Matching ( and ) in a regexp matcher requires escaping #177

Activity

1. lucacasonato commented on Jul 10, 2023

   

   lucacasonato

   on Jul 10, 2023

   Last edited by lucacasonato

   MemberAuthor

   More actions

   Actually I think we need to fix this for sure. Consider these two patterns:

   // valid regexp, but pattern tokenizer thinks there is a nested group that isn't closed
   const pattern = new URLPattern({ pathname: '/([(?])' });
   
   // valid regexp group, valid urlpattern, except that this throws because "Invalid regular expression: /^(?:/([))\]\)$/u: Unterminated character class" 
   const pattern = new URLPattern({ pathname: '/([)])' });
2. wanderview commented on Jul 10, 2023

   

   wanderview

   on Jul 10, 2023

   Member

   More actions

   Yea, seems like a real problem to me. I'm not sure when I will have the bandwidth to look at this, though. Do you have a proposed fix?
3. lucacasonato commented on Jul 10, 2023

   

   lucacasonato

   on Jul 10, 2023

   MemberAuthor

   More actions

   I think we have to keep track of [ and ] in the tokenizer while parsing regexp tokens, and ignore all ( and ) while between a [ and ] in that regexp. This can be a simple boolean as character classes can't be nested (no need to keep track of depth).

   👍React with 👍1Reacted by Ben Kelly
4. jeremyroman commented on Sep 13, 2023

   

   jeremyroman

   on Sep 13, 2023

   Collaborator

   More actions

   character classes can't be nested (no need to keep track of depth).

   Is this true? Given #178 we'll support syntax like [\d--[07]] (every digit except 0 and 7) as part of the UnicodeSets regexp mode.