Abstract:
When developers have more than one API they could potentially use to solve a programming problem, it's often natural for them to start with the easier and familiar, often default, option. Yet, for some tasks, such as manipulating text in the presence of grapheme clusters (e.g. g̈ and 한), the easier API could produce less correct and reliable results. We sought to measure the impact of such misalignment between API usability and correctness. Specifically, we conducted a controlled experiment which shows that user education has a limited effect on helping the programmer choose the appropriate API, when it's not the default and error cases are difficult to imagine. We discuss a few things programming language and SDK designers can consider in order to mitigate the impact of such misalignments.
PPIG 2020 - 31st Annual Workshop
Understanding the Problem of API Usability and Correctness Misalignment