CRAN version R-CMD-check Codecov test coverage

A mocking library for R.


To install the latest CRAN release:


To install directly from the source code in this GitHub repository:

# If you don't have devtools installed yet:

# Then:


Mockery provides the capacity for stubbing out functions and for verifying function calls during testing.


Mockery's stub function will let you stub out a function with another function or simply a return value. Note that if you choose to replace the function with another function, the signatures of the two functions should be compatible.

g = function(y) y
f = function(x) g(x) + 1
test_that('demonstrate stubbing', {
    # replaces 'g' with a function that always returns 100
    # but only when called from f
    stub(f, 'g', 100)

    # this can also be written
    stub(f, 'g', function(...) 100)
    expect_equal(f(1), 101)

Stubbing works with classes of all descriptions and namespaced functions:

# this stubs the 'request_perform' function, but only
# for httr::get, and only when it is called from within this
# test function
stub(httr::GET, 'request_perform', 'some html')

# it is also possible to stub out a namespaced function call
stub(some_function, 'namespace::function', 'some return value')

This also works with R6 classes and methods.


It's possible to specify the depth of stubbing. This is useful if you want to stub a function that isn't called directly by the function you call in your test, but is instead called by a function that that function calls.

In the example below, the function g is both called directly from r, which we call from the test, and from f, which r calls. By specifying a depth of 2, we tell mockery to stub g in both places.

g = function(y) y
f = function(x) g(x) + 1
r = function(x) g(x) + f(x)
test_that('demonstrate stubbing', {
    stub(r, 'g', 100, depth=2)
    expect_equal(r(1), 201)

For more examples, please see the test code contained in this repository.

Comparison to with_mock

Mockery's stub function has similar functionality to testthat's with_mock.

There are several use cases in which mockery's stub function will work, but testthat's with_mock will not.

First, unlike with_mock, it seamlessly allows for mocking out primitives.

Second, it is easy to stub out functions from base R packages with mockery's stub. Because of how with_mock works, you can get into trouble if you mock such functions that the JIT compiler might try to use. These kinds of problems are avoided by stub's design. As of version 2.0.0 of testthat, it will be impossible to mock functions from base R packages with_mock.

The functionality of stub is just slightly different than that of with_mock. Instead of mocking out the object of interest for the duration of some code block, it mocks it out only when it is called from a specified function.


Mock objects allow you to specify the behavior of the function you've stubbed out while also verifying how that function was used.

g = function(y) y
f = function(x) g(x) + 1
test_that('demonstrate mock object usage', {
    # mocks can specify behavior
    mock = mock(100)
    stub(f, 'g', mock)
    result = f(5)
    expect_equal(result, 101)

    # and allow you to make assertions on the mock was treated
    expect_called(mock, 1)
    expect_args(mock, 1, 5)

You can also specify multiple return values

mock = mock(1, "a", sqrt(3))

and access the arguments with which it was called.

mock <- mock()
mock(x = 1)
mock(y = 2)

expect_equal(length(mock), 2)
args <- mock_args(mock)

expect_equal(args[[1]], list(x = 1))
expect_equal(args[[2]], list(y = 2))

Please report bugs and feature requests through github issues.

Try the mockery package in your browser

Any scripts or data that you put into this service are public.

mockery documentation built on Sept. 27, 2023, 1:07 a.m.