Do you want to iterate over the whole thing, or do you have the keys for the node you want to access? If the latter then the code I posted in reply to your question can be used to access the node given the "key path". Otherwise you need to specify your problem a little more accurately.

The following code may help however:

my %data;

@{$data{a}{b}{c}{columns}} = [1, 2, 3];
@{$data{a}{b}{c}{values}} = [4, 5, 6];
@{$data{d}{e}{columns}} = [7, 8, 9];

my %leaves;

walkHash (\%data);

print "columns: ";
print "@$_ " for @{$leaves{columns}};
print "\nvalues: ";
print "@$_ " for @{$leaves{values}};

sub walkHash {
    my $subHash = shift;
    
    for (keys %$subHash) {
        if (/columns/) {
            push @{$leaves{columns}}, @{$subHash->{$_}};
        } elsif (/values/) {
            push @{$leaves{values}}, @{$subHash->{$_}};
        } else {
            walkHash ($subHash->{$_});
        }
    }
}
[download]

Prints:

columns: 1 2 3 7 8 9 
values: 4 5 6
[download]