Testing Documentation

Testing Documentation

Document Revision History

Version	Date	Key Changes	Author
v0.1.0	2026-10-21	Iteration 2: Test Responsibility & Schedule, Unit & Integration Testing (plan and first results)	Yewon Lee

---

1. Test Responsibility & Schedule (Iteration 2)

Who

Whoever writes a feature also writes its tests, in the same PR. Each part has one person who keeps its tests working.

Component (layer)	Files	Maintainer
Backend: route generation & street graph	api/route_algorithm/algorithm.py, street_graph.py	Yewon Lee
Backend: elevation, scoring, LLM briefing	api/route_algorithm/elevation.py, scoring.py, briefing.py, llm.py	Yewon Lee
Backend: traffic signal counting	signal counting in algorithm.py	SoonYoung Park
Backend: accounts API	api/views/auth.py	Anton (log-in, Google, log-out), Yewon Lee (sign-up validation)
Backend: saved routes API	api/views/routes.py, api/models.py	Anton
Backend: place search	api/services/tmap.py, place_translation.py, api/views/places.py	Yewon Lee
Android Model (data layer)	data/remote, data/repository, data/auth	Yewon Lee; elevation in RouteRepository: SoonYoung Park
Android ViewModel	AuthViewModel, MyPageViewModel, route options in RouteViewModel	Yewon Lee
Android ViewModel (elevation)	elevation and slope parts of RouteViewModel	SoonYoung Park
Android View (Compose UI)	ui/**/*Screen.kt, RouteMap.kt	Screen authors (manual testing)

When

• Before merging a PR, the author runs both test suites and everything has to pass. If the PR touches shared code, the reviewer runs them too.
  • Backend: python manage.py test
  • Android: ./gradlew :app:testDebugUnitTest
• At the end of each iteration we run all tests and both coverage reports on the latest iteration-N-demo commit and update the Results below.
• We don't have CI yet. We plan to run the tests on every PR with GitHub Actions in Iteration 3.

---

2. Unit Testing & Integration Testing (starting from Iteration 2)

Plan

Frameworks

Component	Test framework	Mocking / fakes	Coverage tool
Backend (Django)	Django test runner (unittest), DRF APITestCase	unittest.mock, fake street grid (api/tests/fakes.py), recorded TMAP responses (api/tests/data)	coverage.py 7.16.2
Android	JUnit 4, kotlinx-coroutines-test	FakeApiService, fake elevation API, OkHttp MockWebServer, MainDispatcherRule	JaCoCo (Android Gradle plugin)

Coverage goals by architecture (line coverage)

We set these before measuring anything.

Layer	Target	Why
Backend route algorithm (api/route_algorithm)	85%	Main logic of the app (route shape, elevation, score, briefing). Plain Python, easy to test alone.
Backend API & services (api/views, api/services, api/models.py)	80%	Input checks and error responses. Some branches only happen when an outside service fails.
Android ViewModel	60%	Screen state (options, save, rename, delete). Can run on the JVM with fakes.
Android Model (data.*)	50%	Mostly thin Retrofit / DataStore code. The real logic is in the interceptor and repositories.
Android View (Compose UI)	no target	Needs instrumented UI tests. For now we check screens by hand on the emulator.

Scope & exclusions

• Excluded from backend coverage: api/tests/*, api/migrations/* (backend/.coveragerc).
• api/management/commands/build_street_graph.py is counted but has no target. It downloads a 114MB file, so we run it by hand. The code it calls (street_graph.py) has unit tests.
• Android: the Gradle plugin already leaves out generated code (R, BuildConfig). Compose UI has no target.
• Tests never call outside services. Overpass, Open-Meteo, Open-Elevation, Gemini, Google and TMAP are mocked or faked. The backend uses Django's test database.
• api/tests/test_route_flow.py uses the real Seoul street graph. If backend/data/street_graph/seoul_walk.pkl isn't there, it is skipped.

