Array datasets are always written as chunked.

[orenbenkiki]

It seem by default the Python anndata package writes contiguous datasets.

Perhaps an option to control this?

[orenbenkiki]

Thanks!

[orenbenkiki]

BTW, compress=9 is deprecated (according to the latest HDF5 package I'm using, anyway). The warning I get says:

┌ Warning: `compress=9` keyword option is deprecated, use `deflate=9` instead
│   caller = setproperty!(p::HDF5.DatasetCreateProperties, name::Symbol, val::UInt8) at properties.jl:59
└ @ HDF5 ~/.julia/packages/HDF5/Ws1wH/src/properties.jl:59

[ilia-kats]

I'm not sure where this is comping from, to be honest. There are exactly two places in the Muon code that request compression from HDF5, and both are using deflate=....

[orenbenkiki]

In the Muon code I see a lot of compress=..., mainly passed to write and write_metadata... In HDF5 the warning occurs whenever trying to set a compress property of an object. Perhaps because I'm using the latest release, which is old?

[ilia-kats]

Yes, but those are our functions that have nothing to do with HDF5.jl. Interaction with HDF5 happens here and here.

[orenbenkiki]

Using the head revision, I get several WARNING: Wrapping Vararg directly in UnionAll is deprecated when compiling Muon, but no warnings when actually using it. So the compress warning was due to using the ancient released version (which is what I got when using pkg> add Muon). I've switched the project to using the head revision (which also allows me to use the newly added compress flag when writing to a file - nice). I'm also voting for #25 though :-)