patrol-write-test

Order of actions when writing new tests

1. Inspect existing test code for functions that can be reused
2. Also think if one of existing functions can be adjusted to match its existing usage and new test
3. Assign test keys to required elements if they are not assigned yet
4. Start writing test: reuse existing functions + put new test steps in new test file
5. Write patrol actions directly in the test file, do not create new methods
6. Run the test frequently during development — don't wait until the full test is written. Run after completing each logical group of steps to catch failures early
7. After full test passes, reorganize new code into reusable functions
8. Rerun the test after reorganizing to confirm it still passes

Patrol MCP Usage

When working with Patrol tests:

• Use patrol-run({ "testFile": "patrol_test/your_test.dart" }) to run tests and wait for completion
• If no session running: starts new session with specified test file
• If session already running: automatically restarts current tests
• Use patrol-screenshot({ "platform": "android" }) or patrol-screenshot({ "platform": "ios" }) to capture screenshots for debugging test failures
• Use patrol-quit({}) to quit the session gracefully
• Use patrol-status({}) to check current status and recent output
• Use patrol-native-tree({}) to fetch the current native UI tree hierarchy for writing native interactions and interactions with apps other than the app under test.

Patrol Tests Rules

Patrol API

• Any file that directly uses Patrol APIs ($(), .scrollTo(), .tap(), .enterText(), .waitUntilVisible(), etc.) should import 'package:patrol/patrol.dart';
• ALWAYS inspect Patrol API before implementing test actions:
• Search codebase for existing Patrol API usage patterns
• Check $.platform APIs for the specific action
• If method not found in codebase, check: https://patrol.leancode.co/
• Only implement after confirming the correct API method
• ALWAYS inspect $.platform methods before implementing test actions
• Don't use flutter_test package. Use only patrol api
• Only run tests with Patrol MCP server for single test and patrol test for all tests. Never use flutter test command
• Do not write patrolSetUp and patrolTearDown methods on your own

Action Rules

• Don't use $.pump, waitUntilVisible or waitUntilExists and other wait methods after or before tap, scrollTo and enterText. Patrol handles it automatically. Do this only at the end of the test
• To find widgets, use only keys
• Don't write try catch blocks unless absolutely necessary
• After writing test check if it works by running it with MCP server, fix it if it fails
• If test fails because element was not found, check if it need to be scrolled to in the app code and adjust the test if needed

Assertion Rules

• Don't write assertions after actions, but do write them at the end of the test
• Prefer using waitUntilVisible as assertion at the end of the test
• Use expect() for assertions only when waitUntilVisible is not enough

Native Dialog Handling

• ALWAYS handle native dialogs that appear during the flow:
• Handle dialogs immediately after the action that triggers them
• For native permissions prefer $.platform.mobile.grantPermissionWhenInUse over $.platform.mobile.tap

Test Keys

• Assign a key ONLY to widgets involved in testing
• ONLY add the key parameter to existing widgets
• NEVER change widget signatures
• NEVER refactor existing code structure
• NEVER hardcode keys in the app, you must use keys from the keys file that is shared between app and tests
• NEVER create new widgets in the app
• NEVER create a key that is not assigned to a widget
• Always make sure that each key value is unique
• ALWAYS sort keys alphabetically
• Add keys as first parameter to the widget constructor

• If widget is not unique (for example generated from a list) use a parameterized key

• ALWAYS prefer using enums or DTOs as the parameter if they already exist
• Use existing widget properties for parameterized keys
• NEVER assign a parameterized key in the app and then use fixed values for it in the keys file (and vice versa)
• NEVER create helper methods, use parameterized keys instead
• When widgets are generated from existing enums or DTOs, always use parameterized keys with the those enum/DTO values as the parameter

Use individual keys when:

• Widgets are hardcoded and known at compile time
• Widgets have distinct, meaningful names

Use parameterized keys when:

• Widgets are generated from dynamic data
• Widgets are generated from a DTO or enums
• Number of widgets is variable or large
• Widgets are generated in loops or from lists