Wire: Automated Initialization in Go

Wire is a code generation tool that automates connecting components using dependency injection. Dependencies between components are represented in Wire as function parameters, encouraging explicit initialization instead of global variables. Because Wire operates without runtime state or reflection, code written to be used with Wire is useful even for hand-written initialization.

For an overview, see the introductory blog post.

Installing

Install Wire by running:

go get github.com/google/wire/cmd/wire

and ensuring that $GOPATH/bin is added to your $PATH.

Basics

Wire has two core concepts: providers and injectors.

Defining Providers

The primary mechanism in Wire is the provider: a function that can produce a value. These functions are ordinary Go code.

package foobarbaz

type Foo struct {
    X int
}

// ProvideFoo returns a Foo.
func ProvideFoo() Foo {
    return Foo{X: 42}
}

Provider functions must be exported in order to be used from other packages, just like ordinary functions.

Providers can specify dependencies with parameters:

package foobarbaz

// ...

type Bar struct {
    X int
}

// ProvideBar returns a Bar: a negative Foo.
func ProvideBar(foo Foo) Bar {
    return Bar{X: -foo.X}
}

Providers can also return errors:

package foobarbaz

import (
    "context"
    "errors"
)

// ...

type Baz struct {
    X int
}

// ProvideBaz returns a value if Bar is not zero.
func ProvideBaz(ctx context.Context, bar Bar) (Baz, error) {
    if bar.X == 0 {
        return Baz{}, errors.New("cannot provide baz when bar is zero")
    }
    return Baz{X: bar.X}, nil
}

Providers can be grouped into provider sets. This is useful if several providers will frequently be used together. To add these providers to a new set called SuperSet, use the wire.NewSet function:

package foobarbaz

import (
    // ...
    "github.com/google/wire"
)

// ...

var SuperSet = wire.NewSet(ProvideFoo, ProvideBar, ProvideBaz)

You can also add other provider sets into a provider set.

package foobarbaz

import (
    // ...
    "example.com/some/other/pkg"
)

// ...

var MegaSet = wire.NewSet(SuperSet, pkg.OtherSet)

Injectors

An application wires up these providers with an injector: a function that calls providers in dependency order. With Wire, you write the injector's signature, then Wire generates the function's body.

An injector is declared by writing a function declaration whose body is a call to wire.Build. The return values don't matter as long as they are of the correct type. The values themselves will be ignored in the generated code. Let's say that the above providers were defined in a package called example.com/foobarbaz. The following would declare an injector to obtain a Baz:

// +build wireinject
// The build tag makes sure the stub is not built in the final build.

package main

import (
    "context"

    "github.com/google/wire"
    "example.com/foobarbaz"
)

func initializeBaz(ctx context.Context) (foobarbaz.Baz, error) {
    wire.Build(foobarbaz.MegaSet)
    return foobarbaz.Baz{}, nil
}

Like providers, injectors can be parameterized on inputs (which then get sent to providers) and can return errors. Arguments to wire.Build are the same as wire.NewSet: they form a provider set. This is the provider set that gets used during code generation for that injector.

Any non-injector declarations found in a file with injectors will be copied into the generated file.

You can generate the injector by invoking Wire in the package directory:

wire

Wire will produce an implementation of the injector in a file called wire_gen.go that looks something like this:

// Code generated by Wire. DO NOT EDIT.

//go:generate wire
//+build !wireinject

package main

import (
    "example.com/foobarbaz"
)

func initializeBaz(ctx context.Context) (foobarbaz.Baz, error) {
    foo := foobarbaz.ProvideFoo()
    bar := foobarbaz.ProvideBar(foo)
    baz, err := foobarbaz.ProvideBaz(ctx, bar)
    if err != nil {
        return 0, err
    }
    return baz, nil
}

As you can see, the output is very close to what a developer would write themselves. Further, there is little dependency on Wire at runtime: all of the written code is just normal Go code, and can be used without Wire.

Once wire_gen.go is created, you can regenerate it by running go generate.

Advanced Features

The following features all build on top of the concepts of providers and injectors.

Binding Interfaces

Frequently, dependency injection is used to bind a concrete implementation for an interface. Wire matches inputs to outputs via type identity, so the inclination might be to create a provider that returns an interface type. However, this would not be idiomatic, since the Go best practice is to return concrete types. Instead, you can declare an interface binding in a provider set:

type Fooer interface {
    Foo() string
}

type Bar string

func (b *Bar) Foo() string {
    return string(*b)
}

func ProvideBar() *Bar {
    b := new(Bar)
    *b = "Hello, World!"
    return b
}

var BarFooer = wire.NewSet(
    ProvideBar,
    wire.Bind(new(Fooer), new(Bar)))

The first argument to wire.Bind is a pointer to a value of the desired interface type and the second argument is a zero value of the concrete type. Any set that includes an interface binding must also have a provider in the same set that provides the concrete type.

