Mapping-Based Merging of Schemas - Schema Matching and Mapping - page 222

not attributes and are of the same type, then the resulting element is of the original
type.
2. Links integration rule: For any pair of elements that are matched to each other,
if the elements are of the same type, then the links between them are of the same
type.
3. Paths integration rule: If there is a direct path between two nonattribute elements
in one model and an equivalent indirect path in the other model, choose the indi-
rect path. The reason for this is that the indirect path contains more information,
and the direct path can always be inferred from the indirect path.
4. Integration of attributes of corresponding elements: For elements that are already
integrated, if each attribute pair in the original diagrams match, add the attributes
to the resulting schema.
5. Attributes with path integration rule: If there is a direct path between an element
and an attribute in one model and an indirect path between the corresponding
pair in the other model, choose the indirect path.
6. Add rule for elements and links without correspondent: Add all elements that
exist in one model and not in the other.

At a high level, their algorithm first merges the nonattribute elements, then
merges the paths, and finally merges the attributes. The full algorithm is described
in Spaccapietra and Parent [ 1994 ].
This work is notable because it marks a departure of focus from the existing
literature: it concentrates on the fact that type information may conflict (rules 1 and
2) as well as the fact that some relationships that exist explicitly in the input schemas
can be represented implicitly in the resulting schema (rules 3 and 5). This use of
implicit relationships is featured prominently in later work, particularly in generic
schema merging (Sect. 6 ). It is probably not a coincidence that this work, like the
generic work, is not based in the relational model: the more direct representation
of relationships between schema elements allows the authors to deal with complex
relationships without running into the undecidability result in Convent [ 1986 ].

3.4 Rosenthal and Reiner

A contemporary work with Spaccapietra and Parent [ 1994 ] is The Database Design
and Evaluation Workbench [ Rosenthal and Reiner 1994 ]. The workbench allows
users to manipulate schemas, including combining multiple views into one schema.
Their ER-based system largely focuses on “rearrangements,” which transform the
input but leave the content “equivalent” to the input. The primary transformations
that they considered were:

Duplicating the attributes of an entity to propagate to related entities, or alter-
nately removing duplication if necessary.

Simplifying relationships by decomposing them into two simpler relationships.