Gulp plugin to strip inline scripts from HTML- or HTML-like files, stick 'em in a file property

gulp-strip-scripts.js

/* global Buffer, console */
/**
* Strip inline <script> elements and push each
* <script> tag's contents to `file.scripts`
* (Array; each element is the contents of a script element)
*/
'use strict';
var through = require('through2');
var gutil   = require('gulp-util');
var cheerio = require('cheerio');

module.exports = function gulpStripScripts(opts) {

  opts = opts || {};
  opts.property = opts.property || 'scripts';

  return through.obj(function(file, enc, cb) {
    var $, content;

    if (file.isNull()) {
      this.push(file);
      return cb();
    }

    if (file.isStream()) {
      cb(new gutil.PluginError('stripScripts', 'Streaming not supported'));
      return;
    }

    content = file.contents.toString();

    try {
      $ = cheerio.load(content);
    } catch (e) {
      console.warn(e);
      return;
    }
    file[opts.property] = (typeof file[opts.property] !== 'undefined' &&
                           Array.isArray(file[opts.property])) ? file[opts.property] : [];
    $('script').not('[src]').each(function(i, elem) {
      file[opts.property].push($(elem).text());
      content = content.replace($(elem).toString(), '');
    });

    file.contents = Buffer(content);
    this.push(file);
    cb();
  });
};

Would it be more reliable to use Cheerio to transform the entire "DOM" of the file by removing all <script>s without the src attribute instead of iterating over each one and doing a string manipulation each time?

Example:

$('script:not([src])').remove();
content = content.replace($.html());

@erikjung The traversal is necessary to a certain extent because I'm operating on file.contents ultimately, not the parsed DOM ($). Because the file's content string to cheerio DOM to string again is by no means idempotent (and a lot of the files we're processing are fragmentary HBS templates), it doesn't make sense to operate directly on $ (i.e. remove the <script> elements using the cheerio API and then toString at the end). I also need to extract the script but not the tags (ergo the .text() invocation).

However, you're right, I can certainly tighten up that selector! Updated to

    $('script').not('[src]').each(function(i, elem) {
      file[opts.property].push($(elem).text());
      content = content.replace($(elem).toString(), '');
    });

That's a very valid point about file.contents.toString() !== $.html().

To piggy-back on the [src] selector exclusion, it might be nice to allow embedded templates to pass through as well by acknowledging the type attribute in some way...

/** 
 * All scripts excluding:
 * a) those with a "src" attribute, and/or
 * b) those with a "type" attribute having a value other than "text/javascript"
 */
$('script').not('[src], [type][type!="text/javascript"]').each(function(i, elem) {
  // ...
});

To support cases such as this: http://jsbin.com/ciseyicuto/1/edit?html,js