WISeR

Abstract

Mashups are agile applications that aggregate RESTful services, developed by third parties, whose functions are exposed as Web Application Program Interfaces (APIs) within public repositories. From mashups developers’ viewpoint, Web API search may benefit from selection criteria that combine several dimensions used to describe the APIs, such as categories, tags, and technical features (e.g., protocols and data formats). Nevertheless, other dimensions might be fruitfully exploited to support Web API search. Among them, past API usage experiences by other developers may be used to suggest the right APIs for a target application. Past experiences might emerge from the co-occurrence of Web APIs in the same mashups. Ratings assigned by developers after using the Web APIs to create their own mashups or after using mashups developed by others can be considered as well. This article aims to advance the current state of the art for Web API search and ranking from mashups developers’ point of view, by addressing two key issues: multi-dimensional modeling and multi-dimensional framework for selection. The model for Web API characterization embraces multiple descriptive dimensions, by considering several public repositories, that focus on different and only partially overlapping dimensions. The proposed Web API selection framework, called WISeR (Web apI Search and Ranking), is based on functions devoted to developers to exploit the multi-dimensional descriptions, in order to enhance the identification of candidate Web APIs to be proposed, according to the given requirements. Furthermore, WISeR adapts to changes that occur during the Web API selection and mashup development, by revising the dimensional attributes in order to conform to developers’ preferences and constraints. We also present an experimental evaluation of the framework.