mockery

mockery provides the ability to easily generate mocks for golang interfaces using the stretchr/testify/mock package. It removes the boilerplate coding required to use mocks.

NOTE ON Golang 1.18

Mockery currently does not support various 1.18 Golang features, namely generics. PRs are welcome!

Table of Contents

• Installation
  • Github Release
  • Docker
  • Homebrew
  • go install
• Examples
  • Simplest case
  • Next level case
• Return Value Provider Functions
  • Requirements
  • Notes
• Expecter Interfaces
• Mock constructors
• Extended Flag Descriptions
• Mocking interfaces in main
• Configuration
  • Example
• Semantic Versioning
• Stargazers

Installation

Github Release

Visit the releases page to download one of the pre-built binaries for your platform.

Docker

Use the Docker image

docker pull vektra/mockery

Homebrew

Install through brew

brew install mockery
brew upgrade mockery

go install

Alternatively, you can use the go install method:

go install github.com/vektra/mockery/v2@latest

Examples

Simplest case

Given this is in string.go

package test

type Stringer interface {
	String() string
}

Run: mockery --name=Stringer and the following will be output to mocks/Stringer.go:

package mocks

import (
	"github.com/stretchr/testify/mock"

	testing "testing"
)

type Stringer struct {
	mock.Mock
}

func (m *Stringer) String() string {
	ret := m.Called()

	var r0 string
	if rf, ok := ret.Get(0).(func() string); ok {
		r0 = rf()
	} else {
		r0 = ret.Get(0).(string)
	}

	return r0
}

// NewStringer creates a new instance of Stringer. It also registers a cleanup function to assert the mocks expectations.
func NewStringer(t testing.TB) *Stringer {
	mock := &Stringer{}

	t.Cleanup(func() { mock.AssertExpectations(t) })

	return mock
}

Function type case

Given this is in send.go

package test

type SendFunc func(data string) (int, error)

Run: mockery --name=SendFunc and the following will be output to mocks/SendFunc.go:

package mocks

import (
	"github.com/stretchr/testify/mock"

	testing "testing"
)

type SendFunc struct {
	mock.Mock
}

func (_m *SendFunc) Execute(data string) (int, error) {
	ret := _m.Called(data)

	var r0 int
	if rf, ok := ret.Get(0).(func(string) int); ok {
		r0 = rf(data)
	} else {
		r0 = ret.Get(0).(int)
	}

	var r1 error
	if rf, ok := ret.Get(1).(func(string) error); ok {
		r1 = rf(data)
	} else {
		r1 = ret.Error(1)
	}

	return r0, r1
}

// NewSendFunc creates a new instance of SendFunc. It also registers a cleanup function to assert the mocks expectations.
func NewSendFunc(t testing.TB) *SendFunc {
	mock := &SendFunc{}

	t.Cleanup(func() { mock.AssertExpectations(t) })

	return mock
}

Next level case

See github.com/jaytaylor/mockery-example for the fully runnable version of the outline below.

package main

import (
	"fmt"

	"github.com/aws/aws-sdk-go/aws"
	"github.com/aws/aws-sdk-go/service/s3"
	"github.com/jaytaylor/mockery-example/mocks"
	"github.com/stretchr/testify/mock"
)

func main() {
	mockS3 := &mocks.S3API{}

	mockResultFn := func(input *s3.ListObjectsInput) *s3.ListObjectsOutput {
		output := &s3.ListObjectsOutput{}
		output.SetCommonPrefixes([]*s3.CommonPrefix{
			&s3.CommonPrefix{
				Prefix: aws.String("2017-01-01"),
			},
		})
		return output
	}

	// NB: .Return(...) must return the same signature as the method being mocked.
	//     In this case it's (*s3.ListObjectsOutput, error).
	mockS3.On("ListObjects", mock.MatchedBy(func(input *s3.ListObjectsInput) bool {
		return input.Delimiter != nil && *input.Delimiter == "/" && input.Prefix == nil
	})).Return(mockResultFn, nil)

	listingInput := &s3.ListObjectsInput{
		Bucket:    aws.String("foo"),
		Delimiter: aws.String("/"),
	}
	listingOutput, err := mockS3.ListObjects(listingInput)
	if err != nil {
		panic(err)
	}

	for _, x := range listingOutput.CommonPrefixes {
		fmt.Printf("common prefix: %+v\n", *x)
	}
}

