A research poster from the School of Underlying Conditions traces 538 defects through nine years of one laboratory's experiment code and finds them gathered where programming languages meet.
Fourteen experiment codebases break most often at the point where one programming language passes its work to another
Where does a laboratory’s software go wrong? For one unnamed laboratory that studies how bilingual speakers move between their languages, a new research poster answers: wherever the software does the same. Where the Code Switches Language: Defects at programming-language hand-offs in fourteen cued-naming experiment repositories is by Dr Vigdis Emenike, Senior Lecturer, Dr Brynjar Obiora, Senior Research Fellow, and Dr Anouk Mensah, Postdoctoral Fellow in the School of Emergent Priorities.
The School was founded to study what lies underneath things. A good deal of what lies underneath an experiment, it turns out, is a file that two programs each believe the other one understands.
— Dr Vigdis Emenike, Senior Lecturer, School of Underlying Conditions
The team’s reading connects nine years of version history to the 412 occasions on which someone at the laboratory repaired a fault, and follows each repair back to the change that made it necessary. Files in which one programming language exchanges material with another make up 19 per cent of the code and hold 36 per cent of the faults, and in thirteen of the fourteen codebases they were the more fault-prone kind. The poster makes no claim about the laboratory’s published results, none of which were re-examined.
It is the School’s first excursion into software, and the University counts it in the School’s favour that it went in at the joins. Dr Obiora said the fourteenth codebase “has been a great comfort to everyone, and in particular to whoever wrote it”.
The full poster is available from the University’s research repository under an open licence, doi:10.5555/slop.33km6h.