Databases Reference
In-Depth Information
6.2
Extending Schema Mappings to Complex Data Models
We have recently seen research aiming to study the extensions needed to handle
the XML data model for schema mapping [
Arenas and Libkin 2008
;
Amano et al.
2009
], data transformation [
Jiang et al. 2007
], and query rewriting [
Yu and Popa
2004
]. The latter [
Yu and Popa 2004
] starts from proposing novel algorithms to
reformulate target queries against the source schemas, based on the mappings and
on the target constraints. Given that the data is at the sources, such queries need
to be efficiently evaluated and this work considers for the first time both relational
and XML schemas. The presence of the target constraints make the problem ways
more complicated by the fact that the data transformed according to the mapping
needs to be “merged” afterward. A further complication bears from the fact that
the target constraints can enable each other in a recursive way and interact with the
mappings as well. A canonical target instance is defined that takes into account the
presence of target constraints and mappings, and the semantics of query answering
is decided upon this target instance. Moreover, a basic query rewriting algorithm
focuses on only mappings first, and extends to XML queries and XML mappings
the relational techniques for query rewriting using views. The target constraints,
namely the NEGDS, covering XML schema key constraints among the others, are
then considered in a query resolution algorithm. Schema mapping for XML data has
been studied in
Jiang et al.
[
2007
], as an extension of the Clio system. In particular,
data transformations involving such a complex data model require more complex
transformation primitives than previously relational efforts. For instance, a key chal-
lenge arises with XML-to-XML data transformation if the target data is generated
as a hierarchy with multiple levels of grouping (as in Fig.
5.4
b in Sect.
3
). In such
a case, a deep union operator must be natively implemented in the transformation
engine (and this is done in Clio), as XML query languages, such as XQuery, XSLT,
and SQL/XML, are not yet suitable for such task of hierarchically merging XML
trees. Reasoning about the full structure of XML documents and developing a the-
ory of expressive XML schema mapping has been only recently tackled [
Arenas and
Libkin
2008
;
Amano et al. 2009
]. In particular,
Arenas and Libkin
[
2008
] focus on
extending the theory of data exchange to XML, and introduced the XML tree pat-
terns as XML schema mappings. Along the same lines, [
Amano et al. 2009
]present
an analog of source-to-target dependencies for XML schema mappings, discuss
their properties, including their complexity, and present static analysis techniques
for determining the “consistency” between source schemas and target schemas.
The problem of consistency was also dealt with in
Arenas and Libkin
[
2008
], and
in
Amano et al.
[
2009
] it is extended to consider all forms of navigational axes and
joins for XML query languages.
Recently, database vendors are extending their products to support ontologi-
cal reasoning capabilities. Following this direction, research on schema mapping
and query rewriting [
Calı et al. 2009a
,
b
] is focusing on the extension of classical
logical formalisms, such as Datalog, to support query answering over ontologies.
Datalog
C
enriches Datalog with existential quantifiers in the rule head, and allows
a set of restrictions to guarantee efficient ontology querying. In particular, the
