Simplify the installation procedure.
The current installation procedure of EnPT is quite complex and currently also does not include the downloads of the atmospheric lookup tables of SICOR. Apart from that, git-lfs must be installed on the host machine (installation requires admin rights). We have to simplify that for future releases.
Arosics, geoarray and py_tools_ds are on conda-forge now, i.e., can now be installed with a one-liner (conda install -c conda-forge arosics
, see here). This would also be a nice thing for EnPT and seems to be not too much of work because:
- conda-forge auto-creates the feedstock repositories for all dependencies that are not on conda-forge, yet
- It only requires that all the dependencies are at least on PyPI, the setup.py scripts are running properly and all dependencies are correctly declared.
- For future releases, conda-forge uses a bot that auto-creates a pull request in the feedstock repo on conda-forge to bump its version and upload the updated package files.
- conda-forge builds the repo code automatically on Linux, Windows and OSX and for multiple Python version if desired, i.e., the installation process is automatically tested in various environments (which could also be a partial solution for https://gitext.gfz-potsdam.de/EnMAP/sicor/-/issues/56)
Needed steps are:
-
ensure a properly working setup.py script for SICOR -
find a solution for the table downloads of SICOR -
create conda-forge feedstock repositories for: -
SICOR -
sensormapgeo -
pyrsr -
EnPT
-
-
Customize the meta.yaml recipe files for all these packages if needed
The feedstock repository creation should be done all at once as soon as EnPT is added to conda-forge.