Import rasters to a julia array

[rafaqz]

Thanks for putting this package together.

I have a simple need of import tiff rasters with something better than imagemagic. It seems I can do that with this package but I'm not sure of the simplest way to just get a tiff into some kind of Julia Array. My GDAL experience is limited.

[yeesian]

Assuming you have a .tif file, e.g. by running

shell> wget https://github.com/yeesian/ArchGDALDatasets/blob/master/pyrasterio/RGB.byte.tif?raw=true

The following should work:

img = ArchGDAL.registerdrivers() do
    ArchGDAL.read("RGB.byte.tif") do dataset
        ArchGDAL.read(dataset)
    end
end

and yield

791×718×3 Array{UInt8,3}:
[:, :, 1] =
 0x00  0x00  0x00  0x00  0x00  0x00  0x00  0x00  0x00  0x00  0x00  0x00  0x00  …  0x00  0x00  0x00  0x00  0x00  0x00  0x00  0x00  0x00  0x00  0x00  0x00
 0x00  0x00  0x00  0x00  0x00  0x00  0x00  0x00  0x00  0x00  0x00  0x00  0x00     0x00  0x00  0x00  0x00  0x00  0x00  0x00  0x00  0x00  0x00  0x00  0x00
 0x00  0x00  0x00  0x00  0x00  0x00  0x00  0x00  0x00  0x00  0x00  0x00  0x00     0x00  0x00  0x00  0x00  0x00  0x00  0x00  0x00  0x00  0x00  0x00  0x00
 0x00
...

[rafaqz]

Ok thanks somehow I missed that using read twice would return an array.

[yeesian]

Yeah it's not particularly obvious, and It'll be nice to provide a better set of documentation for working with rasters (issue #40).

The first read is for file IO and gives you a handle to the dataset (that contains metadata) without reading in all the values. The second read is for reading in raster values. They are decoupled for out-of-core processing on enormous rasters (often but not exclusively from satellite data) where you'll process raster blocks a chunk at a time.

[rafaqz]

Decoupling that makes sense. But it might be useful to have another package that further wraps ArchGDAL for these simple cases, so people doing basic IOcan totally avoid GDAL and just get a raster array with one command.

[yeesian]

But it might be useful to have another package that further wraps ArchGDAL for these simple cases

Yeah, ArchGDAL is not meant to be end-user facing. It's meant for other packages to build on top of, see e.g. GeoDataFrames.jl

so people doing basic IOcan totally avoid GDAL and just get a raster array with one command.

I agree with you. Does Images.jl not achieve this in the interim?

[rafaqz]

Sure. I was thinking about writing something simple but I'm really not knowledgeable enough about raster data.

Unfortunately Imagemagic mangles some non-standard raster inputs on import, so its not really ok for production work.