Databases Reference
In-Depth Information
This problem has been studied for quite some time, and is the subject of an early
survey [
Batini et al. 1986
].
Batini et al.
[
1986
] categorizes view integration work as
taking one or more of the following steps:
Preintegration:
Deciding which schemas to be integrated, in which order the inte-
gration should occur, and various preferences (e.g., if one of the schemas is
“preferred” over the other).
Comparison of the schemas:
Determining the correspondences and detecting the
possible conflicts. In this context, a conflict is when a concept is represented
differently in the input schemas. For example, a simple conflict might be that
there is an attribute “Last Name” in one schema that is represented by an attribute
“LName” in another schema.
Conforming the schemas:
resolving the conflicts between the schemas; the
authors note that automatic resolution is not typically possible in schema
conformation.
Merging and restructuring:
Now that the schemas are ready to be superimposed,
how should they be combined?
Batini et al.
[
1986
] offers the following qualitative
criteria to decide on the “correctness” of the merged schema:
Completeness and correctness
Minimality
Understandability
These criteria are seen again and again in a number of different guises throughout
the schema merging literature. As far as schema merging is concerned, this catego-
rization is the main contribution of Batini, Lenzerini, and Navathe's paper; the bulk
of the remainder is concentrated on matching. Again, matching (i.e., determining
what concepts in one schema are related to the concepts in another schema) is out-
side the scope of this paper and is surveyed in existing surveys (e.g.,
Rahm and
Bernstein
2001
;
Doan and Halevy 2004
;
Rahm 2011
) (see also Chap. 2). Our work
focuses on the “merging and restructuring.”
The view integration problem was subsequently studied in many areas, including
ER diagrams [
Song et al. 1996
;
Lee and Ling 2003
], XML [
Beeri and Milo 1999
;
Tufte and Maier 2001
;
Yang et al. 2003
], semi-structured data [
Bergamaschi et al.
1999
], relational and object-oriented databases [
Larson et al. 1989
;
Shu et al. 1975
;
Biskup and Convent 1986
;
Navathe and Gadgil 1982
;
Shoval and Zohn 1991
], and
others. The remainder of this section details a few of the schema merging algorithms
in the context of view integration.
3.1
Biskup and Convent
Biskup and Convent
[
1986
] define a formal language for view integration and then
proceed to integrate based on that language. This fairly early work provides a list of
details need to be provided to create a view integration system: