What merlyn says is true enough, although there may be some confusion depending on what version of Perl you run merlyn's example with:

[danger:jandrew:~]$ cat blah.pl 
#!/usr/bin/perl -w
use strict;
my @data = qw/one start two three stop four/;
foreach (grep {not (/^start/../^stop/) =~ /^1?$|E/ } @data) { 
    print "o> $_\n";
}

[danger:jandrew:~]$ perl5.00503 blah.pl 
o> two
o> three

[danger:jandrew:~]$ perl5.6.1 blah.pl 
o> one
o> start
o> two
o> three
o> stop
o> four
[download]

Because, in 5.6+, the not operator with parens acts like a function (returns the negation of what's in the parens), so you'd want another set of parens to properly delimit what we want to negate (or use a unary +):

my @data = qw/one start two three stop four/;
foreach (grep {not ((/^start/../^stop/) =~ /^1?$|E/) } @data) { 
    print "o> $_\n";
}

# or:
my @data = qw/one start two three stop four/;
foreach (grep {not +(/^start/../^stop/) =~ /^1?$|E/ } @data) { 
    print "o> $_\n";
}
[download]

Thanks danger, I understand now. I have been wondering why the two gave such different results when they look so similar. I guess I'd better read the 5.6 docs again - I was assuming that the low precedence of not would make sure that the regex operation completed first - which as you have pointed out is true for versions prior to 5.6

Apologies to merlyn for saying that he was wrong, but I did test his example, and I did try to understand what was going on. I wasn't rude, so I am a bit upset at being --'d for my node.

I have learnt a lot in this node and merlyn's code, with danger's modification is going into the production system today.

Thanks monks.

-- iakobski