Range information is inconsistent for Program nodes with leading whitespace

[not-an-aardvark]

// espree 3.4.3

require('espree').parse('/* foo */ bar /* baz */', { range: true })

Output

{
  "type": "Program",
  "start": 0,
  "end": 23,
  "range": [
    10,
    13
  ],
  "body": [
    {
      "type": "ExpressionStatement",
      "start": 10,
      "end": 13,
      "range": [
        10,
        13
      ],
      "expression": {
        "type": "Identifier",
        "start": 10,
        "end": 13,
        "range": [
          10,
          13
        ],
        "name": "bar"
      }
    }
  ],
  "sourceType": "script"
}

The Program node has start: 0 and end: 23, but its range is [10, 13].

This happens because espree transforms the range property to match Esprima's behavior. The transformation is correct, but we should also update the start and end properties, since they're equivalent in all other cases.

[JamesHenry]

Ah I didn't originally see this issue, but see mine in babylon @babel/parser for more comparisons:

babel/babylon#673

(Was auto-migrated to) babel/babel#6681

[platinumazure]

@not-an-aardvark Is this still an issue?

[fisker]

Is this going to get fixed?

[nzakas]

@fisker can you confirm that this is still an issue?

[fisker]

Yes, I'm sure, I found this during working on Prettier, flow and espree are different from other parsers, so I filed issue on flow facebook/flow#8537, and comment here.

[mdjermanovic]

TemplateElement nodes also have start and end different from range.

[nzakas]

This is already marked as accepted, so any volunteers? 😄

[anikethsaha]

I am working on this. This should probably be a breaking change.