High CPU usage

[flynfish]

When running grunt watch, the process sits around 70% cpu usage on my mac (10.8.2). Is this a known issue?

[shama]

It's not. How many files are you watching and could you post your Gruntfile.js please? Does it happen with just watching or as tasks are ran? Also are you using grunt@0.4.0rc4 and grunt-contrib-watch@0.2.0a?

[flynfish]

Versions:
"grunt": "~0.3.17",
"grunt-contrib-watch": "~0.1.4",
"grunt-contrib": "~0.3.0"

As for the amount of files, since I am using bootstrap there are more 40 less files being watched. There really isn't much of a need for me to watch those though so perhaps that is one of the issues? Looks like it happens just when watching.

module.exports = function(grunt) {

  grunt.initConfig({

    // The lint task will run the build configuration and the application
    // JavaScript through JSHint and report any errors.  You can change the
    // options for this task, by reading this:
    // https://github.com/cowboy/grunt/blob/master/docs/task_lint.md
    lint: {
      files: [
        "public/**/*.js"
      ]
    },

    // The jst task compiles all application templates into JavaScript
    // functions with the underscore.js template function from 1.2.4.  You can
    // change the namespace and the template options, by reading this:
    // https://github.com/tbranyen/build-tasks/tree/master/jst
    //
    // The concat task depends on this file to exist, so if you decide to
    // remove this, ensure concat is updated accordingly.
    jst: {
      "public/js/templates.js": [
        "views/templates/**/*.html"
      ]
    },

    // The concatenate task is used here to merge the almond require/define
    // shim and the templates into the application code.  It's named
    // dist/debug/require.js, because we want to only load one script file in
    // index.html.
    concat: {
      "public/js/debug.js": [
        "public/js/vendor/*.js",
        "public/js/vendor/bootstrap/*.js",
        "public/js/templates.js",
        "public/js/app/*.js",
        "public/js/app/models/*.js",
        "public/js/app/views/*.js",
        "public/js/app/collections/*.js",
        "public/js/app/views/nests/*.js"
      ]
    },

    // This task uses the MinCSS Node.js project to take all your CSS files in
    // order and concatenate them into a single CSS file named index.css.  It
    // also minifies all the CSS as well.  This is named index.css, because we
    // only want to load one stylesheet in index.html.
    mincss: {
      "public/dist/release.css": [
        "public/stylesheets/*.css"
      ]
    },

    // Takes the built require.js file and minifies it for filesize benefits.
    min: {
      "public/dist/release.js": [
        "public/js/debug.js"
      ]
    },

    // less to CSS
    less: {
      compile: {
        files: {
          "public/stylesheets/style.css": "public/stylesheets/style.less",
          "public/stylesheets/nest.css": "public/stylesheets/nest.less",
          "public/stylesheets/style.css": "public/stylesheets/style.less",
          "public/stylesheets/vendor/bootstrap/bootstrap.css": "public/stylesheets/vendor/bootstrap/less/bootstrap.less",
          "public/stylesheets/vendor/fontawesome/font-awesome.css": "public/stylesheets/vendor/fontawesome/less/font-awesome.less"
        }
      }
    },

    watch: {
      scripts: {
        files: '**/*.less',
        tasks: ['less']
      },
      templates: {
        files: 'views/templates/**/*.html',
        tasks: ['jst']
      }
    }

  });

  // The default task will remove all contents inside the dist/ folder, lint
  // all your code, precompile all the underscore templates into
  // dist/debug/templates.js, compile all the application code into
  // dist/debug/require.js, and then concatenate the require/define shim
  // almond.js and dist/debug/templates.js into the require.js file.
  grunt.registerTask("default", "concat min less mincss jst");
  grunt.loadNpmTasks('grunt-contrib');
};

[shama]

Thanks for the info. If you get a chance, could you try the latest grunt and grunt-contrib-watch? Let me know if it still runs up the CPU. We've made a lot of changes since v0.1.4 in the watch. I'll try your setup shortly to see if I can duplicate. Thanks!

[flynfish]

I updated to the lastest grunt and grunt-contrib versions and now the process never gets above 1% cpu, so I would say you guys did some awesome work on that update!

When running watch now and updating a less file, I get this error:

running "less:compile" (less) task
Warning: Object #<Object> has no method 'expandFiles' Use --force to continue.

Aborted due to warnings.

Any ideas on what I need to change to get it to work? (I already update the registerTask default alias to use an array instead of string"

[shama]

Awesome, glad to hear!

To get less working, you'll need to install the Grunt v0.4 version with: npm install git://github.com/gruntjs/grunt-contrib-less.git. In fact all of your plugins will need updating if moving to Grunt v0.4 (https://github.com/gruntjs/grunt/wiki/Upgrading-from-0.3-to-0.4). FWIW, I use Grunt v0.4 exclusively. Just know it is still under development so not all grunt plugins support it yet and things can change.

Another thing to try if staying on Grunt v0.3 for the time being. Try uninstalling grunt-contrib-watch and using the previously built-in watch task in Grunt v0.3. grunt-contrib-watch is intended as a replacement for the watch task in Grunt v0.4. We just had Grunt v0.3 support on the v0.1.x branch to get some early testers. :)

[flynfish]

I tried upgrading to grunt .4 and upgrading the grunt plugins but I ran into some issues I don't have time to figure out. I removed the bootstrap less files from being watched and that dropped the cpu to under 5%.

[shama]

Hopefully the other issues you got figured out :) Otherwise please reopen if one of those is with the watch task, thanks!

[alexbeletsky]

I'm having the same issue on my Windows 8, cpu level near 70%, watching just a few files.

[Jesterovskiy]

Have issue with high load too.

Versions:
"grunt": "~0.4.1",
"grunt-contrib-watch": "~0.4.3",
"grunt-contrib-sass": "~0.3.0"

System: Ubuntu 12.04 LTS

Gruntfile:

...
   watch: {
      sass: {
        files: ['app/stylesheets/{,**/}*.scss'],
        tasks: ['sass:dev']
      },
      html: {
        files: ['app/index.html'],
        tasks: ['html:dev']
      }
    },

  html: {
      dev: {
        options: {
          src: 'app/index.html',
          dest: 'temp/index.html',
          vars: config('development')
        }
      }
    },

    sass: {
      dev: {
        options: {
          style: 'expanded'
        },
        files: {
          'temp/stylesheets/application.css': 'app/stylesheets/application.scss'
        }
      }
    },
...

[shama]

Try running with --verbose to see which files it is watching. Also which version of node?

[Jesterovskiy]

Task is watching 45 scss files and version of node is v0.10.4.

[shama]

I'm curious if the problem is within this lib or just watching in general (as that greatly differs on environments). Are you game to try some examples for me?

Could you first make sure any previous node processes are not running with killall -9 node and then run this example to see if that causes the CPU spike?

var fs = require('fs');
[
  // Please add a bunch more files
  'app/stylesheets/application.scss',
].forEach(function(file) {
  fs.watch(file, function(event, filename) {
    console.log(file + ' was ' + event);
  });
});

[Jesterovskiy]

Not fully understand, how to run this example. Can you explain, please =)

[shama]

Oh that example will watch files using the core node.js api because I want to see if the problem is from this watch task or with any file watching.

Throw that example into a file named test.js. Add a bunch more actual file paths and then do this in your terminal: killall -9 node && node test.js to kill any other running node process and then run the example. Thanks!