colnames-related concerns about summarizedexperiment HOT 5 CLOSED

Thanks Vince.

About 1):

Interestingly we have a similar situation with mcols() on Vector derivatives:

ir <- IRanges(1:3, 5)
mcols(ir) <- DataFrame(stuff=runif(3))
rownames(mcols(ir)) <- letters[1:3]
rownames(mcols(ir))
# NULL

This should be taken care of in SummarizedExperiment 1.17.2 (see commit 495ab00).

Note that when you did colnames(assay(se1)) <- letters[4:6] in your example above the colnames of the assay actually got modified. It's just that you didn't see them when you brought the assay back because by default assay() doesn't bring back the real colnames. You need to use withDimnames=FALSE for that:

> assay(se1, withDimnames=FALSE)
  d e f
A 1 3 5
B 2 4 6

About 2):

Having the colData() setter reorder the rows of the supplied DataFrame "might" sound like a nice feature but it breaks the general assumption that the getter will bring back the original DataFrame so I'm not convinced it's a good idea. Anyway if you want to pursue this please submit the proposal in a separate issue.

from summarizedexperiment.

The latest change broke some edge-case behavior:

library(SummarizedExperiment)
example(SummarizedExperiment) # create some SE's.

assay(se) <- assay(se) # OK.

empty <- se[,0]
assay(empty) <- assay(empty) # fail

from summarizedexperiment.

Thanks @LTLA (when following up on a closed issue, you should re-open)

from summarizedexperiment.

Should be addressed in SummarizedExperiment 1.17.3 (see commit 7ee31d4).

Please re-open or start a new issue if you run into other problems related with commit 495ab00 ("By default assays()/assay() setters reject inconsistent dimnames"). Thanks!

from summarizedexperiment.

Thanks @hpages. I don't have the authority to re-open issues on this repo, so yes, I should have started a new one.

from summarizedexperiment.