*Benjamin Goldberg adds a clever (and i think easier to understand) bit of
code to do this.
Index: perlfaq4.pod
===================================================================
RCS file: /cvs/public/perlfaq/perlfaq4.pod,v
retrieving revision 1.37
diff -u -d -r1.37 perlfaq4.pod
--- perlfaq4.pod 13 Nov 2002 06:04:00 -0000 1.37
+++ perlfaq4.pod 13 Nov 2002 06:32:46 -0000
@@ -593,6 +593,28 @@
@$ = (eval{/$re/},$@!~/unmatched/i);
print join("\n",@$[0..$#$]) if( $$[-1] );
+Benjamin Goldberg offers a Unicode-aware version that replaces
+the markers with wide, private use characters.
+
+ my $str = $_;
+ my @parts;
+ 1 while $str =~ s/BEGIN((?:(?!BEGIN)(?!END).)*)END/
+ push @parts, $1;
+ chr( 0xE000 + $#parts );
+ }ge;
+ my $re = @parts ?
+ sprintf "[%c-%c]", 0xE000, 0xE000 + $#parts :
+ '(?!)';
+ s/($re)/$parts[ord($1) - 0xE000]/g for @parts;
+
+The private use part of Unicode is from 0xE000 to
+0xF8FF---try not to have more than 0x18FF nested parts with
+this code. This is unlikely to work on 5.6.0, due to it's
+limited utf8 handling, unless you change 0xE000 to something
+smaller, like 0x80, which won't require an upgrade from
+bytes to utf8. But this means that it won't work on 8-bit
+data, only on 7-bit data.
+
=head2 How do I reverse a string?
Use reverse() in scalar context, as documented in
