Problem with hash

[shilpam]

I have a hash for storing the userid, managerid pair. The data is something like this:
{UserId} => ManagerId

{1234} => "6789"
{1235} => "6790"
{6789} => "6791"
{1236} => "6789"

I need to insert data for each user. But, the problem is if I try to insert data for the 1'st record (userid:1234), it won't allow me because the data for its manager(i.e. 6789) is not yet inserted. The data for this manager comes later in the hash (i.e. record 3)
Can someone suggest me a way so that I insert data for all the managers first & then their subordinates?

[pelagic]

As still nobody has a really good solution for that hierarchy sort problem I had to hack the following:

#!/usr/bin/perl

use strict;
use Data::Dumper;
my $form = "%30s  %s\n";

my %upgrades;
my %lvl;
my %PofC = (
    1   => 2
,   3   => 2
,   4   => 2
,   2   => 5
,   6   => 5
,   7   => 4
,   5   => undef
,   8   => 4
);
print Dumper \%PofC;

foreach my $child (keys %PofC) {
    $lvl{$child} = 0;
}

my $leftovers = 1;
my $level = 0;
while ($leftovers > 0) {
    %upgrades = ();
    $leftovers = 0;
    foreach my $child (keys %PofC) {
        if (exists $lvl{$PofC{$child}}
        &&  $lvl{$PofC{$child}} == $level)
        {
            $upgrades{$child} = 1;
            ++$leftovers;
        }
    }
    foreach (keys %upgrades) {
        ++$lvl{$_};
    }
    ++$level;
}

print Dumper \%lvl;

printf $form, 'child', 'parent';
foreach my $c ( sort { $lvl{$a} <=> $lvl{$b} } keys %lvl ) {
    printf $form, $c, $PofC{$c};
}


exit;

__OUTPUT__
# %PofC
$VAR1 = {
          1 => 2,
          2 => 5,
          3 => 2,
          4 => 2,
          5 => undef,
          6 => 5,
          7 => 4,
          8 => 4
        };
# %lvl
$VAR1 = {
          1 => 2,
          2 => 1,
          3 => 2,
          4 => 2,
          5 => '0',
          6 => 1,
          7 => 3,
          8 => 3
        };
                         child  parent
                             5  
                             2  5
                             6  5
                             4  2
                             3  2
                             1  2
                             7  4
                             8  4

[shilpam]

Thanks a lot for your solution!!!

[ozone]

one way is to use a recursive sub, that 'walks' up the userid tree until it reaches a node without a manager (or a special id (like 0 or undef)

the other way is do the same as above, but use a stack and a loop

some pseudo code

# simple recursive sub
sub create_users {
my $user_href = shift;
my $userId = shift;

if(defined $user_href->{$userId} and $user_href->{$userId} != 0) {
create_users($user_href, $userId);
} else {
return;
}

# now write the user to the db
write_user_to_db($user_href, $userId

}

I think the above should work :-)

[Zaxo]

Highly data dependent, but it looks like reverse sorting the keys you want will get the managers in first.

On the other hand, I don't see why you need to have a key in place before you duplicate it as a value. Writing,

my %organization = (
    1234 => '6789',
    1235 => "6790',
    6789 => '6791',
    1236 => '6789',
);

works fine. If you mean to have values as references to other hash elements, yes, the order matters, but that brings in concerns about circular references also. (Shouldn't be a problem with a proper tree.)

[thor]

There's something that you're not telling us. Hashes by themselves don't care about the data that's placed in to them. So, you must've imposed that restriction yourself. To which I would reply "don't do that". :) If all of your inserts in to the hash happen at the same time (or to be more precise, happen one after the other without anything in between), I'd do all of my inserts, and then an integrity check at the end. Something simple...like this:

foreach my $emp (values %hash) {
   print "ERROR: $emp not defined\n" unless exists($hash{$emp});
}