IMPORTANT: ALPHA STATUS

Systems of ordinary differential equations as easy as Python.

symodesys is an open-source Python framework for manipulation and analysis of systems of ordinary differential equations. It is tightly built upon Sympy and can be seen as a set of convenience classes and templates for code generation which turns Sympy into a poverful ODE system analysis tool.

The code is generated using templates, and hence no restriction exists on what langauges are used (C/Fortran/C++ used for GSL/ODEPACK/odeint).

Systems of ODE's are appearing in all of the natural sciences. Sympy together with numpy + scipy + matplotlib etc makes python an excellent language to build a versatile common framework for modelling of ODE systems.

Central to symodesys is NOT to reinvent the wheel. The less code put into symodesys the better. Instead it aims to make sympy (as the most promising free-software CAS in the views of the author) an efficient tool for dealing with systems of ODE's.

Symodesys offers (for IVP's):

• General Interactive plotting through Enthought's Chaco library
• Code generation support for: GSL, Sundials and ODEPACK

Other software built using symodesys:

• [TODO] symchemkin - Flexible framework for analysis of systems of chemical reactions

• More test cases where analytic solution is known.
• [TODO] Add options of dynamic step size GSL integrator
• [TODO] Extend ODEPACK support LSODE dense and banded.
• [TODO] Add code generation support for: RODAS (look at assimulo), ODEINT, RKC
• [TODO] Improve code generation to write loops when possible (let compiler unroll if optimal, decalare global const paramters), param structs etc.
• [TODO] Look into generating Jacobian as a sympy.SparseMatrix
• [TODO] Visualization of sensitivity of system by accepting uncertain input parameters (normal distr for starters..)
• [TODO] Optimization of paramters (fiting) to match e.g. numerical data.
• [TODO] Add routines for scaling (and automatic rescaling) in e.g. the IVP class. Note: this is a simple special case of variable transformation and may hence be handled by that code, only convenience routines are needed to be added.
• [TODO] Export of trajectories. (easy)
• [TODO] Add python code writer. (low priority)

(Tested against) Python 2.7 Sympy 0.7.3 Cython 0.19.1

Gnu sciencific library (GSL v. 1.15) Sundials ODEPACK [TODO] RODAS [TODO] rkc [TODO] vode (low priority)

Installing and building symodesys can be a little challenging due to external dependencies. For example, when debugging the setup script I invoke it like:

env DEBUG_DISTUTILS=1 COMPILER_VENDOR=gnu C_INCLUDE_PATH=~/.local/include:$C_INCLUDE_PATH python -m pudb setup.py build

But if you have Sundials 2.5.0 on your systems standard include path and you don't have Intel's fortran compiler installed (and my setup script is clever enough) even pip should do the right thing.

To install Chaco etc. I needed to do:

sudo pip install --upgrade enable --allow-external PIL --allow-unverified PIL
sudo pip install chaco traitsui
apt-get install python-qt4-gl

And prepend python command as e.g. QT_API=pyqt python ode.py.

symodesys is currently in its prototyping stage and hence big changes to API and frequent refactoring is expected. If you are fine with this you may of course fork and I'll happily accept pull requests.

1. Refactor and refine API until convergence is reached (current expected time frame for this: months)
2. Write at least one non-trivial python package based on symodesys
3. Write tests and expand doc strings and comments in code
4. Write proper documentation and doctests (in Sympy style)

opkda1.f fails using ifort 13.0.1 with the following error opkda1.f(9498): error #6633: The type of the actual argument differs from the type of the dummy argument. [RWORK] 1 RWORK(LACOR), IA, JA, IC, JC, RWORK(LWM), RWORK(LWM), IPFLAG, ---------------------------------------------------^ compilation aborted for opkda1.f (code 1) I have not looked into whether this is a real bug detected by the Intel compiler.

• symneqsys

• symvarsub
• pycompilation

Written by Björn Dahlgren. Copyright 2012-2014.

Open Source. Released under the very permissive simplified (2-clause) BSD license. See LICENCE.txt for further details.

Note that you need sundials 2.5.0 (too new for Ubuntu 12.04 LTS) Download from http://computation.llnl.gov/casc/sundials/main.html

e.g.:

mkdir /tmp/sundials_build/ cd /tmp/sundials_build/ cmake-gui ~/Downloads/sundials-2.5.0/

There are many ways to configure Sundials depending on your system. I use Intel C compiler (icc) and ~/.local as prefix, and CMAKE_C_FLAGS=-fPIC. (static linking)

For running a symodesys simulation you need to make sure the include and library dirs are searched, e.g.:

env C_INCLUDE_PATH=~/.local/include:$C_INCLUDE_PATH LIBRARY_PATH=~/.local/lib python codegen.py -i sundials

if you build a shared sundials library at a custom prefix path you must ensure LD_LIBRARY_PATH is also set accordingly.