Solaris “clean” group files (or for that matter any text file)

Question

I am looking to find a way to script a cleaning operation on the /etc/group file on a large number of servers. Looking for users in a particular group more than once.

i.e. somegroup::100000:userA, userB, userC, foo, bar, snafu, userB, userB

Obviously for a small group this is easy and can be done manually, but I have some group files with large numbers of groups and entries.

I thought about writing it up in java since it should be pretty easy, but I would prefer finding something in perl or shell.

Thanks!

Sean

Answer 1

Here's a Perl script to do it.

while (<>) {
    chomp;
    my ($name, $pass, $gid, $users) = split ':';
    my %userset;
    my @users = split ',', $users;
    # create a set (a hash with undef keys) by assigning to hash slice
    @userset{@users} = ();
    $users = join ',', sort(keys(%userset));
    print join(':', $name, $pass, $gid, $users). "\n";
}

Use it like this:

$ ./fixgroups <group >group.tmp
$ diff -u group group.tmp
-somegroup::100000:userA,userB,userC,foo,bar,snafu,userB,userB
+somegroup::100000:bar,foo,snafu,userA,userB,userC
$ mv group.tmp group