[BioC] BUG: median(IRanges::Rle(NA), na.rm=TRUE)) should return NA

[Hervé Pagès]

Hi Malcolm,

Thanks for reporting this. A fix is on its way (in IRanges 1.19.30,
will be in devel). With this version:

   > Rle()[FALSE]
   logical-Rle of length 0 with 0 runs
     Lengths:
     Values :

More generally, most Vector derivatives now support subsetting by a
logical vector longer than 'x' as long as all the out of bounds
subscript elements are FALSE:

   > IRanges(1:2, 4)[c(FALSE, TRUE, FALSE)]
   IRanges of length 1
       start end width
   [1]     2   4     3

With this change:

   > median(Rle(NA), na.rm=TRUE)
   [1] NA

and:

   > median(Rle(integer(0)))
   [1] NA           # NA_integer_

   > median(Rle(logical(0)))
   [1] NA           # logical NA

   > median(Rle())
   [1] NA           # logical NA

The reason the last one returns a logical NA is because the 'Rle()'
call is equivalent to 'Rle(logical(0))'.

Cheers,
H.


On 09/04/2013 07:53 AM, Cook, Malcolm wrote:
> A little further sleuthing suggests that perhaps the more general issues is that
>
>> Rle()[FALSE]
> Error in Rle()[FALSE] : subscript out of bounds
> # but
>> vector()[FALSE][NA]
> [1] NA
>
>
>
>   >-----Original Message-----
>   >From: Cook, Malcolm
>   >Sent: Wednesday, September 04, 2013 9:28 AM
>   >To: 'bioconductor at r-project.org'; 'Hervé Pagès'; 'lawrence.michael at gene.com'
>   >Cc: Chen, Guangbo; Johnston, Jeffrey
>   >Subject: BUG: median(IRanges::Rle(NA),na.rm=TRUE)) should return NA
>   >
>   >Hi,
>   >
>   >I think the Rle functions should preserve semantics from stats.
>   >
>   >Here is at least one case where they do not:
>   >
>   >> median(Rle(NA),na.rm=TRUE)
>   >Error in x[FALSE] : subscript out of bounds
>   >
>   ># but:
>   >
>   >> median(c(NA),na.rm=TRUE)
>   >[1] NA
>   >
>   ># some other border cases:
>   >
>   >> median(c())
>   >NULL
>   ># arguably correct, if you read the docs and appreciate that as(NA,class(c())) is NULL
>   >
>   >> median(Rle())
>   >Error in x[FALSE] : subscript out of bounds
>   ># should probably be NULL
>   >
>   >
>   >Agreed?
>   >
>   >Can be fixed?
>   >
>   >Thanks!
>   >
>   >~ Malcolm Cook
>   >Computational Biology / Shilatifard Lab - Stowers Institute for Medical Research - Kansas City
>   >
>

-- 
Hervé Pagès

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