Do not treat Zarr fill_value as CF _FillValue for coordinate variables

Author: Alexander-Barth

src/variable.jl

@@ -10,9 +10,17 @@ CDM.name(v::ZarrVariable) = Zarr.zname(v.zarray)
 CDM.dimnames(v::ZarrVariable) = Tuple(reverse(v.zarray.attrs["_ARRAY_DIMENSIONS"]))
 CDM.dataset(v::ZarrVariable) = v.parentdataset
 
+function _iscoordvar(v)
+    dn = dimnames(v)
+    if length(dn) == 0
+        return false
+    end
+    return name(v) == first(dn)
+end
+
 function CDM.attribnames(v::ZarrVariable)
     names = filter(!=("_ARRAY_DIMENSIONS"),keys(v.zarray.attrs))
-    if !isnothing(v.zarray.metadata.fill_value)
+    if !isnothing(v.zarray.metadata.fill_value) && !_iscoordvar(v)
         push!(names,"_FillValue")
     end
     return names
@@ -44,6 +52,17 @@ haschunks(v::ZarrVariable) = haschunks(v.zarray)
 eachchunk(v::CFVariable{T,N,<:ZarrVariable}) where {T,N} = eachchunk(v.var)
 haschunks(v::CFVariable{T,N,<:ZarrVariable}) where {T,N} = haschunks(v.var)
 
+"""
+
+    defVar(ds::ZarrDataset,name::SymbolOrString,vtype::DataType,dimensionnames; chunksizes=nothing, attrib = Dict(), fillvalue = nothing)
+
+
+Create a variable `name` in the dataset `ds` with the type `vtype` and the dimension `dimensionnames`.
+
+For coordinate variables, fill values will be used a background value of undefined chunks and not as missing value as coordinate variables cannot have the `_FillValues` in the CF convension as in Zarr v2 format a `fill_value` does not necessarily indicate a missing value.
+
+See also `CommonDataModel.defVar` for more information.
+"""
 function CDM.defVar(ds::ZarrDataset,name::SymbolOrString,vtype::DataType,dimensionnames; chunksizes=nothing, attrib = Dict(), fillvalue = nothing, kwargs...)
     @assert iswritable(ds)
 
@@ -70,6 +89,8 @@ function CDM.defVar(ds::ZarrDataset,name::SymbolOrString,vtype::DataType,dimensi
         kwargs...
     )
 
-    return ZarrVariable{vtype,ndims(zarray),typeof(zarray),typeof(ds)}(
+    zv = ZarrVariable{vtype,ndims(zarray),typeof(zarray),typeof(ds)}(
         zarray,ds)
+
+    return ds[name]
 end

test/Project.toml

@@ -5,9 +5,11 @@ Dates = "ade2ca70-3891-5945-98fb-dc099432e06a"
 DiskArrays = "3c3547ce-8d99-4f5e-a174-61eb10b00ae3"
 NCDatasets = "85f8d34a-cbdd-5861-8df4-14fed0d494ab"
 Test = "8dfed614-e22c-5e08-85e1-65c5234f0b40"
+Zarr = "0a941bbe-ad1d-11e8-39d9-ab76183a1d99"
 
 [compat]
 Aqua = "0.8"
 CommonDataModel = "0.3.6"
 NCDatasets = "0.14"
 julia = "1"
+Zarr = "0.9.2"

test/runtests.jl

@@ -6,5 +6,6 @@ using ZarrDatasets
     include("test_multifile.jl")
     include("test_write.jl")
     include("test_groups.jl")
+    include("test_fillvalue.jl")
     include("test_aqua.jl")
 end

test/test_fillvalue.jl

@@ -0,0 +1,31 @@
+using ZarrDatasets
+using Test
+using Zarr
+
+fname = tempname()
+mkdir(fname)
+
+# CF coordinate variable with Zarr fill_value set
+
+store = Zarr.DirectoryStore(fname)
+zg = zgroup(store, "")
+zarray = zcreate(
+    Int, zg, "lon", 3;
+    fill_value = 9999,
+    attrs = Dict("_ARRAY_DIMENSIONS" => ("lon",)))
+
+ds = ZarrDataset(fname)
+@test eltype(ds["lon"]) == Int
+
+fname = tempname()
+mkdir(fname)
+
+ds = ZarrDataset(fname,"c")
+
+# variable which is not a CF coordinate variable
+v2 = defVar(ds,"foo",Int,(),fillvalue=9999)
+@test eltype(v2) == Union{Missing,Int}
+
+v3 = defVar(ds,"bar",Int16[2,3,4],("time",),fillvalue=9999)
+@test eltype(v3) == Union{Missing,Int16}
+close(ds)