Skip to content

S3 and S4 clash

3 messages · Spencer Graves, Brian Ripley

Spencer Graves
# 
Hello: 

      How can I work around the conflict between the S3 and S4 
illustrated in the example below?  I'm developing a package that 
requires a function in 'stats4', but when 'stats4' is attached, it 
breaks my AIC function.  I could give my AIC function another name so it 
no longer uses the generic dispatch, but I wonder if there is another way. 

      Thanks,
      Spencer Graves
################################
bar.tmp <- structure(1, class = "bar")
AIC.bar <- function(object, ..., k=2) {
  3
}

 > AIC(bar.tmp)
[1] 3
 > library(stats4)
 > AIC(bar.tmp)
Error in UseMethod("logLik") : no applicable method for "logLik"
 > detach("package:stats4")
 > AIC(bar.tmp)
[1] 3
 > objects()
[1] "AIC.bar" "bar.tmp"
 > sessionInfo()
R version 2.6.1 (2007-11-26)
i386-pc-mingw32

locale:
LC_COLLATE=English_United States.1252;LC_CTYPE=English_United 
States.1252;LC_MONETARY=English_United 
States.1252;LC_NUMERIC=C;LC_TIME=English_United States.1252

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base     

loaded via a namespace (and not attached):
[1] stats4_2.6.1
Brian Ripley
# 
I'd say that was pretty clearly a bug in stats4 (which as I recall was 
needed to get around the scoping awkwardnesses of S4).

But could you not write a logLik method for your class?  E.g.
[1] -4.283185
[1] -4.283185
On Thu, 6 Dec 2007, Spencer Graves wrote:

            

      
      
      
    

        

      
      
    

        

  
    

      
      
    

  


  


      

    

    
5 days later

    

      
      

        


  

        

      Spencer Graves
    

    

      
      #
    

  


  

      
Dear Prof. Ripley: 

      Thanks for your help.  I very much appreciate your great 
generosity in both your substantive contributions to the code and 
documentation for R and in responding to so many questions, helping to 
make R what it is today. 

      In the real application that led to the toy example below, I had 
defined

            AIC.factanal <- function(object, ..., k=c(2, "BIC")){ ... }

      1.  If I understand your suggestion, I'd be better NOT defining an 
"AIC.factanal" but instead define only "logLik.factanal" and let methods 
dispatch go through "AIC.default" to "logLik.factanal".  Is this correct? 

      2.  Is it inappropriate to supply other optional values for an 
argument in a generic function like this? 

      3.  If it is OK to specify options for "k" like this, how would 
you suggest I do it?  Use the S4 standard? 

      Thanks again. 
      Spencer Graves

        
Prof Brian Ripley wrote: