[BioC] IRanges coverage integer limit?

[Nicolas Delhomme]

Hi,

I've just discovered that the IRanges coverage function would "overflow" without warnings. Below is an example that reproduce it:

library(IRanges)
rngs <- IRanges(c(1:100),width=100)
coverage(rngs)

'integer' Rle of length 199 with 199 runs
  Lengths:  1  1  1  1  1  1  1  1  1  1  1 ...  1  1  1  1  1  1  1  1  1  1
  Values :  1  2  3  4  5  6  7  8  9 10 11 ... 10  9  8  7  6  5  4  3  2  1

coverage(rngs,weight=1e9)

'integer' Rle of length 200 with 200 runs
  Lengths:           1           1           1 ...           1           1
  Values :  1000000000  2000000000 -1294967296 ...  1000000000           0

runValue(coverage(rngs,weight=1e9))
  [1]  1000000000  2000000000 -1294967296  -294967296   705032704  1705032704
  [7] -1589934592  -589934592   410065408  1410065408 -1884901888  -884901888
...

Clearly, the third position that has a coverage of 3 (not weighted) has a 3e9 weighted one which is > 2^31 (signed integer limit on most machine). I'm just surprised that it is silently ignored.

For NGS, getting a bp coverage > 2^31 is unlikely, although I've already seen extremely high coverage for Ribosomal-like protein that were only 10 order of magnitude away (~2M X). This limits the ranges of weights that can be used (weight as of now can only be integers), i.e. a weight of 100 would already be borderline. 

Is there a way around this, coverage being such a very handy function? I understand that weight being integers probably makes computation faster, but what could be the overhead of allowing numeric instead? And I don't mind looking under the hood if that helps.

Cheers,

Nico

sessionInfo()
R version 2.15.1 (2012-06-22)
Platform: x86_64-apple-darwin9.8.0/x86_64 (64-bit)

locale:
[1] C/UTF-8/C/C/C/C

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

other attached packages:
[1] IRanges_1.15.17    BiocGenerics_0.3.0

loaded via a namespace (and not attached):
[1] stats4_2.15.1 tools_2.15.1 


---------------------------------------------------------------
Nicolas Delhomme

Genome Biology Computational Support

European Molecular Biology Laboratory

Tel: +49 6221 387 8310
Email: nicolas.delhomme at embl.de
Meyerhofstrasse 1 - Postfach 10.2209
69102 Heidelberg, Germany