DEV Community

Cover image for Stop Paying the build_runner Tax: Why I Refuse to Use Mockito in Modern Dart
Google Developer Experts profile image Randal L. Schwartz
Randal L. Schwartz for Google Developer Experts

Posted on

Stop Paying the build_runner Tax: Why I Refuse to Use Mockito in Modern Dart

#dart #flutter #testing #architecture

This is Part 1 of the Dart and Flutter series—practical guides, architectural deep dives, and hard-earned engineering lessons from the field. Each article is completely standalone.

Picture this familiar scene:

You’re deep in the zone. You’re practicing disciplined Test-Driven Development (TDD). You write a failing test, write the minimal code to satisfy it, and refactor. Red, green, refactor. Your fingers are flying.

Then, you add a single new method to your UserRepository interface:

Future<void> updatePreferences(UserPreferences prefs);
Enter fullscreen mode Exit fullscreen mode

You switch back to your test file to stub it, and everything grinds to a dead stop. Your IDE lights up with red squiggly lines because the generated mock doesn't know about updatePreferences yet.

You sigh, drop into the terminal, and type the dreaded incantation:

dart run build_runner build --delete-conflicting-outputs
Enter fullscreen mode Exit fullscreen mode

Now you wait. 15 seconds. 30 seconds. On a large enterprise Flutter codebase, maybe two full minutes. The fan spins up. Your train of thought derails. By the time the builder finishes generating boilerplate Dart files, you've lost your flow state.

Why on earth are we running heavyweight compile-time code generation just to create a mock object for a unit test?

It’s time to stop using mockito in modern Dart. There is a much better way.

How Did We End Up Here?

To be fair to the mockito maintainers, mockito didn’t set out to be slow.

In the pre-null-safety era of Dart, mockito was the undisputed king. You created a mock class with class MockRepo extends Mock implements Repo {}, and Dart’s runtime dynamic invocation (noSuchMethod) handled everything under the hood with zero code generation. It was snappy and painless.

Then came Sound Null Safety in Dart 2.12.

Null safety is one of the best things that ever happened to Dart, but it broke mockito's runtime magic. In a soundly typed null-safe language, if an interface method declares that it returns a non-nullable User, noSuchMethod cannot simply return null while waiting for your when(...) stub to register. Returning null violates the type system and throws an immediate runtime error before your stubbing logic even runs.

Because Dart has deliberately avoided heavy runtime reflection (dart:mirrors is disabled in Flutter for performance and tree-shaking reasons), mockito chose the only route available to it at the time: compile-time code generation via build_runner.

It solved the type-safety problem, but it introduced the dreaded build_runner tax.

The Three Sins of Modern Mockito

Relying on mockito today saddles your codebase with three distinct engineering bottlenecks:

1. The Iteration Tax (Destroying the TDD Flow)

TDD relies on tight feedback loops measured in milliseconds. When every interface change or mock signature update requires running code generation, the feedback loop stretches into tens of seconds. Developers stop running tests frequently. They batch their changes, commit without verifying, and run tests as an afterthought. Velocity plummets.

2. Git Diff Pollution & Merge Conflict Hell

For every test file with mocks, mockito generates a .mocks.dart sibling file. These files are thousands of lines of machine-generated boilerplate.

  • They pollute pull request reviews.
  • They clutter repository search results.
  • When two developers touch overlapping service interfaces on separate branches, merging their generated .mocks.dart files triggers brutal, unreadable merge conflicts.

3. Brittle Annotation Coupling

With mockito, your tests are coupled to global library annotations:

// test/user_service_test.dart
import 'user_service_test.mocks.dart';

@GenerateMocks([HttpClient, UserRepository])
void main() {
  // ...
}
Enter fullscreen mode Exit fullscreen mode

Want to mock a third dependency? You have to edit the annotation list, rerun build_runner, wait for the file system to update, and import the generated symbols. It's clunky and mechanical.

Enter Mocktail: Zero-Config Runtime Mocking

Enter mocktail, created by Felix Angelov (of Bloc and Very Good Ventures fame).

mocktail delivers the exact same intuitive API you know from mockito, but with zero code generation. No build_runner. No .mocks.dart files. No annotations.

How? By leveraging two native features of Dart:

  1. Dart’s implicit interfaces: In Dart, every class implicitly defines an interface. You can implement any class without inheriting its implementation.
  2. Closures for stubbing: Instead of evaluating the mock invocation directly (which triggers the null-safety problem), mocktail wraps the invocation inside a closure: () => mock.getUser(). This defers execution until mocktail can safely intercept and return the registered stub or a fallback value.

