R devel - Jun 1998 - extra arguments to generic functions & bug in model.frame

R developers,

2 things:  a bug in model.frame and a question about setup of generic 
functions. 
 
   I don't understand the following behavior for generic functions:
Suppose I'm working with the cats data in the MASS library and I want to 
create a formula object to model Hwt on Sex:
  # This works:
> formula(Hwt ~ Sex)
Hwt ~ Sex
  # But the following does not:
> formula(Hwt ~ Sex, data = cats)
Error in formula.formula(Hwt ~ Sex, cats) : unused argument to function
  # because I used an extra argument.

The odd thing to me, is that formula() itself allows the
argument:
> formula
function (x, ...) 
UseMethod("formula")

But formula.default does not:
> args(formula.default)
function (x) 

The same thing happens with terms:
> args(terms)
function (x, ...) 
> args(terms.default)
function (x) 

And this caused me a problem when using model.frame() since 
model.frame.default includes the lines:

	    if (!inherits(formula, "terms")) 
	    	    formula <- terms(formula, data = data)

Using 2 arguments in term() is a bug since hitting the line 
will cause an error:

Error in terms.default(formula, data = data) : unused argument to function

  Summary:

1.  There seems to be a bug in model.frame() using terms() with 2 arguments.

2.  What is standard practice for setting up instances of a generic function?
    It seems sensible to me that argument lists should not shrink when 
    going from the generic function to an instance of that function.

Cheers,
Jim

Jim Robison-Cox                 ____________            
Department of Math Sciences    |            |           phone: (406)994-5340
2-214 Wilson Hall               \   BZN, MT |           FAX:   (406)994-1789
Montana State University         |  *_______|
Bozeman, MT 59717                 \_|         e-mail: jimrc@math.montana.edu 


-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-devel mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To:
r-devel-request@stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._

Jim Robison-Cox <jimrc@mathfs.math.montana.edu> writes:
> 1.  There seems to be a bug in model.frame() using terms() with 2
arguments.
Yes.
> 2.  What is standard practice for setting up instances of a generic
function?
>     It seems sensible to me that argument lists should not shrink when 
>     going from the generic function to an instance of that function.
No. There's no limit to what special parameters a specific instance of
a function may implement. So the generic one is generally (x,...).
There's no reason not to allow a specific function to check that it is
getting called with parameters that make sense.

-- 
   O__  ---- Peter Dalgaard             Blegdamsvej 3  
  c/ /'_ --- Dept. of Biostatistics     2200 Cph. N   
 (*) \(*) -- University of Copenhagen   Denmark      Ph: (+45) 35327918
~~~~~~~~~~ - (p.dalgaard@biostat.ku.dk)             FAX: (+45) 35327907
-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-devel mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To:
r-devel-request@stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._

Jim Robison-Cox <jimrc@mathfs.math.montana.edu> writes:
> The odd thing to me, is that formula() itself allows the argument:
> > formula
> function (x, ...) 
> UseMethod("formula")
> 
> But formula.default does not:
> 
> > args(formula.default)
> function (x) 
> 
> The same thing happens with terms:
> > args(terms)
> function (x, ...) 
> > args(terms.default)
> function (x) 
I would say this is a mistake in the definition of terms.default and
formula.default.

It is a good programming practice to ensure that the generic and any
methods have exactly the same calling sequence.  Any such
incompatibilities will have to be eliminated from S-PLUS because S
version 4, the language on which S-PLUS 5.0 will be based, requires
that they have exactly the same calling sequence.  S4 doesn't produce
an error if you define an "old-style" method with a different calling
sequence because it is not recognized as a method.  However, when you
try to use such a method you get an error if the calling sequence does
not match the calling sequence of the generic.

BTW, my version of S-PLUS defines the formula function as
 > formula
 function(object, ...)
 UseMethod("formula")
 > formula.default
 function(object)
We should change the name "x" to "object" to make it easier
for people
to port code.
-- 
Douglas Bates                            bates@stat.wisc.edu
Statistics Department                    608/262-2598
University of Wisconsin - Madison        http://www.stat.wisc.edu/~bates/
-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-devel mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To:
r-devel-request@stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._