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.
- Installation
- Examples
- Return Value Provider Functions
- Expecter Interfaces
- Mock constructors
- Extended Flag Descriptions
- Mocking interfaces in
main
- Configuration
- Semantic Versioning
- Stargazers
Visit the releases page to download one of the pre-built binaries for your platform.
Use the Docker image
docker pull vektra/mockery
Generate all the mocks for your project:
docker run -v "$PWD":/src -w /src vektra/mockery --all
Install through brew
brew install mockery
brew upgrade mockery
Supported, but not recommended: see wiki page and related discussions.
Alternatively, you can use the go install method to compile the project using your local environment:
go install github.com/vektra/mockery/v2@latest
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 testing interface on the mock and a cleanup function to assert the mocks expectations.
func NewStringer(t testing.TB) *Stringer {
mock := &Stringer{}
mock.Mock.Test(t)
t.Cleanup(func() { mock.AssertExpectations(t) })
return mock
}
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 testing interface on the mock and a cleanup function to assert the mocks expectations.
func NewSendFunc(t testing.TB) *SendFunc {
mock := &SendFunc{}
mock.Mock.Test(t)
t.Cleanup(func() { mock.AssertExpectations(t) })
return mock
}
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)
}
}
For mockery to correctly generate mocks, the command has to be run on a module (i.e. your project has to have a go.mod file)
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
})
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.
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.
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)
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{}
factory.Test(t) // so that mock does not panic when a method is unexpected
defer factory.AssertExpectations(t)
After v2.11:
factory := mocks.NewFactory(t)
The constructor sets up common functionalities automatically
- The
AssertExpectations
method is registered to be called at the end of the tests viat.Cleanup()
method. - The testing.TB interface is registered on the
mock.Mock
so that tests don't panic when a call on the mock is unexpected.
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 the 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 . |
--outpkg |
Use --outpkg to specify the package name of the generated 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. |
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.
mockery uses spf13/viper under the hood for its configuration parsing. It is bound to three different configuration sources, in order of decreasing precedence:
- Command line
- Environment variables
- Configuration file
$ 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.
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:
- CLI arguments.
- Parsing of Golang code. New features in the Golang language will be supported in a backwards-compatible manner, except during major version bumps.
- Behavior of mock objects. Mock objects can be considered to be part of the public API.
- Behavior of mockery given a set of arguments.
What the version does not track:
- The interfaces, objects, methods etc. in the vektra/mockery package.
- Compatibility of
go get
-ing mockery with new or old versions of Golang.
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 is the original version of the software, and is no longer supported.
mockery
is currently in v2, which iterates on v1 and includes mostly cosmetic and configuration improvements.
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 callspackage.Load
on each one, which is time-consuming. Moving towards a model where the entire package is loaded at once will dramatically 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 heterogeneous 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.