performance difference between while (each) and foreach in Perl

Question

The hash that I tested contains around 70000 colleges and each college contains around 20 students. I tried it 5 times and following are the results. There is considerable difference in foreach performance and while (each) performance. Why is that so?

Code with while loop:

while ( my ($college_code, $college_info_hr) = each (%{$college_data_hr}) ) {
    while ( my ($student_num, $student_info_hr) = each (%{$college_info_hr->{'students'}}) ) {
        if($student_num < 104000) { ## Delete the info of students before 2004.
            delete $college_info_hr->{'students'}{$student_num};
        }
    }
}

Code with foreach loop:

foreach my $college_code (keys %{$college_data_hr}) {
    foreach my $student_num (keys %{$college_data_hr->{$college_code}{'students'}}) {
        if($student_num < 104000) { ## Delete the info of students before 2004.
            delete $college_data_hr->{$college_code}{'students'}{$student_num};
        }
    }
}

When the number of colleges are 70,000 then the execution times are:

For the code with while loop (Interval time is in seconds):

Interval time: 2.186621

Interval time: 2.058644

Interval time: 2.055645

Interval time: 2.101637

Interval time: 2.124632

For the code with foreach loop: (Interval time is in seconds)

Interval time: 1.341768

Interval time: 1.436751

Interval time: 1.346529

Interval time: 1.302775

Interval time: 1.356765

When the number of colleges are 248,000 then the execution times are:

(execution times for while loop)

Interval time: 9.084427

Interval time: 8.438684

Interval time: 9.329338

Interval time: 9.169687

(execution times for foreach loop)

Interval time: 5.502048

Interval time: 6.386692

Interval time: 5.596032

Interval time: 5.620144

Answer 1

The foreach version only dereferences the $college_data_hr->{$college_code}{'students'} hashref once per college, so is faster than the while version which needs to do it once per student.

The foreach version will likely use more memory though, as it needs to build temporary lists containing the keys for each hash.

Data::Alias might help you speed up the while solution. I've not benchmarked this, but it should be fairly fast...

use Data::Alias;

while ( my ($college_code, $college_info_hr) = each %$college_data_hr ) {
    alias ( my %students = %{$college_info_hr->{'students'}} );
    while ( my ($student_num, $student_info_hr) = each %students ) {
        if ($student_num < 104000) { ## Delete the info of students before 2004.
            delete $students{$student_num};
        }
    }
}