Not quite so simple. Better would be:

my %hash2 = map {/^   "xxxx-(\d+):xxxxxx([^,]+),.+"$/ ? ($1 => $2) : (
+)} @data;
[download]

How would that be "better"?

$ perl -MData::Dumper -le'
my @dc = ( "xxxx-123:xxxxxxABC,z", "xxxx-:xxxxxx,z", "xxxx-456:xxxxxx,
+z", "xxxx-XXX:xxxxxxDEF,z", "xxxx-789:xxxxxxGHI,z" );
my %hash1 = map /^xxxx-(\d+):xxxxxx([^,]+),.+$/, @dc;
my %hash2 = map { /^xxxx-(\d+):xxxxxx([^,]+),.+$/ ? ($1 => $2) : () } 
+@dc;
print Dumper \%hash1, \%hash2;
'
$VAR1 = {
          '123' => 'ABC',
          '789' => 'GHI'
        };
$VAR2 = {
          '123' => 'ABC',
          '789' => 'GHI'
        };
[download]

Grandfather's statement is better because it gracefully(and explicitly) handles the case where the regex does not match.

     "As you get older three things happen. The first is your memory goes, and I can't remember the other two... " - Sir Norman Wisdom

thanks guys, my script works like a charm, but Grandpa's map statement is very stylish(i put it into my /etc/motd for now) thanks!!!!