replace_missing should just do nothing if missingval is allready missing

[rafaqz]

Currently this throws:

MethodError: Cannot `convert` an object of type Missing to an object of type Float64

Which makes no sense

[ayushpatnaikgit]

Hi, I have the following Raster:

julia> dc
84×155×1×123 Raster{Float32,4} unnamed with dimensions: 
  X Mapped Float64[72.7250001018, 72.72916676850001, …, 73.0666667712, 73.0708334379] ForwardOrdered Explicit Intervals crs: EPSG mappedcrs: EPSG,
  Y Mapped Float64[19.495832889299994, 19.491666222599996, …, 18.858332884200003, 18.8541662175] ReverseOrdered Explicit Intervals crs: EPSG mappedcrs: EPSG,
  Band Categorical Int64[1] ReverseOrdered,
  Ti Sampled Int64[201204, 201205, …, 202205, 202206] ForwardOrdered Irregular Points
with missingval: missing

I want to do something like:

dc[1,1,1,1] = missing

I wish to do this because the list of missing observations is provided in a separate file, and I have to manually mark the missing observations.

But get the error that you have mentioned above.
Any suggestions?

I feel if missingval is missing, the Raster should automatically be Union{Missing, something}

Apologies, if this issue wasn't the most suitable place to ask this question.

[rafaqz]

Thanks I just forgot about this! I thought it had been fixed by #102

Edit: sorry I misunderstood the issue. If the missingval is missing but the array eltype is not missing, we should change the array eltype in replace_missing. Probably it shouldn't ever get in that state either - the missingval in your case should be nothing. I'm guessing this has been saved as a netcdf file at some point?

Probably good to make a new issue, pointing to this one, showing how the fix for this problem created another problem.

Also: please include a MWE with your issue, I want to fix the root cause of this - which I think is during loading a netcdf from file.

[ayushpatnaikgit]

Thanks, Rafael,
I'll open a new issue and share the .nc file.