Bug #12728: Negative lookahead does not work for "+" even though works for "@"

• Status: Rejected
• Assignee: -
• Target version: -
• ruby -v: ruby 2.2.5p319 (2016-04-26 revision 54774) [x86_64-cygwin]
• Backport: 2.1: UNKNOWN, 2.2: UNKNOWN, 2.3: UNKNOWN

[rklemme (Robert Klemme)]

I'll attach a test program that shows the effect. Basically, if I have a negative lookahead in the regex like (?!@) and "@" shows up in the proper location I get a mismatch (1. case). This is expected. If I exchange the "@" with a "+" or "[+]" in the regex and a "+" in the input, a match occurs (case 2 and 3). This is the bug. If the "+" or "@" is removed from the string an expected match occurs (case 4 and 5). I was not able to boil this down to a smaller example yet.

[naruse (Yui NARUSE)]

• Status changed from Open to Rejected

In case 2, the regexp just behave as if

t %r{
    (?<!\\)\(                           # outer bracket
    o\+

    (?<!\\) ([+*]|\{\d+,\}) (?!\+)  # inner repetition, non possessive

    .*
    (?<!\\)\)                           # outer bracket
    (?<!\\) (?:[+*]|\{\d+,\}) # unbounded repetition, non possessive

  }x, "f(o++)+"

Of course it matches.
Maybe use should [a-zA-Z0-9]* or something instead of .*.

[rklemme (Robert Klemme)]

• File rx-mini.rb rx-mini.rb added

Yui NARUSE wrote:

In case 2, the regexp just behave as if

t %r{
    (?<!\\)\(                           # outer bracket
    o\+

    (?<!\\) ([+*]|\{\d+,\}) (?!\+)  # inner repetition, non possessive

    .*
    (?<!\\)\)                           # outer bracket
    (?<!\\) (?:[+*]|\{\d+,\}) # unbounded repetition, non possessive

  }x, "f(o++)+"

Of course it matches.

Argh! Stupid me. Yes, the negative lookahead will also match with the closing bracket.

Maybe use should [a-zA-Z0-9]* or something instead of .*.

I included a closing bracket in the negative lookahead, then it works:

$ diff -U3 x1 x2
--- x1 2016-10-01 10:55:50.595060831 +0200
+++ x2 2016-10-01 10:55:44.459048792 +0200
@@ -13,20 +13,19 @@

/x

-MATCH
+NO MATCH
s = 'f(o++)+'
rx = /
(?<!\)( # outer bracket
(.*)

• (?<!\) ([+*]|{\d+,}) (?!+) # inner repetition, non possessive

• (?<!\) ([+*]|{\d+,}) (?!+|)) # inner repetition, non possessive

  (.)
  (?<!\)) # outer bracket
  (?<!\) (?:[+]|{\d+,}) # unbounded repetition, non possessive

/x
-match = #<MatchData "(o++)+" 1:"o+" 2:"+" 3:"">

MATCH
s = 'f(o++)+'

The following .* blinded me for the fact that the closing bracket can match with ) AND fulfill the lookahead like in

irb(main):001:0> /a(?=b)b/.match "abc"
=> #<MatchData "ab">

I am sorry for the hassle.

Kind regards

robert

PS: Attaching the test version that produced output x2.