Inferring Reusability of End-User Programmers’ Code from Low-Ceremony Evidence

While end-user programmers sometimes combine, learn from, or otherwise reuse existing code to quickly create new programs, not all code is equally reusable. Some code is reused by its creator or by others, but other code simply languishes on servers and never provides any help in the creation of subsequent programs. In this paper, we draw on numerous empirical studies of end-user and professional programmers to show that the reusability of code can be inferred on the basis of “low-ceremony” evidence. This evidence is information that is often informal, possibly unreliable, but that can be quickly gathered, interpreted and synthesized without the investment of substantial effort or skill by code producers or consumers. In the studies considered here, it includes information about code’s mass appeal, flexibility, understandability, functional size, authorship, and prior reuses. We summarize a simple machine learning model that has successfully predicted reuse of web macros based on this low-ceremony evidence.