Giant bug eats scientific results

Computational science, by its nature, involves a wide variety of computer simulations. These are done using software that is usually developed by the researchers themselves, on their own computers or sometimes on large HPC machines. As new methods are developed, new software is written and added to the scientific toolbox. Unfortunately, this software is often poorly written and difficult to reuse. The Astrophysical MUltipurpose Software Environment (AMUSE) comprises more than 50 astrophysical simulation codes created by a variety of researchers, making them available from the Python scripting environment in a well-documented manner. Whil this really helps scientists wishing to run simulations, it does not resolve the above-mentioned issues but merely shifts the responsibility for solving them from the end user to the maintainers of AMUSE. In recent years, AMUSE has suffered from a lack of resources for maintenance, which has caused issues for its users. We have been addressing these issues for AMUSE, but the wider problem of software quality in computational science remains, and deserves more attention.