PeteHaitch
commented
7 years ago Stepping through the validity method:
> dim(object)
[1] 28220561 183
> system.time(bsseq:::.checkAssayNames(object, c("Cov", "M")))
user system elapsed
0.001 0.000 0.000
> system.time(class(rowRanges(object)) != "GRanges")
user system elapsed
0 0 0
> system.time(is.null(colnames(object)))
user system elapsed
0 0 0
> system.time(min(assay(object, "M")) < 0)
user system elapsed
7136.154 47650.472 55196.698
> system.time(min(assay(object, "Cov")) < 0)
user system elapsed
10898.65 30457.32 41457.43
> system.time(max(assay(object, "M") - assay(object, "Cov")) > 0.5)
# Killed after > 100,000 seconds
> system.time(!is.null(rownames(assay(object, "M"))) ||
+ !is.null(rownames(assay(object, "Cov"))) ||
+ ("coef" %in% assayNames(object) && !is.null(rownames(assay(object, "coef")))) ||
+ ("se.coef" %in% assayNames(object) && !is.null(rownames(assay(object, "se.coef")))))
user system elapsed
0.007 0.000 0.006
The
validObject()method for BSseq objects is painfully slow when the assays are HDF5-backed. Could make the validity method conditional on the backend of the assay, but it'd be nice to the same, fast validity checking regardless of the assay backend.