lab

Node test utility

Lead Maintainer: Wyatt Preul

lab is sponsored by Joyent. Joyent is currently looking for a Node.js core engineer to hire.

Introduction

lab is a simple test utility for node. Unlike other test utilities, lab uses domains instead of uncaught exception and other global manipulation. Our goal with lab is to keep the execution engine as simple as possible, and not try to build an extensible framework. lab works with any assertion library that throws an error when a condition isn't met.

Command Line

lab supports the following command line options:

• -a, --assert - name of assert library to use. To disable assertion library set to false. Defaults to code.
• --bail - terminate the process with a non-zero exit code on the first test failure. Defaults to false.
• -c, --coverage - enables code coverage analysis.
• --coverage-path - sets code coverage path.
• --coverage-exclude - sets code coverage excludes.
• -C, --colors - enables or disables color output. Defaults to console capabilities.
• -d, --dry - dry run. Skips all tests. Use with -v to generate a test catalog. Defaults to false.
• -D, --debug - print the stack during a domain error event.
• -e, --environment - value to set the NODE_ENV environment variable to, defaults to 'test'.
• -f, --flat - do not perform a recursive load of test files within the test directory.
• -g, --grep - only run tests matching the given pattern which is internally compiled to a RegExp.
• -h, --help - show command line usage.
• -i, --id - only run the test for the given identifier (or identifiers range).
• -I, --ignore - ignore a list of globals for the leak detection (comma separated)
• --inspect - start lab in debug mode using the V8 Inspector.
• -l, --leaks - disables global variable leak detection.
• -L, --lint - run linting rules using linter. Disabled by default.
• --lint-errors-threshold - maximum absolute amount of linting errors. Defaults to 0.
• --lint-warnings-threshold - maximum absolute amount of linting warnings. Defaults to 0.
• -m, --timeout - individual tests timeout in milliseconds (zero disables timeout). Defaults to 2 seconds.
• -M, --context-timeout - default timeouts for before, after, beforeEach and afterEach in milliseconds. Disabled by default.
• -n, --linter - specify linting program file path; default is eslint.
• --lint-fix - apply any fixes from the linter, requires -L or --lint to be enabled. Disabled by default.
• --lint-options - specify options to pass to linting program. It must be a string that is JSON.parse(able).
• -o, --output - file to write the report to, otherwise sent to stdout.
• -p, --parallel - sets parallel execution as default test option. Defaults to serial execution.
• -P, --pattern - only load files with the given pattern in the name.
• -r, --reporter - the reporter used to generate the test results. Defaults to console. Options are:
  • console - text report.
  • html - HTML test and code coverage report (sets -c).
  • json - output results in JSON format.
  • junit - output results in JUnit XML format.
  • tap - TAP protocol report.
  • lcov - output to lcov format.
  • clover - output results in Clover XML format.
  • Multiple Reporters - See Below
  • Custom Reporters - See Below
• -R, --rejections - fail tests on unhandled Promise rejections.
• --shuffle - randomize the order that test scripts are executed. Will not work with --id.
• --seed - use this seed to randomize the order with --shuffle. This is useful to debug order dependent test failures.
• -s, --silence - silence test output, defaults to false.
• -S, --sourcemaps - enables sourcemap support for stack traces and code coverage, disabled by default.
• -t, --threshold - sets the minimum code test coverage percentage to 100%.
• -T, --transform - javascript file that exports an array of objects ie. [ { ext: ".js", transform: (content, filename) => { ... } } ]. Note that if you use this option with -c (--coverage), then you must generate sourcemaps and pass sourcemaps option to get proper line numbers.
• -v, --verbose - verbose test output, defaults to false.
• -V, --version - display lab version information.

Usage

To install locally:

$ npm install --save-dev lab

By default, lab loads all the '*.js' files inside the local 'test' directory and executes the tests found. To use different directories or files, pass the file or directories as arguments:

$ lab unit.js

Test files must require the lab module, and export a test script:

const Lab = require('lab');
const lab = exports.lab = Lab.script();
 
lab.test('returns true when 1 + 1 equals 2', (done) => {
 
    Lab.expect(1 + 1).to.equal(2);
    done();
});

When a test is completed, done(err) must be called, otherwise the test will time out (2 seconds by default) and will fail. The test passes if done() is called once before the timeout, no exception thrown, and no arguments are passed to done(). If no callback function is provided, the test is considered a TODO reminder and will be skipped.

