Skip to content

TESTING.md

Latest commit

 

History

History
201 lines (139 loc) · 4.27 KB

TESTING.md

File metadata and controls

201 lines (139 loc) · 4.27 KB
Raw

Testing Guide

Comprehensive testing documentation for mgrep.

Quick Reference

npm run test           # Run all tests (unit + E2E)
npm run test:unit      # Run unit tests only (Vitest)
npm run test:coverage  # Run tests with coverage report
npm run lint           # Check code style (Biome)

Test Infrastructure

Unit Tests (Vitest)

Located in src/**/*.test.ts. Run with:

npm run test:unit                    # All unit tests
npm run test:unit -- <pattern>       # Filter by pattern
npm run test:unit -- watch_mcp       # Run MCP tests only

E2E Tests (BATS)

Located in test/*.bats. Run with:

npm run test                         # All tests including E2E

Coverage Reports

npm run test:coverage                # Generate coverage report
npm run test:coverage -- --reporter=html  # HTML report in coverage/

Test Categories

MCP Server Tests

The MCP server has comprehensive test coverage across three files:

File Tests Purpose
watch_mcp.test.ts 142 Unit tests: schemas, constants, mocked handlers
watch_mcp.helper.test.ts 21 Helper functions: formatters, extractors
watch_mcp.integration.test.ts 69 Integration tests via TestMCPClient

Run all MCP tests:

npm run test:unit -- watch_mcp

Store Tests

Tests for vector store operations:

npm run test:unit -- store

Provider Tests

Tests for AI providers (OpenAI, Anthropic, Google, Ollama):

npm run test:unit -- providers

Test Utilities

TestStore

In-memory store implementation for testing without Qdrant:

import { TestStore } from "../lib/store.js";

// Automatically used when MGREP_IS_TEST=1
const store = new TestStore();
await store.upsert("file.ts", "content", { path: "/file.ts", hash: "abc" });
const results = await store.search("query", { limit: 10 });

TestMCPClient

Client for testing MCP tools programmatically:

import { TestMCPClient } from "../lib/test-mcp-client.js";

const client = new TestMCPClient();
await client.connect();

// Call tools
const result = await client.callTool("mgrep-search", { query: "test" });

// List resources
const resources = await client.listResources();

// Get prompts
const prompts = await client.listPrompts();
const message = await client.getPrompt("codebase-overview", {});

await client.disconnect();

Writing Tests

Unit Test Pattern

import { describe, expect, it, vi } from "vitest";

describe("MyFeature", () => {
  it("should do something", () => {
    const result = myFunction();
    expect(result).toBe(expected);
  });

  it("should handle errors", () => {
    expect(() => myFunction(invalid)).toThrow("error message");
  });
});

Integration Test Pattern

import { describe, expect, it, beforeAll, afterAll } from "vitest";
import { TestMCPClient } from "../lib/test-mcp-client.js";

describe("MCP Integration", () => {
  let client: TestMCPClient;

  beforeAll(async () => {
    client = new TestMCPClient();
    await client.connect();
  });

  afterAll(async () => {
    await client.disconnect();
  });

  it("should search successfully", async () => {
    const result = await client.callTool("mgrep-search", {
      query: "test query",
    });
    expect(result.isError).toBe(false);
  });
});

Environment Variables

Variable Purpose Default
MGREP_IS_TEST Use TestStore instead of Qdrant undefined
MGREP_TEST_STORE_PATH Path for TestStore file persistence undefined

Debugging Tests

Run Single Test

npm run test:unit -- -t "test name pattern"

Watch Mode

npx vitest watch

Verbose Output

npm run test:unit -- --reporter=verbose

Coverage Targets

File/Area Target Current
src/lib/store.ts 80% ~93%
src/lib/context.ts 80% 100%
src/commands/watch_mcp.ts 80% ~17%*

*MCP server coverage is lower because integration tests exercise TestMCPClient, not the actual server code. All functionality is thoroughly tested.

CI Integration

Tests run automatically on:

  • Pull requests
  • Pushes to main branch

Coverage thresholds can be configured in vitest.config.ts.