gotranx

gotranx is the next generation General ODE translator. The general idea is that you write your ODE in a high level markup language and use gotranx to generate code for solving the ODE in different programming languages. gotranx uses sympy to create a symbolic representation of the ODE which is used to generate the jacobian and numerical schemes.

• Source code: finsberg/gotranx

• Documentation: https://finsberg.github.io/gotranx/

Install

Install with pip

python3 -m pip install gotranx

or for the development version

python3 -m pip install git+https://github.com/finsberg/gotranx

Quick start

Define your ODE in a .ode file, e.g file.ode with the content

states(x=1, y=0)
parameters(a=1.0)

dx_dt = a * y
dy_dt = -x

which defines the ODE system

\[\begin{split} \begin{align} \frac{dx}{dt} &= ay \\ \frac{dy}{dt} &= -x \end{align} \end{split}\]

with the initial conditions \(x(0) = 1\) and \(y(0) = 0\) and the parameter \(a\) with a value of 1.0. Now generate code in python for solving this ODE with the explicit euler scheme using the command

gotranx ode2py file.ode --scheme explicit_euler -o file.py

which will create a file file.py containing functions for solving the ODE. Now you can solve the ode using the following code snippet

import file as model
import numpy as np
import matplotlib.pyplot as plt

s = model.init_state_values()
p = model.init_parameter_values()
dt = 1e-4  # 0.1 ms
T = 2 * np.pi
t = np.arange(0, T, dt)

x_index = model.state_index("x")
x = [s[x_index]]
y_index = model.state_index("y")
y = [s[y_index]]

for ti in t[1:]:
    s = model.explicit_euler(s, ti, dt, p)
    x.append(s[x_index])
    y.append(s[y_index])

plt.plot(t, x, label="x")
plt.plot(t, y, label="y")
plt.legend()
plt.show()

Alternatively, you can use a third-party ODE solver, e.g scipy.integrate.solve_ivp to solve the ODE by passing in the right-hand side function

import file as model
from scipy.integrate import solve_ivp
import numpy as np
import matplotlib.pyplot as plt

s = model.init_state_values()
p = model.init_parameter_values()
dt = 1e-4  # 0.1 ms
T = 2 * np.pi
t = np.arange(0, T, dt)

res = solve_ivp(
    model.rhs,
    (0, T),
    s,
    method="RK45",
    t_eval=t,
    args=(p,),
)

plt.plot(res.t, res.y.T)
plt.legend()
plt.show()

Note that this is a rather artificial example, so check out the demos in the documentation for more elaborate examples.

FAQ

Why should I use gotranx? The main reasons to use gotranx are

1. You want to solve your model using different programming languages (e.g python and C)

2. You want to create a custom numerical scheme that can utilize the symbolic representation of the ODE

3. You would like to share your model in a high level representation (i.e a markup language)

How does it differ from scipy.integrate.solve_ivp? scipy.integrate.solve_ivp is an ODE solver which takes as input a function defining the right-hand. gotranx takes a high level representation of the ODE and can generate code for the right hand side. In other words, you can use scipy.integrate.solve_ivp to solve the ODE and use gotranx to generate the right hand side.

Automated tests

Unit tests

Automated tests can be found in the test folder. To the run the tests please install the test dependencies

python3 -m pip install "gotranx[test]"

or if you have cloned the repo locally you can do

python3 -m pip install ".[test]"

To run the tests you should execute the following command

python3 -m pytest

Also note that the tests are run on every push and pull request to main using GitHub actions.

Linting and formatting

We use pre-commit to run the a set of linters and formatters in order to ensure consistent code style. Developers should install the pre-commit hooks by first installing pre-commit

python3 -m pip install pre-commit

and then install the pre-commit hooks

pre-commit install

To run the hooks on all the files you can do

pre-commit run --all

For further instructions see the contributing guide.

Note also the we run all hooks as a part of our continuous integration, and we are also using pre-commit.ci to update branches automatically that can fix issues automatically.

Performance monitoring

We have defined a set of benchmarks that run on every push to the main branch using codspeed. To monitor the performance over time you can check out the performance report.

To run the benchmarks locally you can install the pytest-codspeed plugin

python3 -m pip install pytest-codspeed

and run

python3 -m pytest tests/ --codspeed

You can find more info at https://docs.codspeed.io/benchmarks/python

License

MIT

Contributing

Contributions are very welcomed, but please read the contributing guide first