.Object at width)
{
    callNextMethod(.Object, area=areaf(), length=length,
                   width=width, ...)
})

setMethod(area, "Rect", function(x) x at area(x))

The signature of initialize is such that one could

setReplaceMethod("length", c("Rect", "numeric"), function(x, value) {
    initialize(x, length=value)
})


so

> a <- new("Rect", length=10, width=5)
> area(a)
expensive
[1] 50
> area(a) # cheap
[1] 50
> b <- a
> area(b)
[1] 50
> length(b) <- 20
> area(a)
[1] 50
> area(b)
expensive
[1] 100

Martin

> 
> Thanks,
> 
> -steve
> 
> On Oct 14, 2009, at 2:57 PM, Benilton Carvalho wrote:
> 
>> If you change 'area' to an environment, you may be able to get
>> something close to what you want.
>>
>> For example:
>>
>> setClass("Square",
>>         representation(
>>                        length='numeric',
>>                        width='numeric',
>>                        area='environment'
>>                        ),
>>         prototype(
>>                   length=0,
>>                   width=0,
>>                   area=new.env()
>>                   )
>>         )
>>
>> setGeneric("area", function(x) standardGeneric("area"))
>> setMethod("area", "Square",
>>          function(x){
>>            if (length(ls(x at area)) == 0){
>>              message("Computing")
>>              assign("area", x at width * x at length, envir=x at area)
>>            }
>>            message("Accessing")
>>            get("area", envir=x at area)
>>          })
>>
>> tmp <- new("Square", length=5, width=10)
>> area(tmp) ## This should show "computing" and "accessing"
>> area(tmp) ## the 2nd call should show 'accessing' only
>>
>>
>> b
>>
>>
>>
>> On Oct 14, 2009, at 3:31 PM, Steve Lianoglou wrote:
>>
>>> Hi,
>>>
>>> I was wondering if there was a way to store the results of a
>>> computationally expensive "getter" call on an S4 object, so that it is
>>> only calculated once for each object.
>>>
>>> Trivial example: let's say I want to cache the "expensive" area
>>> calculation of a square object.
>>>
>>> setClass("Square",
>>>  representation(
>>>    length='numeric',
>>>    width='numeric',
>>>    area='numeric'
>>>  ),
>>>  prototype(
>>>    length=0,
>>>    width=0,
>>>    area=-1
>>>  )
>>> )
>>>
>>> setGeneric("area", function(x) standardGeneric("area"))
>>> setMethod("area", "Square",
>>> function(x) {
>>>  if (x at area == -1) {
>>>    x at area <- x at width * x at height
>>>  }
>>>  x at area
>>> })
>>>
>>> Now the first time I call ``area(my.square)`` it computes
>>> ``my.square at width * my.square at height``, but each subsequent call
>>> returns ``x at area`` since the area computation has already been calc'd
>>> and set for this object.
>>>
>>> Is this possible? I'm guessing the R pass by value semantics is going
>>> to make this one difficult ... is there some S4 reference I missed
>>> that has this type of info from?
>>>
>>> Thanks,
>>> -steve
>>>
>>> -- 
>>> Steve Lianoglou
>>> Graduate Student: Computational Systems Biology
>>>  |  Memorial Sloan-Kettering Cancer Center
>>>  |  Weill Medical College of Cornell University
>>> Contact Info: http://cbio.mskcc.org/~lianos/contact
>>>
>>> ______________________________________________
>>> R-help at r-project.org mailing list
>>> https://stat.ethz.ch/mailman/listinfo/r-help
>>> PLEASE do read the posting guide
>>> http://www.R-project.org/posting-guide.html
>>> and provide commented, minimal, self-contained, reproducible code.
>>
> 
> -- 
> Steve Lianoglou
> Graduate Student: Computational Systems Biology
>   |  Memorial Sloan-Kettering Cancer Center
>   |  Weill Medical College of Cornell University
> Contact Info: http://cbio.mskcc.org/~lianos/contact
> 
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide
> http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.


-- 
Martin Morgan
Computational Biology / Fred Hutchinson Cancer Research Center
1100 Fairview Ave. N.
PO Box 19024 Seattle, WA 98109

Location: Arnold Building M1 B861
Phone: (206) 667-2793