R devel - Oct 2012 - False positive note about ambiguous dispatch

Hi,

In the situation below the note issued by the dispatch algo doesn't
seem right:

   setClass("A", representation(stuff="complex"))
   setAs("ANY", "A", function(from) new("A",
stuff=as.complex(from)))

   > as(6, "A")
   An object of class "A"
   Slot "stuff":
   [1] 6+0i

   > as(6L, "A")
   Note: Method with signature ?numeric#A? chosen for function ?coerce?,
  target signature ?integer#A?.
    "ANY#A" would also be valid
   An object of class "A"
   Slot "stuff":
   [1] 6+0i

First thing is that the note is in contradiction with the output of
selectMethod (seems like it's the method with signature "ANY#A"
that
was chosen, not the method with signature "numeric#A"):

   > selectMethod("coerce", c("integer", "A"))
   Method Definition:

   function (from, to = "A", strict = TRUE)
   new("A", stuff = as.complex(from))

   Signatures:
           from      to
   target  "integer" "A"
   defined "ANY"     "A"

Second, and more importantly, there is no need for a note in the first
place because there is no ambiguity at all (the method for "numeric#A"
itself is inherited from "ANY#A"):

   > selectMethod("coerce", c("numeric", "A"))
   Method Definition:

   function (from, to = "A", strict = TRUE)
   new("A", stuff = as.complex(from))

   Signatures:
           from      to
   target  "numeric" "A"
   defined "ANY"     "A"

showMethod gives me an overview of the situation (which is much
simpler/cleaner than what the note is suggesting):

   > showMethods("coerce")
   Function: coerce (package methods)
   from="ANY", to="A"
   ...
   from="integer", to="A"
       (inherited from: from="ANY", to="A")
   ...
   from="numeric", to="A"
       (inherited from: from="ANY", to="A")

Finally, if I switch the order (i.e. if I do 'as(6L, "A")'
before
'as(6, "A")'), I don't get the note.

I get this with R 2.15.1 and a recent R devel (2012-09-22 r60777).
It looks like a regression because I was not able to reproduce with
R <= 2.14.

Thanks,
H.


-- 
Herv? Pag?s

Program in Computational Biology
Division of Public Health Sciences
Fred Hutchinson Cancer Research Center
1100 Fairview Ave. N, M1-B514
P.O. Box 19024
Seattle, WA 98109-1024

E-mail: hpages at fhcrc.org
Phone:  (206) 667-5791
Fax:    (206) 667-1319

The as() function does some non-standard stuff, and it appears in this 
example to set up methods for coerce() in a way that fools the 
disambiguation code (which is newer than the as() code, I think).

Pending a rewrite of as() (not imminent) the Note has been deactivated 
for the coerce() generic. (rev 60859)

Thanks for the catch.
   John


On 10/1/12 3:20 PM, Herv? Pag?s wrote:
> Hi,
>
> In the situation below the note issued by the dispatch algo doesn't
> seem right:
>
>    setClass("A", representation(stuff="complex"))
>    setAs("ANY", "A", function(from) new("A",
stuff=as.complex(from)))
>
>    > as(6, "A")
>    An object of class "A"
>    Slot "stuff":
>    [1] 6+0i
>
>    > as(6L, "A")
>    Note: Method with signature ?numeric#A? chosen for function ?coerce?,
>   target signature ?integer#A?.
>     "ANY#A" would also be valid
>    An object of class "A"
>    Slot "stuff":
>    [1] 6+0i
>
> First thing is that the note is in contradiction with the output of
> selectMethod (seems like it's the method with signature
"ANY#A" that
> was chosen, not the method with signature "numeric#A"):
>
>    > selectMethod("coerce", c("integer",
"A"))
>    Method Definition:
>
>    function (from, to = "A", strict = TRUE)
>    new("A", stuff = as.complex(from))
>
>    Signatures:
>            from      to
>    target  "integer" "A"
>    defined "ANY"     "A"
>
> Second, and more importantly, there is no need for a note in the first
> place because there is no ambiguity at all (the method for
"numeric#A"
> itself is inherited from "ANY#A"):
>
>    > selectMethod("coerce", c("numeric",
"A"))
>    Method Definition:
>
>    function (from, to = "A", strict = TRUE)
>    new("A", stuff = as.complex(from))
>
>    Signatures:
>            from      to
>    target  "numeric" "A"
>    defined "ANY"     "A"
>
> showMethod gives me an overview of the situation (which is much
> simpler/cleaner than what the note is suggesting):
>
>    > showMethods("coerce")
>    Function: coerce (package methods)
>    from="ANY", to="A"
>    ...
>    from="integer", to="A"
>        (inherited from: from="ANY", to="A")
>    ...
>    from="numeric", to="A"
>        (inherited from: from="ANY", to="A")
>
> Finally, if I switch the order (i.e. if I do 'as(6L,
"A")' before
> 'as(6, "A")'), I don't get the note.
>
> I get this with R 2.15.1 and a recent R devel (2012-09-22 r60777).
> It looks like a regression because I was not able to reproduce with
> R <= 2.14.
>
> Thanks,
> H.
>
>