That Document structure stores a URL and its pagerank for later processing.

In the Sawzall paper, it's used in a program that reports the top URL from a domain, ranked by pagerank:

     proto "document.proto"
     max_pagerank_url:
          table maximum(1) [domain: string] of url: string
                weight pagerank: int;

     doc: Document = input;
     emit max_pagerank_url[domain(doc.url)] <- doc.url
          weight doc.pagerank;