Since TRUST 1.8.3, the ICoCo API is accessible from Python through a SWIG wrapper. This makes it possible to write supervision scripts entirely in Python rather than in C++, with full access to the MEDCoupling Python API for field manipulation and to mpi4py for multi-process coordination.

Setting Up the Python Environment

A dedicated environment script provides the necessary Python paths and MPI bindings. It must be sourced before launching any Python-based ICoCo supervisor:

source $TRUST_ROOT/env_for_python.sh

Verify the availability of the three core packages:

import trusticoco as ti
import medcoupling as mc
from mpi4py import MPI

If all three imports succeed without error, the environment is correctly configured.

Running an ICoCo Case from Python

The Python ICoCo API mirrors the C++ API: the same methods (initialize(), initializeTimeStep(), solveTimeStep(), validateTimeStep(), terminate(), etc.) are available on the Python problem object returned by trusticoco.

The data file preparation for a Python-supervised run is identical to a C++ supervised run: the TRUST data file must declare the ICoCo problem name and must not contain a Solve pb instruction.

Example Case

A fully worked example covering the conjugate heat transfer case from the previous tutorial is available in the TRUST test database:

source $TRUST_ROOT/env_for_python.sh

trust -copy docond_VEF_3D_ICoCo_py

Examine the prepare script in the copied directory for the complete setup procedure and then study the main.py file to see how the Python API is used to drive two coupled problems.

Attention: Even in sequential mode (single-process), a Python ICoCo supervisor must always be launched through mpirun:

mpirun -np 1 python main.py

Launching python main.py directly will fail because the TRUST problem object requires an MPI communicator to be initialized.

Using ICoCo with Python in a Baltik Project

For users who develop their own TRUST-based application through the baltik build system, a dedicated repository provides examples and build scripts for integrating the Python ICoCo wrapper into a custom baltik:

https://github.com/cea-trust-platform/icoco-swig-baltik

This repository documents how to add the trusticoco Python module as a dependency of an existing baltik project and how to structure Python supervisors that interact with custom physics modules.

Summary

The Python interface to ICoCo provides a convenient alternative to C++ supervision scripts, particularly when the coupling logic involves complex data transformations that benefit from the expressiveness of Python and the MEDCoupling Python API. The main points to remember are:

The trusticoco Python module exposes the same interface as the C++ ICoCo API. The MEDCoupling Python API (medcoupling) is used for field construction, interpolation, and manipulation. Multi-process runs require mpi4py and must always be launched with mpirun, regardless of the number of processes. Data file preparation follows the same rules as for C++ supervisors: declare ICoCoProblemName and remove Solve pb.