Return Value Provider Functions

If your tests need access to the arguments to calculate the return values, set the return value to a function that takes the method's arguments as its own arguments and returns the return value. For example, given this interface:

package test

type Proxy interface {
  passthrough(ctx context.Context, s string) string
}

The argument can be passed through as the return value:

import . "github.com/stretchr/testify/mock"

proxyMock := mocks.NewProxy(t)
proxyMock.On("passthrough", mock.AnythingOfType("context.Context"), mock.AnythingOfType("string")).
	Return(func(ctx context.Context, s string) string {
		return s
	})

Requirements

Return must be passed the same argument count and types as expected by the interface. Then, for each of the return values of the mocked function, Return needs a function which takes the same arguments as the mocked function, and returns one of the return values. For example, if the return argument signature of passthrough in the above example was instead (string, error) in the interface, Return would also need a second function argument to define the error value:

type Proxy interface {
  passthrough(ctx context.Context, s string) (string, error)
}

proxyMock := mocks.NewProxy(t)
proxyMock.On("passthrough", mock.AnythingOfType("context.Context"), mock.AnythingOfType("string")).
	Return(
		func(ctx context.Context, s string) string {
			return s
		},
		func(ctx context.Context, s string) error {
			return nil
		},
	)

Note that the following is incorrect (you can't return all the return values with one function):

proxyMock := mocks.NewProxy(t)
proxyMock.On("passthrough", mock.AnythingOfType("context.Context"), mock.AnythingOfType("string")).
	Return(func(ctx context.Context, s string) (string, error) {
		return s, nil
	})

If any return argument is missing, github.com/stretchr/testify/mock.Arguments.Get will emit a panic.

For example, panic: assert: arguments: Cannot call Get(0) because there are 0 argument(s). [recovered] indicates that Return was not provided any arguments but (at least one) was expected based on the interface. Get(1) would indicate that the Return call is missing a second argument, and so on.

Notes

This approach should be used judiciously, as return values should generally not depend on arguments in mocks; however, this approach can be helpful for situations like passthroughs or other test-only calculations.

Expecter Interfaces

New in v2.10.0.

Mockery now supports an "expecter" interface which allows your tests to use type-safe methods to generate call expectations. When enabled through the with-expecter: True mockery configuration, you can enter into the expecter interface by simply calling .EXPECT() on your mock object.

For example, given an interface such as

type Requester interface {
	Get(path string) (string, error)
}

You can use the type-safe expecter interface as such:

requesterMock := mocks.NewRequester(t)
requesterMock.EXPECT().Get("some path").Return("result", nil)
requesterMock.EXPECT().
	Get(mock.Anything).
	Run(func(path string) { fmt.Println(path, "was called") }).
	// Can still use return functions by getting the embedded mock.Call
	Call.Return(func(path string) string { return "result for " + path }, nil)

Mock constructors

New in v2.11.0.

Mockery v2.11 introduces constructors for all mocks. This makes instantiation and mock registration a bit easier and less error-prone (you won't have to worry about forgetting the AssertExpectations method call anymore).

Before v2.11:

factory := &mocks.Factory{}
defer factory.AssertExpectations(t)

After v2.11:

factory := mocks.NewFactory(t)

The AssertExpectations method is registered automatically inside the constructor via t.Cleanup() method.

