v0.14 cannot create variable from an Int64 array with missing values

[sjdaines]

Describe the bug

With NCDatasets v0.14.1, creating a netcdf variable from a Julia Int64 array with missing values causes an ERROR: InexactError

This used to work on NCDatasets v0.12.x

To Reproduce

julia> import NCDatasets
julia>  ds = NCDatasets.NCDataset("test.nc","c")
julia> NCDatasets.defDim(ds, "mydim", 2)

julia> v = NCDatasets.defVar(ds, "test_array_i64_missing", [missing, 2], ("mydim", ))
ERROR: InexactError: trunc(Int32, -9223372036854775806)

(see full output below)

Expected behavior

Should create a variable v with Datatype: Union{Missing, Int64} (Int64)

Examples that do work as expected:

julia> v = NCDatasets.defVar(ds, "test_array_i64", [1, 2], ("mydim", )) # OK
test_array_i64 (2)
  Datatype:    Int64 (Int64)
  Dimensions:  mydim

julia> v = NCDatasets.defVar(ds, "test_array_i32_missing", [missing, Int32(2)], ("mydim", )) # OK
test_array_i32_missing (2)
  Datatype:    Union{Missing, Int32} (Int32)
  Dimensions:  mydim
  Attributes:
   _FillValue           = -2147483647
julia> Array(v)
2-element Vector{Union{Missing, Int32}}:
  missing
 2

Environment

• operating system: Windows 10
• Julia version: julia 1.10.0, official binaries from https://julialang.org/downloads/
• Output of the julia command versioninfo()

julia> versioninfo()
Julia Version 1.10.0
Commit 3120989f39 (2023-12-25 18:01 UTC)
Build Info:
 Official https://julialang.org/ release
Platform Info:
 OS: Windows (x86_64-w64-mingw32)
 CPU: 4 × Intel(R) Core(TM) i5-6300U CPU @ 2.40GHz
 WORD_SIZE: 64
 LIBM: libopenlibm
 LLVM: libLLVM-15.0.7 (ORCJIT, skylake)
 Threads: 1 on 4 virtual cores

• NCDatasets version: v0.14.1
• Output of using Pkg; Pkg.status(mode=PKGMODE_MANIFEST)

Full output

julia> v = NCDatasets.defVar(ds, "test_array_i64_missing", [missing, 2], ("mydim", ))
ERROR: InexactError: trunc(Int32, -9223372036854775806)
Stacktrace:
  [1] throw_inexacterror(f::Symbol, ::Type{Int32}, val::Int64)
    @ Core .\boot.jl:634
  [2] checked_trunc_sint
    @ .\boot.jl:656 [inlined]
  [3] toInt32
    @ .\boot.jl:693 [inlined]
  [4] Int32
    @ .\boot.jl:783 [inlined]
  [5] nc_put_att(ncid::Int32, varid::Int32, name::String, data::Int32)
    @ NCDatasets C:\Users\sd336\.julia\packages\NCDatasets\mKQDc\src\netcdf_c.jl:564 [inlined]
  [6] defAttrib(ds::NCDatasets.Variable{Int64, 1, NCDatasets.NCDataset{Nothing, Missing}}, name::String, data::Int64)
    @ NCDatasets C:\Users\sd336\.julia\packages\NCDatasets\mKQDc\src\attributes.jl:85
  [7] setindex!(a::CommonDataModel.Attributes{NCDatasets.Variable{Int64, 1, NCDatasets.NCDataset{Nothing, Missing}}}, data::Int64, name::String)
    @ CommonDataModel C:\Users\sd336\.julia\packages\CommonDataModel\SVxJP\src\attribute.jl:112
  [8] defVar(ds::NCDatasets.NCDataset{…}, name::String, vtype::DataType, dimnames::Tuple{…}; chunksizes::Nothing, shuffle::Bool, deflatelevel::Nothing, checksum::Nothing, fillvalue::Nothing, nofill::Bool, typename::Nothing, attrib::OrderedCollections.OrderedDict{…})
    @ NCDatasets C:\Users\sd336\.julia\packages\NCDatasets\mKQDc\src\cfvariable.jl:146
  [9] defVar
    @ C:\Users\sd336\.julia\packages\NCDatasets\mKQDc\src\cfvariable.jl:99 [inlined]
 [10] _defVar(ds::NCDatasets.NCDataset{…}, name::String, data::Vector{…}, nctype::Type, vardimnames::Tuple{…}; attrib::Vector{…}, kwargs::@Kwargs{})
    @ CommonDataModel C:\Users\sd336\.julia\packages\CommonDataModel\SVxJP\src\variable.jl:173
 [11] _defVar
    @ CommonDataModel C:\Users\sd336\.julia\packages\CommonDataModel\SVxJP\src\variable.jl:123 [inlined]
 [12] #defVar#5
    @ CommonDataModel C:\Users\sd336\.julia\packages\CommonDataModel\SVxJP\src\variable.jl:101 [inlined]
 [13] defVar(ds::NCDatasets.NCDataset{Nothing, Missing}, name::String, data::Vector{Union{Missing, Int64}}, dimnames::Tuple{String})
    @ CommonDataModel C:\Users\sd336\.julia\packages\CommonDataModel\SVxJP\src\variable.jl:91
 [14] top-level scope
    @ REPL[10]:1
Some type information was truncated. Use `show(err)` to see complete types.

[Alexander-Barth]

Thank you for submitting the detailed bug report.
I am currently testing this fix:
4638a21

[Alexander-Barth]

All tests do pass. I will make a new release.
https://github.com/Alexander-Barth/NCDatasets.jl/actions/runs/7849851817
Please re-open of the issue persists.

[sjdaines]

Fantastic ! Thank you !