Skip to content

testing.md

Latest commit

 

History

History
164 lines (116 loc) · 5.01 KB

testing.md

File metadata and controls

164 lines (116 loc) · 5.01 KB
Raw

Testing

Overview

Three test suites cover the codebase:

Suite Tool Location Command
Backend unit/integration pytest backend/tests/ uv run pytest
Frontend unit/integration vitest frontend/src/**/*.test.jsx npm run test:run
E2E Playwright frontend/e2e/ npm run test:e2e

All three run automatically on every push to main and on every pull request via .github/workflows/tests.yml.

Backend Tests

Framework: pytest Working directory: backend/ Config: backend/pyproject.toml

Running

cd backend
uv run pytest                  # run all tests
uv run pytest tests/test_api_contracts.py  # run a specific file
uv run pytest -k "gap"         # run tests matching a keyword

Test fixture (conftest.py)

The client fixture creates a fastapi.testclient.TestClient and monkeypatches seed_data to a no-op:

@pytest.fixture
def client(monkeypatch):
    monkeypatch.setattr(app_module, "seed_data", lambda: None)
    with TestClient(app_module.app, raise_server_exceptions=False) as test_client:
        yield test_client

Service functions are monkeypatched in individual tests using monkeypatch.setattr.

What is covered

File Coverage
test_api_contracts.py Response shapes for list/get/create endpoints across all resources
test_not_found_contract.py All 404 endpoints return {"error": "not_found", "detail": "..."}
test_error_handlers.py FastAPI exception handler formatting
test_papers_routes.py Paper CRUD, import, BibTeX parse/confirm, duplicate detection, PDF endpoints
test_projects_routes.py Project CRUD, research questions, project-paper links
test_experiment_routes.py Experiment CRUD, CSV import, reorder, duplicate
test_gap_analysis_routes.py Gap analysis endpoint returns shaped GapSuggestion list
test_gap_suggestion_model.py GapSuggestion model validation, ID generation, field defaults
test_chat_and_notes_routes.py Chat and notes CRUD; message persistence patterns
test_service_behaviors.py Service-layer logic: dedup normalization, title casing, author parsing
test_keyword_extraction.py AI keyword extraction endpoint behavior

CI environment variables

Tests run with stub values for external services:

SUPABASE_URL: https://example.supabase.co
SUPABASE_KEY: test-key
OPENAI_API_KEY: test-key

Actual Supabase and OpenAI calls are monkeypatched at the service layer.

Frontend Unit/Integration Tests

Framework: vitest + @testing-library/react Working directory: frontend/

Running

cd frontend
npm run test        # watch mode
npm run test:run    # single pass (used in CI)

Test files

Tests live alongside their source files:

File What is tested
pages/Proposals.smoke.test.jsx Renders fetched proposals list and run details panel
pages/Library.smoke.test.jsx Library table renders items; collection filter working
pages/ProjectDetail.tableview.test.jsx Table vs tree view toggle; column rendering
pages/ProjectDetail.csvimport.test.jsx CSV import modal: parse, preview, confirm flow
pages/ProjectDetail.comparemodal.test.jsx Experiment comparison modal renders diffs
pages/ProjectTasks.tasks.test.jsx Kanban board renders columns and tasks; drag-drop
pages/ProjectReviewDashboard.test.jsx Citation network, timeline, and heatmap computation functions
pages/ProjectReviewDashboard.smoke.test.jsx Page mounts and renders stats panel

Mocking

API modules are vi-mocked at the test level:

vi.mock('../services/api', () => ({
  proposalsApi: { list: vi.fn(), approve: vi.fn() },
}))

Tests render components with MemoryRouter from react-router-dom to avoid needing a real browser router.

E2E Tests (Playwright)

Framework: Playwright Config: frontend/playwright.config.js Test directory: frontend/e2e/ Browser: Chromium only

Running

cd frontend
npm run test:e2e

The Playwright config starts the dev server on port 4173 if not already running:

webServer: {
  command: 'npm run dev -- --host 127.0.0.1 --port 4173',
  port: 4173,
  reuseExistingServer: !process.env.CI,
}

In CI, a fresh server is always started. Locally, an existing server on port 4173 is reused.

What is covered

E2E tests cover user-visible flows: page navigation, basic CRUD interactions, and layout smoke tests. Tests run against the real frontend with a real (or mocked) backend.

CI setup

The e2e CI job depends on both backend and frontend jobs completing successfully before running.

CI Workflow (.github/workflows/tests.yml)

Three parallel jobs:

backend  → pytest (Python 3.11, uv)
frontend → vitest + build check (Node 20)
e2e      → playwright (needs: [backend, frontend])

The e2e job runs after both other jobs succeed, ensuring the build is clean before running browser tests.