Databases Reference
In-Depth Information
Inferring constraints from existing constraints, or alternately removing duplicate
constraints.
Creating keys.
The definition of equivalent, while different in a few details, is very similar to
the notion of information capacity in Sect.
2.1
. One scenario that they tackle is that
of view integration. The authors state that their goals are much more pragmatic
than some of the existing work; as previously discussed [
Batini et al. 1986
;
Biskup
and Convent
1986
;
Casanova and Vidal 1983
], take a more theoretical approach. As
such, Rosenthal and Reiner concentrate on a usable tool: they only detect homonyms
and synonyms, and such conflicts are presented to the user for resolution. They then
perform a duplicate removing union between the two schemas. No mappings are
created between the schemas.
These works represent the type of focus on schema merging present in the
more recent view integration literature. After this point, more of the database
research in schema merging came from data integration (Sect.
4
) and generic schema
management (Sect.
6
).
4
Data Integration
As motivated in the introduction, in data integration, there exists a set of hetero-
geneous, independent sources that contain related data. To have these sources be
queried efficiently, a mediated schema can be created. Because these sources are
independent, heterogeneous, and often change rapidly, it is not possible to then
import all of the data into the mediated schema. Instead, the users query the medi-
ated schema, and then the system translates the queries over the mediated schema
into queries over the sources.
While much of the time this mediated schema is created manually, there exist a
number of works that discuss creating the mediated schema based on the sources.
4.1
Data Warehousing
DWQ is a system for creating a data warehouse from the sources [
Calvanese et al.
1999
].
Calvanese et al.
[
2001
] focuses on the data integration aspects of DWQ. This
paper describes their system on how to use data integration in data warehousing.
One issue with building a data warehouse is that it often has to be highly tuned for
the specific queries; e.g., one might want to have a star schema (i.e., a base “fact”
table from which various dimensions measuring things such as time and location
radiate) for the data warehouse instead of whatever format just happens to be the
merged result of the sources.
