I've used HTML::TreeBuilder and WWW::Mechanize for such purposes before. But there may be something better ... I make no guarantees. :)

use HTML::TableExtract;
my $te = HTML::TableExtract->new;
$te->parse(join('', <>));
foreach my $row ($te->first_table_found->rows) {
  print join(':', @$row), "\n";
}
[download]

In reality, given the entire HTML document, you'd probably need to specify a depth/count in the constructor for H::TE.

Cheers,
Matt

Sometimes it's faster to do custom than learn a new module. Ya, I know, I didn't learn anything new - but the job will be done and the client will be happy.

But htey do look like highly useful modules for the not too distant future - like when this project gets more complicated and my 10-liner breaks!

like when ... my 10-liner breaks!

As many here will tell you, that is almost inevitable when trying to parse HTML on your own. :)
---
It's all fine and dandy until someone has to look at the code.

use FEAR::API -base;

file("your_source_file")
    ->document
    ->html_to_xhtml
    ->xpath("/html/body/table/tr");
doc_filter(use => 'remove_attributes');
print doc->as_string;
template(qq!<tr>\n<td><a>[% city %]</a></td>\n<td>[% cond %]</td>\n<td
+>[% temp %]</td>\n</tr>!);
extract;
print Dumper extresult;
[download]