Regex match information has a somewhat complex interaction with lexical scope. The 'grep' case fails because the scope of the expression is left and re-entered. The 'for' case succeeds because the scope of the body of the 'for' is never effectively left/entered during the running of the loop (where what "effectively" means is intentionally vague).

My evidence to support this claim is that adding another block of scope to the 'for' loop breaks it as well:

use strict;
my $if = 1;
my @list= ("a".."c","DBIC","A".."C","DANCER","a".."c");
my @result;
for (@list) {
    if(  $if  ) {   # Comment this out
        push @result,$_ if /DBIC/.. /DANCER/ and ! //;
    }               # Comment this out
}
print $_, $/ for @result;
__END__
[download]

Comment out the two indicated lines to again see A .. C output. As is, the above code outputs nothing (similar to the 'grep' case).

Looks like a bug to me, whatever special casing .... doesn't happen the same way with grep

$ perl -Mre=debug -le " @F = qw( a b c DBIC A B C DANCER a b c ); for( +@F){print if /DBIC/.. /DANCER/ and ! // } " Compiling REx "DBIC" Final program: 1: EXACT <DBIC> (3) 3: END (0) anchored "DBIC" at 0 (checking anchored isall) minlen 4 Compiling REx "DANCER" Final program: 1: EXACT <DANCER> (4) 4: END (0) anchored "DANCER" at 0 (checking anchored isall) minlen 6 Compiling REx "" Final program: 1: NOTHING (2) 2: END (0) minlen 0 Guessing start of match in sv for REx "DBIC" against "DBIC" Found anchored substr "DBIC" at offset 0... Guessed: match at offset 0 Guessing start of match in sv for REx "DBIC" against "DBIC" Found anchored substr "DBIC" at offset 0... Guessed: match at offset 0 A B C Guessing start of match in sv for REx "DANCER" against "DANCER" Found anchored substr "DANCER" at offset 0... Guessed: match at offset 0 Guessing start of match in sv for REx "DANCER" against "DANCER" Found anchored substr "DANCER" at offset 0... Guessed: match at offset 0 Freeing REx: "DBIC" Freeing REx: "DANCER" Freeing REx: "" [download]

$ perl -Mre=debug -le " @F = qw( a b c DBIC A B C DANCER a b c ); prin +t for grep { /DBIC/.. /DANCER/ and ! // } @F; " Compiling REx "DBIC" Final program: 1: EXACT <DBIC> (3) 3: END (0) anchored "DBIC" at 0 (checking anchored isall) minlen 4 Compiling REx "DANCER" Final program: 1: EXACT <DANCER> (4) 4: END (0) anchored "DANCER" at 0 (checking anchored isall) minlen 6 Compiling REx "" Final program: 1: NOTHING (2) 2: END (0) minlen 0 Guessing start of match in sv for REx "DBIC" against "DBIC" Found anchored substr "DBIC" at offset 0... Guessed: match at offset 0 Guessing start of match in sv for REx "DBIC" against "DBIC" Found anchored substr "DBIC" at offset 0... Guessed: match at offset 0 Matching REx "" against "A" 0 <> <A> | 1:NOTHING(2) 0 <> <A> | 2:END(0) Match successful! Matching REx "" against "B" 0 <> <B> | 1:NOTHING(2) 0 <> <B> | 2:END(0) Match successful! Matching REx "" against "C" 0 <> <C> | 1:NOTHING(2) 0 <> <C> | 2:END(0) Match successful! Guessing start of match in sv for REx "DANCER" against "DANCER" Found anchored substr "DANCER" at offset 0... Guessed: match at offset 0 Guessing start of match in sv for REx "DANCER" against "DANCER" Found anchored substr "DANCER" at offset 0... Guessed: match at offset 0 Freeing REx: "DBIC" Freeing REx: "DANCER" Freeing REx: "" [download]

#!/usr/bin/perl -w
use strict;
use Data::Dumper;

my @list= ("a".."c","DBIC","A".."C","DANCER","a".."c");

my @result;
for (@list) { push @result,$_ if /DBIC/.. /DANCER/ and ! // }

print Dumper \@result;

# uses http://www.perlmonks.org/?node_id=525392
my @result2 = grep { (/DBIC/.. /DANCER/)  =~ /^\d+(?<!^1)$/ }@list;
 
print Dumper \@result2;

__END__
$VAR1 = [
          'A',
          'B',
          'C'
        ];
$VAR1 = [
          'A',
          'B',
          'C'
        ];
[download]

As another note this "xE0" notation is used in other places in Perl. For example, the DBI can return 0E0 as the number of results which is 0 * 10**0 or numeric zero (0 * 1) although this evaluates to "true" when tested in a logical sense, meaning that that the statement "worked" but returned no results (numeric value is zero). Here the line number ending in E0 is the "last one". This is clever but it works.

Good catch, but the whole implementation is annoyingly inconsistent.

I thought I could make this more obvious by implementing functions like nodege(), noflip(), and noflop() to write something like

grep { noedge(/DBIC/ .. /DANCER/)  }@list;

but try guessing which magic behavior hinders that:

  DB<52>  @list= ("a".."c","DBIC","A".."C","DANCER","a".."c");
 
  DB<53> sub ff { print $_[0]}
 
  DB<54>  grep { ff(/DBIC/ .. /DANCER/)  } @list;
0000000000
  DB<55>  grep { my $x=(/DBIC/ .. /DANCER/); ff($x) } @list;
12345E0
[download]

Of course I could also use something like compiled regexes for EDGE, FLIP, FLOP to be able to type something like:

grep { (/DBIC/ .. /DANCER/) !~ EDGE  } @list;

but that doesn't seem to be trivial, too :-(

Thanks anyway!

Cheers Rolf

ARGH!!!!!!!!!

  DB<52>  @list= ("a".."c","DBIC","A".."C","DANCER","a".."c");
 
  DB<53> sub ff { print $_[0]}
 
  DB<54>  grep { ff(/DBIC/ .. /DANCER/)  } @list;
0000000000
  DB<55>  grep { my $x=(/DBIC/ .. /DANCER/); ff($x) } @list;
12345E0
[download]

ff() evaluates the range operator in list context!

applying prototypes solves this "problem":

  DB<182> @list= ("a".."c","DBIC","A".."C","DANCER","a".."c");
 
  DB<183> sub ff ($) { print $_[0]}
 
  DB<184> grep { ff(/DBIC/ .. /DANCER/)  } @list;
12345E0
[download]

Cheers Rolf ( ...banging his head against the table...)

That doesn't work if DBIC appears multiple times (which it can).