Meck

A mocking library for Erlang.

Features
Examples
Use
Manual Build
Caveats
Contribute

Features

Dynamic return values using sequences and loops of static values
Compact definition of mock arguments, clauses and return values
Pass through: call functions in the original module
Complete call history showing calls, return values and exceptions
Mock validation, will invalidate mocks that were not called correctly
Throwing of expected exceptions that keeps the module valid
Throws an error when mocking a module that doesn't exist or has been renamed (disable with option non_strict)
Support for Hamcrest matchers
Automatic backup and restore of cover data
Mock is linked to the creating process and will unload automatically when a crash occurs (disable with option no_link)
Mocking of sticky modules (using the option unstick)

Examples

Here's an example of using Meck in the Erlang shell:

Eshell V5.8.4  (abort with ^G)
1> meck:new(dog, [non_strict]). % non_strict is used to create modules that don't exist
ok
2> meck:expect(dog, bark, fun() -> "Woof!" end).
ok
3> dog:bark().
"Woof!"
4> meck:validate(dog).
true
5> meck:unload(dog).
ok
6> dog:bark().
** exception error: undefined function dog:bark/0

Exceptions can be anticipated by Meck (resulting in validation still passing). This is intended to be used to test code that can and should handle certain exceptions indeed does take care of them:

5> meck:expect(dog, meow, fun() -> meck:exception(error, not_a_cat) end).
ok
6> catch dog:meow().
{'EXIT',{not_a_cat,[{meck,exception,2},
                    {meck,exec,4},
                    {dog,meow,[]},
                    {erl_eval,do_apply,5},
                    {erl_eval,expr,5},
                    {shell,exprs,6},
                    {shell,eval_exprs,6},
                    {shell,eval_loop,3}]}}
7> meck:validate(dog).
true

Normal Erlang exceptions result in a failed validation. The following example is just to demonstrate the behavior, in real test code the exception would normally come from the code under test (which should, if not expected, invalidate the mocked module):

8> meck:expect(dog, jump, fun(Height) when Height > 3 ->
                                  erlang:error(too_high);
                             (Height) ->
                                  ok
                          end).
ok
9> dog:jump(2).
ok
10> catch dog:jump(5).
{'EXIT',{too_high,[{meck,exec,4},
                   {dog,jump,[5]},
                   {erl_eval,do_apply,5},
                   {erl_eval,expr,5},
                   {shell,exprs,6},
                   {shell,eval_exprs,6},
                   {shell,eval_loop,3}]}}
11> meck:validate(dog).
false

Here's an example of using Meck inside an EUnit test case:

my_test() ->
    meck:new(my_library_module),
    meck:expect(my_library_module, fib, fun(8) -> 21 end),
    ?assertEqual(21, code_under_test:run(fib, 8)), % Uses my_library_module
    ?assert(meck:validate(my_library_module)),
    meck:unload(my_library_module).

Pass-through is used when the original functionality of a module should be kept. When the option passthrough is used when calling new/2 all functions in the original module will be kept in the mock. These can later be overridden by calling expect/3 or expect/4.

Eshell V5.8.4  (abort with ^G)
1> meck:new(string, [unstick, passthrough]).
ok
2> string:strip("  test  ").
"test"

It's also possible to pass calls to the original function allowing us to override only a certain behavior of a function (this usage is compatible with the passthrough option). passthrough/1 will always call the original function with the same name as the expect is defined in):

Eshell V5.8.4  (abort with ^G)
1> meck:new(string, [unstick, passthrough]).
ok
2> meck:expect(string, strip, fun
    ("foo") -> "bar";
    (String) -> meck:passthrough([String])
end).
ok
3> string:strip("  test  ").
"test"
4> string:strip("foo").
"bar"
5> meck:unload(string).
ok
5> string:strip("foo").
"foo"

Use

Meck is best used via Rebar 3. Add Meck to the test dependencies in your rebar.config:

{profiles, [{test, [{deps, [meck]}]}]}.

Manual Build

Meck uses Rebar 3. To build Meck go to the Meck directory and simply type:

rebar3 compile

In order to run all tests for Meck type the following command from the same directory:

rebar3 eunit

Documentation can be generated through the use of the following command:

rebar3 edoc

Test Output

Normally the test output is hidden, but if EUnit is run directly, two things might seem alarming when running the tests:

Warnings emitted by cover
An exception printed by SASL

Both are expected due to the way Erlang currently prints errors. The important line you should look for is All XX tests passed, if that appears all is correct.

Caveats

Meck will have trouble mocking certain modules since Meck works by recompiling and reloading modules. Since Erlang have a flat module namespace, replacing a module has to be done globally in the Erlang VM. This means certain modules cannot be mocked. The following is a non-exhaustive list of modules that can either be problematic to mock or not possible at all:

erlang
os
crypto
compile
global
timer (possible to mock, but used by some test frameworks, like Elixir's ExUnit)

Also, a meck expectation set up for a function f does not apply to the module- local invocation of f within the mocked module. Consider the following module:

-module(test).
-export([a/0, b/0, c/0]).

a() ->
  c().

b() ->
  ?MODULE:c().

c() ->
  original.

Note how the module-local call to c/0 in a/0 stays unchanged even though the expectation changes the externally visible behaviour of c/0:

3> meck:new(test, [passthrough]).
ok
4> meck:expect(test,c,0,changed).
ok
5> test:a().
original
6> test:b().
changed
6> test:c().
changed

Common Test

When using meck under Erlang/OTP's Common Test, one should pay special attention to this bit in the chapter on Writing Tests:

init_per_suite and end_per_suite execute on dedicated Erlang processes, just like the test cases do.

Common Test runs init_per_suite in an isolated process which terminates when done, before the test case runs. A mock that is created there will also terminate and unload itself before the test case runs. This is because it is linked to the process creating it. This can be especially tricky to detect if passthrough is used when creating the mock, since it is hard to know if it is the mock responding to function calls or the original module.

To avoid this, you can pass the no_link flag to meck:new/2 which will unlink the mock from the process that created it. When using no_link you should make sure that meck:unload/1 is called properly (for all test outcomes, or crashes) so that a left-over mock does not interfere with subsequent test cases.

Contribute

Patches are greatly appreciated! For a much nicer history, please write good commit messages. Use a branch name prefixed by feature/ (e.g. feature/my_example_branch) for easier integration when developing new features or fixes for meck.

Should you find yourself using Meck and have issues, comments or feedback please create an issue here on GitHub.

Meck has been greatly improved by many contributors!

Donations

If you or your company use Meck and find it useful, a sponsorship or donations are greatly appreciated!