R help - Dec 2003 - expressing functions

# Why does expressing one function

require(ctest)
t.test

# return only

function (x, ...)
UseMethod("t.test")
<environment: namespace:ctest>

# but expressing another function

shapiro.test

# returns more complete code?

function (x)
{
     DNAME <- deparse(substitute(x))
     x <- sort(x[complete.cases(x)])
     n <- length(x)
     if (n < 3 || n > 5000)
         stop("sample size must be between 3 and 5000")
     rng <- x[n] - x[1]
     if (rng == 0)
         stop("all `x[]' are identical")
     if (rng < 1e-10)
         x <- x/rng
     n2 <- n%/%2
     sw <- .C("swilk", init = FALSE, as.single(x), n, n1 =
as.integer(n),
         as.integer(n2), a = single(n2), w = double(1), pw = double(1),
         ifault = integer(1), PACKAGE = "ctest")
     if (sw$ifault && sw$ifault != 7)
         stop(paste("ifault=", sw$ifault, ". This should not
happen"))
     RVAL <- list(statistic = c(W = sw$w), p.value = sw$pw, method = 
"Shapiro-Wilk normality test",
         data.name = DNAME)
     class(RVAL) <- "htest"
     return(RVAL)
}
<environment: namespace:ctest>

-- 

Richard E. Remington III
Statistician
KERN Statistical Services, Inc.
PO Box 1046
Boise, ID 83701
Tel: 208.426.0113
KernStat.com

On Tue, 9 Dec 2003, Remington, Richard wrote:
> # Why does expressing one function
> 
> require(ctest)
> t.test
> 
> # return only
> 
> function (x, ...)
> UseMethod("t.test")
> <environment: namespace:ctest>
> 
> # but expressing another function
> 
> shapiro.test
> 
> # returns more complete code?
[...]

False hypothesis: both are the complete code.

You are not understanding (S3-style) generic functions: see any good book
on R/S.

(`An Introduction to R' is based on notes that predate them, but they are
covered in more detail in the draft R Language manual.  `The reader is
referred to the official references for a complete discussion of this
mechanism.' which I think means Chambers & Hastie, 1992,)

-- 
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

Remington, Richard wrote:
> # Why does expressing one function
> 
> require(ctest)
> t.test
> 
> # return only
> 
> function (x, ...)
> UseMethod("t.test")
> <environment: namespace:ctest>
> 
> # but expressing another function
> 
> shapiro.test
> 
> # returns more complete code?
> 
> function (x)
> {
>     DNAME <- deparse(substitute(x))
>     x <- sort(x[complete.cases(x)])
>     n <- length(x)
>     if (n < 3 || n > 5000)
>         stop("sample size must be between 3 and 5000")
...

Short answer:  Unless you're programming your own functions, you don't 
need to worry about that.

Long answer:  Because the first is generic - it looks at what kind of 
data you're testing (two vectors, a formula, whatever, ...) and calls 
the appropriate sub-function.  shapiro.test does not; it just takes one 
data format, and  stops in its tracks if that's not what you've
provided.

The ideas behind this are documented in "Writing R Extensions" 
(R-exts.pdf) which is supplied with binary R distributions, and is 
available from CRAN.  See chapter 6, "Generic functions and methods",
in
the version that accompanies R-1.8.1.

Cheers

Jason
-- 
Indigo Industrial Controls Ltd.
http://www.indigoindustrial.co.nz
64-21-343-545
jasont at indigoindustrial.co.nz

Perhaps what should be added to the previous answers is
that you can find out where the real work is done like
this:

require(ctest)
t.test
methods(t.test)
ctest:::t.test.default
ctest:::t.test.formula

If the class of the first argument to t.test is formula
then t.test.formula gets invoked so that's where the
real work is done; otherwise, t.test.default
gets invoked so that's where the real work is done.

--- 
 
Remington, Richard wrote:
> # Why does expressing one function
> 
> require(ctest)
> t.test
> 
> # return only
> 
> function (x, ...)
> UseMethod("t.test")
> <environment: namespace:ctest>
> 
> # but expressing another function
> 
> shapiro.test
> 
> # returns more complete code?
> 
> function (x)
> {
> DNAME <- deparse(substitute(x))
> x <- sort(x[complete.cases(x)])
> n <- length(x)
> if (n < 3 || n > 5000)
> stop("sample size must be between 3 and 5000")