Ragged arrays can't be read back in

[MaHaWo]

The following code derived from the tutorial fails with an error message related to type conversions:

 z = Zarr.zcreate(Vector{Int}, 4, path="./test")

z[1:3] = [[1,3,5],[4],[7,9,14]];

z[:] # this works fine 

z = nothing 

z = Zarr.zopen("./test")

z[1] # this fails 

the error message reads:

MethodError: Cannot `convert` an object of type Vector{Int64} to an object of type Int64
The function `convert` exists, but no method is defined for this combination of argument types.

Closest candidates are:
  convert(::Type{T}, !Matched::EzXML.NodeType) where T<:Integer
   @ EzXML ~/.julia/packages/EzXML/2xuPL/src/node.jl:36
  convert(::Type{T}, !Matched::EzXML.ReaderType) where T<:Integer
   @ EzXML ~/.julia/packages/EzXML/2xuPL/src/streamreader.jl:74
  convert(::Type{T}, !Matched::AbstractChar) where T<:Number
   @ Base char.jl:185
  ...


Stacktrace:
  [1] setindex!(A::Memory{Int64}, x::Vector{Int64}, i1::Int64)
    @ Base ./genericmemory.jl:243
  [2] unsafe_copyto!(dest::Memory{Int64}, doffs::Int64, src::Memory{Vector{Int64}}, soffs::Int64, n::Int64)
    @ Base ./genericmemory.jl:153
  [3] unsafe_copyto!
    @ ./genericmemory.jl:133 [inlined]
  [4] _copyto_impl!
    @ ./array.jl:308 [inlined]
  [5] copyto!
    @ ./array.jl:294 [inlined]
  [6] copyto!
    @ ./array.jl:319 [inlined]
  [7] zuncompress!(data::Vector{Int64}, compressed::Vector{UInt8}, c::Zarr.BloscCompressor, f::Tuple{Zarr.VLenArrayFilter{Int64}})
    @ Zarr ~/.julia/packages/Zarr/3QSdj/src/Compressors.jl:37
  [8] uncompress_raw!(a::Vector{Int64}, z::Zarr.ZArray{Int64, 1, Zarr.BloscCompressor, Zarr.DirectoryStore}, curchunk::Vector{UInt8})
    @ Zarr ~/.julia/packages/Zarr/3QSdj/src/ZArray.jl:261
  [9] uncompress_to_output!(aout::Vector{Int64}, output_base_offsets::Tuple{Int64}, z::Zarr.ZArray{Int64, 1, Zarr.BloscCompressor, Zarr.DirectoryStore}, chunk_compressed::Vector{UInt8}, current_chunk_offsets::Tuple{Int64}, a::Vector{Int64}, indranges::Tuple{UnitRange{Int64}})
    @ Zarr ~/.julia/packages/Zarr/3QSdj/src/ZArray.jl:270
 [10] readblock!(aout::Vector{Int64}, z::Zarr.ZArray{Int64, 1, Zarr.BloscCompressor, Zarr.DirectoryStore}, r::CartesianIndices{1, Tuple{UnitRange{Int64}}})
    @ Zarr ~/.julia/packages/Zarr/3QSdj/src/ZArray.jl:178
 [11] readblock!
    @ ~/.julia/packages/Zarr/3QSdj/src/ZArray.jl:247 [inlined]
 [12] getindex_disk
    @ ~/.julia/packages/DiskArrays/WgDSb/src/indexing.jl:54 [inlined]
 [13] getindex(a::Zarr.ZArray{Int64, 1, Zarr.BloscCompressor, Zarr.DirectoryStore}, i::Int64)
    @ DiskArrays ~/.julia/packages/DiskArrays/WgDSb/src/indexing.jl:312
 [14] top-level scope

I'm using Julia v1.11.6 with Zarr.jl v0.9.4

[publik-void]

I have the exact same issue here on Julia v1.11.6, Zarr.jl v0.9.6, Ubuntu 24.04.3 LTS.

[publik-void]

Python's zarr (v2.18.7) can read the array back correctly, so it seems that the data is written correctly and the problem is on the reading side of Zarr.jl.

[publik-void]

It seems that the fix would be to define:

Zarr.sourcetype(::Zarr.VLenArrayFilter{T}) where T = Vector{T}

I tested this with a range of numeric types (bools, uints, ints, floats, complexes) to see if both ragged and non-ragged arrays are read back correctly, and it appears to work. I would leave it up to the package maintainers to verify it fully.