Abstract
Open APIs are software intermediaries that make it possible for application programs to interact with data and processes, which can both be viewed as forms of services. In many scenarios, when one wants to obtain or publish a new service, one would like to check whether the new functionality can simply be obtained by suitably composing existing services. In this paper we study this problem by distinguishing between the two forms of services, that we call data-centric and process-centric, respectively. In the former, each API is an abstraction of a query specified on a data source, and composition amounts to building a new query by using the available APIs as views over the data. In the latter, each API abstracts a process made up by sequences of atomic actions, and composition means realizing a new process by suitably using the APIs exposed by the available services. We make the assumption that the semantics of services is specified by means of one of the most basic formalisms used in Computer Science, namely, regular languages. As a result, we get a rich analysis framework, where composition shows similarities to conformant and conditional planning. We describe composition principles and automated synthesis techniques for each of the two settings.