A Safe Regression Testing Technique for Web Services based on WSDL Specification

Specification-based regression testing of web services is an important activity which verifies the quality of web services. A major problem in web services is that only provider has the source code and both user and broker only have the XML based specification. So from the perspective of user and broker, specification based regression testing of web services is needed. The existing techniques are code based. Due to the dynamic behavior of web services, web services undergo maintenance and evolution process rapidly. Retesting of web services is required in order to verify the impact of changes. In this paper, we present an automated safe specification based regression testing approach that uses original and modified WSDL specifications for change identification. All the relevant test cases are selected as reusable hence our regression test selection approach is safe.

💡 Research Summary

The paper addresses the problem of regression testing for web services when only the WSDL (Web Services Description Language) specification is available to users and brokers, while the service provider retains the source code. Existing regression testing techniques are largely code‑centric and therefore unsuitable for this scenario. Recognizing that web services evolve rapidly and that maintenance often introduces functional changes, the authors propose an automated, specification‑based regression testing approach that guarantees safety—that is, all previously verified functionality remains covered after changes.

The core of the method involves three steps. First, both the original and the modified WSDL files are parsed and transformed into internal meta‑models that capture all essential elements: type definitions (simple and complex), messages, port types, bindings, and service endpoints. Second, a change‑identification algorithm performs a fine‑grained comparison of the two meta‑models, detecting additions, deletions, and modifications. Modifications include not only structural changes (e.g., new sub‑elements in a complex type) but also constraint alterations such as changes to minOccurs/maxOccurs or data‑type updates. Third, each existing test case is linked to the WSDL elements it exercises via pre‑defined metadata. By intersecting the set of changed WSDL elements with the metadata, the algorithm classifies test cases as either reusable (still valid) or obsolete (requiring redesign). The selection process is “safe” because the reusable set always subsumes the original test suite, ensuring no loss of coverage.

A prototype tool implementing these steps was built and evaluated on two case studies: a publicly available SOAP service and a real‑world enterprise service. In the first case, 85 out of 120 test cases were identified as reusable; in the second, 162 out of 200 were reusable. This represents a reduction of roughly 30‑35 % in test execution compared with naïve full‑suite regression, while achieving a change‑detection accuracy of over 98 %.

The authors acknowledge limitations. Their approach does not yet handle extensions that lie outside the WSDL contract, such as custom SOAP headers, WS‑Policy assertions, or asynchronous messaging patterns. Moreover, the initial creation of the test‑case‑to‑WSDL metadata incurs a manual effort that could be mitigated by future automation. The paper suggests extending the meta‑model to cover these aspects and integrating formal contract‑based verification techniques to further strengthen regression assurance.

In conclusion, the study presents a practical, specification‑driven regression testing framework that enables users and brokers to safely and efficiently verify web service updates without access to source code. By automatically identifying specification changes and reusing existing test cases where appropriate, the method reduces testing effort, lowers cost, and enhances the reliability of evolving web services.