Human-friendly unit testing for C++11

Writing tests

With bandit, you create a command line application that runs all registered tests, reports the result on stdout, and then returns with exit code 0 if all tests passed, or exit code greater than 0 if some tests did not pass.

Entry point

Every application needs an entry point. For bandit it should look like this:

#include <bandit/bandit.h>

int main(int argc, char* argv[]) {
  return bandit::run(argc, argv);

The run() function will run all registered tests and report the result.

If you build and run this, bandit will report the following:

$ specs

Could not find any tests.

This means bandit is running fine, we just have not added any tests for it to run yet. No tests failed, so the exit code is 0.

Adding tests

Once you have a command line application with an entry point that delegates to bandit, it is time to start writing some tests.

To make sure everything is working, we begin by creating a failing test. That way, we know that our tests actually get executed and can verify that they fail for the right reasons.

We add a new .cpp file to our project with the following content:

#include <bandit/bandit.h>

using namespace snowhouse;
using namespace bandit;

go_bandit([]() {
  describe("our first test", []() {
    it("should fail", [&]() {
      AssertThat(5, Equals(6));

If we compile and run this, we get the following response:

$ specs
There were failures!

our first test should fail:
/<path to our project>/first.spec.cpp:8: Expected: equal to 6
Actual: 5

Test run complete. 1 tests run. 0 succeeded. 1 failed.

This means bandit was able to find our test and execute it. A non-zero exit code is returned.

What we did

After including bandit and using its namespace by default, we tell bandit that this file includes tests that we want to be part of the execution by using the go_bandit() construct.

Inside go_bandit() we can now write our tests.

Bandit takes some inspiration from frameworks like RSpec, where our tests take the form of specifications that describe how a component should work. These specifications can be executed to verify that your code fulfills them.

This is conveyed by the ‘describe()/it()’ syntax.

describe() specifies what we are testing and it() specifies what should hold true for that thing.

A slightly more elaborate example:

describe("a calculator", []() {
  calculator calc;

  it("can add", [&]() {
    AssertThat(calc.add(2, 3), Equals(5));

  it("can subtract", [&]() {
    AssertThat(calc.subtract(2, 3), Equals(-1));

Here we describe a calculator and specify that it can add and subtract.

Note that describe()s can be nested arbitrarily. This can be useful to specify particular parts of a class or to specify the behavior of a class in different contexts. (In contrast to RSpec, bandit does not provide context logic.)


Bandit uses the snowhouse assertion framework to verify that things work as expected. Snowhouse contains a lot of helpers aiming at creating readable tests.