xspectrampoline

Installing the XSPEC model library can be as easy as

pip install xspectrampoline

and that's what this package provides.

Usage

This package is indented only for providing the XSPEC model binaries, and is not a wrapper around the model library. The minimal wrapper code that exists is code you would have to write anyway, but beyond that there may are too many ways the models could be put together for Python usage. That's an exercise left to the reader.

Whilst you think about that, here's how you can use xspectrampoline

# Importing calls FNINIT in the libXS
import xspectrampoline

# Get a handle to the minimal wrapper
lib_xspec = xspectrampoline.get_libraries()

# Get the function by the symbol name
powerlaw = lib_xspec.get_model("C_powerlaw")

# Setup an energy, parameter, and output arrays
energy = np.linspace(0.01, 10.0, 100, dtype=np.float64)
parameters = np.array([3.0], dtype=np.float64)
output = np.zeros(len(energy) - 1, dtype=np.float64)
error = np.zeros(len(energy) - 1, dtype=np.float64)

# Invoke the model
powerlaw(energy, parameters, output, error)

# Then do things with `output`
...

Overwriting which libraries are used

By default xspectrampoline bundles everything that is needed to use the XSPEC model library. But what if you already have HEASOFT / XSPEC installed and want to use your local version? xspectrampoline checks if the HEADAS environment variable is set and determines the library locations accordingly. If you want to use your local version, just make sure HEADAS points to the right location and your linker paths are configured via headinit.

xspectrampoline sets the HEADAS environment variable if it's not already set. You can also access the path it uses to resolve the libraries using

import xspectrampoline

headas_path = xspectrampoline.get_HEADAS()
print(headas_path)

How does it work

This package is possible all thanks to Julia's BinaryBuilder.jl who have done all the hard work for making compiled libraries relatively easy to share around. The XSPEC model library was cross compiled into LibXSPEC_jll.jl for Linux and MacOS, and xspectrampoline is basically those artifacts bundled for distribution with Python.

For developers

To bundle things locally, ensure you have setuptools, wheel, and build installed:

python3 -m pip install --upgrade setuptool wheel build

You can then use the Makefile targets to automate the build commands. The Makefile fetches the BinaryBuilder.jl artifacts, unpacks them, and runs the necessary Python packaging commands to create the wheels. Those will then be available under ./dist in your local directory.

make

You can test the packaging by installing from the wheel:

python -m pip install ./dist/xspectrampoline-0.1.0-py3-none-linux_x86_64.whl

selecting the correct platform tag as needed.

There are also handy targets for doing this:

make install    # install the wheel on your system

make clean      # delete all files (include dist) that were generating during
                # the packaging process

Testing locally

Because you need to test the locally installed version of xspectrampoline, you need to make sure you run pytest outside of the root directory. After you've installed the wheel, the easiest way to do this is

cd test && pytest -vvv .