From Towards computational reproducibility: researcher perspectives on the use and sharing of software:
… we conducted a survey to better understand the characteristics of research software as well as how it is created, used, and shared by researchers. … Differences between researchers from computer science and other disciplines related to the knowledge of best practices of software creation and sharing were not statistically significant.
The article reports the results from a survey for researchers with the title “Understanding researcher needs and values related to software”. The authors advertised the survey “through blog posts, social media, and data- and research-related e-mailing lists and listservs.”. There is an example advertisement here. The paper notes that “it is likely that our responses were influenced by our use of computer science and neuroscience-focused e-mail lists.”. Sure enough, of 215 valid responses, 18% were from “Computer Science”, 14% from Biology, and 13% from Psychology, with fewer responses from Engineering, Mathematics, Physics and various other fields.
Section 1 of the survey has the title “Characteristics of Your Code and Software”, so people who filled in the survey beyond the initial few questions were probably also authors of software.
Generally, there were few large differences in responses from computer scientists and those from other fields.
The one big difference was that computer scientists were a lot more likely to “create executables” (82% compared to 50%), perhaps reflecting a tendency of computer scientists to use compiled languages such as Java and C++.
Computer scientists were a little better than their peers in responses to most questions about coding and maintenance hygiene. Relevant questions were “Have you received training in coding conventions or best practices?” (62% CS vs 52% non-CS); “Do you generate documentation for your code?” (54% vs 61%); “When you share your code or software, do you provide information about dependencies?” (90% vs 68%); “Do you use a version control system (e.g., Git)?” (97% vs 78%).