An API Testing Method

This paper presents a technique to test APIs that combines aspects of two published software testing methods, namely Markov modeling and category partitioning. Markov modeling provides a basis for model based testing. Markov modeling establishes the context for generating API calls and call sequences within a single test case. For modeling purposes, each combination of parameter values for each function call is a unique “input.” Category partitioning helps select parameter values and effective combinations of multiple parameters for individual API calls. Small examples demonstrate these techniques and two case study summaries illustrate its effectiveness. One case under laboratory conditions established proof-ofconcept and the other applicability to a large commercial API. Some aspects of these techniques are manually intensive and suggest a need for automation.