hpages
commented
4 months ago This functionality was added last week (in SparseArray 1.5.15):
> library(SparseArray)
> SVT_SparseArray(dim=c(8000, 2500, 300, 5), type="integer")
<8000 x 2500 x 300 x 5 SparseArray> of type "integer" [nzcount=0 (0%)]:
,,1,1
[,1] [,2] [,3] [,4] ... [,2497] [,2498] [,2499] [,2500]
[1,] 0 0 0 0 . 0 0 0 0
[2,] 0 0 0 0 . 0 0 0 0
... . . . . . . . . .
[7999,] 0 0 0 0 . 0 0 0 0
[8000,] 0 0 0 0 . 0 0 0 0
...
,,300,5
[,1] [,2] [,3] [,4] ... [,2497] [,2498] [,2499] [,2500]
[1,] 0 0 0 0 . 0 0 0 0
[2,] 0 0 0 0 . 0 0 0 0
... . . . . . . . . .
[7999,] 0 0 0 0 . 0 0 0 0
[8000,] 0 0 0 0 . 0 0 0 0See ?SVT_SparseArray for more information.
You'll need Bioconductor 3.20 (current devel) for that. Use BiocManager::install(version="devel") to upgrade your installation to BioC devel.
H.
Requesting a simple
emptySparseArray(dim = c(x_1, ..., x_n), type = desired_type)functionality. This is possible as a straightforward wrapper of a function that already exists in the package.I need to pre-allocate an enormous empty 4D-array that is too large to fit in RAM, so I can't use sparse matrix constructors (not high enough dimension) or dense array constructors (too big for RAM). I need the storage type to be integer for memory efficiency.
If I use
x=poissonSparseArray(..., lambda=0), thenx$typeprints outinteger, but the same is not true forrandomSparseArray(..., density=0).It is counterintuitive not to be able to initialize an "empty" (all zeros) array and the easiest solution would be to write a wrapper around poissonSparseArray to be able to create empty sparse arrays of arbitrary dimension. There is likely an even simpler solution as well.
For now I am going to use
poissonSparseArray(..., lambda=0). If this is suboptimal or if I am missing something, then I apologize, but I have hunted through the vignettes and manual for quite some time.