Why is it recommended to use concat then uglify when the latter can do both?

Question

I keep seeing the recommendation for making JS files ready for production to be concat then uglify.

For example here, in on of Yeoman's grunt tasks.

By default the flow is: concat -> uglifyjs.

Considering UglifyJS can do both concatenation and minification, why would you ever need both at the same time?

Thanks.

Answer 1

Running a basic test to see if there is a performance difference between executing concat and then uglify vs. just uglify.

package.json

{
  "name": "grunt-concat-vs-uglify",
  "version": "0.0.1",
  "description": "A basic test to see if we can ditch concat and use only uglify for JS files.",
  "devDependencies": {
    "grunt": "^0.4.5",
    "grunt-contrib-concat": "^0.5.0",
    "grunt-contrib-uglify": "^0.6.0",
    "load-grunt-tasks": "^1.0.0",
    "time-grunt": "^1.0.0"
  }
}

Gruntfile.js

module.exports = function (grunt) {

    // Display the elapsed execution time of grunt tasks
    require('time-grunt')(grunt);
    // Load all grunt-* packages from package.json
    require('load-grunt-tasks')(grunt);

    grunt.initConfig({
        paths: {
            src: {
                js: 'src/**/*.js'
            },
            dest: {
                js: 'dist/main.js',
                jsMin: 'dist/main.min.js'
            }
        },
        concat: {
            js: {
                options: {
                    separator: ';'
                },
                src: '<%= paths.src.js %>',
                dest: '<%= paths.dest.js %>'
            }
        },
        uglify: {
            options: {
                compress: true,
                mangle: true,
                sourceMap: true
            },
            target: {
                src: '<%= paths.src.js %>',
                dest: '<%= paths.dest.jsMin %>'
            }
        }
    });

    grunt.registerTask('default', 'concat vs. uglify', function (concat) {
        // grunt default:true
        if (concat) {
            // Update the uglify dest to be the result of concat
            var dest = grunt.config('concat.js.dest');
            grunt.config('uglify.target.src', dest);

            grunt.task.run('concat');
        }

        // grunt default
        grunt.task.run('uglify');
    });
};

In src, I've put a bunch of JS files, including the uncompressed source of jQuery, copied several times, spread around into subfolders. Much more than what a normal site/app usually has.

Turns out the time it takes to concat and compress all of these files is essentially the same in both scenarios.
Except when using the sourceMap: true option on concat as well (see below).

On my computer:

grunt default      : 6.2s (just uglify)
grunt default:true : 6s   (concat and uglify)

It's worth noting that the resulting main.min.js is the same in both cases.
Also, uglify automatically takes care of using the proper separator when combining the files.

The only case where it does matter is when adding sourceMap: true to the concat options.
This creates a main.js.map file next to main.js, and results in:

grunt default      : 6.2s (just uglify)
grunt default:true : 13s  (concat and uglify)

But if the production site loads only the min version, this option is useless.

I did found a major disadvantage with using concat before uglify.
When an error occurs in one of the JS files, the sourcemap will link to the concatenated main.js file and not the original file. Whereas when uglify does the whole work, it will link to the original file.

Update:
We can add 2 more options to uglify that will link the uglify sourcemap to concat sourcemap, thus handling the "disadvantage" I mentioned above.

    uglify: {
        options: {
            compress: true,
            mangle: true,
            sourceMap: true,
            sourceMapIncludeSources: true,
            sourceMapIn: '<%= paths.dest.js %>.map',
        },
        target: {
            src: '<%= paths.src.js %>',
            dest: '<%= paths.dest.jsMin %>'
        }
    }

But it seems highly unnecessary.

Conclusion

I think it's safe to conclude that we can ditch concat for JS files if we're using uglify, and use it for other purposes, when needed.

Answer 2

In the example you mention, which I'm quoting below, the files are first concatenated with concat and then uglified/minified by uglify:

{
  concat: {
    '.tmp/concat/js/app.js': [
      'app/js/app.js',
      'app/js/controllers/thing-controller.js',
      'app/js/models/thing-model.js',
      'app/js/views/thing-view.js'
    ]
  },
  uglifyjs: {
    'dist/js/app.js': ['.tmp/concat/js/app.js']
  }
}

The same could be achieved with:

{
  uglifyjs: {
    'dist/js/app.js': [
      'app/js/app.js',
      'app/js/controllers/thing-controller.js',
      'app/js/models/thing-model.js',
      'app/js/views/thing-view.js'
    ]
  }
}

Typically, the task clean would then run after tasks that write to a temporary folder (in this example concat) and delete whatever content is in that folder. Some people also like to run clean before tasks like compass, to delete things like randomly named image sprites (which are newly generated every time the task runs). This would keep wheels turning even for the most paranoid.

This is all a matter of preference and workflow, as is with when to run jshint. Some people like to run it before the compilation, others prefer to run it on compiled files.

Complex projects with an incredible amount of JavaScript files - or with a increasingly broad number of peers & contributors, might choose to concatenate files outside uglify just to keep things more readable and maintainable. I think this was the reasoning behind Yeoman's choice of transformation flow.

uglify can be notoriously slow depending of the project's configuration, so there might be some small gain in concatenating it with concat first - but that would have to be confirmed.

concat also supports separators, which uglify doesn't as far as README.md files are concerned.

concat: {
  options: {
    separator: ';',
  }
}