[R] 2 setGeneric's, same name, different method signatures

[Martin Morgan]

On 2/14/2013 8:57 AM, Greg Minshall wrote:
> hi.  below is a small test case (hopefully minimal, though i'm still a
> bit confused about initializers).
>
> i would have guessed (and maybe i still would have been right) that one
> could re-use the name of a generic function for functions with different
> numbers of arguments.  in the case below, class A's bB() queries the
> status of a single A object, so bB(A) (where here "A" is an instance of
> class A, just for "clarity") returns a value.  class B's bB() compares
> values in two B objects, and returns the minimum, so bB(B1, B2) are its
> methods.
>
> after loading the file, i see the method for A's bB has disappeared (as
> measured by showMethods("bB"), as well as trying bB(A).  if i have R
> re-parse the setGeneric/setMethod  A's bB(), then B's bB() disappears.
>
> somehow my code, or my model of how things work, is wrong.  any ideas
> what am i missing?
>
> cheers, Greg Minshall
> ----
> setClass("A",
>           representation(
>             x="numeric"));
>
> setMethod(
>    f="initialize",
>    signature="A",
>    definition=function(.Object) {
>      .Object at x <- 23;
>      return(.Object)
>    });
>
> setGeneric("bB", function(me) standardGeneric("bB"));
> setMethod(
>    "bB",
>    signature("A"),
>    definition=function(me) {
>      return (new("B", me at x))});
>
> setClass("B", representation(
>    bx="numeric"));
>
> setMethod(
>    "initialize",
>    signature("B"),
>    definition=function(.Object, x) {
>      .Object at bx <- x;
>      return(.Object);
>    });
>
> setGeneric("bB", function(b1, b2) standardGeneric("bB"));

Hi Greg --

this setGeneric is over-writing the first, as would

f = function() "first"
f = function() "second"
f() # "second"

If you'd like to dispatch on a single argument, then

setGeneric("one", function(x, ...) standardGeneric("one"))
setMethod("one", "A", function(x, ...) "A-method")
setMetohd("one", "B", function(x, y, ...) "B-method")

The '...' in the generic allow you to add arguments that are 'picked off' by 
methods. The user could provide any value for y, not only an object of class "B".

If you'd like to dispatch sometimes on two arguments then

setGeneric("two", function(x, y, ...) standardGeneric("two"))
setMethod("two", c("A", "ANY"), function(x, y, ...) "A,ANY-method")
setMethod("two", c("B", "B"), function(x, y, ...) "B,B-method")

then two(new("A")), two(new("A"), new("A")) and two(new("A"), new("B")) end up 
in A,ANY,two-method while two(new("B"), new("B")) ends up in "B,B,two-method". 
Other combinations are errors. One might instead not define A,ANY but instead

setMethod("two", c("A", "missing"), function(x, y, ...) "A,missing-method")

and then two(new("A"), new("A")) would be an error.

Multiple dispatch is complicated, and perhaps best to avoid if possible.

It's possible to write a generic and methods that dispatch on '...', with the 
requirement that all classes are the same; this in the spirit of comparing two 
B's and returning the smaller; see ?dotsMethods though again this is not a 
trivial use case.

Hope that helps enough.

Martin

> setMethod(
>    "bB",
>    signature("B", "B"),
>    definition=function(b1, b2) {
>      return(new("B", min(b1 at bx, b2 at bx)))});
>
> ______________________________________________
> 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.
>


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