Hervé Pagès

Hi, I've implemented the ReshapedHDF5Array class in HDF5Array 1.13.3 (HDF5Array 1.13.3 belongs to BioC 3.10, which is the current development version of Bioconductor). The ReshapedHDF5Array class is a DelayedArray subclass...

Oops, my bad sorry. I meant to have this latest version of HDF5Array (1.13.3) depend on DelayedArray >= 0.11.3, not on DelayedArray >= 0.11.4. Sorry for that. (FWIW I have...

This 2-step reshaping is a consequence of a restriction on the kind of reshaping supported by the `ReshapedHDF5Array()` constructor. The restriction is documented in `?ReshapedHDF5Array`: Also please note that arbitrary...

I see. Interesting use case. Note that if the 1000 3D datasets were stored as a single 4D dataset (512x512x512x1000) then things would be much easier. It would just be...

Something that maybe would help is the ability to "stack" the 1000 3D datasets into a 4D dataset. That would need to be a new array operation so we would...

Only reason I don't check for duplicate entries in the `nzindex` slot at the moment is because it's kind of expensive. Also these duplicates should not occur in the context...

I'm just seeing your last post now. Yes, nice trick to use a combination of `aperm`/`arbind` to achieve binding along an arbitrary dimension. That's what my proposed `abind()` above would...

Also, try to use intermediate realizations to circumvent some of the current limitations. For example after you've stacked the 1000 datasets together, you won't be able to reshape the result...

SparseArraySeed objects are now deprecated in favor of COO_SparseArray objects (in **DelayedArray** >= 0.31.5), but it seems that COO_SparseArray objects had the same issue. The issue is now fixed in...

Sounds like an interesting project. Thanks for the offer. I would love to get more involved but I'm overbooked with too many other projects going on at the moment. Also...