Proposal: String representation/RleMatrix about variantannotation HOT 6 OPEN

Ideally we would use a light weight representation (like the proposed RleListMatrix) that is more computable than a delimited string.

from variantannotation.

I'm not sure whether or not the Rle benefit will work across lists, but if it can then that may solve all of the problems.

from variantannotation.

I think it would in this case, because of how the Rle compression helps the character case.

from variantannotation.

Does it work in that context because a matrix is just a vector with a dim so you can actually compress long runs of sequential elements? I'm not sure I see how that manifests for the 'vector of lists' case but perhaps something clever can be done in the unlisting/relisting sense to take advantage of that.

from variantannotation.

It will be efficient as long as the RleList is a CompressedList, meaning that internally there is just a flat vector that is lazily partitioned into list elements.

from variantannotation.

In that case it may even be slightly more performant than the string version as there would be even fewer unique values over which a run may compress. Nice!

from variantannotation.