requirements
Requirements Management in Effect
Overview
The third type parameter in Effect<A, E, R> represents requirements - services and dependencies the effect needs to run:
Effect<Success, Error, Requirements>;
// ^^^^^^^^^^^^ Services needed
Effect uses a powerful dependency injection system based on Context and Layer.
The primary reason to define services is testability. Every external dependency (API calls, databases, file systems, third-party SDKs) MUST be wrapped in a Context.Tag service so that tests can provide a test implementation instead of hitting real systems. This is how Effect achieves 100% test coverage — business logic depends only on service interfaces, and tests swap in test layers that control all I/O.
Defining Services
Using Effect.Tag (Recommended)
import { Effect, Context } from "effect";
// Define service interface and tag together
class UserRepository extends Context.Tag("UserRepository")<
UserRepository,
{
readonly findById: (id: string) => Effect.Effect<User, UserNotFound>;
readonly save: (user: User) => Effect.Effect<void>;
}
>() {}
// Using the service
const program = Effect.gen(function* () {
const repo = yield* UserRepository;
const user = yield* repo.findById("123");
return user;
});
// Type: Effect<User, UserNotFound, UserRepository>
Alternative: Context.Tag Directly
interface UserRepository {
readonly findById: (id: string) => Effect.Effect<User, UserNotFound>;
}
const UserRepository = Context.Tag<UserRepository>("UserRepository");
Using Services
const program = Effect.gen(function* () {
const userRepo = yield* UserRepository;
const emailService = yield* EmailService;
const user = yield* userRepo.findById(userId);
yield* emailService.send(user.email, "Welcome!");
});
Creating Layers
Layers are recipes for building services:
Layer.succeed - Simple Value
const LoggerLive = Layer.succeed(Logger, {
log: (msg) => Effect.sync(() => console.log(msg)),
});
Layer.effect - Effect-Based Construction
const ConfigLive = Layer.effect(
Config,
Effect.gen(function* () {
const env = yield* Effect.sync(() => process.env);
return {
apiUrl: env.API_URL ?? "http://localhost:3000",
debug: env.DEBUG === "true",
};
}),
);
Layer.scoped - Resource with Lifecycle
const DatabaseLive = Layer.scoped(
Database,
Effect.gen(function* () {
const pool = yield* Effect.acquireRelease(createPool(), (pool) => Effect.promise(() => pool.end()));
return {
query: (sql) => Effect.promise(() => pool.query(sql)),
};
}),
);
Layer.function - From Function
const HttpClientLive = Layer.function(HttpClient, (baseUrl: string) => ({
get: (path) => Effect.tryPromise(() => fetch(baseUrl + path)),
}));
Providing Dependencies
Effect.provide - Provide Layer
const program = getUserById("123");
const runnable = program.pipe(Effect.provide(AppLive));
await Effect.runPromise(runnable);
Effect.provideService - Provide Single Service
const runnable = program.pipe(
Effect.provideService(UserRepository, {
findById: (id) => Effect.succeed(mockUser),
save: (user) => Effect.void,
}),
);
Composing Layers
Layer.merge - Combine Independent Layers
const InfraLive = Layer.merge(DatabaseLive, LoggerLive);
Layer.provide - Layer Dependencies
const UserRepositoryLive = Layer.effect(
UserRepository,
Effect.gen(function* () {
const db = yield* Database;
return {
findById: (id) => db.query(`SELECT * FROM users WHERE id = ${id}`),
};
}),
);
const FullUserRepo = UserRepositoryLive.pipe(Layer.provide(DatabaseLive));
Layer.provideMerge - Provide and Keep
const Combined = UserRepositoryLive.pipe(Layer.provideMerge(DatabaseLive));
Building Application Layers
Typical Pattern
const InfraLive = Layer.mergeAll(DatabaseLive, LoggerLive, HttpClientLive);
const RepositoryLive = Layer.mergeAll(UserRepositoryLive, OrderRepositoryLive).pipe(Layer.provide(InfraLive));
const ServiceLive = Layer.mergeAll(UserServiceLive, OrderServiceLive).pipe(Layer.provide(RepositoryLive));
const AppLive = ServiceLive.pipe(Layer.provide(InfraLive));
Layer Memoization
Layers are memoized by default - each service is created once:
const AppLive = Layer.mergeAll(UserServiceLive, OrderServiceLive).pipe(Layer.provide(DatabaseLive));
Fresh Layers (No Memoization)
const FreshDatabase = Layer.fresh(DatabaseLive);
Default Services
Effect provides default implementations for common services:
const program = Effect.gen(function* () {
const now = yield* Clock.currentTimeMillis;
const random = yield* Random.next;
});
Overriding Defaults
import { TestClock } from "effect";
const testProgram = program.pipe(Effect.provide(TestClock.layer));
Testing with Services (CRITICAL for 100% Coverage)
Every service MUST have a test layer. This is how you achieve complete test coverage without hitting real external systems.