Tests can be organized into experiments:

lab.experiment('math', () => {
 
    lab.test('returns true when 1 + 1 equals 2', (done) => {
 
        Lab.expect(1 + 1).to.equal(2);
        done();
    });
});

If you need to perform some async actions before or after executing the tests inside an experiment, the before() and after() methods can be used. To execute code before or after each test in an experiment, use beforeEach() and afterEach().

lab.experiment('math', () => {
 
    lab.before((done) => {
 
        // Wait 1 second 
        setTimeout(() => {
 
            done();
        }, 1000);
    });
 
    lab.beforeEach((done) => {
 
        // Run before every single test 
        done();
    });
 
    lab.test('returns true when 1 + 1 equals 2', (done) => {
 
        Lab.expect(1 + 1).to.equal(2);
        done();
    });
});
 

test(), before(), beforeEach(), after() and afterEach() also support returning promises instead of using the done callback:

lab.experiment('math', () => {
 
    lab.before(() => {
 
        const promise = aFunctionReturningAPromise();
 
        return promise;
    });
 
    lab.test('returns true when 1 + 1 equals 2', () => {
 
        return aFunctionReturningAPromise()
            .then((aValue) => {
 
                Lab.expect(aValue).to.equal(expectedValue);
            });
    });
});

Both test() and experiment() accept an optional options argument which must be an object with the following optional keys:

• timeout - set a test or experiment specific timeout in milliseconds. Defaults to the global timeout (2000ms or the value of -m).
• parallel - sets parallel execution of tests within each experiment level. Defaults to false (serial execution).
• skip - skip execution. Cannot be overridden in children once parent is set to skip.
• only - marks all other tests or experiments with skip.

You can also append .only(…) or .skip(…) to test and experiment instead of using the options flags:

lab.experiment('with only', () => {
 
    lab.test.only('only this test will run', (done) => {
 
        Lab.expect(1 + 1).to.equal(2);
        done();
    });
 
    lab.test('another test that will not be executed', (done) =>  {
 
        done();
    });
});

The test() callback has a note() function attached to it that can be used to attach notes to the test case. These notes are included in the console reporter at the end of the output. For example, if you would like to add a note with the current time, your test case may look like the following:

lab.test('attaches notes', (done) => {
 
    Lab.expect(1 + 1).to.equal(2);
    done.note(`The current time is ${Date.now()}`);
    done();
});

Multiple notes can be appended for the same test case by simply calling note() repeatedly.

The test() callback provides a second onCleanup argument which is a function used to register a runtime cleanup function to be executed after the test completed. The cleanup function will execute even in the event of a timeout. Note that the cleanup function will be executed as-is without any timers and if it fails to call it's next argument, the runner will freeze.

lab.test('cleanups after test', (done, onCleanup) => {
 
    onCleanup((next) => {
 
        cleanup_logic();
        return next();
    });
 
    Lab.expect(1 + 1).to.equal(2);
    done();
});

Additionally, test() options support a plan setting to specify the expected number of assertions for your test to execute. This setting should only be used with an assertion library that supports a count() function, like code. plan may not work with parallel test executions

lab.experiment('my plan', () => {
 
    lab.test('only a single assertion executes', { plan: 1 }, (done) => {
 
        Lab.expect(1 + 1).to.equal(2);
        done();
    });
});

before(), after(), beforeEach(), afterEach() accept an optional options argument which must be an object with the following optional keys:

• timeout - set a specific timeout in milliseconds. Disabled by default or the value of -M.

lab.experiment('math', { timeout: 1000 }, () => {
 
    lab.before({ timeout: 500 }, (done) =>  {
 
        doSomething();
        done();
    });
 
    lab.test('returns true when 1 + 1 equals 2', { parallel: true }, (done) =>  {
 
        Lab.expect(1 + 1).to.equal(2);
        done();
    });
});

The script([options]) method takes an optional options argument where options is an object with the following optional keys:

• schedule - if false, an automatic execution of the script is disabled. Automatic execution allows running lab test scripts directly with node without having to use the cli (e.g. node test/script.js). When using lab programmatically, this behavior is undesired and can be turned off by setting schedule to false. If you need to see the output with schedule disabled you should set output to process.stdout. Defaults to true.
• cli - allows setting command line options within the script. Note that the last script file loaded wins and usage of this is recommended only for temporarily changing the execution of tests. This option is useful for code working with an automatic test engine that run tests on commits. Setting this option has no effect when not using the CLI runner. For example setting cli to { ids: [1] } will only execute the first test loaded.

