Database Reference
In-Depth Information
1
Introduction and Technical Preliminaries
1.1
Historical Background
Database schema mappings are so important in information integration that many
mapping formalisms have been proposed for different tasks. A schema mapping is
a high-level
declarative specification
of the relationship between two schemas, it
specifies how data structured under one schema, called source schema, is to be con-
verted into data structured under possibly different schema, called the target schema.
It the last decade, schema mappings have been fundamental components for both
data exchange and data integration. In this work, we will consider the declarative
schema mappings between relational databases. A widely used formalism for spec-
ifying relational-to-relational schema mappings is that of tuple generating depen-
dencies (tgds). In the terminology of data integration, tgds are equivalent to global-
and-local-as-view (GLAV) assertions. Using a language that is based on tgds for
specifying (or 'programming') database schema mappings has several advantages
over lower-level languages, such as XSLT scripts of Java programs, in that it is
declarative and it has been widely used in the formal study of the semantics of data
exchange and data integration. Declarative schema mapping formalisms have been
used to provide formal semantics for data exchange [
21
], data integration [
38
], peer
data management [
25
,
29
], pay-as-you-go integration systems [
72
], and model man-
agement operators [
5
].
Indeed, the use of higher-level declarative language for 'programming' schema
mappings is similar to the goal of model management [
4
,
63
]. One of the goals in
model management is to reduce programming effort by allowing a user to manipu-
late higher-level abstractions, called models, and mappings between models (in this
case, models and mappings between models are database schemas and mappings
between schemas). The goal of model management is to provide an algebra for ex-
plicitly manipulating schemas and mappings between them. A whole area of model
management has focused on such issues as mapping composition [
22
,
43
,
67
] and
mapping inversion [
20
,
23
].
