karma-benchmark

karma-benchmark

A Karma plugin to run Benchmark.js over multiple browsers with CI compatible output.

☁️ Installation

npm install --save-dev benchmark karma-benchmark

🕵🏾‍♀️ Reporters

• karma-benchmark-json-reporter by @etpinard
• karma-benchmark-plotly-reporter by @etpinard
• karma-benchmarkjs-reporter by @FormidableLabs
• karma-benchmark-reporter by @lazd

⚖️ Configuration

In karma.conf.js, add 'benchmark' to the list of frameworks:

module.exports = (config) => {
  config.set({
    autoWatch: false,
    browsers: ['Chrome'],
    concurrency: 1,
    files: ['bench/**/*.bench.js'],
    frameworks: ['benchmark'],
    singleRun: true
  });
};

Terminal Reporting

Now let's add karma-benchmarkjs-reporter by @FormidableLabs to report results to the Terminal:

npm install --save-dev karma-benchmarkjs-reporter

In karma.conf.js, add 'benchmark' to the list of reporters:

module.exports = (config) => {
  config.set({
    autoWatch: false,
    browsers: ['Chrome'],
    concurrency: 1,
    files: ['bench/**/*.bench.js'],
    frameworks: ['benchmark'],
+   reporters: ['benchmark'],
    singleRun: true
  });
};

JUnit Reporting

To feed our data into Continuous Integration, we can use the karma-junit-reporter.

npm install --save-dev karma-junit-reporter

In karma.conf.js, add junit to the list of reporters and configure the reporter accordingly:

module.exports = (config) => {
  config.set({
    autoWatch: false,
    browsers: ['Chrome'],
    concurrency: 1,
    files: ['bench/**/*.bench.js'],
    frameworks: ['benchmark'],
+   junitReporter: {
+     outputDir: 'reports',
+     outputFile: 'benchmark.xml'
+   },
-   reporters: ['benchmark'],
+   reporters: ['benchmark', 'junit'],
    singleRun: true
  });
};

Data Visualisation Reporting

With a free plot.ly account, we can generate visualisations using the karma-benchmark-plotly-reporter by @etpinard.

npm install --save-dev karma-benchmark-plotly-reporter

In karma.conf.js, add benchmark-plotly to the list of reporters and configure the reporter accordingly:

module.exports = (config) => {
  config.set({
    autoWatch: false,
+   benchmarkPlotlyReporter: {
+     username: '<your username>',
+     apiKey: '<your api key>',
+     cloudFilename: 'plotly-example',
+     imageFilename: 'plotly-example.png'
+   },
    browsers: ['Chrome'],
    concurrency: 1,
    files: ['bench/**/*.bench.js'],
    frameworks: ['benchmark'],
    junitReporter: {
      outputDir: 'reports',
      outputFile: 'benchmark.xml'
    },
-   reporters: ['benchmark', 'junit'],
+   reporters: ['benchmark', 'benchmark-plotly', 'junit'],
    singleRun: true
  });
};

👩🏻‍🔬 Writing Benchmarks

Benchmarks can be written by using the original Benchmark.js API, but a wrapper API is also provided by karma-benchmark in the form of the suite and benchmark globals. The karma-benchmark API aims to make the process of writing Benchmarks feel familiar to users of Jasmine or Jest.

In this example, a suite is defined that pits _.each against the native Array.forEach method:

suite('Array iteration', () => {
  benchmark('_.each', () => {
    _.each([1, 2, 3], (el) => {
      return el;
    });
  });

  benchmark('native forEach', () => {
    [1, 2, 3].forEach((el) => {
      return el;
    });
  });
});

Suite options

Suite options are the same as in Benchmark.js with one exception: onStart and onComplete can be set at the suite level.

See the Benchmark.js Suite constructor API docs for a full list of options.

suite(
  'Array iteration',
  () => {
    benchmark('_.each', () => {
      _.each(this.list, (number) => {
        return number;
      });
    });

    benchmark('native forEach', () => {
      this.list.forEach((number) => {
        return number;
      });
    });
  },
  {
    onCycle(event) {
      var suite = this;
      var benchmark = event.target;
      console.log('Cycle completed for ' + suite.name + ': ' + benchmark.name);
    },
    onStart() {
      this.list = [5, 4, 3];
    },
    onComplete() {
      this.list = null;
    }
  }
);

Benchmark options

Benchmark options are the same as in Benchmark.js. If setup and teardown are passed to benchmark(), they will override setup and teardown from the suite. Pass null or undefined to remove them.

See the Benchmark.js Benchmark constructor API docs for a full list of options.

suite('Iteration', () => {
  benchmark(
    '_.each with array',
    () => {
      _.each(this.list, (number) => {
        return number;
      });
    },
    {
      setup() {
        this.list = ['a', 'b', 'c'];
      },
      teardown() {
        delete this.list;
      }
    }
  );

  benchmark(
    '_.each with object',
    () => {
      _.each(this.list, (number) => {
        return number;
      });
    },
    {
      setup() {
        this.list = { 0: 'a', 1: 'b', 2: 'c' };
      },
      teardown() {
        delete this.list;
      }
    }
  );
});

Running only a specific benchmark or suite

To run only a specific benchmark, use benchmark.only() or bbenchmark() instead of benchmark():

benchmark.only(() => {
  // Only this benchmark will run
  // bbenchmark() does the same thing
});

benchmark(() => {
  // This benchmark won't run
});

The same applies to suites with suite.only() and ssuite().

Skipping benchmarks & suites

To skip a benchmark, use benchmark.skip() or xbenchmark() instead of benchmark():

benchmark.skip(() => {
  // This benchmark won't run
  // xbenchmark() does the same thing
});

benchmark(() => {
  // This and all other benchmarks will run
});

The same applies to suites with suite.skip() and xsuite().

🙋🏿‍♂️ Get Help

There are few ways to get help:

1. For bug reports and feature requests, open issues :bug:
2. For direct and quick help, you can use Gitter :rocket:

Changelog

1.0.4 (2019-07-22)

Bug Fixes

• npm: update dependencies (4adac7e)