No tests for type: 'ref'

[mathiasbynens]

There are currently no tests/examples for atoms with type: 'ref' (see createRef()).

[mathiasbynens]

Also, let’s rename this to reference now that we still can. Currently it’s the only type name that is an abbreviation. :(

[mathiasbynens]

The reason there are no tests for type: 'ref' is probably because back-references are parsed as if they were simple values.

$ regjsparser '(s)\1'
{
    "type": "alternative",
    "body": [
        {
            "type": "group",
            "behavior": "normal",
            "body": [
                {
                    "type": "value",
                    "kind": "symbol",
                    "codePoint": 115,
                    "range": [
                        1,
                        2
                    ],
                    "raw": "s"
                }
            ],
            "range": [
                0,
                3
            ],
            "raw": "(s)",
            "matchIdx": 1
        },
        {
            "type": "value",
            "kind": "octal",
            "codePoint": 1,
            "range": [
                3,
                5
            ],
            "raw": "\\1"
        }
    ],
    "range": [
        0,
        5
    ],
    "raw": "(s)\\1"
}

Back-references should be preserved in the AST.

Ref. mathiasbynens/regexpu#4
Ref. bnjmnt4n/regjsgen#5

[jviereck]

This is strange and seems like a regression. Testing the ecma suite with the full RegExpJS engine worked fine last year during JSConf, so I am sure it worked somewhen^^

@mathiasbynens, I am heads down in other stuff at the moment, but I will talk about RegExp.JS at Amsterdam.JS and will make sure this is fixed before then ;)

[mathiasbynens]

Thanks for the info, @jviereck! I’ll try to look into it as well.

[bnjmnt4n]

This seems to be related to a variable called lastMatchClosed. If you look in parseDecimalEscape, detection of references depends on lastMatchClosed, but lastMatchClosed remains at 0 throughout the script.