Schema Translation - Information Systems Reengineering, Integration and Normalization - page 84

+LHUDUFKLFDOVFKHPD 0DSSHGUHODWLRQDOVFKHPD

DFFW
QDPH

*$$

5HODWLRQV
*$$ DFFW QDPH
*$% DFFWPHWHU
*$& DFFWELOOPR QHWBFKDUJH

ELOOPR
QHWBFKDUJH

*$% *$&

PHWHU

Fig. 3.2 A hierarchical schema mapped to a relational schema

3.3 Indirectly Translating a Network Model

into a Relational Model

In much of the published literature on schema translation by direct translation as-
sumptions have generally been made on the semantics of the database. There is
always a chance that the translated schema may not encapsulate the original de-
signer's idea. This problem occurs because there are so many possible relational
schema that can be derived from a known hierarchical schema or a network schema
and the translation analyst makes many very primitive assumptions (for example,
the direct translation of a hierarchical schema or a network schema into a compos-
ite key of the relational model by concatenating its parent record key with its own
key). However, there are exceptions such that the child record is fully internally
identified, which can be transformed directly to a primary key of a logical relational
schema. As a result, the translated relational model may be incorrect.
When a company's existing database system needs to be upgraded into a new
model such as relational, object-oriented, or XML, the current nonrelational data
models must be translated into the new models. To translate from one model to an-
other involves not just data structure transformation, but also the transfer of seman-
tics. Very often, semantics are lost once a conceptual model has been mapped into a
logical model because the former is richer in semantics than the latter. Thus, schema
translations between logical schema such as hierarchical, network, relational, object
oriented, and XML are done by mapping them back to a higher semantic model of
the EER model.
To solve the problem in a logical manner, we need users as the domain and rela-
tion integrity experts for the nonrelational schema. They can provide information
on the semantics of the data; that is, their domain values and constraints in the da-
tabase. A knowledge acquisition system can assist the user to confirm the translated
EER model by enforcing the database integrity heuristic rules such as functional
dependencies (FDs) and inclusion dependencies (IDs) in the translation. The re-
sultant conceptual model can meet the heuristic rule requirements in the existing
nonrelational schema. Even though there are many possible EER models that can
be constructed from a known logical schema, the translated EER model should be
the one closest to the user's expectation.