Dependencies during installation of pyroSAR

[sid-ks]

Hi, just a quick question.

Why does pyroSAR install matplotlib, jupyter, notebook and other such dependencies when doing a pip3 install?

Is there a way to make it more lean, with only pure dependencies of pyroSAR being installed?

If so, can you tell me how, if not, please explain why these are required.

Im quite a fan of pyroSAR and how actively it is maintained and updated. Thanks for your efforts! :)

[sid-ks]

Hi, just a quick question.

Why does pyroSAR install matplotlib, jupyter, notebook and other such dependencies when doing a pip3 install?

Is there a way to make it more lean, with only pure dependencies of pyroSAR being installed?

If so, can you tell me how, if not, please explain why these are required.

Im quite a fan of pyroSAR and how actively it is maintained and updated. Thanks for your efforts! :)

Any comments on this please? Thank you. :)

[johntruckenbrodt]

Hi @sid-ks, thanks for addressing this issue. pyroSAR depends on spatialist for general spatial data handling and several other convenience functionalities. This package has grown to be quite a large toolbox over the years and some functionality is demonstrated in a Jupyter notebook spatialist_tutorial.ipynb. The dependencies are mainly needed for a visualization class RasterViewer, which is part of spatialist and whose use is also demonstrated in the notebook.
So basically, these dependencies are not directly needed by pyroSAR. However, excluding them would mean I would have to split up spatialist to outsource the visualization components. I am developing pyroSAR and spatialist almost alone and the more packages I need to maintain the more work needs to be done, which unfortunately I cannot justify at the moment.
I agree though that making the package more lean is important. I will leave this open and try to find time for it.

[sid-ks]

Wow, really appreciate your time and efforts. I would be very glad to help out in someway, but its quite hectic at work even for me, and also being a beginner I wouldn't wanna slow down your schedule.
Thanks for keeping it open, and I look forward to the time when this can be implemented.

Great package ! Have a nice day! :)

[johntruckenbrodt]

Thanks mate! I had a quick search of how to solve this with little effort. With pip you have the PEP 508 convention. For the conda package the best approach seems to be to split spatialist into spatialist-base (core functionality) and spatialist (meta package with optional packages). This is for example done by matplotlib.
If you like you could modify the requirements files of spatialist and send me a pull request. There are two, but I think they could also be combined into one using the PEP 508 convention. For example by adding two sections testing and visualization. Once this is done I can set up the conda packages.