Extended Flag Descriptions

The following descriptions provide additional elaboration on a few common parameters.

flag name	description
--name	The --name option takes either the name or matching regular expression of interface to generate mock(s) for.
--all	It's common for a big package to have a lot of interfaces, so mockery provides --all. This option will tell mockery to scan all files under the directory named by --dir ("." by default) and generates mocks for any interfaces it finds. This option implies --recursive=true.
--recursive	Use the --recursive option to search subdirectories for the interface(s). This option is only compatible with --name. The --all option implies --recursive=true.
--output	mockery always generates files with the package mocks to keep things clean and simple. You can control which mocks directory is used by using --output, which defaults to ./mocks.
--inpackage and --keeptree	For some complex repositories, there could be multiple interfaces with the same name but in different packages. In that case, --inpackage allows generating the mocked interfaces directly in the package that it mocks. In the case you don't want to generate the mocks into the package but want to keep a similar structure, use the option --keeptree.
--filename	Use the --filename and --structname to override the default generated file and struct name. These options are only compatible with non-regular expressions in --name, where only one mock is generated.
--case	mockery generates files using the casing of the original interface name. This can be modified by specifying --case underscore to format the generated file name using underscore casing.
--print	Use mockery --print to have the resulting code printed out instead of written to disk.
--exported	Use mockery --exported to generate public mocks for private interfaces.
--with-expecter	Use mockery --with-expecter to generate EXPECT() methods for your mocks. This is the preferred way to setup your mocks.

Mocking interfaces in main

When your interfaces are in the main package you should supply the --inpackage flag. This will generate mocks in the same package as the target code avoiding import issues.

Configuration

mockery uses spf13/viper under the hood for its configuration parsing. It is bound to three different configuration sources, in order of decreasing precedence:

1. Command line
2. Environment variables
3. Configuration file

Example

$ export MOCKERY_STRUCTNAME=config_from_env
$ echo $MOCKERY_STRUCTNAME
config_from_env
$ grep structname .mockery.yaml
structname: config_from_file
$ ./mockery showconfig --structname config_from_cli | grep structname
Using config file: /home/ltclipp/git/vektra/mockery/.mockery.yaml
structname: config_from_cli
$ ./mockery showconfig  | grep structname
Using config file: /home/ltclipp/git/vektra/mockery/.mockery.yaml
structname: config_from_env
$ unset MOCKERY_STRUCTNAME
$ ./mockery showconfig  | grep structname
Using config file: /home/ltclipp/git/vektra/mockery/.mockery.yaml
structname: config_from_file

By default it searches the current working directory for a file named .mockery.[extension] where [extension] is any of the recognized extensions.

Semantic Versioning

The versioning in this project applies only to the behavior of the mockery binary itself. This project explicitly does not promise a stable internal API, but rather a stable executable. The versioning applies to the following:

1. CLI arguments.
2. Parsing of Golang code. New features in the Golang language will be supported in a backwards-compatible manner, except during major version bumps.
3. Behavior of mock objects. Mock objects can be considered to be part of the public API.
4. Behavior of mockery given a set of arguments.

What the version does not track:

1. The interfaces, objects, methods etc. in the vektra/mockery package.
2. Compatibility of go get-ing mockery with new or old versions of Golang.

Development Efforts

v2 is in a soft change freeze due to the complexity of the software and the fact that functionality addition generally requires messing with logic that has been thoroughly tested, but is sensitive to change.

v1

v1 is the original version of the software, and is no longer supported.

v2

mockery is currently in v2, which iterates on v1 and includes mostly cosmetic and configuration improvements.

v3

v3 will include a ground-up overhaul of the entire codebase and will completely change how mockery works internally and externally. The highlights of the project are:

• Moving towards a package-based model instead of a file-based model. mockery currently iterates over every file in a project and calls package.Load on each one, which is time consuming. Moving towards a model where the entire package is loaded at once will dramtically reduce runtime, and will simplify logic. Additionally, supporting only a single mode of operation (package mode) will greatly increase the intuitiveness of the software.
• Configuration-driven generation. v3 will be entirely driven by configuration, meaning:
  • You specify the packages you want mocked, instead of relying on it auto-discovering your package. Auto-discovery in theory sounds great, but in practice it leads to a great amount of complexity for very little benefit.
  • Package- or interface-specific overrides can be given that change mock generation settings on a granular level. This will allow your mocks to be generated in a heterogenous manner, and will be made explicit by yaml configuration.
• Proper error reporting. Errors across the board will be done in accordance with modern Golang practices
• Variables in generated mocks will be given meaningful names.

Stargazers