[Bioc-devel] a pattern to be avoided? mcols(x)$y <- z

[Hervé Pagès]

Hi Vince,

This issue was reported here a couple of weeks ago:

   https://github.com/Bioconductor/GenomicRanges/issues/11

Internally $<- uses something like:

   do.call(DataFrame, list(DF1, DF2))

to combine the metadata columns. However in some situations
the do.call(DataFrame, list(...)) form is **very** inefficient
compared to the more direct DataFrame(...) form:

   library(S4Vectors)
   DF1 <- DataFrame(a=Rle(11:1999, 1011:2999), b=5)
   DF2 <- DataFrame(c=Rle(12:2000, 1011:2999))
   system.time(DF12 <- do.call(DataFrame, list(DF1, DF2)))
   #   user  system elapsed
   #  4.476   0.000   4.476
   system.time(DF12b <- DataFrame(DF1, DF2))
   #   user  system elapsed
   #  0.002   0.000   0.001
   identical(DF12, DF12b)
   # [1] TRUE

@Michael: Any idea what's going on?

Thanks,
H.


On 10/03/2018 07:01 AM, Vincent Carey wrote:
> The following comes up in use of Fdb.InfiniumMethylation.hg19::getPlatform
> 
> 
> debug: mcols(GR)$channel <- Rle(as.factor(mcols(GR)$channel450))
> 
> Browse[3]> system.time(uu <- Rle(as.factor(mcols(GR)$channel450)))
> 
>     user  system elapsed
> 
>    0.020   0.003   0.022
> 
> Browse[3]> system.time(mcols(GR)$channel <-
> Rle(as.factor(mcols(GR)$channel450)))
> 
>     user  system elapsed
> 
>   47.263   0.067  47.373
> 
> Browse[3]> GR$channel[1]
> 
> factor-Rle of length 1 with 1 run
> 
>    Lengths:    1
> 
>    Values : Both
> 
> Levels(3): Both Grn Red
> 
> Browse[3]> system.time(GR$channel <- Rle(as.factor(mcols(GR)$channel450)))
> 
>     user  system elapsed
> 
>    0.058   0.006   0.065
> 
> 
> Presumably the mcols()$<- copies/rewrites a lot of data needlessly?
> 

-- 
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 using fredhutch.org
Phone:  (206) 667-5791
Fax:    (206) 667-1319