Data Normalization - Information Systems Reengineering, Integration and Normalization - page 356

9.3 Normalize XML Schema with Minimum Fragments

This section aims to design an XML schema with minimum reference pointers and
redundancy freed for better performance.
In XML data modeling, there are two aspects of considerations. The normal one
parent many child elements tree structure, and pointers referenced elements struc-
ture. The former is more efficient than the latter. However, in order to implement
certain data semantics such as many-to-many cardinality and partial participation,
pointers must be used among elements for their connectivity (data navigation) to
each other. Therefore, if we consider each branch of elements under the root ele-
ment is a fragment, then the minimum fragmented elements XML design is more
efficient than fragmented elements XML design.
The benefit of such design is illustrated in Fig. 9.15 which shows the differ-
ence for an XML schema with or without minimum fragments. The XML schema
with minimum fragments can perform better than the XML schema without mini-
mum fragments. Figure 9.15 a shows a DTD Graph without reference pointers. Fig-
ure 9.15 b shows DTD Graph with reference pointer such that an idref referring to
an id. The effectiveness of deleting element A is same in both sides because element
B will be dangling with integrity error, that is, an idref without an id is an integrity
constraint error.
Following this conceptual model tradition, we seek for algorithms to derive good
XML schemas when the intended usage is for data storage in native XML databases
with their child elements stored in minimum fragments. A fragment is defined as
an element and its child elements immediately under root element. Fragmented ele-

DTD Graph

Root

Root

A1 +

B1

B A

A1

A

idref

id

*

B

B1

XML Document

<Root>
<A Attr_A= Ϙ A1">
<B Attr_B= Ϙ B1">
</Root>

<Root>
<A Attr_A= Ϙ A1" id= Ϙ 1">
<B Attr_B= Ϙ B1" idref= Ϙ 1">
</Root>

a Minimum fragmented DTD without reference pointer b Fragmented DTD with reference pointer

Fig. 9.15 XML database with and without minimum fragments