Reproducible commands

Backend (Python 3.14, from backend/):

python -m venv venv && source venv/bin/activate
pip install -r requirements.txt -r requirements-dev.txt
python manage.py migrate
python manage.py build_street_graph        # needed for the 3 integration tests; otherwise they are skipped
GOOGLE_CLIENT_ID=test coverage run manage.py test
coverage report                            # coverage html for an HTML report

GOOGLE_CLIENT_ID just has to exist (any value works). Other API keys aren't needed because outside calls are mocked.

Android (Gradle 9.5.0 wrapper, JDK 17+, from android_app/):

./gradlew :app:testDebugUnitTest
./gradlew :app:createDebugUnitTestCoverageReport
# report: app/build/reports/coverage/test/debug/index.html

Results

Iteration 2: 2026-10-21, commit 496a542 (branch iteration-2/test/app-unit-tests, PR #21)

Component	Frameworks	Target	Actual	Date	Command / Report
Backend route algorithm	Django test runner, mock, coverage.py	85%	88.0% (574/652) ✅	10/21	coverage run manage.py test
Backend API & services	Django test runner, mock, coverage.py	80%	87.4% (347/397) ✅	10/21	coverage run manage.py test
Android Model	JUnit, MockWebServer, JaCoCo	50%	69.5% (191/275) ✅	10/21	./gradlew :app:createDebugUnitTestCoverageReport
Android ViewModel	JUnit, coroutines-test, JaCoCo	60%	54.6% (288/527) ❌	10/21	./gradlew :app:createDebugUnitTestCoverageReport
Android View	—	—	6.3% (82/1311)	10/21	same

Totals: backend 85% (925/1094 statements), Android 26.5% (561/2113 lines).

Test counts

• Backend: 103 tests, 103 passed, 0 failed, 0 skipped. This is with the Seoul graph built. Without it, the 3 tests in test_route_flow.py are skipped.
• Android: 34 tests, 34 passed, 0 failed, 0 skipped.

Area	Tests	What is checked
Backend test_algorithm.py	11	Loop / point-to-point / shortest-path fallback, 3 distinct options ranked by score, measured climb and signals in the briefing, routes without elevation, traffic signal counting
Backend test_generate_route.py	15	Route API validation, languages, end point, options with their own temporary route IDs, map server errors (503)
Backend test_elevation.py	12	50m sampling, total climb, uphill sections, batching, lookup errors, score formula
Backend test_street_graph.py	7	Walking filter, building from an OSM file, traffic signals, cutting out an area, Overpass fallback
Backend test_route_flow.py (integration)	3	Start point → loop → elevation / signals → scoring → briefing → save → saved list on the real Seoul graph; point-to-point; no elevation
Backend auth / routes / places / LLM / briefing / crossings	55	test_auth 8, test_route_names 5, test_places 11, test_place_translation 8, test_llm 7, test_briefing 10, test_crossings 6
Android RouteViewModelOptionsTest	10	Input validation, options, switching, save state per option, retry, reset
Android MyPageViewModelTest	7	Rename, swipe-to-delete with rollback on failure, reset
Android AuthInterceptorTest	6	Token header, expired-session detection
Android ServerMessageTest	4	Server error → localized message mapping
Android RouteRepositoryTest, RouteGeometryUtilsTest, ExampleUnitTest	7	Slope calculation and route splitting for slope colors

Android ViewModel is below target (54.6%, target 60%)

• AuthViewModel is at 0%. It makes its repository from the Android Application and DataStore, so it can't run in a plain JVM test.
• The elevation summary text in RouteViewModel has no tests yet.
• In Iteration 3 we will pass the repository into AuthViewModel (or add Robolectric) and test sign-up checks and session handling. SoonYoung will add tests for the elevation summary.

Coverage reports

Backend (coverage.py):

Android (JaCoCo):

Android unit test results:

---

3. AI Module Testing (starting from Iteration 3)

To be completed in Iteration 3.

4. Acceptance Testing (starting from Iteration 3)

To be completed in Iteration 3 (plan) and Iteration 4 (results).

5. Alpha & Beta Testing (Iteration 5)

To be completed in Iteration 5.