A survey of scientific software development

Abstract

Software for scientific research purposes has received increased attention in recent years. Case studies have noted development practices, limitations, and problems in the development of scientific software. However, applicability of the results of these studies to improving the wider scientific software development practices is not known. This paper presents a survey of 60 scientific software developers. The survey was conducted online from August--September 2009, and aims to identify where improvements to scientific software practices can be made. While our results generally confirm previous work, we have found some notable differences. The use of IDEs and version control tools among the surveyed scientific software developers has increased, and trace-ability of scientific software is not as important to scientific software developers as it is to scientific software users. Documentation also appears to be more widely produced than previous studies indicate. However, there remains room for improvement in tool use, documentation, testing, and verification activities for scientific software development.