Fix gobbleGroup not working with sequence expressions, i.e. "(1, 2)"

[6utt3rfly]

• make the expressions[] parsing a separate function so it can be reused
• instead of parsing a single expression, let it parse multiple expressions
• Add test to compare against eprisma, which uses 'SequenceExpression' instead of 'Compound' in that case

Fixes #92

(it also allows for future arrow-function support, versus the code/comment from this PR )

[LeaVerou]

Hey, thanks! Are there any existing expressions that would be parsed differently now or does this only enable parsing for expressions that would previously throw?

[6utt3rfly]

Hey, thanks! Are there any existing expressions that would be parsed differently now or does this only enable parsing for expressions that would previously throw?

I pushed another commit to add a check for !nodes.length to return false so it matches previous behavior. So now the only difference should be that it will return a SequenceExpression node instead of throwing an Unclosed ( error.

The gobbleGroup code only gets run when the parser sees an open parentheses with no preceding identifier token (otherwise it would call gobbleArguments). So things like: (gobbleGroup) or identifer[1, (gobbleGroup)] or identifier((gobbleGroup)), etc. It also handles nested () properly now too: (((1))). I tried coming up with other expressions and comparing against the previous behavior but couldn't see any change.

[LeaVerou]

Thank you.
What about things like (1 + 2)? Is that still parsed the same?

[6utt3rfly]

You bet! Old/PR both respond the same, with a single node:

{
  "type": "BinaryExpression",
  "operator": "+",
  "left": {
    "type": "Literal",
    "value": 1,
    "raw": "1"
  },
  "right": {
    "type": "Literal",
    "value": 2,
    "raw": "2"
  }
}

[LeaVerou]

LGTM. Merging. Thanks again for working on this!