Predicting the Install, Compile, and Test Attributes of a Library

The task of selecting a third-party library is a common one, but it can be challenging due to the many factors that a developer has to consider. Previous research has mainly focused on specific attributes. In this paper, we take a comprehensive approach by identifying the most desirable attributes and examining how they are related to one another. Our work is split into three parts. First, we conduct a developer survey to rank their preferences on a set of 30 attributes from existing work, grouped into four categories: build, documentation, source code, and contribution. To confirm our results, in the second part of the study, we then mined 104,364 NPM libraries to find that the build (i.e., installation, compiling, and testing) of a library is important. Finally, in the third part of the study, we find that static source code attributes (i.e., number of files, repository size, and directories) correlate and predict the success of installing, building, and testing a library. Our work lays the groundwork on what attributes of a library are useful for predicting a successful build of a library.