kmerize with hg38 seems to create an integer overflow

[omsai]

> library(kmap)
> mappable("hg38")
 Removing non-standard DNA bases and chopping into 36-mers ...
 Error in .Call2("Integer_fancy_mseq", lengths, offset, rev, PACKAGE = "S4Vectors") :
   negative length vectors are not allowed
 Timing stopped at: 0.872 0.012 120.9
 > traceback()
 27: .Call(.NAME, ..., PACKAGE = PACKAGE)
 26: .Call2("Integer_fancy_mseq", lengths, offset, rev, PACKAGE = "S4Vectors")
 25: S4Vectors:::fancy_mseq(width(x), offset = start(x) - 1L)
 24: as.integer(IRanges(rep(0L, length(n)), width = n))
 23: as.integer(IRanges(rep(0L, length(n)), width = n))
 22: .local(x, width, step, ...)
 21: slidingWindows(ranges(x), width, step)
 20: slidingWindows(ranges(x), width, step)
 19: .local(x, width, step, ...)
 18: slidingWindows(granges(views), kmer)
 17: slidingWindows(granges(views), kmer)
 16: eval(lhs, parent, parent)
 15: eval(lhs, parent, parent)
 14: slidingWindows(granges(views), kmer) %>% unlist()
 13: kmerize(., kmer = kmer)
 12: function_list[[k]](value)
 11: withVisible(function_list[[k]](value))
 10: freduce(value, `_function_list`)
 9: `_fseq`(`_lhs`)
 8: eval(quote(`_fseq`(`_lhs`)), env, env)
 7: eval(quote(`_fseq`(`_lhs`)), env, env)
 6: withVisible(eval(quote(`_fseq`(`_lhs`)), env, env))
 5: stddna_from_genome(bsgenome %>% as("BSgenomeViews"), BPPARAM) %>%
        kmerize(kmer = kmer)
 4: eval(code, env)
 3: system.time(eval(code, env))
 2: timeit(sprintf("Removing non-standard DNA bases and chopping into %d-mers",
        kmer), views <- stddna_from_genome(bsgenome %>% as("BSgenomeViews"),
        BPPARAM) %>% kmerize(kmer = kmer))
 1: mappable("hg38")
 > 

This is a fairly messy stack trace. Will remove magrittr::%>% to clean up.

[omsai]

The fault is from kmerize(), particularly from window.start

> views
 BSgenomeViews object with 1649 views and 0 metadata columns:
                  seqnames              ranges strand                       dna
                     <Rle>           <IRanges>  <Rle>            <DNAStringSet>
      [1]             chr1 [  10001,   207666]      * [TAACCCTAAC...AAGTACAGAT]
      [2]             chr1 [ 257667,   297968]      * [GATTCATGGC...CCAGCTGGAT]
      [3]             chr1 [ 347969,   535988]      * [GATCCCAATT...GTTAGAACTG]
      [4]             chr1 [ 585989,  2702781]      * [GATCCTTGAA...AACAGAATTC]
      [5]             chr1 [2746291, 12954384]      * [AGCCTGTACA...AAAGGAATTC]
      ...              ...                 ...    ...                       ...
   [1645] chrUn_KI270756v1      [11350, 79590]      * [TTCCATTCGA...TCCATTCTAT]
   [1646] chrUn_KI270757v1      [    1, 10219]      * [GTTTTTGAAC...GGAATGGAAT]
   [1647] chrUn_KI270757v1      [14265, 16789]      * [TTAATGGAAT...GGAGTGGAGT]
   [1648] chrUn_KI270757v1      [20006, 44211]      * [GGGAACGGAG...CTGAGAAACT]
   [1649] chrUn_KI270757v1      [46351, 71251]      * [TTTTGATTTG...CACATCACAA]
   -------
   seqinfo: 455 sequences (1 circular) from hg38 genome
 > kmap:::kmerize(views)
 Error in .Call2("Integer_fancy_mseq", lengths, offset, rev, PACKAGE = "S4Vectors") :
   negative length vectors are not allowed
 > slidingWindows(ranges(views), width = 36, step = 1)
 Error in .Call2("Integer_fancy_mseq", lengths, offset, rev, PACKAGE = "S4Vectors") :
   negative length vectors are not allowed
 > traceback()
 8: .Call(.NAME, ..., PACKAGE = PACKAGE)
 7: .Call2("Integer_fancy_mseq", lengths, offset, rev, PACKAGE = "S4Vectors")
 6: S4Vectors:::fancy_mseq(width(x), offset = start(x) - 1L)
 5: as.integer(IRanges(rep(0L, length(n)), width = n))
 4: as.integer(IRanges(rep(0L, length(n)), width = n))
 3: .local(x, width, step, ...)
 2: slidingWindows(ranges(views), width = 36, step = 1)
 1: slidingWindows(ranges(views), width = 36, step = 1)
 > getMethod("slidingWindows", "Ranges")
 Method Definition:

 function (x, width, step = 1L, ...)
 {
     .local <- function (x, width, step = 1L)
     {
         if (!isSingleNumber(width))
             stop("'width' must be a single, non-NA number")
         if (!isSingleNumber(step))
             stop("'step' must be a single, non-NA number")
         if (any(width < 0L))
             stop("some 'width' are negative")
         if (any(step < 0L))
             stop("some 'step' are negative")
         n <- ceiling(pmax(width(x) - width, 0L)/step) + 1L
         window.starts <- as.integer(IRanges(rep(0L, length(n)),
             width = n)) * step + 1L
         windows <- restrict(IRanges(window.starts, width = width),
             end = rep(width(x), n))
         windows.abs <- shift(windows, rep(start(x), n) - 1L)
         relist(windows.abs, PartitioningByWidth(n))
     }
     .local(x, width, step, ...)
 }
 <environment: namespace:IRanges>

 Signatures:
         x
 target  "Ranges"
 defined "Ranges"
 >

IRanges(rep(0L, length(n)), width = n) works, but exploding it with as.integer() raises the error.

Will try splitting the granges in kmerize into more manageable chunks. One can always split a view range that's too large into overlapping ranges.