Installation

From PyPI

You can install landau using pip:

pip install landau

From Source

To install the latest version from source, clone the repository and install it in editable mode:

git clone https://github.com/eisenforschung/landau.git
cd landau
pip install -e .

Building packages

If you want to build the package yourself:

pip install build
python -m build --sdist

Caching landau objects with fleche

fleche keys its cache on a content digest of every argument. Installing landau registers digest hooks for its types under fleche’s fleche entry point group, so landau objects can be passed to cached functions with no further setup — no imports, no add_hook calls:

from fleche import fleche

from landau.calculate import calc_phase_diagram


@fleche
def diagram(phases, Ts, mu):
    return calc_phase_diagram(phases, Ts, mu)

Calling diagram again with phases rebuilt from the same data hits the cache, in this interpreter and in the next one. landau does not depend on fleche; the hook module is imported only by fleche’s entry point loader.

One case is deliberately refused rather than digested: a fitted Interpolation returned by an interpolator that builds it from a closure (SplineFit, StitchedFit, SoftplusFit, the Whitney interpolators). fleche digests a function from its code object alone, and two fits of one interpolator share that code object, so caching on them would return one result for two different curves. Pass the Interpolator and its samples into the cached function and fit there instead:

@fleche
def free_energy(interpolator, temperatures, energies, T):
    return interpolator.fit(temperatures, energies)(T)