The code is oriented towards maximum extensivity and reusability. A lot of concepts are implemented as relatively powerful meta-concepts with explicit implementations (which inherit a lot of the meta-concepts functionalities).

There is still room for more abstraction, but without a concrete need and a forseeable set of use-cases I do consider it fruitless.

The scripts are largely undocumented and might carry crippling bugs. Explanation, documentation and usage examples will be added upon request (and at the discretion of the author).

• a general framework for algrorithms, where a chain of operators is applied to the wave is supplied
• the multislice algortihms, a simple projection and a single scattering algorithm are implemented
• the following parametrisations for the electron scattering potential are supplied (Weickenmeier & Kohl (both), Kirkland and Peng-Dudarev)
• the transfer-function is constructed by applying a Debye-Waller factor, a ROI-patching method is implemented (for large fields of view)
• the propagator is implemented paraxially both in real and Fourier space
• some operators have been ported via Reikna on GPU

• a general framework for iterative reconstruction using general linear least squares solvers is supplied
• the linear operator is split in a projector which constructs the tensor virtually from the transversal repetitions of a kernel and the kernel
• the kernel (the minimal representation of the tensor) can be computed as either a ray (as in the Radon transform) or a Fresnel propagation convolution kernel
• the projectors can construct the tensor for the 2d problem from a 2d kernel (as in the in-plane ray transform) or the 3d problem from a 2d kernel (as in the ray transform with an aligned tilt axis) or semi-3d kernel (as in the Fresnel kernel with an aligned tilt axis)
• the operations are based on COO-formatted sparse arrays
• a powerful system for the hybrid existence of objects on disk and in memory is supplied for on-demand loading and persistence of arrays in memory (in preparation for a CSR/CSC-model for the arrays)

• coordinate transformations in 2d and 3d, chaining of transformations
• Fourier transformation related functions
• a 2d and a 3d interpolation algorithm with COO-sparse in- and output (only linear interpolation)
• calculation of Laplace kernels, optimized of azimuthal uniform gain

• an atomic add for use in Cython even for complex data types)
• a Jupyter progress bar for tracking Python iterator objects
• a slice player for Matplotlib
• an atom viewer for VisPy using the array of atoms type from /Scattering/Potentials/AtomicPotential
• a norm for matplotlib for colourmaps symmetric around 0 (or some other value)
• ...

• Python3 https://www.python.org/
• NumPy http://www.numpy.org/
• SciPy https://www.scipy.org/
• NumExpr https://github.com/pydata/numexpr
• h5py http://www.h5py.org/
• Cython http://cython.org/
• Reikna https://reikna.readthedocs.org/en/latest/
• Matplotlib http://matplotlib.org/
• VisPy http://vispy.org/
• pyFFTW https://hgomersall.github.io/pyFFTW/index.html (optional)

As you normally would: git clone git@github.com:JKrehl/Electrons.py.git and python setup.py install or python setup.py develop.