I used Grunt and grunt-contrib-watch in my project.

Although grunt-contrib-watch works fine when the number of files is small, but as it grows up, grunt-contrib-watch starts using too much CPU. For example, when grunt-contrib-watch is watching about 1,000 files, node always uses 10% of my CPU.

This is known issue. Look at the FAQ:

Another reason if you're watching a large number of files could be the low default interval. Try increasing with options: { interval: 5007 }. Please see issues #35 and #145 for more information.

What causes high CPU usage?

grunt-contrib-watch uses gaze module to watch files.

This module uses not only fs.watch() (native API) but also fs.watchFile() which periodically executes fs.statSync() (read Difference between fs.watch() and fs.watchFile() for details).

Why does gaze use fs.watchFile()? chokidar/README.md says fs.watch() had many problems.

Node.js fs.watch:

• Doesn't report filenames on mac.
• Doesn't report events at all when using editors like TextMate2 on mac.
• Sometimes report events twice.
• Has only one non-useful event: rename.
• Has a lot of other issues

How do I avoid it?

Use grunt-este-watch instead of grunt-contrib-watch.

Let's read grunt-este-watch/README.md.

What's wrong with official grunt-contrib-watch?

It's slow and buggy, because it uses combination fs.fileWatch and fs.watch, for historical reason. From Node 0.9.2+, fs.watch is ok.

github.com/steida/este Needs maximum performance and stability, so that's why I had to create yet another Node.js file watcher. This watcher is continuously tested on Mac, Linux, Win platforms.

I'm using grunt-este-watch now. It uses no CPU when watching. Nice!

Example

Here is my Gruntfile.js:

module.exports = function(grunt) {
  grunt.initConfig({
    pkg: grunt.file.readJSON('package.json'),
    shell: {
      jekyll_build: {
        command: 'jekyll build'
      },
      // snip
    },
    // snip
    esteWatch: {
      options: {
        dirs: ['./', '_posts/*/', '_layouts', '_includes',
               'javascript/**/', 'apollo/tutorial',
               '_plugins/**/', 'stylesheets', 'javascripts'],
        livereload: {
          enabled: false
        }
      },
      '*': function(filepath) { return 'shell:jekyll_build' }
    }
  });

grunt-este-watch's configuration is quite different from grunt-contrib-watch's.

grunt-este-watch requires directory list to watch. When any file under watched directory is modified, callback function is called.

My callback function returns 'shell:jekyll_build', so grunt executes shell:jekyll_build when watched files are modified. We can check filepath argument to determine the tasks returned.

'*' means all extensions. We can register callback to specified extension. There is a good example on grunt-este-watch/README.md.

    coffee: function(filepath) {
      var files = [{
        expand: true,
        src: filepath,
        ext: '.js'
      }];
      grunt.config(['coffee', 'app', 'files'], files);
      grunt.config(['coffee2closure', 'app', 'files'], files);
      return ['coffee:app', 'coffee2closure:app'];
    },
    // snip
    css: function(filepath) {
      if (grunt.option('stage')) {
        return 'cssmin:app';
      }
    }

Enjoy!