| In World Wide Web applications, a user request may produce a large result set. This result set is typically not returned to the user all at once. Instead, the user receives the first page of the result, and may request further pages one by one. This thesis describes a technique for providing this type of service when the result is produced on the back end by a relational database.; Disconnected Browsing is a technique for computing an SQL query progressively, N tuples at a time, on demand. The underlying base query is transformed into a page query. The first invocation of the page query produces the first page of N tuples of the result, the second produces the second N tuples, and so forth. In order to avoid returning tuples already presented, the page query is fully ordered, and parameterized by the last tuple on the previous page (the next-page tuple). Accordingly, the technique is almost stateless, since only a single tuple needs to be retained between invocations of the page query.; To perform efficiently, the query order must be efficiently computable. The thesis describes how to search for such efficient orders when the base query is unordered or only partially ordered. For efficiency, the technique must also be able to eliminate tuples before and after the currently requested page as early as possible. A predicate based on the next-page tuple eliminates Tuples after the current page. A Top N operator eliminates Tuples after the current page. The thesis describes techniques for making efficient use of both of these query features. |
