From: Stefan Lang <perfectly.normal.hacker@...>
Date: 2008-03-08T09:41:28+09:00
Subject: Re: Bug in regex engine ?? Must be...

2008/3/3, D. Krmpotic <david.krmpotic@gmail.com>:
> Hi,
>
>  I'm using Ruby 1.8.6, and I just discovered something rather
>  interesting, here is a test:
>
>  require 'test/unit'
>
>  class TestRegexBug < Test::Unit::TestCase
>
>   def test_bug
>
>     hours = "pon-èet"
>
>     assert(hours =~ /[è]et/i)
>     assert(hours =~ /èet/i)
>     assert(hours =~ /-èet/i)
>     assert(hours =~ /[cè]et/i)
>     assert(hours =~ /-[è]et/i)
>
>   end
>
>  end
>
>  As you can see, this only happens with unicode letters... (the last test
>  fails).. I'm used to the fact that //i doesn't work for unicode chars
>  and I already know that you need two dots to match one of these.. But
>  this problem is different and weirder, because what triggers it is a
>  minus sign before the square brackets.. if you remove either the '-' or
>  '[]' from the regex, it works..

In the regex [è] is a character class with _two_ bytes. So
Ruby tries to match a minus followed by _one_ of the bytes
out of "è" followed by "et".  So the regex would match
"pon-\304et" or "pon-\215et", but not "pon-\304\215et".

Stefan