I guess it's a situation where the unit tests were missing to cover an
important situation ("important" as in "it's important that the returned
object be always sorted"). ;-)
Cheers,
H.
On 13/10/2021 17:46, Pariksheet Nanda wrote:
Hi Hervé,
On 10/13/21 12:43 PM, Hervé Pagès wrote:
On 12/10/2021 15:43, Pariksheet Nanda wrote:
The function in question is:
replace_unstranded <- function (gr) {
idx <- strand(gr) == "*"
if (length(idx) == 0L)
^^^^^^^^^^^^^^^^^
Not related to the "internal logical NA value has been modified" error
but shouldn't you be doing '!any(idx)' instead of 'length(idx) == 0L'
here?
Indeed. Although in a roundabout way the result somehow satisfied the
unit tests, idx is a poor choice of name because it's really a mask, and
your suggestion of OR-ing the mask FALSE values with any() is more
intuitive. The name is_unstranded might be less cryptic than mask.
Applying your suggestion of the correct condition uncovered a bug where
return(gr) was returning the unsorted value, which is inconsistent with
the behavior of the final statement returns a sorted value. So changed
to return(sort(gr)) for a consistent contract.
Fixed in f6892ea
Best,
H.
return(gr)
sort(c(
gr[! idx],
`strand<-`(gr[idx], value = "+"),
`strand<-`(gr[idx], value = "-")))
}
Pariksheet
--
Hervé Pagès
Bioconductor Core Team
hpages.on.git...@gmail.com
_______________________________________________
Bioc-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/bioc-devel
