oscode: Oscillatory ordinary differential equation solver
Oscode is a C++ tool with a Python interface that solves oscillatory ordinary differential equations efficiently. It is designed to deal with equations of the form
Oscode makes use of an analytic approximation of x(t) embedded in a stepping procedure to skip over long regions of oscillations, giving a reduction in computing time. The approximation is valid when the frequency changes slowly relative to the timescales of integration, it is therefore worth applying when this condition holds for at least some part of the integration range.
For the details of the numerical method used by oscode, see Citation.
Basic requirements for using the C++ interface:
- C++11 or later
- Eigen (a header-only library included in this source)
The strictly necessary Python dependencies are automatically installed when you use pip or the setup.py. They are:
The optional dependencies are:
pyoscode can be installed via pip
pip install pyoscode
or via the setup.py
git clone https://github.com/fruzsinaagocs/oscode cd oscode python setup.py install --user
git clone https://github.com/fruzsinaagocs/oscode cd oscode pip install .
You can then import
pyoscode from anywhere. Omit the
--user option if
you wish to install globally or in a virtual environment. If you have any
difficulties, check out the FAQs - Installation section below.
You can check that things are working by running tests/ (also ran by Travis continuous integration):
oscode is a header-only C++ package, it requires no installation.
git clone https://github.com/fruzsinaagocs/oscode
and then include the relevant header files in your C++ code:
#include "solver.hpp" #include "system.hpp"
Try the following quick examples. They are available in the examples.
|Introduction to pyoscode:|
|Scipy 2020 lecture notebook:|
|Introduction to oscode:||examples/burst.cpp|
|To plot results from burst.cpp:||examples/plot_burst.py|
To compile and run:
g++ -g -Wall -std=c++11 -c -o burst.o burst.cpp g++ -g -Wall -std=c++11 -o burst burst.o ./burst
Documentation is hosted at readthedocs.
To build your own local copy of the documentation you can run:
cd pyoscode/docs make html
If you use
oscode to solve equations for a publication, please cite:
- Efficient method for solving highly oscillatory ordinary differential equations with applications to physical systems,
- Dense output for highly oscillatory numerical solutions
Any comments and improvements to this project are welcome. You can contribute by:
- Opening and issue to report bugs and propose new features.
- Making a pull request.
You can get help by submitting an issue or posting a message on Gitter.
- Eigen import errors:
pyoscode/_pyoscode.hpp:6:10: fatal error: Eigen/Dense: No such file or directory #include <Eigen/Dense> ^~~~~~~~~~~~~
Try explicitly including the location of your Eigen library via the
CPLUS_INCLUDE_PATHenvironment variable, for example:
CPLUS_INCLUDE_PATH=/usr/include/eigen3 python setup.py install --user # or CPLUS_INCLUDE_PATH=/usr/include/eigen3 pip install pyoscode
/usr/include/eigen3should be replaced with your system-specific eigen location.
Many thanks to Will Handley, Lukas Hergt, Anthony Lasenby, and Mike Hobson for their support and advice regarding the algorithm behind oscode. There are many packages without which some part of oscode (e.g. testing and examples) wouldn't run as nicely and smoothly, thank you all developers for making and maintaining these open-source projects. A special thanks goes to the devs of exhale for making the beautiful C++ documentation possible.
- set minimally required numpy version: numpy>=1.20.0
- drop Python 2.7 support, instead support 3.8 and 3.9 in addition to 3.7
- paper accepted to JOSS
- Fixed getting correct numpy include directories
- Added pyproject.toml to handle build dependencies (numpy)
- Dense output
- Arrays for frequency and damping term need not be evenly spaced
- Automatic C++ documentation on readthedocs
- Eigen included in source for pip installability
- First pip release :)
- Bug that occurred when beginning and end of integration coincided corrected
- Automatic detection of direction of integration
- Memory leaks at python interface fixed
- C++ documentation added