Clean up NULL / Nothing / Missing semantics

[yeesian]

Came up in a few places (e.g. #177) so opening this issue to make sure we have a systematic approach across the entirety of this package. See also https://discourse.julialang.org/t/whats-the-difference-between-null-na-null-nothing/6468/6.

[yeesian]

See discussion in #179 (comment), in particular this FAQ entry:

To represent missing data in the statistical sense (NA in R or NULL in SQL), use the missing object.

As described by @visr,

The NULL in the example of #177 seems to represent missing data to me. Especially if we see Features as rows in a Table / records in a dataset.

findfieldindex, if it was written in Julia, should probably return nothing instead of -1, like findfirst. But the index truly doesn't exist, whereas the data can be viewed as missing (the field name exists after all).

A counterpoint by @rafaqz in #179 (comment) is that

I'm not sure about the choice to return missing when there is no nodata value. I found nothing more appropriate as it suggests that there simply is no nodata value. ArchGDAL is not also returning an array of Union{T,Missing}.

[rafaqz]

This is the most important part for me:

help?> nothing
search: nothing Nothing isnothing

  nothing

  The singleton instance of type Nothing, used by convention when there is no value to return (as in a C void
  function) or when a variable or field holds no value.

help?> missing
search: missing Missing missingval missingmask MissingException ismissing nonmissingtype skipmissing

  missing

  The singleton instance of type Missing representing a missing value.

Returning missing when there is no nodatavalue in a raster is not really aligned with the julia meanings of missing and nothing. There are simply no nodata values in the file at all, as opposed to some nodata values with the value of missing. This is like findfirst returning nothing when the value simply isn't in the searched object.

[visr]

I agree with @rafaqz that getnodatavalue should return nothing if there is no nodata value. So in #192 (comment) we're not disagreeing, just talking about different functions. Per function we'll have to decide which one is more appropriate.

Actually in this commit: 83b7025

I would argue that findfieldindex should return nothing, because there is actually no index, whereas getfield should return missing, since the field exists for other features, it's just missing/NULL for that feature. That is what I meant with #179 (comment).

[yeesian]

I would argue that findfieldindex should return nothing, because there is actually no index, whereas getfield should return missing, since the field exists for other features, it's just missing/NULL for that feature.

Thanks! See 36c5969