setClass question

Seems that methods are being attributed a bit strangely in undoc().
Here's what I get with, I think, current r-devel and SparseM.

R> library(tools)
R> undoc("methods")
R> library(SparseM)
[1] "SparseM library loaded"
R> undoc("SparseM")
Undocumented code objects:
  %x% as.matrix diag diag<- diff t
Undocumented S4 methods:
  \S4method{initialize}{traceable} \S4method{initialize}{signature}
  \S4method{initialize}{environment}
R> undoc("methods")
Undocumented S4 methods:
  \S4method{initialize}{matrix.csr} \S4method{initialize}{matrix.coo}
Warning message:
package:methods is required by package:SparseM (still attached) in:
detach(pos = pos)

Notice that methods for initialize() seem to be cross-attributed to
"methods" and "SparseM".  (And despite the message, there IS
documentation for both  method?initialize("traceable") and
method?initialize("matrix.csr"), but of course in the other package.)

If it's relevant, after the undoc("methods"), the order of the search
list has been altered, with methods in position 2 (maybe related to the
warning message when it was detached from its previous position?).

The "Undocumented code objects" messages are "real", but not your
fault--we'll have to figure what to do about them.  You created methods
for "diag" (I assume), which is a non-generic in base.  The result is to
create a generic version of diag(), which is assigned in your package.
(In general, with namespaces, I don't think it's possible to assign the
generic back into the original package, even just into the exported
environment.)

R> find("diag")
[1] "package:SparseM" "package:base"

This creates a formal "conflict" but in my understanding not one that
should be reported, nor should you be expected to document the function
diag(), as opposed to methods created for it.

[Just got back to e-mail after 3 days away.  Will look at this and your
other mail, probably not until tomorrow.]

Thanks,
 John

Roger Koenker wrote:

I have another S4 methods query -- this time about undocumented objects:
I'm now getting the following warning from R CMD check SparseM:

* checking for missing documentation entries ... WARNING
Undocumented code objects:
  %x% as.matrix diag diag<- diff t
Undocumented S4 methods:
  \S4method{coerce}{ANY,array} \S4method{coerce}{ANY,call}
  \S4method{coerce}{ANY,character} \S4method{coerce}{ANY,complex}
  \S4method{coerce}{ANY,environment} \S4method{coerce}{ANY,expression}
  \S4method{coerce}{ANY,function} \S4method{coerce}{ANY,integer}
  \S4method{coerce}{ANY,list} \S4method{coerce}{ANY,logical}
  \S4method{coerce}{ANY,matrix} \S4method{coerce}{ANY,name}
  \S4method{coerce}{ANY,numeric} \S4method{coerce}{ANY,single}
  \S4method{coerce}{ANY,ts} \S4method{coerce}{ANY,vector}
  \S4method{coerce}{ANY,NULL}
  \S4method{coerce}{vector,matrix.diag.csr}
  \S4method{diff}{matrix.csr} \S4method{initialize}{traceable}
  \S4method{initialize}{signature} \S4method{initialize}{environment}

A couple of these are directly related to the package, but the rest
seem to be items that I wouldn't have thought that I should be responsible
for documenting.  I'm wondering whether the coerce items are related to
my earlier query.  Even supposing that they were due to some blunder
of mine, I'm still puzzled by the initialize items which were present
even before I began to mess with my new setClass strategy.  Again,
many thanks for any suggestions.

url:    www.econ.uiuc.edu/~roger/my.html        Roger Koenker
email   rkoenker@uiuc.edu                       Department of Economics
vox:    217-333-4558                            University of Illinois
fax:    217-244-6678                            Champaign, IL 61820

--
John M. Chambers                  jmc@bell-labs.com
Bell Labs, Lucent Technologies    office: (908)582-2681
700 Mountain Avenue, Room 2C-282  fax:    (908)582-3340
Murray Hill, NJ  07974            web: http://www.cs.bell-labs.com/~jmc