CLI Testing

The Compose CLI uses Vitest as its test runner.

Tests follow the same boundaries as the implementation: module behavior is tested through modules, workflow behavior through pipelines, external behavior through adapters, and standalone logic through utilities.

Test Structure

The cli/test directory mirrors cli/src. Keep each feature's test, harness, and fixtures together.

cli/test/
├── adapters/
│   └── hashingAdapter.test.ts
├── modules/
│   └── deployGeneration/
│       ├── fixtures/
│       │   └── expected/
│       │       └── Deploy.s.sol
│       ├── harness.ts
│       └── module.test.ts
├── pipelines/
│   └── infoPipeline/
│       ├── harness.ts
│       └── infoPipeline.test.ts
└── utils/
    └── regex.test.ts

Use *.test.ts for test files. Use harness.ts when a test needs reusable context, files, adapters, or prerequisite state.

Configuration

The CLI test setup uses:

• vitest.config.ts for the Node test environment and test/**/*.test.ts discovery.
• tsconfig.test.json to type-check src, tests, and Vitest types without emitting test files into dist.
• The test script in cli/package.json to run vitest run.

Harnesses

A harness prepares the environment required by a test. It may create a context, populate prerequisite state, write a temporary project, and expose cleanup.

export type ExampleHarness = {
  ctx: ComposeContext;
  projectRoot: string;
  cleanup(): Promise<void>;
};

export async function createExampleHarness(): Promise<ExampleHarness> {
  const projectRoot = await fs.mkdtemp(
    path.join(os.tmpdir(), "compose-cli-example-"),
  );
  const ctx = Context.create();
  ctx.param.projectRoot = projectRoot;

  return {
    ctx,
    projectRoot,
    cleanup: () => fs.rm(projectRoot, { recursive: true, force: true }),
  };
}

Use the operating system's temporary directory for generated test projects. Always clean it in a finally block so a failed assertion does not leave files behind.

Fixtures and Snapshot Files

Use fixtures for input or expected files that are easier to review as complete artifacts. Generated Solidity and TypeScript files should normally be compared directly with a snapshot file.

An exact comparison makes formatting and content changes visible in review:

expect(generated).toBe(expected);

Update a snapshot file intentionally when generated output changes. Do not normalize whitespace or replace the comparison with a hash unless formatting is explicitly outside the behavior under test.

Module Tests

Modules are the main place to test CLI behavior.

• Import and execute the real module action.
• Use a small harness to create the required ctx and filesystem state.
• Mock only the ctx.param, ctx.config, and ctx.state fields read by the module.
• Assert the resulting module state and any generated files.
• Keep harnesses, fixtures, and mocks under cli/test.

The deploy generation test builds a complete ERC-20 context through its harness, executes the real module, and compares the generated script with an expected-output snapshot:

const harness = await createDeployGenerationHarness();

try {
  const result = await DeployGenerationModule.generateDeployScript(
    harness.ctx,
    harness.scriptRoot,
  );
  const state = result.state.generateDeployScript as ModuleState<GeneratedDeployScriptState>;
  const generated = await fs.readFile(state.result!.outputPath, "utf8");
  const expected = await fs.readFile(expectedDeployScriptPath, "utf8");

  expect(state.success).toBe(true);
  expect(generated).toBe(expected);
} finally {
  await harness.cleanup();
}

Required Previous State

Some module actions depend on state produced by an earlier action.

• Create focused mock state when the prerequisite is small and its behavior is not under test.
• Call the real prerequisite action when it is easier and clearer than reproducing its result.
• Use a pipeline test when module order or the complete workflow is part of the behavior.
• Avoid reproducing a large production state object in every test; move shared setup into the feature harness.

For example, a deploy generation harness provides the selected catalog, scaffold map, and project parameters because those are direct inputs to DeployGenerationModule. An info pipeline harness instead writes a real compose.json and facet source because loading and scanning are part of that pipeline's behavior.

Pipeline Tests

Pipeline tests exercise a workflow through the real pipeline entry point.

• Build a realistic starting context with a harness.
• Let the pipeline call its real modules in production order.
• Assert meaningful final state, scanned data, generated files, or displayed output.
• Keep internal module edge cases in module tests.

The info pipeline test creates a temporary local project, scans a CounterFacet, and checks the resulting selector and storage metadata:

const harness = await createInfoPipelineHarness();
const output = vi.spyOn(console, "log");

try {
  const result = await InfoPipeline.execute(harness.ctx);
  const state = result.state.infoProject as ModuleState<ComposeProjectInfo>;
  const facet = state.result?.diamonds[0]?.facets[0];

  expect(state.success).toBe(true);
  expect(facet?.selectors).toEqual(["getValue()"]);
  expect(facet?.storageSlots[0]?.slot).toBe("counter");
  expect(
    output.mock.calls.flat().some((value) => String(value).includes("info-example")),
  ).toBe(true);
} finally {
  output.mockRestore();
  await harness.cleanup();
}

Spying on console.log without replacing its implementation records the output while preserving the command's normal display behavior.

Adapter Tests

Test adapters separately when their external or runtime behavior is the target. Assert the complete adapter result and keep higher-level interpretation outside the adapter test.

For example, HashingAdapter returns a full Keccak-256 digest. Deriving a four-byte selector belongs to selector logic, not the hashing adapter test:

expect(HashingAdapter.keccak256("transfer(address,uint256)")).toBe(
  "0xa9059cbb2ab09eb219583f4a59a5d0623ade346d962bcd4e46b11da047c9049b",
);

When a module requires an adapter and adapter resolution is not under test, pass a focused mock implementing the real interface. When resolution is part of the workflow, resolve it as the production pipeline does and guard the partial result:

const deps = await DependencyResolver.resolve([
  { key: DependencyKey.Hashing },
]);

if (!deps.hashing) {
  throw new Error("Hashing adapter was not resolved");
}

const result = await ValidationModule.detectSelectorCollisions(ctx, {
  hashing: deps.hashing,
});

expect(result.state.validationSelectorCollisions?.success).toBe(true);

Utility Tests

Utility tests cover small, standalone logic directly. Prefer behavior-oriented cases over checking implementation details.

const value = "Facet[0].getValue(address,uint256) + $slot?";
const pattern = new RegExp(`^${escapeRegExp(value)}$`);

expect(pattern.test(value)).toBe(true);

Running Tests

Run the suite from the CLI package:

cd cli
npm run test

Type-check the tests separately when changing harnesses or test-only types:

npx tsc -p tsconfig.test.json

Checklist

• Test files mirror the corresponding src boundary.
• Module behavior is covered through real module actions.
• Complex setup is isolated in a feature harness.
• Temporary projects are removed in a finally block.
• Generated code is compared with a reviewable snapshot file.
• Pipeline tests cover real workflow behavior and meaningful final state.
• Adapter tests assert adapter behavior without leaking higher-level logic.
• Utility tests cover standalone behavior directly.
• npm run test and the test TypeScript check pass.