Kamodo provides a functional interface for space weather analysis, visualization, and knowledge discovery, allowing many problems in scientific data analysis to be posed in terms of function composition and evaluation. We'll walk through its general features here.

Kamodo objects

Users primarily interact with models and data through Kamodo objects.

from kamodo import Kamodo

Function registration

Kamodo objects are essentially python dictionaries storing variable symbols as keys and their interpolating functions as values. New functions may be registered either at the initialization of the Kamodo object or later using dictionary bracket syntax.

kamodo = Kamodo('$x = t^2$')
kamodo['g'] = 'y-1'

Function composition

Kamodo automatically composes functions through specifying on the right-hand-side.

kamodo['f'] = 'g(x)'

Here we have defined two functions , , and the composition . Kamodo was able to determine that is implicitly a function of even though we did not say so in 's declaration.

Function evaluation

Kamodo uses sympy's lambdify function to turn the above equations into highly optimized functions for numerical evaluation. We may evaluate for using "dot" notation:


where the return type is a numpy array. We could also have passed in a numpy array and the result shares the same shape:

import numpy as np
t = np.linspace(-5, 5, 100000)
result = kamodo.f(t)
assert(t.shape == result.shape)

Unit conversion

Kamodo automatically handles unit conversions. Simply declare units on the left-hand-side of expressions using bracket notation.

kamodo = Kamodo('mass[kg] = x', 'vol[m^3] = y')

Unless specified, Kamodo will assign the units for newly defined variables:

kamodo['rho'] = 'mass/vol'

We may override the default behavior by simply naming the our chosen units in the left hand side.

kamodo['rho[g/cm^3]'] = 'mass/vol'


Kamodo will raise an error if the left and right-hand-side units are incompatible.

Even though generated functions are unitless, the units are clearly displayed on the lhs. We think this is a good trade-off between performance and legibility.

We can verify that kamodo produces the correct output upon evaluation.

assert(kamodo.rho(3,8) == (3*1000.)/(8*100**3))

Variable naming conventions

Kamodo allows for a wide array of variable names to suite your problem space, including greek, subscripts, superscripts.

kamodo = Kamodo(
    'rho = ALPHA+BETA+GAMMA',
    'rvec = t',
    'fprime = x',
    'xvec_i = xvec_iminus1 + 1',
    'F__gravity = G*M*m/R**2',

For more details on variable names, see the Syntax section.


Many functions can not be written as simple mathematical expressions - they could represent simulation output or observational data. For this reason, we provide a @kamodofy decorator, which turns any callable function into a kamodo-compatible variable and adds metadata that enables unit conversion.

from kamodo import kamodofy, Kamodo
import numpy as np

@kamodofy(units = 'kg/m^3', citation = 'Pembroke et. al, 2018')
def rho(x = np.array([3,4,5]), y = np.array([1,2,3])):
    """A function that computes density"""
    return x+y

kamodo = Kamodo(rho = rho)
kamodo['den[g/cm^3]'] = 'rho'


kamodo.rho.meta # PyHC standard
{'units': 'kg/m^3',
 'citation': 'Pembroke et. al, 2018',
 'equation': None,
 'hidden_args': []}
kamodo.rho.data # PyHC standard
array([4, 6, 8])

Original function doc strings and signatures passed through

Help on function rho in module __main__:

rho(x=array([3, 4, 5]), y=array([1, 2, 3]))
    A function that computes density

    citation: Pembroke et. al, 2018
lhs rhs symbol units
rho(x, y) rho None rho(x, y) kg/m^3
den(x, y) den rho(x, y)/1000 den(x, y) g/cm^3


Kamodo graphs are generated directly from function signatures by examining the structure of both output and input arguments.

from plotting import plot_types
plot_type function
out_shape arg_shapes
(1,) ((N, M), (N, M), (N, M)) 3d-parametric <function surface at 0x12f27f620>
(N,) ((N,),) 1d-line <function line_plot at 0x11e265950>
((N,), (N,), (N,)) 3d-line-scalar <function line_plot at 0x11e265950>
(N, 2) ((N,),) 2d-line <function line_plot at 0x11e265950>
((N, 2),) 2d-vector <function vector_plot at 0x12f27f400>
(N, 3) ((N,),) 3d-line <function line_plot at 0x11e265950>
((N, 3),) 3d-vector <function vector_plot at 0x12f27f400>
(N, M) ((N,), (M,)) 2d-contour <function contour_plot at 0x12f27f488>
((N, M), (N, M)) 2d-contour-skew <function contour_plot at 0x12f27f488>
((N, M), (N, M), (N, M)) 3d-parametric-scalar <function surface at 0x12f27f620>
((1,), (N, M), (N, M)) 3d-plane <function plane at 0x12f27f598>
((N, M), (1,), (N, M)) 3d-plane <function plane at 0x12f27f598>
((N, M), (N, M), (1,)) 3d-plane <function plane at 0x12f27f598>
(N, M, 1) ((1,), (N,), (M,)) 3d-plane <function plane at 0x12f27f598>
((N,), (1,), (M,)) 3d-plane <function plane at 0x12f27f598>
((N,), (M,), (1,)) 3d-plane <function plane at 0x12f27f598>

Kamodo uses plotly for visualization, enabling a rich array of interactive graphs and easy web deployment.

import plotly.io as pio
from plotly.offline import iplot,plot, init_notebook_mode
@kamodofy(units = 'kg/m^3')
def rho(x = np.linspace(0,1, 20), y = np.linspace(-1, 1, 40)):
    """A function that computes density"""
    x_, y_ = np.meshgrid(x,y)
    return x_*y_

kamodo = Kamodo(rho = rho)

We will generate an image of this function using plotly

fig = kamodo.plot('rho')
pio.write_image(fig, 'images/Kamodo_fig1.svg')


See the Visualization section for detailed examples.

Latex I/O

Even though math is the language of physics, most scientific analysis software requires you to learn new programing languages. Kamodo allows users to write their mathematical expressions in LaTeX, a typesetting language most scientists already know:

kamodo = Kamodo('$rho[kg/m^3] = x^3$', '$v[cm/s] = y^2$')
kamodo['p[Pa]'] = '$\\rho v^2$'

The resulting equation set may also be exported as a LaTeX string for use in publications:

print(kamodo.to_latex() + '\n.')
\begin{equation}\rho{\left(x \right)} [kg/m^3] = x^{3}\end{equation}\begin{equation}v{\left(y \right)} [cm/s] = y^{2}\end{equation}\begin{equation}p{\left(x,y \right)} [Pa] = \frac{\rho{\left(x \right)} v^{2}{\left(y \right)}}{10000}\end{equation}

Simulation api

Kamodo offers a simple api for functions composed of each other.

Define variables as usual (order matters).

kamodo = Kamodo()
kamodo['y_iplus1'] = 'x_i + 1'
kamodo['x_iplus1'] = 'y_i - 2'

Now add the update attribute to map functions onto arguments.

kamodo.x_iplus1.update = 'x_i'
kamodo.y_iplus1.update = 'y_i'

Create a simulation with initial conditions

simulation = kamodo.simulate(x_i = 0, steps = 5)
simulation #an iterator of arg, val dictionaries
<generator object simulate at 0x12fdd5a98>

Run the simulation by iterating through the generator.

import pandas as pd
pd.DataFrame(simulation) # pandas conveniently iterates the results for display
y_i x_i
0 NaN 0
1 1.0 -1
2 0.0 -2
3 -1.0 -3
4 -2.0 -4
5 -3.0 -5