Quick Fixes for novice programmers: effective but under-utilised

Professional software development environments typically provide "quick fixes" for common program errors: a common solution to an error that can be enacted with a single click. For example, an "unknown type" error can be fixed by adding an import for the type, or an "unknown variable" error can be fixed by changing a misspelt identifier to match the name of an already-declared variable. The BlueJ environment recently added some support for quick fixes, and in this paper we use the associated Blackbox dataset to investigate how they are used by novice programmers. We use a combination of automatic filtering of over 100 000 quick fixes, and manual categorisation of 900 programming session fragments. We find that acceptance of suggestions ranges from 1-17% for different quick fix types, usually within 3-5 seconds, and many manually performed alternatives are similar to the actions of unselected quick fixes. Furthermore, we find that users are faster and are more able to make productive progress when quick fix support is available, especially for fixing imports. This data can be used to inform design of future quick fix systems. There are pedagogical arguments for and against providing such a feature to novice programmers and we provide some initial discussion on the matter.