Automated RPC Interface Diff Framework for Efficient Testing

To ensure RPC service stability and minimize bugs reaching production, the article proposes an automated interface testing approach that replaces manual Java‑based test development with a diff‑driven solution.

The current manual method suffers from four main drawbacks: it requires Java expertise, frequent maintenance due to API changes, difficulty judging large responses, and limited coverage under tight iteration schedules.

The proposed solution automatically captures request parameters from online logs via an interceptor, stores them as JSON on an FTP server, and uses these real‑world inputs to generate Java test classes through a Velocity‑style template.

Key steps include: (1) providing a cases file describing target interfaces; (2) parsing service source files to collect import information; (3) compiling the contract JAR to obtain method signatures; (4) matching cases with signatures and generating .java files; (5) compiling the generated sources and creating a TestNG suite; (6) executing the suite against both production and sandbox endpoints; and (7) retrieving and comparing results from the database.

Result comparison is performed by converting returned values—whether primitives, collections, or objects—into JSON and using a JSONDiffUtil utility to highlight differences.

The framework, illustrated with flow diagrams and example code snippets, requires only a Git branch URL and a cases file, making it applicable to any RPC service without additional manual test code.

In practice, the diff framework enables rapid validation of strategies such as advertising recall, where offline data is insufficient and A/B testing is noisy; by diffing live and sandbox responses, teams can assess effectiveness before full rollout.

Extensions include automatic task triggering via MQ messages and configurable templates for different testing scenarios, further reducing human intervention.