API library-based identification and documentation of usage patterns

Abstract

Application programming interfaces (APIs) are important sources for supporting source code reuse as each API provides a large set of pre-implemented functionalities that support programmers to achieve their daily work in different contexts. However, APIs provide huge number of classes and methods that hinder programmers to understand and use APIs. Numerous client-based approaches have been proposed for facilitating APIs usage through identifying frequent usage pattern. Although they represent significant efforts for helping APIs understanding, the client applications are not available for either newly released APIs libraries or APIs that are not widely used. In this paper, a non-client-based approach for frequent usage patterns identification and documentation is proposed. The approach incorporates hierarchical clustering algorithm and API's source code information. An experimental evaluation is conducted using four widely used APIs. For all studied APIs, the results show that the proposed approach is comparable with client-based approaches in terms of usage patterns cohesion.