WebSQL Query

define context begin=0, end=P;
select e.href, d.title, e.context
from Document d such that "www/~gus/websql/demo-cascon/ullman-pubs.html" = d,
Anchor e such that base=d
where not (e.label contains "shorter" or e.href contains "html");
Jeffrey D. Ullman -- Papers Answering Queries Using Templates With Binding Patterns. In PODS 1995, with Anand Rajaraman and Shuky Sagiv. Shows how to construct queries from given query forms, i.e., views, that specify binding patterns.
Jeffrey D. Ullman -- Papers The TSIMMIS Approach to Mediation: Data Models and Languages. A survey of TSIMMIS in the NGITS Symposium, Naharia, Israel, June, 1995; many coauthors.
Jeffrey D. Ullman -- Papers Querying Semistructured, Heterogeneous Information (with Dallan Quass, Anand Rajaraman, Shuky Sagiv, and Jennifer Widom). Presents the LOREL query language, its motivation, and semantics. Also, a A shorter Version that appeared in DOOD '95.
Jeffrey D. Ullman -- Papers Medmaker: A Mediation System Based on Declarative Specifications. Appears in ICDE '96; with Yannis Papakonstantinou and Hector Garcia-Molina. Describes an "object logic" for generating mediators and its semantics. A shorter version.
Jeffrey D. Ullman -- Papers A Query Translation Scheme for Rapid Implementation of Wrappers (with Yannis Papakonstantinou, Ashish Gupta, and Hector Garcia-Molina). Allows query templates to be a recursively defined family of queries. Uses the containment test for a conjunctive query in a datalog program to find a best match between a query and a template. A shorter version that appeared in DOOD '95.
Jeffrey D. Ullman -- Papers Integrating Information by Outerjoins and Full Disjunctions (with Anand Rajaraman). Appears in 1996 PODS. The full disjunction of a collection of relations is the natural join of all the connected tuples, padded out with nulls. When can we compute the full disjunction by using the outerjoin operator on the relations in some order? The answer involves a venerable concept of Ron Fagin's called "gamma acyclicity." In addition to providing an interesting and useful algorithm for outerjoining, this paper reminds us of the often forgotten value of investigating natural concepts (like gamma-acyclicity) for their own sake.
Jeffrey D. Ullman -- Papers Information Integration Using Limited External Processors (with A. Y. Levy and A. Rajaraman). Solves the problem of whether a conjunctive query can be expressed in terms of views, where the (possibly infinite set of) views are those generated by the expansion of a Datalog program. Appears in 1996 PODS.
Jeffrey D. Ullman -- Papers Representative Objects: Concise Representations of Semi-Structured Hierarchical Data (with Svetlozar Evtimov, Sudarshan Chawathe, and Janet Wiener). A representative object tells us about the possible paths (label sequences) in OEM objects. We relate the construction of representative objects to finite automaton construction, and we give some simplifications that contain less information but are more compact.
Jeffrey D. Ullman -- Papers Information Integration Using Logical Views Invited paper for ICDT '97. Surveys relevant database theory and compares the use of this theory in Information Manifold (AT Labs) and Tsimmis (Stanford).
Jeffrey D. Ullman -- Papers Implementing Data Cubes Efficiently (with Venky Harinarayan and Anand Rajaraman). Appears in 1996 SIGMOD; winner: best paper award. Data cubes and related decision-support systems invite us to consider the materialization of a selected subset of a very large number of views. We model the problem as a lattice of views and show that the greedy algorithm of selecting for materialization those views that, given what views we have already materialized, offer the most improvement in average response time is within 63% of optimal in all cases. In many realistic cases, we can show the difference between the greedy and optimal solutions is essentially nothing.
Jeffrey D. Ullman -- Papers Index Selection for OLAP (with Himanshu Gupta, Venky Harinarayan and Anand Rajaraman). This follow-on to the above paper looks at the problem of incorporating indexes into the theory. The 63% bound seems to be based on a monotonicity condition that when we pick one view to materialize, it does not increase the benefit of any other view. Thinking of indexes on a view as another view, monotonicity is violated. When you materialize the underlying view, the value of its indexes jump up. In this paper we give several work-arounds, where we look at a view and some of its indexes to materialize together, in ways that don't make the selection process grow exponentially or high-degree polynomially. We show bounds on performance compared with the optimum that approach the 63% bound.
Jeffrey D. Ullman -- Papers Efficient Implementation of Data Cubes Via Materialized Views A survey of the field for the 1996 KDD conference.
Jeffrey D. Ullman -- Papers Constraint Checking With Partial Information (PODS 1994 paper with coauthors Ashish Gupta, Shuky Sagiv, and Jennifer Widom).
Jeffrey D. Ullman -- Papers The ``Lagunita-II report on the future of database research (coedited with Avi Silberschatz, Mike Stonebraker).
Jeffrey D. Ullman -- Papers A Survey of Research in Deductive Database Systems (with Raghu Ramakrishnan; appears in J. Logic Programming, May, 1995, pp. 125-149.
Jeffrey D. Ullman -- Papers The Database Approach to Knowledge Representation A brief survey of how the Datalog and Conjunctive-Query technology developed by the PODS community can be applied to problems in knowledge representation, especially information integration. Appears in 1996 AAAI.