{
  my $count = 0;
  s/(\d+_([a-z]+))/++$count > 2 ? $1 : $2/eg;
}
[download]

-- Randal L. Schwartz, Perl hacker

s/^\d_(\w+)\/\d_(\w+)/$1\/$2/;
[download]

Hint: use anchors.

With your example, it appears you want to get rid of the first two digits and their following underscores that appear in the string. A regex that does that is:

$string =~ s/^\d_(\D+?)\d_(.*)/$1$2/;
[download]

the caret "^" at the beginning of this regex anchors things to the beginning of the string. You then make judicious use of captures to get the desired result.

Philosophy can be made out of anything. Or less -- Jerry A. Fodor

Actually this will not work if the string includes numbers in the part that you match with \D. I think you have to "unroll the loop" if you want to avoid (.*?):

#!/bin/perl -w
use strict;
foreach( <DATA>)
  { s{^\d+_                          # start then number(s) then _
       ((?:\D*(?:\d(?!\d*_))?)*)     # non_digits then optionnaly
                                     #   a digit _not_ followed by dig
+its and _
                                     # repeat, rince...
       \d+_                          # the second digit _ sequence
       (.*)}                         # the rest of the string
     {$1$2}x;
    print;
  }

__DATA__
1_abc/2_deg/bla_30_31_blah
1_ab1c/2_deg/bla_30_31_blah
1_ab1c/22_deg/bla_30_31_blah
1_a2b33c/22_deg/bla_30_31_blah
a_ab1c/b_deg/bla_30_31_blah
[download]

Note that your regexp will not work properly for the second string. I also added a line where the regexp does not match as this can show bugs when a badly written regexp takes for ever to match on a string that it does not match.

Frankly it this case, and despite all the warnings around, I would use (.*?) and write the regexp as:

s/^\d_(.*?)\d_(.*)/$1$2/;

I would be happy to get enlightened as to why this is not safe by the way.

    s#(^|/)\d_#$1#g;
[download]

$in='1_abc/2_deg/bla_30_31_blah';

#one-liner with split, nicely complicated (including backdoor regex)
$out=join '/', (split /_|\//, $in, 4)[1,3];

#This works if the second number is a single digit 
$out=substr $in, index( $in, '_') + 1;
substr( $out, index( $out, '_') - 1, 2)='';
[download]

Cheers,

Jeroen
"We are not alone"(FZ)
Update: Just thought of one substr use that doesn't break on multiple digits...:

$out=substr $in, index( $in, '_') + 1;
substr( $out, $us = index( $out, '/') + 1, index( $out, '_')- $us + 1)
+='';
[download]

s/(\d+_)(.*)?(\d+_)(.*)/$1$2/;
[download]

With what you have above, the replace should be $2$4, since you have 4 grouping operators there. Or you could just have "\d+_", or use "(?:\d+_)" as to use $1$2 as others have done later in this thread.

---

Dr. Michael K. Neylon - mneylon-pm@masemware.com || "You've left the lens cap of your mind on again, Pinky" - The Brain

Here is a fun (and powerful) way to only do some substitutions:

my $str= "1_abc/2_def/bla_30_31_blah";
my $count= 0;
while(  $str =~ /\d+_/g  ) {
    last   if  2 < ++$count;
    $str =~ s///;
}
print "\$str=($str)\n";
# $str=(abc/def/bla_30_31_blah)
[download]

my $str= "1_abc/2_def/bla_30_31_blah"; 
my $count= 0; 
while(  $str =~ /\d+_/g  ) { 
    last   if  2 < ++$count; 
    $str =~ s///; 
} 
print "\$str=($str)\n";
[download]

I believe that middle s/// is not doing what you think it is. It's going back to the beginning of the string, looking for the first match that matches the most recent successful regex match, and then removing it.

I think you lucked out on this one. Had your replacement been s//49_/, you would have seen only the first one changed, since the first replacement string would have been matched again on the second while-pass.

Oh, and even more so... I don't think your //g in the while is actually inching along! I think it's scanning from the beginning again as well, since you modify the scalar within the loop. So you just did this the hard way:

s/\d+_// for 1..2;
[download]

But that's not exactly what was asked for originally. It just happens to be good for this problem and this input data.

Well, it's what was literally asked for, but it's not very generalizable. {grin}

-- Randal L. Schwartz, Perl hacker

Thanks. I was surprised when that appeared worked.

Well, it would have been powerful... (:

        - tye (but my friends call me "Tye")

Instead of removing the first two occurrences, why not just remove the first occurrence, twice?

s/(^|/)\d_//;
s/(^|/)\d_//;
[download]