[Bioc-devel] My "==" method breaks existing "==" method for signature c("Rle", "vector')

Tue May 27 01:37:29 CEST 2014

Thanks for the suggested work-around, Martin. In order to define the method on the group generic 'Ops' rather than '==' I will need to generalise .MTuples.compare to the 'Arith', 'Compare' and 'Logic' sub-groups listed in ?Ops, won't I? I'll try to implement the comparison methods for MTuples via the 'Ops' group generic but I want to first make sure that I'm not missing out on getting some of these "for free". The rest of this post might be tangential to the issue I first raised, but it also explains why I implemented '==' rather than on the group generic 'Ops'.

The MTuples class extends GRanges:

setClass('MTuples', representation(extraPos = "matrix"), contains = "GRanges", validity = .valid.MTuples)

I defined .MTuples.compare because while MTuples is based on GRanges, the GRanges 'compare' method is inappropriate for MTuples (see https://github.com/PeteHaitch/cometh/blob/8be01b8b37b6288db5493f53188a193a423ab69f/R/methods-MTuples-class.R if interested). Based on reading ?compare, I'd hoped that the binary "parallel" comparison operators (<=, ==, !=, >=, < and >) would work out-of-the-box. But I already found that I needed to include explicit method definitions for '<=' and '=='  because otherwise these (and the other binary comparison operators) would dispatch on the GRanges methods rather than the MTuples methods. This is because the GRanges '==' and '<=' methods are explicitly implemented as:

setMethod("==", signature(e1="GenomicRanges", e2="GenomicRanges"), function(e1, e2) { .GenomicRanges.compare(e1, e2) == 0L })
setMethod("<=", signature(e1="GenomicRanges", e2="GenomicRanges"), function(e1, e2) { .GenomicRanges.compare(e1, e2) <= 0L })

I don't understand why these explicit definitions are necessary; the 'compare' method for GRanges is already defined so why don't these work out-of-the-box, which is what I expected based on my reading of ?compare (specifically, point 1 in the Note sub-section). Because these are explicitly defined, and not simply defined via inheritance, I must also explicitly define '<=' and '==' methods for MTuples (although I do get '>=', '!=', '<' and '>' for free, which is great!). Unfortunately, as you can see, the MTuples '==' method breaks '==' for other signatures. Hope that wasn't too tangential.

Cheers,
Pete

On 27/05/2014, at 2:44 AM, Martin Morgan <mtmorgan at fhcrc.org> wrote:

> On 05/25/2014 09:39 PM, Peter Hickey wrote:
>> The "==" method that I have defined for my class, MTuples, is breaking the "==" method for signature c("Rle", "vector"). I discovered this when working on something quite unrelated, namely, I couldn't resize IRanges with "fixed = end" when my package was loaded. The attached code highlights the initial problem.
>> 
>> The error message and traceback were a little hairy - could someone please help me figure out what's going wrong?
>> 
> 
> I don't have an immediate answer. A work-around is to define the method on the group generic 'Ops' rather than '==', though maybe the generalization of .MTuples.compare to other members of the Ops family is not easy?
> 
> setMethod("Ops", c("MTuples", "MTuples"), function(e1, e2) TRUE)
> 
> Martin
> 
>> Thanks,
>> Pete
>> 
>> ------------------------------------------------------------------------------------------------------------------
>> # Mimics the initial problem
>> # Fresh R session
>> library(IRanges)
>> ir <- IRanges(start = 11, end = rep.int(20, 5))
>> 
>>  # Works when repeatedly called
>> resize(ir, 1, 'end') # Works
>> resize(ir, 1, 'end') # Works
>> 
>> # The method definition that breaks things
>> # I haven't included the MTuples class definition or the .MTuples.compare function
>> # But that shouldn't matter in order to highlight the problem, should it?
>> setMethod("==", c("MTuples", "MTuples"), function(e1, e2) {
>>   .MTuples.compare(e1, e2) == 0L
>> })
>> 
>> # No longer works
>> resize(ir, 1, 'end')
>> ------------------------------------------------------------------------------------------------------------------
>> 
>> 
>> I managed to simplify the reproducible example to the following, but I can't figure out what's going wrong:
>> ------------------------------------------------------------------------------------------------------------------
>> # A simplified example
>> # Fresh R session
>> library(IRanges)
>> fix <- Rle('end', 10)
>> 
>> # Works when repeatedly called
>> fix == 'end' # Works
>> fix == 'end' # Works
>> 
>> # The method definition that breaks things
>> # I haven't included the MTuples class definition or the .MTuples.compare function
>> # But that shouldn't matter in order to highlight the problem, should it?
>> setMethod("==", c("MTuples", "MTuples"), function(e1, e2) {
>>   .MTuples.compare(e1, e2) == 0L
>> })
>> 
>> # Works the first time, but not the second
>> fix == 'end' # Works
>> fix == 'end' # Breaks
>> ------------------------------------------------------------------------------------------------------------------
>> 
>> The same problem occurs if the vector is numeric and not character, e.g. using 7 instead of 'end'.
>> 
>> When this breaks I get the error:
>> ------------------------------------------------------------------------------------------------------------------
>> Error in Rle(values = callGeneric(runValue(e1)[which1], runValue(e2)[which2]),  :
>>   error in evaluating the argument 'values' in selecting a method for function 'Rle': Error in as.character(call[[1L]]) :
>>   cannot coerce type 'builtin' to vector of type 'character'
>> ------------------------------------------------------------------------------------------------------------------
>> 
>> My session info is:
>> ------------------------------------------------------------------------------------------------------------------
>> R version 3.1.0 (2014-04-10)
>> Platform: x86_64-apple-darwin10.8.0 (64-bit)
>> 
>> locale:
>> [1] en_AU.UTF-8/en_AU.UTF-8/en_AU.UTF-8/C/en_AU.UTF-8/en_AU.UTF-8
>> 
>> attached base packages:
>> [1] parallel  stats     graphics  grDevices utils     datasets  methods
>> [8] base
>> 
>> other attached packages:
>> [1] IRanges_1.22.7      BiocGenerics_0.10.0
>> 
>> loaded via a namespace (and not attached):
>> [1] stats4_3.1.0
>> ------------------------------------------------------------------------------------------------------------------
>> 
>> 
>> --------------------------------
>> Peter Hickey,
>> PhD Student/Research Assistant,
>> Bioinformatics Division,
>> Walter and Eliza Hall Institute of Medical Research,
>> 1G Royal Parade, Parkville, Vic 3052, Australia.
>> Ph: +613 9345 2324
>> 
>> hickey at wehi.edu.au
>> http://www.wehi.edu.au
>> 
>> 
>> ______________________________________________________________________
>> The information in this email is confidential and intend...{{dropped:8}}
>> 
>> _______________________________________________
>> Bioc-devel at r-project.org mailing list
>> https://stat.ethz.ch/mailman/listinfo/bioc-devel
>> 
> 
> 
> -- 
> Computational Biology / Fred Hutchinson Cancer Research Center
> 1100 Fairview Ave. N.
> PO Box 19024 Seattle, WA 98109
> 
> Location: Arnold Building M1 B861
> Phone: (206) 667-2793

--------------------------------
Peter Hickey,
PhD Student/Research Assistant,
Bioinformatics Division,
Walter and Eliza Hall Institute of Medical Research,
1G Royal Parade, Parkville, Vic 3052, Australia.
Ph: +613 9345 2324

hickey at wehi.edu.au
http://www.wehi.edu.au

______________________________________________________________________
The information in this email is confidential and intend...{{dropped:6}}