generic framework and xarray extension for computer model simulations

python, xarray, modelling, simulation, framework, simulation-framework
pip install xarray-simlab==0.5.0


xarray-simlab: xarray extension for computer model simulations

Build Status Coverage Status Documentation Status Citation

xarray-simlab is a Python library that provides both a generic framework for building computational models in a modular fashion and a xarray extension for setting and running simulations using the xarray's Dataset structure. It is designed for fast, interactive and exploratory modeling.

xarray-simlab is well integrated with other libraries of the PyData ecosystem such as dask and zarr.

In a nutshell

The Conway's Game of Life example shown below is adapted from this blog post by Jake VanderPlas.

  1. Create new model components by writing compact Python classes, i.e., very much like dataclasses:
import numpy as np
import xsimlab as xs

class GameOfLife:
    world = xs.variable(
        dims=('x', 'y'), intent='inout', encoding={'fill_value': None}

    def run_step(self):
        nbrs_count = sum(
            np.roll(np.roll(self.world, i, 0), j, 1)
            for i in (-1, 0, 1) for j in (-1, 0, 1)
            if (i != 0 or j != 0)
        self._world_next = (nbrs_count == 3) | (self.world & (nbrs_count == 2))

    def finalize_step(self):
        self.world[:] = self._world_next

class Glider:
    pos = xs.variable(dims='point_xy', description='glider position')
    world = xs.foreign(GameOfLife, 'world', intent='out')

    def initialize(self):
        x, y = self.pos

        kernel = [[1, 0, 0],
                  [0, 1, 1],
                  [1, 1, 0]]

        self.world = np.zeros((10, 10), dtype=bool)
        self.world[x:x+3, y:y+3] = kernel
  1. Create a new model just by providing a dictionary of model components:
model = xs.Model({'gol': GameOfLife,
                  'init': Glider})
  1. Create an input xarray.Dataset, run the model and get an output xarray.Dataset:
input_dataset = xs.create_setup(
    clocks={'step': np.arange(9)},
    input_vars={'init__pos': ('point_xy', [4, 5])},
    output_vars={'gol__world': 'step'}

output_dataset = input_dataset.xsimlab.run(model=model)
>>> output_dataset
Dimensions:     (point_xy: 2, step: 9, x: 10, y: 10)
  * step        (step) int64 0 1 2 3 4 5 6 7 8
Dimensions without coordinates: point_xy, x, y
Data variables:
    init__pos   (point_xy) int64 4 5
    gol__world  (step, x, y) bool False False False False ... False False False
  1. Perform model setup, pre-processing, run, post-processing and visualization in a functional style, using method chaining:
import matplotlib.pyplot as plt

with model:
         input_vars={'init__pos': ('point_xy', [2, 2])}
         col='step', col_wrap=3, figsize=(5, 5),
         xticks=[], yticks=[],
         add_colorbar=False, cmap=plt.cm.binary)



Documentation is hosted on ReadTheDocs: http://xarray-simlab.readthedocs.io


3-clause ("Modified" or "New") BSD license, see License file.

xarray-simlab uses short parts of the code of the xarray, pandas and dask libraries. Their licenses are reproduced in the "licenses" directory.


This project is supported by the Earth Surface Process Modelling group of the GFZ Helmholtz Centre Potsdam.


If you use xarray-simlab in a scientific publication, we would appreciate a citation.