gulp-spawn-mocha

This is a plugin for gulp which runs Mocha tests in a separate process from the gulp process. Each time tests are run a new child process is created meaning the test environment always starts cleanly, i.e., globals are reset as are non-enumerable properties defined on native prototypes via Object.defineProperty. This also means that if your tests crash the node process (e.g., process.exit(-1).) then an error event is emitted rather than your whole gulp process crashing (good for watching).

Usage

Usage is according to this API:

1stream.pipe(mocha({
2  // options
3}))

This plugin uses mocha version ^3.0.0. The major version of this plugin will match the major version of mocha, which is a peer dependency of this plugin.

The plugin accepts these special options:

• bin: a path to a mocha executable to use instead of the one this plugin looks for by default. This is useful if you want to use a fork of mocha which goes by a different name or a different executable altogether.
• env: the environment variables that the child process will have access to (key-value pairs, see child_process::fork). These variables are merged with your current environment variables and sent to the mocha executable.
• cwd: the working directory for the child process. This can be used to put files that the test creates or reads from the working directory in a specific directory, instead of the directory where you are running gulp from.
• execPath: an alternative execution path to the Node.js instance. If not specified, by default, child_process::fork will spawn the new Node.js instances using the process::execPath of the parent process.

All other options are properly prefixed with either - or -- and passed to the mocha executable. Any arguments which do not take a value (e.g., c, colors, or debug) should just have a value of true. Any arguments which have dashes in the name can be specified by using camelCase (i.e., debugBrk for --debug-brk, inlineDiffs for --inline-diffs, etc) so you don't have to use strings for the argument names. Please note that the gc option must be specified as exposeGc (please see issue #21). For an example, see this plugin's very own gulpfile.js:

1const DEBUG = process.env.NODE_ENV === 'debug',
2      CI = process.env.CI === 'true';
3
4var gulp = require('gulp'),
5    mocha = require('./lib');
6
7gulp.task('test', function () {
8  return gulp.src(['test/*.test.js'], {read: false})
9    .pipe(mocha({
10      debugBrk: DEBUG,
11      r: 'test/setup.js',
12      R: CI ? 'spec' : 'nyan',
13      istanbul: !DEBUG
14    }));
15});
16
17gulp.task('default', ['test'], function () {
18  gulp.watch('{lib,test}/*', ['test']);
19});

With this setup the nyan reporter will be used in development and the spec reporter will be used in CI (Travis sets the CI environment variable to true automatically).

The default task will execute tests and watch for changes and execute tests whenever a change is detected.

Conditional Arguments

If the value of an argument is falsy (but not 0) then it will not be passed to mocha. This is useful, for example, if you want to enable debugging only when a certain environment variable is true. Example:

1const DEBUG = process.env.NODE_ENV === 'debug';
2stream.pipe(mocha({
3    debugBrk: DEBUG,
4    istanbul: !DEBUG
5}));

Custom Environment Variables

As mentioned above an object provided underneath the env options key will allow you to specify a custom environment. This is useful, for example, to run your tests in a different NODE_ENV than the default. Such a gulp task would look like this:

1var gulp = require('gulp'),
2    mocha = require('gulp-spawn-mocha');
3
4gulp.task('test', function() {
5  return gulp
6    .src(['test/*.test.js'])
7    .pipe(mocha({
8      env: {'NODE_ENV': 'test'}
9    }));
10});

These variables are merged with your current environment variables and sent to the mocha executable.

Code Coverage

Because of the nature of this plugin launching an external process to run tests, the standard coverage plugins for gulp will not work with this module. Starting in version 0.4.0 Istanbul is included in order to enable code coverage reports without having to instrument code on disk. You can use it by passing the istanbul option.

Set istanbul to true if you want to use all the default settings:

1gulp.task('test', function() {
2  return gulp
3    .src(['test/*.test.js'])
4    .pipe(mocha({
5      istanbul: true
6    }));
7});

This will launch a process equivilant to:

istanbul cover -- _mocha

The default settings of Istanbul output to a directory in the cwd called coverage.

If you want to pass options to Istanbul, you can do that as well:

1gulp.task('test', function() {
2  return gulp
3    .src(['test/*.test.js'])
4    .pipe(mocha({
5      istanbul: {
6        dir: 'path/to/custom/output/directory'
7      }
8    }));
9});

This will launch a process equivilant to:

istanbul cover --dir path/to/custom/output/directory -- _mocha

This will output to a directory called path/to/custom/output/directory.

Istanbul, like mocha, supports a custom bin option so you can use a custom fork of Istanbul:

1gulp.task('test', function() {
2  return gulp
3    .src(['test/*.test.js'])
4    .pipe(mocha({
5      istanbul: {
6        dir: 'path/to/custom/output/directory',
7        bin: require.resolve('isparta') + '/bin/isparta'
8      }
9    }));
10});

This will launch a process equivilant to:

./node_modules/isparta/bin/isparta cover --dir path/to/custom/output/directory -- _mocha

Publishing Coverage Reports

Assuming you are using Travis for CI and Coveralls for publishing code coverage reports it is very easy to automatically have Travis publish to Coveralls when tests are run successfully. First make sure you install and save the coveralls module as a dev dependency:

npm i --save-dev coveralls

Then edit your .travis.yml to have an after_success command:

1language: node_js
2node_js:
3  - "0.11"
4  - "0.10"
5after_success: ./node_modules/.bin/coveralls --verbose < coverage/lcov.info

The coveralls module requires no additional configuration to publish to Coveralls as long as both Travis and Coveralls are configured for the same public repository. See node-coveralls for more details.

Output Reports to a File

You can pass output option to write a report to a writeable stream. If output is a string then a writeable stream will be created with output as its path. Note, if you are using istanbul, your reports content may contain istanbul's result.

Use file path:

1gulp.task('test', function () {
2  return gulp.src(['test/*.test.js'], {read: false})
3    .pipe(mocha({
4      debugBrk: DEBUG,
5      r: 'test/setup.js',
6      R: CI ? 'spec' : 'nyan',
7      istanbul: !DEBUG,
8      output: 'result.log'
9    }));
10});

Use file stream:

1gulp.task('test', function () {
2  return gulp.src(['test/*.test.js'], {read: false})
3    .pipe(mocha({
4      debugBrk: DEBUG,
5      r: 'test/setup.js',
6      R: CI ? 'spec' : 'nyan',
7      istanbul: !DEBUG,
8      output: fs.createWriteStream('result.log', {flags: 'w'})
9    }));
10});

This or gulp-mocha?

The original gulp-mocha is fine in most circumstances. If you need your tests to run as a separate process (or a separate process is simply your preference for the reasons specified above) or you need to use a custom version of Mocha (e.g., a fork with bug fixes or custom functionality) then you should use this plugin.

License

MIT