To make lab look like BDD:

const Lab = require('lab');
const lab = exports.lab = Lab.script();
 
const describe = lab.describe;
const it = lab.it;
const before = lab.before;
const after = lab.after;
const expect = Lab.expect;
 
describe('math', () => {
 
    before((done) => {
 
        done();
    });
 
    after((done) => {
 
        done();
    });
 
    it('returns true when 1 + 1 equals 2', (done) => {
 
        expect(1 + 1).to.equal(2);
        done();
    });
});

To make lab look like TDD:

const Lab = require('lab');
const lab = exports.lab = Lab.script();
 
const suite = lab.suite;
const test = lab.test;
const expect = Lab.expect;
 
suite('math', () => {
 
    test('returns true when 1 + 1 equals 2', (done) => {
 
        expect(1 + 1).to.equal(2);
        done();
    });
});

To use source transforms, you must specify a file with the -T command line option that tells Lab how to do the transformation. You can specify many extensions with different transform functions such as .coffee or .jsx. A sample file using the babel transpiler and the CoffeeScript compiler could look like:

const Babel = require('babel-core');
const Coffee = require('coffee-script');
const Btoa = require('btoa');
 
module.exports = [
    { ext: '.js', transform: (content, filename) => {
 
        // Make sure to only transform your code or the dependencies you want 
        if (filename.indexOf('node_modules') === -1) {
            const result = Babel.transform(content, { sourceMap: 'inline', filename, sourceFileName: filename });
            return result.code;
        }
 
        return content;
    } },
    { ext: '.coffee', transform: (content, filename) => {
 
        // Again, make sure to only transform your code or the dependencies you want 
        if (filename.indexOf('node_modules') === -1) {
            const result = Coffee.compile(content, {
                sourceMap: true,
                inline: true,
                sourceRoot: '/',
                sourceFiles: [filename]
            });
 
            // append source map to end of compiled JS 
            return result.js +
              '\n//# sourceMappingURL=data:application/json;base64,' +
              Btoa(unescape(encodeURIComponent(result.v3SourceMap)));
        }
 
        return content;
    } }
];

Sometimes you want to disable code coverage for specific lines, and have the coverage report omit them entirely. To do so, use the $lab:coverage:(off|on)$ comments. For example:

// There is no way to cover this in node 0.10 
/* $lab:coverage:off$ */
if (typeof value === 'symbol') {
    return '[' + value.toString() + ']';
}
/* $lab:coverage:on$ */
 

.labrc.js file

lab supports a .labrc.js configuration file for centralizing lab settings.
The .labrc.js file can be located in the current working directory, any directory that is the parent of the current working directory, or in the user's home directory. The .labrc.js file needs to be able to be required by Node.js. Therefore, either format it as a JSON file or with a module.exports that exports an object with the keys that are the settings.

Below is an example of a .labrc.js file to enable linting and test coverage checking:

module.exports = {
    coverage: true,
    threshold: 90,
    lint: true
};

.labrc.js setting precedent

The .labrc.js file will override the lab default settings. Any options passed to the lab runner will override the settings found in .labrc.js. For example, assume you have the following .labrc.js file:

module.exports = {
    coverage: true,
    threshold: 100
};

If you need to reduce the coverage threshold for a single run, you can execute lab as follows:

lab -t 80

.labrc.js available settings

The .labrc.js file supports configuration keys that are named with the long name of the command line settings. Therefore, if you need to specify an assert library, you would export a key named "assert" with the desired value.

Extending the linter

lab uses a shareable eslint config, and a plugin containing several hapi specific linting rules. If you want to extend the default linter you must:

1. Add eslint-plugin-hapi and eslint-config-hapi as dependencies in your package.json. You must add both the plugin and the config because eslint treats them as peer dependencies. For more background, see eslint/eslint#3458 and eslint/eslint#2518.

2. In your project's eslint configuration, add "extends": "eslint-config-hapi". eslint will automatically infer the eslint-config-, so technically you can just write "extends": "hapi".

Your project's eslint configuration will now extend the default lab configuration.

