Regexp parse error - dollar escape /\$/ about duktape HOT 7 CLOSED

This is probably caused by the dollar escape $ inside the brackets. Ecmascript E5/E5.1 IdentityEscape production (which is the only applicable one AFAIK) is:

IdentityEscape ::
  SourceCharacter but not IdentifierPart
  <ZWJ>
  <ZWNJ>

Because "$" is a valid IdentifierPart, the escape $ is not valid as far as I understand, and \u0024 should be instead:

var a = /[\\\^\u0024\*\+\?\|\[\]\(\)\.\{\}]/g;

There are other non-standard regexp forms that nevertheless appear in existing code. A $ escape is already accepted outside a CharacterClass but not inside it. There's an ongoing task item to allow more non-standard regexps either by default or through a feature option.

from duktape.

Thanks for this detailed explanation. I agreed, but this runs nevertheless on Chrome, Firefox and v8 command line without any configuration parameters.

from duktape.

I know - there are actually a lot of behaviors that are not standard but still work in various other engines. Being fully compatible with other engines is not a practical goal because the non-standard features of other engines are not even compatible with each other (for instance function statements, a non-standard feature, have two competing semantics; Duktape adopts the V8 semantics here).

The approach with Duktape is that the standard behavior is implemented first, and non-standard behaviors are adopted when they either fulfill an important role (like the .stack property of errors) or are crucial to compatibility. Non-standard regexp syntax is one of the latter, and it's on the work list for future releases.

from duktape.

Ok, thanks!

from duktape.

I'll add dollar escape support inside character classes into the next release as it is an easy change (and dollar escape outside is already supported). There will still unfortunately be some real world regexps that won't parse - e.g. something similar to this came up with recent Emscripten (and also the test262 test suite I think):

/^function \(\) {.*}$/

Adding support to these requires a bit more work because backtracking is needed to handle the ambiguous unquoted braces. This change will be done at some point, because these regexps do crop up with existing code bases.

from duktape.

Ok, the "dollar escape inside character class" is now in the master branch, so your original regexp should now work :-)

from duktape.

Thanks!

from duktape.