identify unique records

[tempest92]

hi,

can someone help please. i have a file in the following format -

serverName:username:uidgrp

e.g.

s1:bob:999:staff
s2:bob:888:staff
s3:bob:999:staff
s1:jon:777:staff
s2:jon:666:staff
s3:jon:666:staff

so i would like to read through this file and for the records that there is only 1 of i.e. jon:777 but there are other jon:xxx records in the file, write the jon:777 record to a file.

thanks

 

[columb]

This really requires a higher language than Unix scripting. Personally I'd do it in perl, I'm sure there are some who would do it in awk.

The (untested) perl version would go something like

my %users;
open FH, "datafile";
while (<FH>)
  {
  my ( $server, $username, $uid, $group ) = split /:/;
  exists $users{$username} and push @$users{$username}, "$serfer:$username:$uid:$group" or $users{$username} = [ "$serfer:$username:$uid:$group" ];
  }
close FH;
foreach my $user ( keys %users )
  {
  $#{@$users{$user}} or next;
  foreach ( @$users{$user} )
    { print "$_\n"; }
  }

Ceci n'est pas une signature
Columb Healy

 

[sbrews]

This will work... to an extent

sort -t: -k2 /some/file|uniq -f 1 -u > /some/newfile

- the input file is sorted on the 2nd field (username)
- the sorted output is passed to uniq, which will skip the
first field (server) and spit out only the lines that
are not repeated. IE s1:jon:777:staff

The catch here is that if the pgrp changes, IE its something other than staff, then uniq doesnt generate the desired output.

As columb said, you really want to use awk/perl/other higher level language to process this.