Python-FSPS no longer has any compile time prerequisites (besides a working Fortran compiler), but (if you’re not installing using git, see the “development version” section below) it does require a clone of the FSPS project for the data files. To get this set up, you can clone that repository using:
export SPS_HOME="/path/where/you/want/to/download/fsps" git clone https://github.com/cconroy20/fsps.git $SPS_HOME
Where, on different systems, you might have to specify the SPS_HOME environment variable. Python-FSPS requires that variable to be set and it will fail to import if it is set incorrectly.
Python-FSPS is built against specific versions of the FSPS Fortran API and data files, so it is important that you have a recent version of FSPS through git. Currently Python-FSPS is built against FSPS v3.2.
Installing stable version¶
Python-FSPS is available as a package on PyPI that you can install using pip:
python -m pip install fsps
Choosing Stellar Libraries¶
FSPS can use several different stellar isochrone and spectral libraries, but switching between these libraries in python-FSPS requires (re-)installing python-FSPS with specific compiler flags. The available libraries are described in more detail in the FSPS documentation, but their names are:
Stellar Isochrone libraries
BPASS (in this case the spectral library and SSP parameters cannot be changed)
Stellar spectral libraries
Dust emission libraries
Changing any of these libraries requires switching off the relevant default, which for isochrones is MIST and for spectra is MILES, and switching on the desired library. As an example, you can change to Padova isochrones and the BaSeL low resolution synthetic stellar library by re-installing:
pip uninstall fsps FFLAGS="-DMIST=0 -DPADOVA=1 -DMILES=0 -DBASEL=1" python -m pip install fsps --no-binary fsps
where the –no-binary fsps flag is required to force building from source.
Installing development version¶
Python-FSPS is being actively developed on GitHub so you can got there to get the most recent development version. You can do this by cloning the python-fsps repository and building:
git clone --recursive https://github.com/dfm/python-fsps.git cd python-fsps python -m pip install .
Flags can be prepended to change the stellar libraries as described above. This repository includes FSPS as a submodule, so if you forget the –recursive flag above, you can get the submodule by running the following commands in the root directory of the Python-FSPS repository:
git submodule init git submodule update
If you install Python-FSPS using this method, you don’t actually need a separate FSPS clone and you can just set the SPS_HOME variable as:
It is recommended that you install using pip even for a local clone. In the root directory of the repository, run:
python -m pip install .
Starting with version 0.5.0, Python-FSPS no longer supports “development” or “editable” builds, but the builds should be cached and re-used if you run pip install multiple times.
Running the unit tests¶
The unit tests are implemented in the tests subdirectory, and can be executed using nox:
python -m pip install nox python -m nox
or pytest directly:
python -m pip install ".[test]" python -m pytest -n 2 -v tests/tests.py
where -n 2 specifies the number of parallel processes to use. The tests can be quite slow, but also memory intensive, so it can be useful to run them in parallel, but don’t overdo it and run out of memory!