Ignoring files in linting

Since eslint is used to lint, you can create an .eslintignore containing paths to be ignored:

node_modules/*
**/vendor/*.js

Only run linting

In order to run linting and not to execute tests you can combine the dry run flag with the lint flag.

lab -dL

Running a custom linter

If you would like to run a different linter, or even a custom version of eslint you should pass the -n or --linter argument with the path to the lint runner. For example, if you plan to use jslint, you can install lab-jslint then pass --linter node_modules/lab-jslint.

Integration with an assertion library

Using the --assert argument allows you to integrate Lab with your favorite assertion library. By default the assertion library code is included and integrated. However, you can override this behavior by setting the CLI --assert argument or changing the assert option when executing report. Whatever assertion library you specify is imported and assigned to the Lab.assertions property. Here is an example using lab --assert code:

// Testing shortcuts 
const expect = Lab.assertions.expect;
const fail = Lab.assertions.fail;
 
 
describe('expectation', () => {
 
    it('should be able to expect', (done) => {
 
        expect(true).to.be.true();
 
        done();
    });
 
    it('should be able to fail (This test should fail)', (done) => {
 
        fail('Should fail');
 
        done();
    });
 
});

If you use the Code assertion library Lab will let you know if you have any missing assertions. An example of this is:

describe('expectation', () => {
 
    it('Test should pass but get marked as having a missing expectation', (done) => {
 
        // Invalid and missing assertion - false is a method, not a property! 
        // This test will pass. 
        Lab.expect(true).to.be.false;
 
        done();
    });
 
});

This is an invalid test but it will pass as the .false assertion was not actually called. Lab will report the number of incomplete assertions, their location in your code and return a failure of the tests.

Debuggers

lab can be started with the option --inspect which will run it with the V8 Inspector.

This debugger can be accessed using the URL that is printed in the console, or used in association with a few Chrome extensions (Node.js V8 Inspector, NIM, etc).

As you may know, if your tests are associated with the command npm test, you can already run npm test -- --inspect to run it with the inspector and avoid creating another command. If you want to listen on a specific port for the inspector, pass --inspect={port}.

lab also has automatic support for the WebStorm debugger, just start a normal debugging session on your npm test script.

Best practices

• Install lab as a global module:

$ npm install -g lab

• Add lab as a dev dependency to your project's package.json along with a test script:

{
  "name": "example",
  "version": "1.0.0",
  "dependencies": {
  },
  "devDependencies": {
    "lab": "5.x.x"
  },
  "scripts": {
    "test": "lab -t 100",
    "test-cov-html": "lab -r html -o coverage.html"
  },
  "licenses": [
    {
      "type": "BSD",
      "url": "http://github.com/hapijs/lab/raw/master/LICENSE"
    }
  ]
}

Note that npm test will execute lab with the -t 100 option which will require 100% code coverage. Run npm run test-cov-html and check the coverage.html file to figure out where coverage is lacking. When coverage is below the threshold, the CLI will exit with code 1 and will result in an npm Error message.

• Run your tests with

$ npm test

Multiple Reporters

Multiple reporters can be specified by providing multiple reporter options.

$ lab -r console -r html

If any output -o is provided, they must match the same number of provided reporter options. The reporters would be paired with an output based on the order in which they were supplied. When specifying multiple outputs, use stdout to send a particular reporter to stdout.

$ lab -r console -o stdout -r html -o coverage.html -r lcov -o lcov.info -r json -o data.json

Multiple reporters of the same kind are also supported.

$ lab -r console -o stdout -r console -o console.log

Custom Reporters

If the value passed for reporter isn't included with Lab, it is loaded from the filesystem. If the string starts with a period ('./custom-reporter'), it will be loaded relative to the current working directory. If it doesn't start with a period ('custom-reporter'), it will be loaded from the node_modules directory, just like any module installed using npm install.

Reporters must be a class with the following methods: start, test and end. options are passed to the class constructor upon initialization.

See the json reporter for a good starting point.

Excluding paths from coverage reporting

The --coverage-exclude argument can be repeated multiple times in order to add multiple paths to exclude. By default the node_modules and test directories are excluded. If you want to exclude those as well as a directory named public you can run lab as follows:

lab -c --coverage-exclude test --coverage-exclude node_modules --coverage-exclude public

Acknowledgements

lab initial code borrowed heavily from mocha, including the actual code used to render the coverage report into HTML. lab coverage code was originally adapted from blanket which in turn uses falafel.