The first place to go is `man perlre`: you should get rid of all these [ and ] characters when they're redundant..

for( qw{
   2312112
   abc-erte
} ){
   print "$_ ",
      m<^(?:
        \d{5,8}  #5 to 8 digits
           |     # OR
        \w{3}    # 3 alpha chars 
                 # (maybe you want to replace \w)
        -        # one hyphen
        .{4}     # 4 chars
      )$>x
     ? 'matches' : 'doesn\'t match', "\n";
}
[download]

update: typos at end of regexp

#!/usr/bin/perl

use strict; 
use warnings;


my @test = ( 123, 12345, 12345678, "abc-abc", "abc-abcd" );

foreach my $match ( @test )
{
  if ( $match =~ m/\d{5,8}|\w{3}-\w{4}/ )
  {
    print "$match match\n";
  }
  else 
  {
    print "$match doesn't match\n";
  }
}
[download]

m/[\d]{5,8}|[:alpha:]{3}[-]{1}[:alpha:]{4}/g
[download]

I think you should use [[:alpha:]], not [:alpha:]. I'm not sure, though

- Yes, I reinvent wheels.
- Spam: Visit eurotraQ.

No you got that right. Be unsure no longer. The named POSIX classes are nice for keeping things clear especially when you consider character sets and Unicode. [A-Za-z0-9_] is sufficient for US-ASCII but it's wordy and easier to mistype than [:alphanum:]. The nice thing is that if you start moving beyond US-ASCII then you've already taken care of some of th work up front. It's also more in line with what I gather will be the expected idiom for perl6. The idea here is that it is infeasable for you to actually specify what is a "word" character for whatever character set you're using in your regular expression.