Core data structures (MEDCoupling)

Overview

The MEDCoupling library gather the core structures used everywhere in the MED world.

• Arrays of data
• Array indexing and numbering
• Meshes
• Fields

Design philosophy

The structures described above (arrays, meshes and fields), with the algorithms associated to it form the core of the MEDCoupling library (libmedcoupling.so or medcoupling.dll). It is fully written in C++ and wrapped in Python.

The MEDCoupling C++ library implements a data structure which is the result of the following tradeoff:

• compliant with code coupling:
  • fields data structure containing enough information to perform a well defined interpolation
  • data structure exchangeable through processes, as well in parallel (SPMD paradigm - sometimes called ParaMEDMEM), as in a distributed paradigm (using CORBA).
• minimize as much as possible the number of external software prerequisites needed to use it (the core of the library, MEDCoupling, can be built as a standalone tool).
• light and agile enough to:
  • maximize the amount of possible algorithms being applied on it
  • ease the implementation of other projects using MEDCoupling as a prerequisite.
• large enough to be used for MED file I/O (thanks to the MEDLoader part).
• compliant with VTK visualization data structures
• integrate HPC constraints (compact structures, limitation of copies and launching of CPU consuming algorithms only when absolutely needed ).
• compliant with the ICoCo API

The MEDCoupling also implements a set of algorithms linked to this data structure. The functionalities section gives an insight of the avalaible algorithm.