Regex should be working but not in PrismJS?

I have this:

Prism.languages.insertBefore("javascript", "comment", {
    "def": {
        pattern: /(var\s+|(?!^),\s*|(?!^),\s*^|[^\\:]\/\/.*\n\s*)((?!\d)[a-z0-9_$]+)(?= *[=,;])/,
        lookbehind: !0
    }
})

This should highlight all variable identifiers in the declaration code block.

It works on Regex101: https://regex101.com/r/uF4oY4/46

Why doesn’t it work with this? The y in var x, y and y = undefined aren’t highlighted.

var x;
var y;
//or
var x, y; //with multiple variables declared

var x,
    y = undefined;

The y isn’t wrapped in any class.

---

P.S. I know, it falsely highlights this, but that’s not the point:

x, y;

[Golmote]

The problem is related to the parts (?!^) in your regexp. Prism breaks the string into pieces progressively as it matches patterns, so here are the strings that are tested, in order:

1. First, Prism will test the entire text for your `def` pattern:

   var x;
   var y;
   //or
   var x, y; //with multiple variables declared
   
   var x,
       y = undefined;
   

   This will match var x

2. Then Prism will test the following string. Notice how the first match has been removed.

   ;
   var y;
   //or
   var x, y; //with multiple variables declared
   
   var x,
       y = undefined;
   

   It will match var y

3. Then we have this:

   ;
   //or
   var x, y; //with multiple variables declared
   
   var x,
       y = undefined;
   

   It will match var x

4. Then we have this:

   , y; //with multiple variables declared
   
   var x,
       y = undefined;
   

   And here is the problem: your regexp can’t match because of the (?!^) clauses that fail. Indeed, the way Prism works, the comma is actually at the beginning of the string.

I hope it is more clear.

[Golmote]

Well… CodeMirror reads the code character by character to determine the tokens. This way of doing things allows it to understand the code. See the code source of the JS mode in CodeMirror: http://codemirror.net/mode/javascript/javascript.js.

Prism, on the other side, uses regexps to determine the tokens. So it is not aware of the actual meaning of the code. This was a design choice made by @LeaVerou, to allow for easier addition of new languages and small code size. See the code source of the JS component in Prism: https://github.com/PrismJS/prism/blob/gh-pages/components/prism-javascript.js