Towards property-based testing of RESTful web services

Abstract

Developing APIs as Web Services over HTTP implies adding an extra layer to software, compared to the ones that we would need to develop an API distributed as, for example, a library.This additional layer must be included in testing too, but this implies that the software under test has an additional complexity due both to the need to use an intermediate protocol in tests and to the need to test compliance with the constraints imposed by that protocol: in this case the constraints defined by the REST architectural style.On the other hand, these requirements are common to all the Web Services, and because of that, we should be able to abstract this aspect of the testing model so that we can reuse it in testing any Web Service.In this paper, as a first step towards automating the testing of Web Services over HTTP, we describe a practical mechanism and model for testing RESTful Web Services without side effects and give an example of how we successfully adapted that mechanism to test two different existing Web Services: Storage Room by Thriventures and Google Tasks by Google. For this task we have used Erlang together with state machine models in the property-based testing tool Quviq QuickCheck, implemented using the statem module.

Automated Test Generation for REST APIs: No Time to Rest Yet

“…The approaches proposed by Chakrabarti and Kumar [23] and by Reza and Gilst [68] rely on an XML-format specification for REST API testing. The approaches developed by Seijas et al [47], by Fertig and Braun [30], and by Pinheiro et al [59] perform model-based testing. We did not include these techniques in our study because they use a tool-specific REST API specification [23,68] that requires non-trivial information to be provided by the user, rely on formal specifications [30,47,59], or target specific problems [24,35].…”