PS: I've committed a workaround that seems to fix the symptoms ( rev.
47020/1 for 2.9.0 and 2.8-patched), but it would be good to fix the real
problem.
John Chambers wrote:
> There were a bunch of red herrings in your example, but the end result
> is at least as weird.
>
> Near as I can figure, it's a bug in the evaluator when assigning a local
> function call object that is then modified. I'll look a bit at the
> code, but someone more expert in the innards of the evaluator might be
> needed.
>
> All you need is to do one of the setGeneric() calls. The clue is to
> look at a utility function in the methods package that constructs a
> function call to .Method.
>
> The bug is that after the first call, the definition of the utility
> function has been modified to contain the results of the call. So all
> the arguments from successive calls accumulate!
>
> The minimal example is below. Look at the "call <-" line the second
> time the function .makeDotsCall is printed: it has the arguments from
> order() in the expression on the right hand side.
>
> At a guess, for some reason object "call" isn't marked correctly in the
> NAMED() field, so the evaluation of, e.g., call$na.last <-
> as.name("na.last") overwrites the expression in the definition of
> .makeDotsCall.
>
> John
>
> > methods:::.makeDotsCall
> function (formals)
> {
> call <- quote(.Method(...))
> if (length(formals) > 1) {
> idots <- match("...", formals)
> for (what in formals[-idots]) {
> eval(substitute(call$NAME <- as.name(WHAT), list(NAME =
> as.name(what),
> WHAT = what)))
> }
> }
> call
> }
> <environment: namespace:methods>
>
> > setGeneric("order", signature="...",
> + function (..., na.last=TRUE, decreasing=FALSE)
> + standardGeneric("order"))
> Creating a generic for "order" in package ".GlobalEnv"
> (the supplied definition differs from and overrides the implicit
> generic in package "base": Signatures differ: (...), (na.last, decreasing))
> [1] "order"
>
> > methods:::.makeDotsCall
> function (formals)
> {
> call <- quote(.Method(..., na.last = na.last, decreasing = decreasing))
>
>
> Michael Lawrence wrote:
>
>> There seems to be a bug arising when using multiple S4 generics with "..."
>> as the signature.
>>
>> The following code works as expected:
>>
>> ##############################################################
>>
>>
>>
>>> setGeneric("rbind", function(..., deparse.level=1)
>>>
>>>
>> standardGeneric("rbind"),
>> + signature = "...")
>> Creating a generic for "rbind" in package ".GlobalEnv"
>> (the supplied definition differs from and overrides the implicit generic
>> in package "base": Signatures differ: (...), (deparse.level))
>> [1] "rbind"
>>
>>
>>> rbind(1)
>>>
>>>
>> [,1]
>> [1,] 1
>>
>>
>>> rbind
>>>
>>>
>> standardGeneric for "rbind" defined from package ".GlobalEnv"
>>
>> function (..., deparse.level = 1)
>> standardGeneric("rbind")
>> <environment: 0x83862f4>
>> Methods may be defined for arguments: ...
>> Use showMethods("rbind") for currently available ones.
>>
>>
>>> removeGeneric("rbind")
>>>
>>>
>> [1] TRUE
>>
>>
>>> rbind
>>>
>>>
>> function (..., deparse.level = 1)
>> .Internal(rbind(deparse.level, ...))
>> <environment: namespace:base>
>>
>>
>>> setGeneric("order", signature="...",
>>>
>>>
>> + function (..., na.last=TRUE, decreasing=FALSE)
>> + standardGeneric("order"))
>> Creating a generic for "order" in package ".GlobalEnv"
>> (the supplied definition differs from and overrides the implicit generic
>> in package "base": Signatures differ: (...), (na.last, decreasing))
>> [1] "order"
>>
>>
>>> rbind(1)
>>>
>>>
>> [,1]
>> [1,] 1
>>
>>
>>> setGeneric("rbind", function(..., deparse.level=1)
>>>
>>>
>> standardGeneric("rbind"),
>> + signature = "...")
>> Creating a generic for "rbind" in package ".GlobalEnv"
>> (the supplied definition differs from and overrides the implicit generic
>> in package "base": Signatures differ: (...), (deparse.level))
>> [1] "rbind"
>>
>>
>>> rbind(1)
>>>
>>>
>> [,1]
>> [1,] 1
>>
>> ##########################################################
>>
>> That is, rbind(1) always returns the expected result. The rbind() generic is
>> set, then removed, the order() generic is set, and the rbind() generic is
>> set again. No problems.
>>
>> But in a new session, this code breaks, where the rbind() generic is not set
>> before the order() generic:
>>
>> ########################################################
>>
>>
>>
>>> setGeneric("order", signature="...",
>>>
>>>
>> + function (..., na.last=TRUE, decreasing=FALSE)
>> + standardGeneric("order"))
>> Creating a generic for "order" in package ".GlobalEnv"
>> (the supplied definition differs from and overrides the implicit generic
>> in package "base": Signatures differ: (...), (na.last, decreasing))
>> [1] "order"
>>
>>
>>> rbind(1)
>>>
>>>
>> [,1]
>> [1,] 1
>>
>>
>>> setGeneric("rbind", function(..., deparse.level=1)
>>>
>>>
>> standardGeneric("rbind"),
>> + signature = "...")
>> Creating a generic for "rbind" in package ".GlobalEnv"
>> (the supplied definition differs from and overrides the implicit generic
>> in package "base": Signatures differ: (...), (deparse.level))
>> [1] "rbind"
>>
>>
>>> rbind(1)
>>>
>>>
>> Error in .Method(..., na.last = na.last, decreasing = decreasing,
>> deparse.level = deparse.level) :
>> object "na.last" not found
>>
>> ##############################################################
>>
>> It looks like something is trying to send the parameters of order() to an
>> rbind() method. Weird..
>>
>> Thanks,
>> Michael
>>
>>
>>
>>> sessionInfo()
>>>
>>>
>> R version 2.9.0 Under development (unstable) (--)
>> i686-pc-linux-gnu
>>
>> locale:
>> C
>>
>> attached base packages:
>> [1] stats graphics grDevices utils datasets methods base
>>
>> [[alternative HTML version deleted]]
>>
>> ______________________________________________
>> R-devel@r-project.org mailing list
>> https://stat.ethz.ch/mailman/listinfo/r-devel
>>
>>
>>
>
> [[alternative HTML version deleted]]
>
> ______________________________________________
> R-devel@r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-devel
>
>
[[alternative HTML version deleted]]
______________________________________________
R-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel
