Database Reference
In-Depth Information
its properties in detail [12]. At the language level, GeRoMe uses source-to-target
extensional mappings which are based on second-order tuple generating depen-
dencies (SO tgs) [8] that are specified over the schema elements and their roles
[13] to express the relationships between heterogeneous schemas represented us-
ing different data models. A conjunctive query posed over an integration schema
expressed using the same formalism as the extensional mappings, is rewritten
into a query over the sources using composition of the conjunction of source-
to-target mappings between the source schemas and the integration schema and
the query itself. The predicates of the resulting query, which is expressed over all
the source schemas, are partitioned according to the corresponding sources and
are then translated into the corresponding source-specific query language (SQL
or XQuery) to be evaluated [13]. In contrast, rather than using composition we
have illustrated an approach for expansion of a query posed over an integration
schema using query unfolding [11] and presented in detail the rewriting of
SMql
(sub-) queries that are associated with specific sources into the source-specific
query languages (SQL and XQuery).
In contrast, Automed utilises a lower-level hypergraph data model consisting
of edges, nodes and constraints to represent schemas expressed in heterogeneous
data models including XML [15]. Relationships between different schemas are
expressed by a number of low level transformations between them, e.g., removing
a node or an edge, that can be combined to form more complex transformations.
The approach is called both as view (BAV) and the transformations are specified
in such a way that they are reversible and that both local as view (LAV) and
global as view (GAV) mappings can be derived between an integration schema
and source schemas from the BAV transformations [16,7]. A query over an in-
tegration schema or any of the source schemas can be expressed in Automed's
IQL query language, a comprehension-based functional query language, that is
reformulated into a query over the (other) sources schemas using a combination
of LAV and GAV query processing techniques over the BAV transformations
[17]. However, no detail is provided on how the IQL query posed over the source
schemas is rewritten into the source-specific query languages, which is the main
contribution of our approach presented here.
6 Conclusions
Complementing the model management platform MISM we have presented
SMql
,
a query language and its algebra over the MISM supermodel. We have illus-
trated an approach for expanding queries over multiple sources and presented
an approach for rewriting
SMql
queries into the corresponding source specific
queries posed over the sources to be queried. To add query rewriting capabilities
for other data models for which MISM already provides support, such as, the
object-relational model, the same approach as presented here for XSD and the
relational model can be followed, i.e., gather the information according to the
structure of the corresponding query language and use the information on the
correspondences between the model-specific model and the source-model inde-
pendent supermodel to create the specific target query. To include other models
