classEscape SyntaxError

[mdjermanovic]

Module build failed (from ./node_modules/babel-loader/lib/index.js):
SyntaxError: /home/vsts/work/1/s/lib/linter/config-comment-parser.js: classEscape at position 12
    a-zA-Z0-9\-/]+):
              ^
    at bail (/home/vsts/work/1/s/node_modules/regjsparser/parser.js:1118:13)

ref eslint/eslint#12660

[mdjermanovic]

The regex in question is: /([a-zA-Z0-9\-/]+):/gu

That \- could be an unnecessary escape, but v8 doesn't complain about it.

[jviereck]

Firefox does this:

(/([a-zA-Z0-9\-/]+):/gu).source == "([a-zA-Z0-9\\-/]+):"

[mdjermanovic]

Firefox does this:

(/([a-zA-Z0-9\-/]+):/gu).source == "([a-zA-Z0-9\\-/]+):"

Why that doesn't look correct, \\ on the right side is a string escape?

With the u flag, \- seems to be explicitly allowed in character classes as a ClassEscape? (not IdentityEscape)

[mysticatea]

/([a-zA-Z0-9\-/]+):/gu is a valid regular expression, but regjsparser raises wrong syntax error. (https://runkit.com/embed/3a5p2ec8y9fl)

See the above link that throws on a native regular expression.

[chriseppstein]

This is breaking ember applications. Will there be a fix released soon?

Workaround for yarn users:

{
  "resolutions": {
    "regjsparser": "0.6.0"
  }
}

[jviereck]

Looking at this and will keep you posted.

[jviereck]

Reduction:

regjsparser.parse('[\\-/]', '') // Works (1)
regjsparser.parse('[\\-/]', 'u') // Fails (2)

Also note, that

// Executed in Firefox 71
r = new RegExp('[\\-/]')
r.exec('\\') // -> null
r.exec('-') // -> Array ['-']
r.exec('/') // -> Array ['/']

Note that the parsing of regjsparser.parse('[\\-/]', '') returns the correct AST matching the output required to what the test against the real engine in Firefox is producing.

From this, it seems the problem is when using the 'u' flag for parsing the string.

[jviereck]

@chriseppstein , I am about to board my NYC -> ZHR bound plane and cannot commit to have a fix done soon. In case this is time critical, can you maybe do a workaround in the meantime by avoid using the "u" flag?

[chriseppstein]

@jviereck Thank you for looking into this. It's a transitive dependency of regexpu-core and I'm pretty sure the point of using it is the u flag 😬

[jviereck]

Thanks to the quick PR from @JLHwung this was fixed in #99.

RegExp provided as error case in the first issue is parsing now upstream: http://www.julianviereck.de/regjsparser/#%2F%28%5Ba-zA-Z0-9%5C-%2F%5D%2B%29%2Fug

A new v0.6.2 was drafted and pushed to NPM: https://www.npmjs.com/package/regjsparser/v/0.6.2

[jviereck]

@chriseppstein @mdjermanovic please let us know in case there are still problems on this one!

On a personal note, it's useful for me to know what's the impact of a reported bug is to give it the proper priorities. Reading that this breaks Ember applications made this a P0 for me.

[JLHwung]

@jviereck As a side note, astexplorer.net has supported regjsparser: https://astexplorer.net/#/gist/ef57a380fe2b8297bb9c1ea563ee5025/ba427130bca471ec269cea94f07c489462cdd7fc