Here is how you declare a mock in mocktail:

import 'package:mocktail/mocktail.dart';

// No annotations. No build_runner. Just pure, clean Dart.
class MockUserRepository extends Mock implements UserRepository {}
class MockHttpClient extends Mock implements HttpClient {}
Enter fullscreen mode Exit fullscreen mode

That’s it. You define the class right in your test file (or in a shared test_helpers.dart), hit save, and run your tests instantly.

Side-by-Side: Mockito vs. Mocktail

The migration from mockito to mocktail requires practically zero mental re-learning:

Feature mockito mocktail 🍹
Code Generation Requires build_runner None (Pure runtime)
Declaration @GenerateMocks([UserRepo]) class MockUserRepo extends Mock implements UserRepo {}
Synchronous Stub when(repo.name).thenReturn('Randal') when(() => repo.name).thenReturn('Randal')
Asynchronous Stub when(repo.fetch()).thenAnswer((_) async => user) when(() => repo.fetch()).thenAnswer((_) async => user)
Argument Matchers anyNamed('id'), anyString any(named: 'id'), any()
Verification verify(repo.login()).called(1) verify(() => repo.login()).called(1)

Notice the only primary syntax difference: mocktail passes the invocation as an anonymous function () => mock.method().

That simple closure is the secret sauce that bypasses the need for compile-time code generation.

The One "Gotcha": Custom Type Fallbacks

To be an objective engineer, you must evaluate the trade-offs. mocktail has exactly one gotcha, and once you understand it, it takes 5 seconds to address.

The Problem

When you use flexible argument matchers like any() on a method that accepts a non-nullable custom type:

when(() => mockRepo.saveUser(any())).thenAnswer((_) async => true);
Enter fullscreen mode Exit fullscreen mode

Dart’s sound type system insists that something matching the type User must be passed into saveUser during internal matcher registration. Under the hood, mocktail cannot synthesize an instance of your custom User class out of thin air. If it passed null, Dart would throw a type error.

If you don't tell mocktail how to satisfy that type, it will throw an informative exception:

Bad state: A test tried to use any() or captureAny() on a User which was not registered.
Register a fallback value using `registerFallbackValue`.
Enter fullscreen mode Exit fullscreen mode

The Solution

You define a dummy Fake and register it in setUpAll:

import 'package:flutter_test/flutter_test.dart';
import 'package:mocktail/mocktail.dart';

// 1. Create a Fake implementation that satisfies the type checker
class FakeUser extends Fake implements User {}

void main() {
  setUpAll(() {
    // 2. Register it once before any tests run
    registerFallbackValue(FakeUser());
  });

  test('saves user profile', () async {
    final repo = MockUserRepository();
    when(() => repo.saveUser(any())).thenAnswer((_) async => true);

    final service = UserService(repo);
    final result = await service.updateProfile(User(id: '123', name: 'Randal'));

    expect(result, isTrue);
    verify(() => repo.saveUser(any())).called(1);
  });
}
Enter fullscreen mode Exit fullscreen mode

That’s all there is to it. A two-line Fake registration in setUpAll completely eliminates the need to run build_runner on your entire project.

How to Migrate Without Boiling the Ocean

If you’re staring at an existing codebase with dozens of build_runner-powered .mocks.dart files, you don't need to rewrite everything in a single frantic weekend PR.

Take an incremental approach:

  1. Add mocktail to dev_dependencies: 
   dev_dependencies:
     mocktail: ^1.0.4
Enter fullscreen mode Exit fullscreen mode
  1. Migrate one test file at a time:
    • Whenever you open an existing test file to add or modify a test, migrate just that file.
    • Delete the .mocks.dart import.
    • Replace @GenerateMocks with simple class MockX extends Mock implements X {} declarations.
    • Wrap your when and verify invocations in () => ....
    • Delete the obsolete .mocks.dart file from disk.
  2. Commit the diff: Notice how clean and readable your git diff is without hundreds of lines of autogenerated noise.

The Verdict

The official Flutter documentation and cookbooks still mention mockito largely due to historical inertia. But we no longer live in 2020.

In modern Dart, testing should be fast, expressive, and friction-free. Waiting on build_runner just to verify that a service method was called with the right argument is a waste of your machine's CPU cycles and your mental bandwidth.

Switch to mocktail. Your TDD cycle will thank you.

What's your take?

Are you still running build_runner for your mocks, or have you already made the switch to mocktail (or hand-rolled fakes)? Let me know in the comments below!

Randal L. Schwartz is a Google Developer Expert (GDE) for Dart & Flutter and veteran software architect.

Top comments (0)