If necessary, you can also bind one interface to another:

type FooerPlus interface {
  Fooer
  Bar() String
}

func ProvideFooerPlus() FooerPlus {
  ...
}

var FooerPlusAsFooer = wire.NewSet(
  ProvideFooerPlus,
  wire.Bind(new(Fooer), *new(FooerPlus)))

Struct Providers

Structs can also be marked as providers. Instead of calling a function, an injector will fill in each field using the corresponding provider. For a given struct type S, this would provide both S and *S. For example, given the following providers:

type Foo int
type Bar int

func ProvideFoo() Foo {
    // ...
}

func ProvideBar() Bar {
    // ...
}

type FooBar struct {
    Foo Foo
    Bar Bar
}

var Set = wire.NewSet(
    ProvideFoo,
    ProvideBar,
    FooBar{})

A generated injector for FooBar would look like this:

func injectFooBar() FooBar {
    foo := ProvideFoo()
    bar := ProvideBar()
    fooBar := FooBar{
        Foo: foo,
        Bar: bar,
    }
    return fooBar
}

And similarly if the injector needed a *FooBar.

Binding Values

Occasionally, it is useful to bind a basic value (usually nil) to a type. Instead of having injectors depend on a throwaway provider function, you can add a value expression to a provider set.

type Foo struct {
    X int
}

func injectFoo() Foo {
    wire.Build(wire.Value(Foo{X: 42}))
    return Foo{}
}

The generated injector would look like this:

func injectFoo() Foo {
    foo := _wireFooValue
    return foo
}

var (
    _wireFooValue = Foo{X: 42}
)

It's important to note that the expression will be copied to the injector's package; references to variables will be evaluated during the injector package's initialization. Wire will emit an error if the expression calls any functions or receives from any channels.

For interface values, use InterfaceValue:

func injectReader() io.Reader {
    wire.Build(wire.InterfaceValue(new(io.Reader), os.Stdin))
    return nil
}

Cleanup functions

If a provider creates a value that needs to be cleaned up (e.g. closing a file), then it can return a closure to clean up the resource. The injector will use this to either return an aggregated cleanup function to the caller or to clean up the resource if a provider called later in the injector's implementation returns an error.

func provideFile(log Logger, path Path) (*os.File, func(), error) {
    f, err := os.Open(string(path))
    if err != nil {
        return nil, nil, err
    }
    cleanup := func() {
        if err := f.Close(); err != nil {
            log.Log(err)
        }
    }
    return f, cleanup, nil
}

A cleanup function is guaranteed to be called before the cleanup function of any of the provider's inputs and must have the signature func().

Alternate Injector Syntax

If you grow weary of writing return foobarbaz.Foo{}, nil at the end of your injector function declaration, you can instead write it more concisely with a panic:

func injectFoo() Foo {
    panic(wire.Build(/* ... */))
}

Best Practices

The following are practices we recommend for using Wire. This list will grow over time.

Distinguishing Types

If you need to inject a common type like string, create a new string type to avoid conflicts with other providers. For example:

type MySQLConnectionString string

Options Structs

A provider function that includes many dependencies can pair the function with an options struct.

type Options struct {
    // Messages is the set of recommended greetings.
    Messages []Message
    // Writer is the location to send greetings. nil goes to stdout.
    Writer io.Writer
}

func NewGreeter(ctx context.Context, opts *Options) (*Greeter, error) {
    // ...
}

var GreeterSet = wire.NewSet(Options{}, NewGreeter)

Provider Sets in Libraries

When creating a provider set for use in a library, the only changes you can make without breaking compatibility are:

• Change which provider a provider set uses to provide a specific output, as long as it does not introduce a new input to the provider set. It may remove inputs. However, note that existing injectors will use the old provider until they are regenerated.
• Introduce a new output type into the provider set, but only if the type itself is newly added. If the type is not new, it is possible that some injector already has the output type included, which would cause a conflict.

All other changes are not safe. This includes:

• Requiring a new input in the provider set.
• Removing an output type from a provider set.
• Adding an existing output type into the provider set.

Instead of making one of these breaking changes, consider adding a new provider set.

As an example, if you have a provider set like this:

var GreeterSet = wire.NewSet(NewStdoutGreeter)

func DefaultGreeter(ctx context.Context) *Greeter {
    // ...
}

func NewStdoutGreeter(ctx context.Context, msgs []Message) *Greeter {
    // ...
}

func NewGreeter(ctx context.Context, w io.Writer, msgs []Message) (*Greeter, error) {
    // ...
}

You may:

• Use DefaultGreeter instead of NewStdoutGreeter in GreeterSet.
• Create a new type T and add a provider for T to GreeterSet, as long as T is introduced in the same commit/release as the provider is added.

You may not:

• Use NewGreeter instead of NewStdoutGreeter in GreeterSet. This both adds an input type (io.Writer) and requires injectors to return an error where the provider of *Greeter did not require this before.
• Remove NewStdoutGreeter from GreeterSet. Injectors depending on *Greeter will be broken.
• Add a provider for io.Writer to GreeterSet. Injectors might already have a provider for io.Writer which might conflict with this one.

