Decode & back to & in JavaScript [duplicate]

Most answers given here have a huge disadvantage: if the string you are trying to convert isn't trusted then you will end up with a Cross-Site Scripting (XSS) vulnerability. For the function in the accepted answer, consider the following:

htmlDecode("<img src='dummy' onerror='alert(/xss/)'>");

The string here contains an unescaped HTML tag, so instead of decoding anything the htmlDecode function will actually run JavaScript code specified inside the string.

This can be avoided by using DOMParser which is supported in all modern browsers:

function htmlDecode(input) {
  var doc = new DOMParser().parseFromString(input, "text/html");
  return doc.documentElement.textContent;
}

console.log(  htmlDecode("<img src='myimage.jpg'>")  )    
// "<img src='myimage.jpg'>"

console.log(  htmlDecode("<img src='dummy' onerror='alert(/xss/)'>")  )  
// ""

This function is guaranteed to not run any JavaScript code as a side-effect. Any HTML tags will be ignored, only text content will be returned.

Compatibility note: Parsing HTML with DOMParser requires at least Chrome 30, Firefox 12, Opera 17, Internet Explorer 10, Safari 7.1 or Microsoft Edge. So all browsers without support are way past their EOL and as of 2017 the only ones that can still be seen in the wild occasionally are older Internet Explorer and Safari versions (usually these still aren't numerous enough to bother).

Do you need to decode all encoded HTML entities or just & itself?

If you only need to handle & then you can do this:

var decoded = encoded.replace(/&/g, '&');

If you need to decode all HTML entities then you can do it without jQuery:

var elem = document.createElement('textarea');
elem.innerHTML = encoded;
var decoded = elem.value;

Please take note of Mark's comments below which highlight security holes in an earlier version of this answer and recommend using textarea rather than div to mitigate against potential XSS vulnerabilities. These vulnerabilities exist whether you use jQuery or plain JavaScript.

EDIT: You should use the DOMParser API as Wladimir suggests, I edited my previous answer since the function posted introduced a security vulnerability.

The following snippet is the old answer's code with a small modification: using a textarea instead of a div reduces the XSS vulnerability, but it is still problematic in IE9 and Firefox.

function htmlDecode(input){
  var e = document.createElement('textarea');
  e.innerHTML = input;
  // handle case of empty input
  return e.childNodes.length === 0 ? "" : e.childNodes[0].nodeValue;
}

htmlDecode("<img src='myimage.jpg'>"); 
// returns "<img src='myimage.jpg'>"

Basically I create a DOM element programmatically, assign the encoded HTML to its innerHTML and retrieve the nodeValue from the text node created on the innerHTML insertion. Since it just creates an element but never adds it, no site HTML is modified.

It will work cross-browser (including older browsers) and accept all the HTML Character Entities.

EDIT: The old version of this code did not work on IE with blank inputs, as evidenced here on jsFiddle (view in IE). The version above works with all inputs.

UPDATE: appears this doesn't work with large string, and it also introduces a security vulnerability, see comments.