Abstract
To achieve any meaningful task of a certain complexity, developers need to use APIs. Learning to use them is both time consuming and cognitively demanding. We propose to leverage a formal description of API usage as temporal patterns to help developers make sense of the complexities of working with APIs. To achieve this, we propose to deploy recommender systems at various points of the development process that make these patterns useful when most needed. In this paper, we illustrate the approach on a non trivial, real world running example from Android development. The example allows us to articulate a research agenda for leveraging API usage patterns during: (a) testing and compilation times by recommending potential violations of the patterns; (b) coding time by recommending API method calls; and (c) API delivery by recommending improvements to documentation.