An improved python package installation experience is now available out of the box for extension developers: slicer.packaging.
This feature was merged about a month ago and is available in the preview version of Slicer.
The existing slicer.util.pip_install has been expanded in functionality to support a more interactive experience: a GUI like the one shown below when blocking, or the possibility to run without blocking the application.
Additionally, new functions have been added:
| Function | Purpose |
|---|---|
slicer.packaging.load_requirements(path) |
Parse a requirements.txt into Requirement objects. |
slicer.packaging.load_pyproject_dependencies(path) |
Parse [project.dependencies] from a pyproject.toml |
slicer.packaging.pip_check(reqs) |
Are the requirements already satisfied? |
slicer.packaging.pip_install(...) |
Install, with various modes and hooks. |
slicer.packaging.pip_ensure(reqs, requester="...") |
High-level workflow: check, prompt, install with progress, and offer restart if needed. |
The bottom line for extension developers is: Use slicer.packaging.pip_ensure! If you need lower level control, you can instead use slicer.packaging.pip_check and slicer.packaging.pip_install directly.
See the following for more details:
- the python package management section of the script repository
- the
slicer.packagingAPI docs - the Kitware blog post