HDF5 Database

Preserving generated models is crucial for this package. Several technologies were considered with primary concerns being:

• Wide-spread use
• Simple structure
• Cross-platform compatibility
• Programming-language-agnostic

Initially, SQLite was considered a good option for simple tabular data. However, SWAP's diverse input and output formats would challenge SQLite, requiring unconventional use of RDBMS. Why force a cylinder through a square hole?

A better option, widely used in science, is the HDF5 file format. It accepts almost any format, is compatible with multiple languages (R, Python, Fortran), and has a simple, folder-like structure. Thus, pySWAP builds on the h5py library.

The goal is to make the HDF5 database easily shareable with modellers using different programming languages. Currently, pySWAP saves and retrieves Python-specific binary formats using pickle. However, more formats can be added according to the community needs. Using pickle objects with some metadata attached is a very straightforward approach and allowes for easy storage and retrieval of the entire models and their results. To demonstrate how this works, Follow the HDF5 database tutorial.