Assuming that you need to figure out the columns dynamically, here's one way to do it:
use warnings;
use strict;
use XML::LibXML;
use Text::CSV;
my $dom = XML::LibXML->load_xml(location => 'example.xml');
my %items;
for my $product ($dom->findnodes('/products/product')) {
my $sku = $product->findvalue('sku');
for my $downloads ($product->findnodes('downloads')) {
for my $group ($downloads->findnodes('group')) {
for my $group_items ($group->findnodes('group_items')) {
for my $item ($group_items->findnodes('item')) {
my $name = $item->findvalue('name');
my $url = $item->findvalue('url');
$items{$sku}{$name} = $url;
}
}
}
}
}
my @columns = sort keys %{{map {$_=>1} map {keys %$_} values %items}};
my $csv = Text::CSV->new({binary=>1, auto_diag=>2, eol=>$/,
always_quote=>1 });
$csv->print(select, ["sku", @columns]);
for my $sku (sort keys %items) {
$csv->print(select, [ $sku, map { $items{$sku}{$_} } @columns ]);
}
Note that this does assume that SKUs and the item names per SKU are unique (Update: in other words, that later <item>s don't overwrite earlier ones in the %items hash). Also, note that unless you need to separate out the <downloads>, <group>, or <group_items> elements, you don't need all those nested loops, this produces the same output:
for my $product ($dom->findnodes('/products/product')) {
my $sku = $product->findvalue('sku');
for my $item ( $product->findnodes(
'downloads/group/group_items/item') ) {
my $name = $item->findvalue('name');
my $url = $item->findvalue('url');
$items{$sku}{$name} = $url;
}
}