Database Reference
In-Depth Information
generally, TA is not isomorphic in DB to A
Σ D (T A) (in fact, T(TA)
=
TA
TA
TA
=
TA
TTA
=
A
Σ D (T A)
T(A
Σ D (T A)) , i.e., T(TA)
=
T(A
Σ D (T A)) ).
Cartesian product
×:
Set
Set is translated into the matching operation (tensor
product)
DB .
This translation is based on observations that any n -ary algebraic operator o i
Σ R is represented as a function (an arrow in Set )
⊗:
DB
−→
o i : T P X n
T P X which using
the n -fold Cartesian product
×···× T P X as domain, while such an operator
in the DB category is represented by an atomic view-mapping f o i ={
T P X
q }∪{
q j :
α(r i 1 )
for each subset of relational sym-
bols r i 1 ,...,r in A α A }: TA TA , where α # : T P X TA is the unique
extension (of the initial algebra semantics in the diagram in the previous section)
of an R-algebra (i.e., an assignment) α
×···×
α(r in )
α # (o i (r i 1 ,...,r in ))
|
:
X
A
TA for the relational symbols
in X .
Thus, this algebraic operator o i is translated into an arrow from TA into TA .
n
In fact, if we replace
×
by
in the n -fold
T P X
×···× T P X , we obtain (for a
n
n
⊗···⊗
=
∩···∩
=
=
simple database A )
TA
TA
T(TA)
T(TA)
T(TA)
TA .
Any disjoint union endofunctor X
_
:
Set
Set (used for a construction of the
syntax endofunctor X
Σ R :
Set
Set ) is translated into the DB endofunctor
A Σ D :
DB
−→
DB (the coproduct
in Set is translated by the coproduct
in DB ).
We have only finitary operators o i
Σ R and hence to model them we can
use only the finite databases, i.e., finite simple objects A
Ob DB . Thus, the
o i : A ar(o i )
set of all functions obtained from the signature Σ R ,
TA , can
be internalized in DB by a homomorphism o i Σ R :
Σ D (A)
TA , where,
= o i Σ R X ar(o i ) , here in DB we have
analogously as in Set where Σ R (X)
= o i Σ R A ar(o i ) .
Consequently, for a finite A we obtain
Σ D (A)
ar (o i )
A
A =
Σ D (A)
=
⊗···⊗
TA
=
TA,
o i
Σ R
o i
Σ R
and hence the endofunctor Σ D is a restriction of the power-view endofunctor
T
DB to finite objects in DB .
It is well known [ 14 ] that for any monoidal category C with a monoidal product
:
DB
,
P OP
any two functors F 1 :
C and F 2 :
P
C have a tensor functorial product
F 1 P F 2 = p P (F 1 p)
(F 2 p) .
In what follows, we will consider the simple databases only: these results then
are extended to all complex databases, by considering that they are all finite
(co)products (equivalent, up to isomorphism, to a disjoint union) of the simple
databases.
Search WWH ::




Custom Search