So far, I haven't found a good generic approach to web scraping. The approach I usually use has condensed into App::scrape, which, while useful in itself, is more the general toolkit I use:

It uses

• HTML::TreeBuilder::XPath for parsing the HTML and querying the tree as XPath expressions
• HTML::Selector::XPath for querying the tree using CSS selectors
• LWP::Simple for page fetching

Of course, for the more complex tasks, LWP::Simple has to be replaced by WWW::Mechanize or something speaking Javascript, but if I need more complex navigation, I found no approach to a library or framework that makes this easier other than writing code for the WWW::Mechanize API directly.