Fellow Monasterians,

As a web developer I spend lots of time with HTML::Template and it's requirement for AoH in it's <tmpl_loop ...> function, which I use frequently. Many of the lists I loop in the browser need to be grouped first. It's great when I can use GROUP BY when querying the database, but often I don't have that luxury. My question is, what's the best way to group a list?

My example has several entries for a few users that I want to group, including sums of hours worked and a monetary extension.

my $allusers = [ {
                  'user'      => 'Sarah',
                  'duration'  => 2,
                  'amount'    => 200
               },
               {
                  'user'      => 'William',
                  'duration'  => 1,
                  'amount'    => 100
               },
               {
                  'user'      => 'Michael',
                  'duration'  => 3,
                  'amount'    => 300
               },
               {
                  'user'      => 'Michael',
                  'duration'  => 3,
                  'amount'    => 300
               },
               {
                  'user'      => 'William',
                  'duration'  => 7,
                  'amount'    => 700
               },
               {
                  'user'      => 'Sarah',
                  'duration'  => 5,
                  'amount'    => 500
               } ];
[download]

My first attempt sorts first and then iterates over the list—not especially efficient and a little goofy with the -1 counter:

$allusers = [ sort { $a->{'user'} cmp $b->{'user'} } @$allusers ];
            
my ($temp, $grouped);
my $ctr = -1; 

for ( @{ $allusers } ) {
   if ( $temp && $temp eq $_->{'user'} ) {
      $grouped->[$ctr]{'amount'}    += $_->{'amount'};
      $grouped->[$ctr]{'duration'}  += $_->{'duration'};
   } else {
      $ctr++;
      $temp = $grouped->[$ctr]{'user'} = $_->{'user'};
      $grouped->[$ctr]{'amount'}       = $_->{'amount'};
      $grouped->[$ctr]{'duration'}     = $_->{'duration'};
      
   }
}
[download]

I reworked it to eliminate the sort and cut the processing time basically in half:

my $allkeys;
for  ( @{ $allusers } ) {
    my $user = $_->{'user'};
    if ( !exists $allkeys->{$user} ) {
        my $attributes = { 'amount'    => $_->{'amount'},
                           'duration'  => $_->{'duration'}
                         };
        $allkeys->{$user} = $attributes;
        
    } else {
        my $refAttributes              = $allkeys->{$user};
        $refAttributes->{'amount'}     += $_->{'amount'};
        $refAttributes->{'duration'}   += $_->{'duration'};
    }
}

my $finalArray;
while ((my $key, my $value) = each %{$allkeys}) {
   my $hash = {  'user'       => $key,
                 'amount'     => $value->{'amount'},
                 'duration'   => $value->{'duration'}
               };
   push @{$finalArray}, $hash;
}
[download]

OUTPUT for both methods are the same:

print Dumper ($finalArray);
$VAR1 = [
          {
            'amount' => 600,
            'user' => 'Michael',
            'duration' => 6
          },
          {
            'amount' => 800,
            'user' => 'William',
            'duration' => 8
          },
          {
            'amount' => 700,
            'user' => 'Sarah',
            'duration' => 7
          }
        ];
[download]

But even the second way seems a bit tedious with that second while loop to build the final array. Is there a better way?