speed-measure-webpack-v5-plugin

The credit for this package goes miles and miles to Stephen Cook - Speed Measure Webpack Plugin. The only work of mine is to migrate it to Webpack v5 and adding an option to exclude certain plugins in the wrap config.

Speed Measure Plugin

^{(for Webpack v5)}

The first step to optimising your webpack build speed, is to know where to focus your attention.

This plugin measures your webpack build speed, giving an output like this:

Install

npm install --save-dev speed-measure-webpack-v5-plugin

or

yarn add -D speed-measure-webpack-v5-plugin

Requirements

SMP requires at least Node v6. But otherwise, accepts all webpack versions (1, 2, 3, and 4).

Usage

Change your webpack config from

const webpackConfig = {
  plugins: [new MyPlugin(), new MyOtherPlugin()],
};

to

const SpeedMeasurePlugin = require("speed-measure-webpack-v5-plugin");

const smp = new SpeedMeasurePlugin();

const webpackConfig = smp.wrap({
  plugins: [new MyPlugin(), new MyOtherPlugin()],
});

and you're done! SMP will now be printing timing output to the console by default.

But if you choose to integrate it within your script all the time, you may face the following:

• The build fails at the end (in case of prod)
• Some plugin features such as HMR in react don't work.

To resolve this, you would have to exclude these plugins from the meassure (Learn more about why it happens): Just you can do by passing an optional array parameter of the name of the plugin class to exclude.

For example, to exlcude, ReactRefreshPlugin and MiniCSSExtractPlugin send the list as:

smp.wrap(config, ['ReactRefreshPlugin', 'MiniCSSExtractPlugin'])

And voila! You are now free to use this in production.

Check out the examples folder for some more examples.

Options

Pass these into the constructor, as an object:

const smp = new SpeedMeasurePlugin(options);

options.disable

Type: Boolean
Default: false

If truthy, this plugin does nothing at all.

{ disable: !process.env.MEASURE } allows opt-in measurements with MEASURE=true npm run build.

options.outputFormat

Type: String|Function
Default: "human"

Determines in what format this plugin prints its measurements

• "json" - produces a JSON blob
• "human" - produces a human readable output
• "humanVerbose" - produces a more verbose version of the human readable output
• If a function, it will call the function with the JSON blob, and output the response

options.outputTarget

Type: String|Function
Default: console.log

• If a string, it specifies the path to a file to output to.
• If a function, it will call the function with the output as the first parameter

options.pluginNames

Type: Object
Default: {}

By default, SMP derives plugin names through plugin.constructor.name. For some plugins this doesn't work (or you may want to override this default). This option takes an object of pluginName: PluginConstructor, e.g.

const uglify = new UglifyJSPlugin();
const smp = new SpeedMeasurePlugin({
  pluginNames: {
    customUglifyName: uglify,
  },
});

const webpackConfig = smp.wrap({
  plugins: [uglify],
});

options.loaderTopFiles

Type: Number
Default: 0

You can configure SMP to include the files that take the most time per loader, when using outputFormat: 'humanVerbose'. E.g., to show the top 10 files per loader:

const smp = new SpeedMeasurePlugin({
  outputFormat: "humanVerbose",
  loaderTopFiles: 10,
});

options.compareLoadersBuild

Type: Object
Default: {}

This option gives you a comparison over time of the module count and time spent, per loader. This option provides more data when outputFormat: "humanVerbose".

Given a required filePath to store the build information, this option allows you to compare differences to your codebase over time. E.g.

const smp = new SpeedMeasurePlugin({
  compareLoadersBuild: {
    filePath: "./buildInfo.json",
  },
});

options.granularLoaderData (experimental)

Type: Boolean
Default: false

By default, SMP measures loaders in groups. If truthy, this plugin will give per-loader timing information.

This flag is experimental. Some loaders will have inaccurate results:

• loaders using separate processes (e.g. thread-loader)
• loaders emitting file output (e.g. file-loader)

We will find solutions to these issues before removing the (experimental) flag on this option.

What Happens under the Hood?

Loader times are simple to measure, we can directly consume the callback and measure the before and after time. For Plugins, it's a bit tricky. We have to wrap each Plugin into a Proxy Plugin which has the functions to get the before and after times. Which is why, we wrap the configuration.

A problem that the users faced in the original speed-measure-webpack-plugin, is that plugins such as MiniCSSExtractPlugin throw an exception when they are invoked not by the original class, but a proxy - this results in build failure. (Issues like these for example)

To resolve this, we would have to exclude these plugins from being wrapped. And since the support for the original package has seemingly ended since the last 2 years, it inspired me to make this new package with the feature to exlude such Plugins.

If you're still curious on getting the timings of all the plugins, you can choose to not pass the array and let the build fail to get the complete timings.

FAQ

What does general output time mean?

This tends to be down to webpack reading in from the file-system, but in general it's anything outside of what SMP can actually measure.

What does modules without loaders mean?

It means vanilla JS files, which webpack can handle out of the box.

Contributing

Contributors are welcome! 😊

Please check out the CONTRIBUTING.md.

License

MIT