MULTIDATABASE QUERYING BY CONTEXT
Date
2000-10-16
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
The overwhelming acceptance of the SQL standard \cite{Date94} has
curtailed continuing research work in relational database query languages and
processing. Since all commercial relational database systems conform with the
SQL standard, there is little motivation for developing new query languages.
Despite its benefits and wide-spread acceptance, SQL is not a perfect
query language. Complex database schema challenge even experienced database
users during query formulation. As increasing numbers of less sophisticated
users access numerous data sources within an organization or across the
Internet, their ability to accurately construct queries with the appropriate
structure and semantics diminishes. SQL can be hard to use as it provides
only physical access transparency not logical transparency. That is, a user
is responsible for mapping the semantics of their query to the semantics and
structure of the database. Although graphical tools for query construction
and high-level programming languages mask some of the complexity, the notion
of querying by structure is intrinsic to most forms of data access. In
this work, we overview a new query language developed in conjunction with our
integration architecture for automatically integrating relational schema.
Although the major focus of this work is on database interoperability, the
contribution of this paper is a language for specifying queries on the
integrated view produced. The complexities of querying across database
systems and resolving conflicts are too numerous to be fully described here,
so this paper will discuss querying the integrated view of a single database.
The integration architecture integrates database schema information
into a context view (CV). The context view is a high-level view of database
semantics which allows logically and physically transparent access to the
underlying data source(s). Since this context view is an entirely new way of
organizing and categorizing database information, a new query language is
developed. However, we demonstrate that the context view has similar
properties as the Universal Relational Model and thus can benefit from its
associated algorithms and ideas. By allowing the user to query by
context and semantic connotation, a whole new level of query complexity arises.
Mapping of queries from semantic concepts to physical tables, fields, and
relationships must be automatically performed. We will demonstrate that
specific relational calculus expressions or SQL queries can be generated from
abstract concepts which are rigorous enough for use in industrial applications
and systems. Specifically, SQL generation and join discovery are overviewed.
Thus, the query language can be mapped to SQL allowing backwards compatibility
with existing systems. Notes: Joint released technical report. Released as
TR-00-16 for the University of Manitoba, and 2000-663-15 for the University of
Calgary.
Description
Keywords
Computer Science