snews2/snoshews

A code to compute the MSW effect in supernovae

SNOWSHEWS: Supernova NeutrinO SHockwave Effects With SNEWPY

Image by dreamstudio.ai

Installation Instructions

1. You will need the packages python-devel (in Linux), pybind11 and setuptools

2. Modify setup.py to use the correct libraries and paths.

3. To compile enter

sudo python3 setup.py install

4. If you don't want to sudo you may want to use the option

--install-lib=destination/directory/

Troubleshooting:

6. You may have to set the PYTHONPATH environment variable to your PWD and/or wherever the SNOSHEWS module was installed

7. If the script cannot find the module you may need to put the *.so library in the same directory as your script. The *.so library is in one of the subfolders in the build directory.

8. SNOSHEWS uses OpenMP. You may want to set the OMP_NUM_THREADS environment variable to a reasonable number for your machine.

snoshews's Issues

add data files as `SNOSHEWS.data` submodule?

The repo includes two sample files (profiles00585_stp.d and profiles00585_Ye.d) which are hard-coded in SNOSHEWS_example.py; so the user has to manually download them and ensure they are in the current working directory.

Unlike in BEMEWS (where the same PREM data files will be used everytime), these data files will need to be different for each progenitor, so users will generally need to provide their own data files. I wonder if it would still be nice to include those files as a submodule[*] to make it easier to run the demo script? Definitely not a high priority, though; we’ve got some more important work to do to get snewpy v2.0, EMEWS and SNOSHEWS out to users first.

[*] I.e. add them to the installable package (sdist or wheel), so users automatically get them when they run pip install. We would then be able to access the files via importlib.resources, similar to the SNOwGLoBES data files.

Publish releases to PyPI

In addition to building SNOSHEWS in GitHub Actions (see #4 and #5), we should add a workflow to automatically publish releases to PyPI. This can reuse much of the machinery developed for BEMEWS in SNEWS2/BEMEWS#10.

Build on Linux and Windows as part of GitHub Actions

In #4, I’ve focussed on getting this to build on macOS; and the workflow is currently skipping Linux and Windows which I can’t test on my machine. @jpkneller and @Sheshuk, could you help with this?

Ideally, this would only require modifying a single install step per OS, to

1. install any OS-specific dependencies
2. export the LIBOMP_INCLUDE environment variable

… and everything else should remain unchanged.
(In practice, I suspect it won’t go quite as smoothly; but we should be able to figure this out.)