PatternLens: Inferring evolutive patterns from web API usage logs

Abstract

The use of web Application Programming Interfaces (WAPIs) has experienced a boost in recent years. Developers (i.e., WAPI consumers) are continuously relying on third-party WAPIs to incorporate certain features into their applications. Consequently, WAPI evolution becomes more challenging in terms of the service provided according to consumers’ needs. When deciding on which changes to perform, besides several dynamic business requirements (from the organization whose data are exposed), WAPI providers should take into account the way consumers use the WAPI. While consumers may report various bugs or may request new endpoints, their feedback may be partial and biased (based on the specific endpoints they use). Alternatively, WAPI providers could exploit the interaction between consumers and WAPIs, which is recorded in the WAPI usage logs, generated while consumers access the WAPI. In this direction, this paper presents PatternLens, a tool with the aim of supporting providers in planning the changes by analyzing WAPI usage logs. With the use of process mining techniques, this tool infers from the logs a set of usage patterns (e.g., endpoints that are frequently called one after the other), whose occurrences imply the need for potential changes (e.g., merging the two endpoints). The WAPI providers can accept or reject the suggested patterns, which will be displayed together with informative metrics. These metrics will help providers in the decision-making, by giving them information about the consequences of accepting/rejecting the suggestions.