Database Reference
In-Depth Information
Therefore, we need to consider the possible interactions of the new source with each
of the existing sources, and this limits the ability of the GAV approach to scale to a
large collection of sources. In the Local-As-View (LAV) approach, the descriptions
of the sources are given in the opposite direction. That is, the contents of a data
source are described as a query over the global schema relation.
Notice that GLAV includes as special cases both LAV (when each assertion is of
the form
q
) and GAV (when each assertion is
of the form
q
G
(
x
,
z
)
=
r
g
(
x
,
z
)
, for some relation
r
g
in
(
x
)
=
r
s
(
x
)
, for some relation
r
s
in
S
S
) data integration mapping
in which the views are sound. The GLAV system in data integration corresponds to
the data exchange setting. Both data exchange and data integration use the certain
answers as the standard semantics of queries over the target (global) schema. How-
ever, it should be noted that there are important differences: in data integration, we
do not materialize a target (global) schema and hence the source instances are used
to compute the certain answers by using the query-rewriting algorithms. In contrast,
in our generalization of the data exchange setting, it may not be feasible to cou-
ple applications as in data integration because a global schema is contemporarily
a source schema for other schemas. This may occur, for instance, in peer-to-peer
applications that must share data, yet maintain a high degree of autonomy. Hence,
queries over a given schema may have to be answered using its materialized instance
alone.
Moreover, each integrity constraint for the global schema
G
G
in
Σ
T
will be one of
the possible cases:
1. A
tuple-generating dependency
(
tgd
) of the form
z
ψ
(
x
,
z
)
.
∀
x
(
∃
y
φ
(
x
,
y
)
=⇒ ∃
(4.3)
G
G
In the Data Integration setting, we will consider only the class of weakly-full
tgd for which query answering is decidable, i.e., when the right-hand side of the
implication has no existentially quantified variables, and if each
y
i
∈
y
appears
at most once on the left side), for example, in the inclusion constraints (IC) and
its more specific case of the foreign-key (FK) constraints.
2. An
equality-generating dependency
(
egd
)
x
φ
G
(
x
)
x
2
)
(x
1
.
∀
=⇒
=
(4.4)
where the formulae
φ
G
(
x
)
and
ψ
G
(
x
,
y
)
are conjunctions of atomic formulae
over
, and
x
1
,
x
2
are among the variables in
x
, used, for example, for the key-
constraints (PK) and the exclusion constraints (EC).
Let us consider the most general case of GLAV mapping:
G
Definition 30
For
a
general
GLAV
data
integration/exchange
system
I
=
G
,
S
,
M
where
G
=
(S
T
,Σ
T
)
is the global and
S
=
(S
S
,
∅
)
is the source database
schema and
a set of tgds which map a view of source database into a view of
global database, we define the following graph
G
M
:
I
