Database Reference
In-Depth Information
As an example, consider an instance with the following tuples:
FLIGHT
(
Paris, Santiago, AirFrance,
2320)
GEO
(
Santiago, Chile, 5.3M
).
Then the following is a solution:
ROUTES
(
⊥
1
,
Pa ris
,
Santiago
)
INFO FLIGHT
(
⊥
2
,
AirFrance
)
SERVES
(
AirFrance
,
Santiago
,
Chile
,
⊥
1
, 2320,
⊥
3
).
M
The setofallsolutionsfor
S
under
is denoted by S
OL
M
(
S
):
S
OL
M
(
S
)=
T
Σ
t
.
∈
I
NST
(R
t
)
|
(
S
,
T
)
|
=
Σ
st
and
T
|
=
It shall often be convenient for us to view the semantics of a schema mapping as a binary
relation, containing source-target pairs that it relates. In fact, when we deal with metadata
management, we often viewmappings this way. Thus, the semantics of a mapping
M
is a
binary relation
M
between I
NST
(R
s
) and I
NST
(R
t
) (i.e.,
M
⊆
I
NST
(R
s
)
×
I
NST
(R
t
))
defined as
⎧
⎨
⎫
⎬
S
∈
I
NST
(R
s
)
,
M
=
(
S
,
T
)
T
∈
I
NST
(R
t
)
,
.
⎩
⎭
T
∈
S
OL
M
(
S
)
3.3 Query answering and rewriting
Given a query
Q
over the target schema R
t
, the notion of query answering is that of certain
answers, which are true in all solutions for a given source instance
S
, and thus do not
depend on a particular target instance that is materialized.
Thus, certain answers are defined as
certain
M
(
Q
,
S
)=
T
∈
Q
(
T
)
.
S
OL
(
S
)
M
However, one needs to compute
certain
M
(
Q
,
S
) using just one specific materialized so-
lution, say
T
0
. In general, there is no reason why simply running
Q
on
T
0
and computing
Q
(
T
0
) would yield
certain
M
(
Q
,
S
). Thus, instead one tries to find another query
Q
that,
