Slicer in a python pipeline (but not in the slicer python interpretor)

Development
1

I would like to use slicer in a python pipeline that will be executed on a large cluster in a batch mode.

I succeed to execute the python code on my laptop:

  • thanks to the Slicer3D JupyterKernel kernel extention.
  • thanks to a external python env, throw jupyter (after having executing this command jupyter-kernelspec install "/home/jacquemi/usr/local/Slicer-5.0.2-linux-amd64/NA-MIC/Extensions-30822/SlicerJupyter/share/Slicer-5.0/qt-loadable-modules/JupyterKernel/Slicer-5.0" --replace --user)

But I did not succeed to install slicer (using pip or whatever) in an external Python virtualenv without using jupyter. As I need to execute my pipeline into a cluster in batch mode, I cannot use neither Jupyter nor the 3D slicer python interpreter

So is there a way to install slicer Python library in a virtual
venv and to load (and use) it in a simple python session (not a jupyter one) ?

This documentation indicates that : “Python scripting and development of new Slicer modules in Python does not require building the application either”
.

2

No, Slicer needs to use Slicer’s python build for all the dependencies to work. You should be able to run Slicer from a directory available to your cluster node. You can use xvfb to get an X session if it’s not available on your node. This requires some memory (real or virtual) on the node, but this is usually manageable and small compared to a large computing task requirement.

That means you can use Slicer’s python to develop Slicer modules, but you still need Slicer itself.

As a future goal we’d like to support installing parts of Slicer as packages for standard python systems but this isn’t possible currently.