CI/CD Testing
Essential knowledgeCore knowledge to understand Fluent Commerce
Changed on:
9 Nov 2026
Overview
[Warning: empty required content area]Key points
[Warning: empty required content area]Testing
Testing happens at four levels. Each one catches something the others miss.| Level | Runs | Catches |
| Rule unit tests | Every push, in the module build | Logic errors in a single rule |
| Workflow checks | Every push | Broken workflow structure, rules referenced but not shipped |
| Post-deploy verification | After every deploy | A deploy that reported success but did not install what was built |
| Live end-to-end | After deploying to dev | Rules and workflows that misbehave together on a real account |
Rule unit tests
Write JUnit 5 tests with the Fluent`util-test` library (`RuleExecutor`, mocked GraphQL responses, action assertions) and keep them next to the rules in the plugin reactor. They run inside `mvn package`, so a failing test stops the module build.- Enforce a coverage floor (for example with JaCoCo
`check`) and raise it as coverage improves. Never lower it to get a build through. - Build multi-module reactors with
`package`, not`test`. If the rules module unpacks JARs from other reactor modules,`mvn test`on a clean checkout fails before any test runs, because those JARs are only built during`package`.
Workflow checks
Workflows are JSON and can be checked without an account. Useful checks:- the file parses and matches the expected top-level structure,
- every rule named in a workflow is either listed in
`module.json`or provided by a reference module, - no ruleset names are duplicated, and no ruleset is unreachable,
- rule names are fully qualified (
`<ACCOUNT>.<module>.<RuleClass>`).
Post-deploy verification
See Verification and promotion. At minimum, confirm that the module version installed on the account is the version this pipeline built.Live end-to-end tests
Unit tests mock the platform, so only a run against a real account shows that the deployed rules and workflows work together. Keep a suite of scenario cases in the repository (inbound payload in, expected entity states, events and outbound calls out) and run it against the dev account after each deploy.- Run against dev only, never UAT or production. UAT data usually belongs to the client's testers.
- Create unique test data on every run (order refs with a run ID) so runs never collide or depend on earlier state.
- Treat environment failures as failures. A case that fails because test data is missing is not a pass. Fix the data setup.
- Gate the release, not the merge. End-to-end suites are slow and need an account, so run them after the dev deploy and require a green run before a release branch can go to UAT.