golang-samber-do

Persona: You are a Go architect setting up dependency injection. You keep the container at the composition root, depend on interfaces not concrete types, and treat provider errors as first-class failures.

Using samber/do for Dependency Injection in Go

Type-safe dependency injection toolkit for Go based on Go 1.18+ generics.

Official Resources:

• pkg.go.dev/github.com/samber/do/v2
• do.samber.dev
• github.com/samber/do/v2

This skill is not exhaustive. Please refer to library documentation and code examples for more information. Context7 can help as a discoverability platform.

DO NOT USE v1 OF THIS LIBRARY. INSTALL v2 INSTEAD:

go get -u github.com/samber/do/v2

Core Concepts

The Injector (Container)

import "github.com/samber/do/v2"

injector := do.New()

Service Types

• Lazy (default): Created when first requested
• Eager: Created immediately when the container starts
• Transient: New instance created on every request
• Value: Pre-created value, no instantiation

Provider Functions

Services MUST be registered via provider functions:

type Provider[T any] func(i Injector) (T, error)

Basic Usage

1. Define and Register Services

Follow "Accept Interfaces, Return Structs":

// Register a service (lazy by default)
do.Provide(injector, func(i do.Injector) (Database, error) {
    return &PostgreSQLDatabase{connString: "postgres://..."}, nil
})

// Register a pre-created value
do.ProvideValue(injector, &Config{Port: 8080})

// Register a transient service (new instance each time)
do.ProvideTransient(injector, func(i do.Injector) (*Logger, error) {
    return &Logger{}, nil
})

// Register an eager service (created immediately at startup)
do.ProvideValue(injector, &Config{Port: 8080})

2. Invoke Services

The container MUST only be accessed at the composition root:

// Invoke with error handling
db, err := do.Invoke[Database](injector)

// MustInvoke panics on error (use when confident service exists)
db := do.MustInvoke[Database](injector)

3. Service Dependencies

func NewUserService(i do.Injector) (UserService, error) {
    db := do.MustInvoke[Database](i)
    cache := do.MustInvoke[Cache](i)
    return &userService{db: db, cache: cache}, nil
}

do.Provide(injector, NewUserService)

4. Implicit Aliasing (Preferred)

Register a concrete type and invoke as an interface without explicit aliasing:

// Register concrete type
do.Provide(injector, func(i do.Injector) (*PostgreSQLDatabase, error) {
    return &PostgreSQLDatabase{}, nil
})

// Invoke directly as interface (implicit aliasing)
db := do.MustInvokeAs[Database](injector)

5. Named Services

Register multiple services of the same type:

do.ProvideNamed(injector, "primary-db", func(i do.Injector) (*Database, error) {
    return &Database{URL: "postgres://primary..."}, nil
})

mainDB := do.MustInvokeNamed[*Database](injector, "primary-db")

Package Organization

Use do.Package() to organize service registration by module:

// infrastructure/package.go
var Package = do.Package(
    do.Lazy(func(i do.Injector) (*postgres.DB, error) {
        cfg := do.MustInvoke[*Config](i)
        return postgres.Connect(cfg.DatabaseURL)
    }),
    do.Lazy(func(i do.Injector) (*redis.Client, error) {
        cfg := do.MustInvoke[*Config](i)
        return redis.NewClient(cfg.RedisURL), nil
    }),
)

// main.go
injector := do.New(infrastructure.Package, service.Package)

Full Application Setup

func main() {
    injector := do.New(
        infrastructure.Package,
        repository.Package,
        service.Package,
        transport.Package,
    )

    server := do.MustInvoke[*http.Server](injector)
    go server.ListenAndServe()

    _ = injector.ShutdownOnSignalsWithContext(context.Background(), os.Interrupt)
}

Best Practices

1. Depend on interfaces, not concrete types — lets you swap implementations in tests without touching production code
2. Each service should have one job — services with multiple responsibilities are harder to test and harder to replace
3. Keep dependency trees shallow — chains beyond 3-4 levels make initialization order fragile and errors harder to trace
4. Handle errors in provider functions — a silently failing provider creates a broken service that crashes later in unexpected places
5. Use scopes to organize services by lifecycle — request-scoped services prevent leaks, global services prevent redundant initialization

For scopes, lifecycle management, struct injection, and debugging, see Advanced Usage.

For testing patterns (cloning, overrides, mocks), see Testing.

Quick Reference

Registration

Invocation

Cross-References

• → See samber/cc-skills-golang@golang-dependency-injection skill for DI concepts, comparison, and when to adopt a DI library
• → See samber/cc-skills-golang@golang-structs-interfaces skill for interface design patterns
• → See samber/cc-skills-golang@golang-testing skill for general testing patterns