Simple Test Layer (Stateless)
Use Layer.succeed for services that don't need to track state:
const EmailServiceTest = Layer.succeed(EmailService, {
send: (to, subject, body) => Effect.void, // No-op in tests
sendBulk: (recipients, subject, body) => Effect.void,
});
Stateful Test Layer (Repositories)
Use Layer.effect with Ref for services that need to maintain state:
const UserRepositoryTest = Layer.effect(
UserRepository,
Effect.gen(function* () {
const store = yield* Ref.make<Map<string, User>>(new Map());
return {
findById: (id: string) =>
Effect.gen(function* () {
const users = yield* Ref.get(store);
return yield* Option.match(Option.fromNullable(users.get(id)), {
onNone: () => Effect.fail(new UserNotFound({ userId: id })),
onSome: Effect.succeed,
}).pipe(Effect.flatten);
}),
save: (user: User) => Ref.update(store, (m) => new Map(m).set(user.id, user)),
};
}),
);
Composing Test Layers
const TestEnv = Layer.mergeAll(UserRepositoryTest, EmailServiceTest, PaymentGatewayTest);
Using Test Layers with @effect/vitest
import { it, expect, layer } from "@effect/vitest";
import { Effect } from "effect";
layer(TestEnv)("UserService", (it) => {
it.effect("should create user and send welcome email", () =>
Effect.gen(function* () {
const repo = yield* UserRepository;
const email = yield* EmailService;
const user = new User({ id: "1", name: "Alice", email: "alice@test.com" });
yield* repo.save(user);
yield* email.send(user.email, "Welcome!", "Hello Alice");
const found = yield* repo.findById("1");
expect(found.name).toBe("Alice");
}),
);
});
Combining Test Layers with Property Testing
The ultimate testing pattern — service test layers control all I/O, Arbitrary generates all data:
layer(TestEnv)("UserService Properties", (it) => {
it.effect.prop("should save and retrieve any valid user", [Arbitrary.make(User)], ([user]) =>
Effect.gen(function* () {
const repo = yield* UserRepository;
yield* repo.save(user);
const found = yield* repo.findById(user.id);
expect(found).toEqual(user);
}),
);
});
Best Practices
- Define service interface with Tag - Keeps interface and tag together
- ALWAYS create a test layer for every service - This is required, not optional. Without test layers, your code is untestable.
- Wrap ALL external dependencies in services - API calls, database queries, file I/O, third-party SDKs, email, caches, queues — everything external MUST go through a service
- Use Layer.scoped for resources - Ensures proper cleanup
- Compose layers bottom-up - Infrastructure → Repositories → Services
- Keep layers focused - One service per layer typically
- Name convention -
*Livefor production layers,*Testfor test layers (e.g.,UserRepositoryLive,UserRepositoryTest) - Use
Layer.effectwithReffor stateful test layers - Repositories and caches need state tracking - Combine test layers with Arbitrary - Services control I/O, Arbitrary generates data — together they enable 100% coverage
Additional Resources
For comprehensive requirements management documentation, consult ${CLAUDE_PLUGIN_ROOT}/references/llms-full.txt.
Search for these sections:
- "Managing Services" for service patterns
- "Managing Layers" for layer composition
- "Layer Memoization" for sharing services
- "Default Services" for built-in services
More from andrueandersoncs/claude-skill-effect-ts
schema
This skill should be used when the user asks about "Effect Schema", "Schema.Struct", "Schema.decodeUnknown", "data validation", "parsing", "Schema.transform", "Schema filters", "Schema annotations", "JSON Schema", "Schema.Class", "Schema branded types", "encoding", "decoding", "Schema.parseJson", or needs to understand how Effect handles data validation and transformation.
13testing
This skill should be used when the user asks about "Effect testing", "@effect/vitest", "it.effect", "it.live", "it.scoped", "it.layer", "it.prop", "Schema Arbitrary", "property-based testing", "fast-check", "TestClock", "testing effects", "mocking services", "test layers", "TestContext", "Effect.provide test", "time testing", "Effect test utilities", "unit testing Effect", "generating test data", "flakyTest", "test coverage", "100% coverage", "service testing", "test doubles", "mock services", or needs to understand how to test Effect-based code.
13traits
This skill should be used when the user asks about "Effect Equal", "Effect Hash", "Equivalence", "Order", "structural equality", "custom equality", "comparing objects", "sorting", "Equal.equals", "Hash.hash", "Equivalence.make", "Order.lessThan", "comparable types", or needs to understand how Effect handles equality, hashing, and ordering of values.
12configuration
This skill should be used when the user asks about "Effect Config", "environment variables", "configuration management", "Config.string", "Config.number", "ConfigProvider", "Config.nested", "Config.withDefault", "Config.redacted", "sensitive values", "config validation", "loading config from JSON", "config schema", or needs to understand how Effect handles application configuration.
12concurrency
This skill should be used when the user asks about "Effect concurrency", "fibers", "Fiber", "forking", "Effect.fork", "Effect.forkDaemon", "parallel execution", "Effect.all concurrency", "Deferred", "Queue", "PubSub", "Semaphore", "Latch", "fiber interruption", "Effect.race", "Effect.raceAll", "concurrent effects", or needs to understand how Effect handles parallel and concurrent execution.
11observability
This skill should be used when the user asks about "Effect logging", "Effect.log", "Effect metrics", "Effect tracing", "spans", "telemetry", "Metric.counter", "Metric.gauge", "Metric.histogram", "OpenTelemetry", "structured logging", "log levels", "Effect.logDebug", "Effect.logInfo", "Effect.logWarning", "Effect.logError", or needs to understand how Effect handles logging, metrics, and distributed tracing.
10