Escape literal [ in regexp

[josh]

Regexps parsing regexps always causing trouble.

It'd be nice if we could escape the literal [ in within the character class here to improve compatibility with "stricter ES5 regexp engines".

For context, I'm trying to get the CoffeeScript parser running under the Duktape interpreter. It has a stricter regexp parser svaarala/duktape#86 svaarala/duktape#74

/cc @sstephenson @jashkenas

[jashkenas]

Huh. I thought that /[[]/ was legit in JavaScript:

(/[[]/).test("one two [ three")
=> true

"escaping" these won't have any adverse consequences?

[josh]

I thought that /[[]/ was legit in JavaScript

"SourceCharacter but not one of \ or ] or -" from http://www.ecma-international.org/ecma-262/5.1/#sec-15.10.1 the regular expression grammar seems to exclude it.

Most other language's require it to be escaped too. /[[]/ is a syntax error in Ruby and Python. But seems to be accepted by Perl. My guess is most browsers have been just using a PCRE backed engine so its mostly kinda just worked.

"escaping" these won't have any adverse consequences?

Seems to still match any literal [ usage.

(/[\[]/).test("one two [ three")
// true

[michaelficarra]

SourceCharacter but not one of \ or ] or -

@josh: That allows [. There's no harm in escaping it, but there's also no reason to do so, even back to ES3. You should tell Duktape to fix its parser or RegExp engine.

[jashkenas]

@michaelficarra GitHub garbled that. It actually disallows [ as well.

[michaelficarra]

No, I don't think it does. I looked at the production in the spec.

[jashkenas]

Ah, you're right. I was confused by the "PatternCharacter" bit.

This is actually the relevant rule:

ClassAtomNoDash ::
    SourceCharacter but not one of \ or ] or -
    \ ClassEscape

I'll revert.

[josh]

Sorry, my mistake. The character class stuff actually seems fine, L840 doesn't need to be changed. The parse error is actually on the closing ] L844

Thats a literal ] thats unescaped. That seems to be excluded in the PatternCharacter set.

Would you guys reconsider just changing L844? or am I still wrong about the grammar 😉

[michaelficarra]

That sounds right. Please open a new PR.

[josh]

See #3893.