Larch uses a plugin architecture to implement most of its features. The reading and writing of various kinds of file, most math functions, all functionality for moving motors and reading detectors, and all functionality for managing the processing and analysis of synchrotron data -- these are all managed via plugins.
A plugin is a file written in python with a few lines of special code which allow the Larch interpreter to recognize it as a plugin. While the Larch document includes a section on programming with Larch and the many plugins that ship with Larch can serve as examples, it is a bit challenging to get started writing Larch plugins. There are some strange and surprising aspects to Larch plugins.
After writing the feffrunner and diffkk plugins as well as the
unit testing framework for the
feff85exafs project, I
decided to write this tutorial in hopes of lowering the barrier for
the next participant in Larch. To Matt's credit, he has built a
powerful platform for doing chores at the beamline and for dealing
with data obtained at the beamline. There is real potential for Larch
to become part of the synchrotron infrastructure. That will be
particularly true if lots of people do neat, new things with Larch.
In this tutorial you will learn how plugins are used by Larch and what goes into creating a new plugin. We will start with a toy problem, allowing us to understand what distinguishes a Larch plugin from any other file containing python code. We will also see how to incorporate the plugin into Larch for testing and use.
Then we will build a substantial plugin which implements the MBACK algorithm by Tsu-Chien Weng, et al.. This is a way of normalizing XANES data by matching the measured spectrum to tabulated values of the atomic cross sections.
Finally we will examine the diffkk plugin, which implements a
differential Kramers-Kronig transform used to generate the
energy-dependent atomic scattering factors for elements in real
materials starting from a XAS measurement. This will demonstrate how
to use object-oriented programming as part of your Larch plugin as
well as provide an example of using NumPy,
python's package for scientific computing.