Vectorization as an independant package
The vectorisation builder relies on the "grass" library to do some of its processing. As mentioned in #628, there are currently no alternatives to grass in python.
One possibility for iota2 to move forward with the deployment of a version with modern versions of python would be to put the vectorisation part aside, in an independent package and freeze its dependencies. Vectorisation could no longer be used in the same environment as the one containing the updated version of iota2.
This raises some questions:
- the creation of a conda package dedicated to vecto with fixed dependencies for python 3.9 must necessarily go through a new repository. Indeed, maintaining everything in the same repository seems complicated (the API of dependencies can change etc...).
- The documentation in the iota2 repository should no longer mention vectorization (as it is no longer accessible).
- find a name for this package/project "scaled_vectorisation", "large_semantic_vectorisation", "iota2_vectorisation"...?
- same license as iota2?
- If no contributions are envisaged for this code, why not just say that vectorisation is accessible in iota2 for such and such versions? (vectorisation has however recently benefited from compatibility with Slurm).
Some things need to be done before creating the conda de vecto package
- Make an inventory of the dependencies (libraries) needed to create the package.
- Make an inventory of the code in iota2 needed for the vecto.
- answer the above questions