As such, you should pick the output types in a library provider set carefully. In general, prefer small provider sets in a library. For example, it is common for a library provider set to contain a single provider function along with a wire.Bind to the interface the return type implements. Avoiding larger provider sets reduces the likelihood that applications will encounter conflicts. To illustrate, imagine your library provides a client for a web service. While it may be tempting to bundle a provider for *http.Client in a provider set for your library's client, doing so would cause conflicts if every library did the same. Instead, the library's provider set should only include the provider for the API client, and let *http.Client be an input of the provider set.

Mocking

There are two approaches for creating an injected app with mocked dependencies. Examples of both approaches are shown here.

Approach A: Pass mocks to the injector

Create a test-only injector that takes all of the mocks as arguments; the argument types must be the interface types the mocks are mocking. wire.Build can't include providers for the mocked dependencies without creating conflicts, so if you're using provider set(s) you will need to define one that doesn't include the mocked types.

Approach B: Return the mocks from the injector

Create a new struct that includes the app plus all of the dependencies you want to mock. Create a test-only injector that returns this struct, give it providers for the concrete mock types, and use wire.Bind to tell Wire that the concrete mock types should be used to fulfill the appropriate interface.

Frequently Asked Questions

How does Wire relate to other Go dependency injection tools?

Other dependency injection tools for Go like dig or facebookgo/inject are based on reflection. Wire runs as a code generator, which means that the injector works without making calls to a runtime library. This enables easier introspection of initialization and correct cross-references for tooling like guru.

How does Wire relate to other non-Go dependency injection tools (like Dagger 2)?

Wire's approach was inspired by Dagger 2. However, it is not the aim of Wire to emulate dependency injection tools from other languages: the design space and requirements are quite different. For example, the Go compiler does not support anything like Java's annotation processing mechanisms. The difference in languages and their idioms necessarily requires different approaches in primitives and API.

Why is Wire part of Go Cloud?

Wire was designed to reduce the toil in setting up the multitude of dependencies often found when interacting with common cloud providers. However, Wire is not coupled to Go Cloud or cloud servers: Wire is a standalone tool.

The Go Cloud team is investigating splitting out Wire as its own repository. Follow #513 for updates.

Why use pseudo-functions to create provider sets or injectors?

In the early prototypes, Wire directives were specially formatted comments. This seemed appealing at first glance as this meant no compile-time or runtime impact. However, this unstructured approach becomes opaque to other tooling not written for Wire. Tools like gorename or guru would not be able to recognize references to the identifiers existing in comments without being specially modified to understand Wire's comment format. By moving the references into no-op function calls, Wire interoperates seamlessly with other Go tooling.

What if my dependency graph has two dependencies of the same type?

This most frequently appears with common types like string. An example of this problem would be:

type Foo struct { /* ... */ }
type Bar struct { /* ... */ }

func newFoo1() *Foo { /* ... */ }
func newFoo2() *Foo { /* ... */ }
func newBar(foo1 *Foo, foo2 *Foo) *Bar { /* ... */ }

func inject() *Bar {
	// ERROR! Multiple providers for *Foo.
	wire.Build(newFoo1, newFoo2, newBar)
	return nil
}

Wire does not allow multiple providers for one type to exist in the transitive closure of the providers presented to wire.Build, as this is usually a mistake. For legitimate cases where you need multiple dependencies of the same type, you need to invent a new type to call this other dependency. For example, you can name OAuth credentials after the service they connect to. Once you have a suitable different type, you can wrap and unwrap the type when plumbing it through Wire. Continuing our above example:

type OtherFoo Foo

func newOtherFoo() *OtherFoo {
	// Call the original provider...
	foo := newFoo2()
	// ...then convert it to the new type.
	return (*OtherFoo)(foo)
}

func provideBar(foo1 *Foo, otherFoo *OtherFoo) *Bar {
	// Convert the new type into the unwrapped type...
	foo2 := (*Foo)(otherFoo)
	// ...then use it to call the original provider.
	return newBar(foo1, foo2)
}

func inject() *Bar {
	wire.Build(newFoo1, newOtherFoo, provideBar)
	return nil
}

Why does Wire forbid including the same provider multiple times?

Wire forbids this to remain consistent with the principle that specifying multiple providers for the same type is an error. On the surface, Wire could permit duplication, but this would introduce a few unintended consequences:

• Wire would have to specify what kinds of duplicates are permissible: are two wire.Value calls ever considered to be the "same"?
• If a provider set changes the function it uses to provide a type, then this could break an application, since it may introduce a new conflict between another provider set that was specifying the "same" provider.

As such, we decided that the simpler behavior would be for this case to be an error, knowing we can always relax this restriction later. The user can always create a new provider set that does not have the conflicting type. A proposed subtract command would automate the toil in this process.

Should I use Wire for small applications?

Probably not. Wire is designed to automate more intricate setup code found in larger applications. For small applications, hand-wiring dependencies is simpler.

Who is using Wire?

Wire is still fairly new and doesn't have a large user base yet. However, we have heard a lot of interest from Go users wanting to simplify their applications. If your project or company uses Wire, please let us know by either emailing us or sending a pull request amending this section.