Description
Issue Description
The current behavior of the install.py script in the tudatpy causes issues when adding new modules. Specifically, using the -e flag only creates symbolic links for .py and .pyi files that exist at the time of installation. As a result, any new module added after the initial installation requires a full uninstall (using uninstall.py) and reinstall to create symbolic links for the new files. This implementation issue hinders the iterative development workflow.
Steps to Reproduce
- Install the
tudatpypackage usinginstall.py -e. - Add a new module or script to
tudatpy. - Observe that the new module is not linked or recognized in the environment.
- Uninstall and reinstall
tudatpyto correctly register the new module.
Expected Behavior
Newly added Python modules should be automatically recognized without requiring reinstallation, as long as the symbolic links point to the appropriate subdirectories.
Current Behavior
New modules are not automatically included via symbolic links, requiring a full reinstall of the package to make them accessible.
Possible Solution or Implementation
@alfonsoSR proposes:
- Modify
install.pyto create symbolic links not just to the roottudatpydirectory, but to its internal submodules (tudatpy/astro,tudatpy/numerical_simulation, etc.). This would enable updates to individual modules without needing a complete reinstall, as long as the module structure is preserved. - Alternatively, mirror the original behavior where
kernel.soand source files reside in the same directory (tudat-bundle/tudatpy), but note that this was changed to conform to preferences against storing compiled extensions next to source files.
Additional Context
The .so files are excluded from GitHub via .gitignore, and the current setup requires placing kernel.so in the tudat-bundle/build directory. A workaround involving creating the tudatpy directory in the conda prefix and linking each script individually can be error-prone.
Metadata
Metadata
Assignees
Type
Projects
Status