As a one liner...

perl -pi.bak -e 's/^cn:(.+)\n$/cn: $1\nuid: $1\n/;' file.name
[download]

See perlrun, perlreintro, and perlretut for more information on this one liner's technique.

The -p switch causes Perl to iterate through all the files in @ARGV, line by line, reading each line into $_, and then at the end of each loop iteration printing $_. The -i switch specifies "inline editing", which is implemented by opening a tempfile for output, and the input file for...input. After each iteration, $_ is printed to the temp file. Then at the end of script execution the original file is renamed with '.bak' appended to its name, and the tempfile is renamed to what the original file's name was. So when you're all done, it looks like your original file got modified, and its pre-modification version was saved with the .bak filename extension.

Update: I just realized this doesn't solve the issue of the uid: already existing in some places. Sorry.

The one-liner, perl -ni -e'print; print "uid:", $1, $/ if /^cn:(.*)/;' foo.file

Update: Argh, I missed that bit, too. Off the top of my head, perl -ni -e'print; if (/^cn:(.*)/) { my $uid = "uid:$1\n"; my $nl = <>; print $nl =~ /^uid:/ ? "": $uid, $nl}' foo.file Untested.

use strict;
use warnings;

open OUTPUT, '>', 'output.txt' or die;

my @record;
my $uid_flag = 0;

while (<DATA>){
  chomp;
  unless ( /^$/ ){
    push @record, $_;
    $uid_flag++ if /^uid/;
  }
  else{
    for ( @record ){
      print OUTPUT "$_\n";
        if ( /^cn/ and ! $uid_flag ){
          my ( $field, $value ) = split ':';
          print OUTPUT "uid:$value\n";
        }
    }
    @record = ();
    $uid_flag = 0;
    print OUTPUT "\n";
  }
}
close OUTPUT;

__DATA__
dn: cn=MTOP,cn=users,dc=myco,dc=com
orcldefaultprofilegroup: cn=myco,cn=portal_groups,cn=groups,dc=myco,dc
+=com
cn: MTOP
orclactivestartdate: 20031028060637Z
objectclass: top
objectclass: person
objectclass: inetOrgPerson
objectclass: organizationalPerson
objectclass: orclUser
objectclass: orclUserV2
sn: Joe
givenname: Blow
o: myo
userpassword: {MD4}oLezu+AceuHBDEBHFuVDwg==

dn: cn=me,cn=users,dc=polk,dc=com
orcldefaultprofilegroup: cn=user_grp,cn=portal_groups,cn=groups,dc=myc
+o,dc=com
userpassword: {MD4}oLezu+AceuHBDEBHFuVDwg==
orclactivestartdate: 20031030114555Z
objectclass: top
objectclass: person
objectclass: inetOrgPerson
objectclass: organizationalPerson
objectclass: orclUser
objectclass: orclUserV2
cn: me
uid: me
sn: Administrator
givenname: me
o: myco
[download]

dn: cn=MTOP,cn=users,dc=myco,dc=com
orcldefaultprofilegroup: cn=myco,cn=portal_groups,cn=groups,dc=myco,dc
+=com
cn: MTOP
uid: MTOP
orclactivestartdate: 20031028060637Z
objectclass: top
objectclass: person
objectclass: inetOrgPerson
objectclass: organizationalPerson
objectclass: orclUser
objectclass: orclUserV2
sn: Joe
givenname: Blow
o: myo
userpassword: {MD4}oLezu+AceuHBDEBHFuVDwg==

dn: cn=me,cn=users,dc=polk,dc=com
orcldefaultprofilegroup: cn=user_grp,cn=portal_groups,cn=groups,dc=myc
+o,dc=com
userpassword: {MD4}oLezu+AceuHBDEBHFuVDwg==
orclactivestartdate: 20031030114555Z
objectclass: top
objectclass: person
objectclass: inetOrgPerson
objectclass: organizationalPerson
objectclass: orclUser
objectclass: orclUserV2
cn: me
uid: me
sn: Administrator
givenname: me
o: myco
[download]

{
  local $/ = ""; # set to empty string -- see perldoc perlvar

  while (<>) {
     if ( /(\ncn: (\S+))\n(\S+)/ ) {
        my ( $cnline, $cnval, $nexttag ) = ( $1, $2, $3 );
        if ( $nexttag ne "uid" ) {
           s/$cnline/$cnline\nuid: $cnval/;
        }
     } else {
        warn "Record $. has no cn value\n";
     }
     print;
  }
}
[download]

The line:

if ( $nexttag ne "uid" ) {
[download]

needs to be:

if ( $nexttag ne "uid:" ) {
[download]

(tested!)
Picky or what!
I'm not familiar with the 'LDAP Data Interchange Format' (to say the least) but is 'uid' always bound to follow 'cn'?
wfsp

Thanks for the fix (and the test).

I'm not familiar with the 'LDAP Data Interchange Format' (to say the least) but is 'uid' always bound to follow 'cn'?

I don't know either, but if it varies, reading the data one whole record at a time will make it a lot easier to handle the variation.