From: ts <decoux@...>
Date: 2000-09-02T22:46:07+09:00
Subject: [ruby-talk:4747] Re: Possible regex bug?

>>>>> "D" == Dave Thomas <Dave@thomases.com> writes:

D> So, in this respect, Perl and Ruby do the same thing.

 There is a difference between perl and ruby.

 If you look at simple example you'll probably not see it, for example :

pigeon% cat b.rb
#!/usr/bin/ruby
'foo'.scan(/o?/) do
   p "match <#$&>"
end
pigeon% b.rb
"match <>"
"match <o>"
"match <o>"
"match <>"
pigeon% 

pigeon% cat a.pl
#!/usr/bin/perl -l
print "match <$&>" while 'foo' =~ /o?/g;
pigeon% 

pigeon% a.pl
match <>
match <o>
match <o>
match <>
pigeon%

 But when you write regexp with non-greedyness you see the difference :

pigeon% cat b.rb
#!/usr/bin/ruby
'foo'.scan(/o??/) do
   p "match <#$&>"
end
p 'bar'.gsub(/\w??/) do |x|
   "<#{x}>"
end
pigeon% 

pigeon% b.rb
"match <>"
"match <>"
"match <>"
"match <>"
"<>b<>a<>r<>"
pigeon% 

pigeon% cat a.pl
#!/usr/bin/perl -l
print "match <$&>" while 'foo' =~ /o??/g;
$_ = 'bar';
s/\w??/<$&>/g;
print;
pigeon% a.pl
match <>
match <>
match <o>
match <>
match <o>
match <>
<><b><><a><><r><>
pigeon% 

 The difference seems because perl has complex rule to resolve zero-length
 match (a zero-length match don't move the "cursor").

 For ruby an expression like /\w??/ in global context is more like :
   1) try the next expression in the regexp (in this case return with an
      empty match)
   2) move the "cursor"
   3) go to 1

p.s.: the examples are extract from perlre


Guy Decoux