gdal library is too complicated
Closed this issue · 7 comments
I suggest we migrate to imageio which is far simpler to install and maintain. It does not need the complicated binary / gcc built dependencies of gdal.
I might not be able to recommend this for the processing of satellite images etc because gdal does give access to the geo transform data but for our demonstrations in stripy it works fine.
There are some examples in the lavavu texture mapping for reference sake.
But cartopy requires gdal libraries...
Yes, and scipy are deprecating their imread function, so alternatives like imageio are now necessary!
Python gdal is not really needed for cartopy ... the missing functionality if we switch to imageio is the geotransform extraction from a geotiff. That's the kind of thing that is not important for demonstrating stripy even if it is vital for the eventual applications.
Ah - I was getting confused with GEOS. But that's a much simpler installation than GDAL. The geotransformation stuff is probably not even that important considering all our geotiffs use a global extent. Removing GDAL should trim down the docker image size too.
One last comment and then I am going to close this. The PIL library does allow us to extract the metadata from a tif and access the geotransforms that are embedded. It is not as elegant as the gdal solution but is helpful if we want to include a demo that uses non-global data.