Database Reference
In-Depth Information
it is easy to verify in DB that Tg = g η OP
B
g
η OP
B
, from Tg g
φ g
f =
η O B
η O C
η O B
=
g
f
=
φ(g)
φ(f)
φθ 1 θ(g)
φθ 1 θ(f)
=
K 1 θθ 1 (g T )
K 1 θθ 1 (f T ) =
K 1 (g T )
K 1 (f T ).
=
Thus, each arrow f T :
A
−→
B in DB T is “internalized” in DB by its represen-
K 1 (f T )
φθ 1 (f T )
η O B
θ 1 (f T )
B , where θ 1 (f T )
tation f
=
=
:
A
−→
:
A
−→
TB is a program equivalent to the database mapping f
:
A
−→
B , i.e.,
θ 1 (f T )
f .
The functor K is faithful. In fact, for any two arrows f T ,h T :
A
−→
B
in DB T , K 1 (f T )
K 1 (h T ) implies f T =
h T . Indeed, from K 1 (f T )
K 1 (h T )
=
=
we obtain φθ 1 (f T )
φθ 1 (h T ) , and if we apply a bijection φθ 1 , we obtain
=
φθ 1 φθ 1 (f T )
φθ 1 φθ 1 (h T ) , i.e., θθ 1 (f T )
θθ 1 (h T ) , or equivalently,
=
=
h T (the compositions θθ 1 and φφ 1 are the identity functions).
Let us show that K is an isomorphism: from the adjunction (F T ,G T T T )
f T =
:
DB T , where F T
is identity, F T
θφ 1 , we obtain that F T
DB
−→
K
=
I DB T and
K
F T =
I DB . Thus, the functor K is an isomorphism of DB and Kleisli category
DB T .
Notice that from Theorem 12 in Sect. 8.1.1 , DB T (Υ,A)
TA Υ
A Υ
=
=
TA
and DB T (Υ,f T )
1 (f T )
1 (f T ) .
id Υ
=
Thus, K 0 (A)
DB T (Υ,A) , and for f T :
B , K 1 (f T )
=
A
TA
=
A
=
is 1
φθ 1 (f T ) = η OP
θ 1 (f T ) : A B , and from the fact that Tg
cod(g) g =
η OP
cod(g) g for any arrow g in DB , we obtain K 1 (f T )
DB T (Υ,f T ) . Consequently,
we obtain the natural isomorphism ϕ :
DB T (Υ, _ ) K .
:
I DB −→
Remark It is easy to verify that a natural isomorphism η
T of the monad
:
−→
(T,η,μ) is equal to the natural transformation η
K
G T (by considering that
DB is defined by G T =
T 0
G T :
DB T −→
and for any f T :
−→
A
B in DB T ,
G T (f T )
1 (f T )
TB ).
Hence, the functor F T has two different adjunctions: the universal adjunction
(F T ,G T T T ) which defines the same monad (T,η,μ) , and this particular (for
the DB category only) isomorphism's adjunction (F T ,K,η I I ) which defines the
trivial identity monad.
μ B
:
TA
−→
We are now ready to define the semantics of queries in DB category and the
categorial definition of query equivalence . This is important in the context of the
Database integration/exchange and for the theory of query-rewriting [ 8 ].
When we define a mapping (i.e., an arrow or morphism) f
B between
two simple databases A and B , we define implicitly the information flux f , i.e.,
the set of views of A transferred by this mapping into B . Thus, in the context of
the query-rewriting, we consider only the queries (i.e., view-maps) whose resulting
:
A
−→
Search WWH ::




Custom Search