Information Technology Reference
In-Depth Information
2 State of the Art
Transformations are usually related to the functional requirements aspects of
database schemas. Transforming a source schema must (should) preserve its
information capacity
4
in such a way that the eventual DDL code completely
translates the semantics of the conceptual schema. The use of transformations in
the context of schema quality mainly concerns non-functional requirements, and,
in this context, it has been rather limited. However, a few authors have already
considered processes in which a local set of objects in a schema is replaced by
another one in a way that improves some quality properties of the schema.
A first major (historical) proposal is the relational schema normalization pro-
cess [4], based on functional dependencies mainly in order to remove redundan-
cies at the data level. Though the term
transformation
was not used at that
time, normalization decomposition actually makes use of semantics-preserving
transformations
5
. These transformations can also be used to influence the perfor-
mance of the database. Leaving the semantics of data unaltered but improving
its redundancy or performance state, relational normalization clearly contribute
to make the schema meet non-functional requirements. In [5], the authors stud-
ied the impact of relationships types attributes on the clarity of ER schemas. In
a similar way in [6], Gemino and Wand have analysed the difference between the
use of the mandatory and optional properties, also in ER schemas. Though these
papers naturally called for substitution techniques to improve the readability of
schemas, the authors did not push their analysis to this point.
Only a few authors have explicitly used semantics-preserving transformations
for improving the schema quality. Among them, we can underline the framework
of Assenova and Johanesson [7] for dealing with understandability of conceptual
schemas. They assign qualitative quality scores to a set of transformations and
propose to use them in order to improve schema quality. Rauh and Stickel [8] also
use transformations in the context of conceptual schemas in order to normalize
them and therefore to improve their quality.
The framework we propose is close to the work of Assenova and Johanesson [7].
Yet, we paid particular attention to genericity, referring to the possibility to use
the framework on different abstraction levels, different paradigms and consider-
ing different quality criteria. Also, we did not associate quality preferences to
the transformations themselves but to the structures.
3 Framework Reminder
In section 1, we introduced the main principles of our framework. In this section,
we present some detail of the framework, based on reference [1], where the reader
can find an extended description.
4
A discussion on semantics preserving transformation and information capacity can
be found in [3].
5
The concept of semantics preservation is a bit more complex in this context since
data preservation and functional dependency preservation may conflict.
