Information Technology Reference
In-Depth Information
Fig. 2.
A form-based interface describing a “Person” and its associated conceptual
schema, using the GER representation
2
-
the processing of the validated constraints and dependencies, once there are
no other potential constraints or dependencies left.
The acquisition of data samples progressively restricts the set of potential con-
straints, while conversely, validating constraints also restricts the future data
samples that will be encodable.
4.2
Initialisation
Before beginning the interaction with the end-users, we start by initialising an
empty set of tuples and defining the initial sets of validated, potential and dis-
carded constraints and dependencies for each entity type associated with a given
form. The
validated
constraints are initially the same than the previously
de-
clared
constraints. From these initial validated constraints, the
potential
domains
of value, cardinalities and existence constraints are initialised using
induction
.
Typically, if a given component is considered optional so far, it could actually
be mandatory if there is no tuple with this component empty (whereas the op-
posite is not possible). Similarly, an attribute declared as textual could be of
any other type, while an attribute declared as real could only be restricted to
the type integer. In the same way, any subset of optional components for which
no existence constraint has been declared should be submitted to existence con-
straint elicitation. Consider for instance that the attribute
Zip code
has been
declared optional and textual in Fig. 2. Further examination should therefore
check whether this attribute is not mandatory and whether its value domain is
not restricted to integers, reals, dates, ...
Regarding functional dependencies, the ideal process should lead us to build
a set of data samples and dependencies so that each entity type of the underly-
ing conceptual schema becomes an
Armstrong relation
.Reachingsuchastateis
2
Note that the GER notation uses the
participation
interpretation rather than the
look-across
interpretation (as in UML). In the given example, the notation therefore
indicates that same contact details may apply to more than one person.
