Database Reference
In-Depth Information
The translation
S
r
= I
NL
D
OC
(
S
,
D
s
) was shown in
Figures 15.3
(a) and
15.3
(b). To
translate the mapping, we need to translate
π
(
x
,
y
,
z
) into conjunctive queries
π
(
x
,
y
) and
Q
π
(
x
,
y
) and
Q
π
(
x
,
y
,
z
); the st-tgd in I
NL
M
AP
(
M
) will then be
Q
π
(
x
,
y
)
→∃
zQ
π
(
x
,
y
,
z
)
.
π
is
We have already shown the translation
Q
π
(
x
,
y
) in
Example 15.7
. The translation of
id
w
R
r
(
id
r
)
.
∧
R
writer
(
id
wr
,
id
r
,
y
)
∧
Q
π
(
x
,
y
,
z
)=
∃
id
r
∃
id
wr
∃
∧
R
work
(
id
w
,
id
wr
,
x
,
z
)
When we apply the resulting mapping I
NL
M
AP
(
) to the instance
S
r
in
Figures 15.3
(a)
and
15.3
(b), we get the following canonical universal solution:
M
rId
writerID
rID
@name
⊥
2
⊥
6
⊥
10
⊥
1
⊥
2
Kleinberg
R
r
:
R
writer
:
⊥
5
⊥
6
Tardos
⊥
9
⊥
10
Hungerford
workID
writerID
@title
@year
⊥
3
⊥
1
'Algorithm Design'
⊥
4
R
work
:
⊥
7
⊥
5
⊥
8
'Algorithm Design'
⊥
11
⊥
9
'Algebra'
⊥
12
Note that this canonical universal solution does not coincide with an inlining of the
canonical XML solution but rather contains one, as explained in
Proposition 15.10
: notice,
for instance, that there are three distinct nulls in the relation for the root.
Now, if we have an XML query, for instance
Q
(
x
)=
r
/
writer
(
x
) asking for writer
names, it will be translated into a relational query
Q
r
= I
NL
Q
UERY
(
Q
,
D
t
) as follows:
R
writer
(
id
wr
,
id
r
,
x
)
.
When this query is evaluated over the canonical universal solution above (as
Algorithm
15.6
prescribes), it produces the set
id
wr
R
r
(
id
r
)
Q
r
(
x
)=
∃
id
r
∃
∧
{
Kleinberg
,
Tardos
,
Hungerford
}
, which of course is
certain
M
(
Q
,
S
).
