Research has made big steps in the field of database theory, leading to large technological improvements of the existing systems. These systems ensure efficiency and suitable computational power to handle the so-called “big data”. In particular, relational DBMSs provide high performance on read/write operations, and support a query language sufficiently expressive such as SQL. Furthermore, there exist logic-based systems - the so-called Deductive Database Systems (DDS) - that provide support to recursive queries, enabling complex reasoning capabilities and allowing querying of databases, using Datalog-like rules. Among the available DDS systems, one of the most known implementation is DLV. In particular, there is an extension of DLV - called DLVDB - that merges the potential of a DDS with the well-known optimization techniques implemented in the current DBMSs, improving rea- soning efficiency on big data sets. DLVDB can evaluate a Datalog program having EDBs stored on a relational DBMS. Unfortunately, it can not handle database types in a satisfying way, and it does not support all DBMSs. Moreover, it is difficult to integrate DLVDB into a real system, due to the lack of an API that simplifies the interoperability with third-party software. In this thesis, we present two new software systems that address these problems. The first one is called RelGrounder, which is an evaluator of stratified Datalog programs with EDBs stored on relational databases. Unlike DLVDB , it is database-independent, and can manage the data types. The second one is called RESTdlv, a RESTful web service for remote calls to RelGrounder, DLVDB and DLV.

1. Dipartimento di Matematica e Informatica
A System for Stratified Datalog Programs
SUPERVISOR
Prof. Francesco Ricca
STUDENT
Simone Spaccarotella
138937

2. Recursion
DLV
Semi
Naive
Method
Fixpoint
DLVDB
Datalog
Tp
Operator
Stratification

3. 
DLVDB
 is DBMS-dependent
 does not handle the Data-Types properly

It lacks a Standard API for DLV/DLVDB

4. Recursion
DLV
DLVDB
Semi
Naive
Method
Rel
Grounder
Datalog
Fixpoint
Tp
Operator
Stratification
DLV
Wrapper
RESTdlv
REST
Remote
API
Front-End
Web
Service

5. 
RelGrounder
 A new Evaluator of Stratified Datalog Programs
 Developed in Java
 DBMS-independent
 Handles Data Types

RESTdlv
 An umbrella project that cover RESTdlv WS and RESTdlv Client
 … and SudokuLogic (a PoC for RESTdlv WS)
 DLVWrapper uses RESTdlv Client

6. RelGrounder
Datalog
INPUT
Translation
Layer
Persistence
Layer
Mapping
EXECUTE
DB
SQL
Statement
SQL
Statement
Datalog
Rule

8. 
Non-ground Atom  Relation Schema
 Predicate Symbol  Relation Name
 Terms  Attributes

Ground Atoms  Tuples

XML Mapping File

11. evaluator = EvaluatorInjector.getInstance(properties);
evaluator.init(program);
evaluator.storeProgramFacts(programScanner.getFacts());
evaluator.storeMappingFacts();
// for each strongly-connected component
while (programScanner.hasNextComponent()) {
StronglyConnectedComponent component = null;
component = programScanner.nextComponent();
evaluator.evaluateExitRules(component);
evaluator.evaluateRecursiveRules(component);
}
evaluator.commit();

12. p(X,Y,Z) :- ....., ....., .....
INSERT INTO p(X,Y,Z) (
SELECT ....
FROM ....
[WHERE] ....
)

13. p(X,Y) :- t(X,Y), s(Y), Y >= 10, Y <= 50
SELECT t.X, t.Y
FROM t, s
WHERE t.Y = s.Y AND t.Y >= 10 AND t.Y <= 50

14. rich(X) :- worker(X,_), salary(X,Y), Y >= 2500.
subsidy(X,500) :- worker(X,_), not rich(X).
INSERT INTO subsidy(X) (
SELECT worker.X, 500
FROM worker
WHERE worker.X NOT IN (
SELECT rich.X
FROM rich
)
)

15. INSERT INTO q(X,Y) (
SELECT t.X, t.Y
FROM table as t
)
X
Y
7
2
13
9
2
0
9
11
5
10
11
3
2
0
11
3

16. INSERT INTO q(X,Y) (
SELECT t.X, t.Y
FROM table as t
EXCEPT ( SELECT * FROM q )
)
X
Y
7
2
13
9
0
9
11
5
10
3
2
0
11
3

17. INSERT INTO q(X,Y) (
SELECT t.X, t.Y
FROM table as t
WHERE NOT EXISTS (
SELECT X, Y FROM q
WHERE t.X = q.X AND t.Y = q.Y
)
)

18. INSERT INTO q(X,Y) (
SELECT t.X, t.Y
FROM table as t
WHERE NOT EXISTS (
SELECT X, Y FROM q
WHERE t.X = q.X AND t.Y = q.Y
)
)
table
X
Y
7
2
13
9
2
9
11
5
10
11

19. INSERT INTO q(X,Y) (
SELECT DISTINCT t.X, t.Y
FROM table as t
WHERE NOT EXISTS (
SELECT X, Y FROM q
WHERE t.X = q.X AND t.Y = q.Y
)
)
table
X
Y
7
2
13
9
2
9
11
5
10
11

20. INSERT INTO q(X,Y) (
SELECT DISTINCT ......
FROM ......
UNION
SELECT DISTINCT ......
FROM ......
)
Source 1
Source 2
X
Y
X
Y
7
2
13
9
9
11
5
10
10
2
4
7
14
11
3
9

21. Tk-1
T0
T1
Tk-2
∆T0
n=0
∆T1
n=1
...
∆Tk-1
n = k-1

Fixpoint

22. t(X,Y) :- t(X,Z), t(Z,Y), …, t(Y,X), q(X,Y).
∆Tk = ∆Tk-1 Tk-1 … Tk-1
Tk-2 ∆Tk-1 … Tk-1
……………
Tk-2 Tk-2 … ∆Tk-1
Qu
Qu
Q

23. t(X,Y) :- t(X,Z), t(Z,Y), …, t(Y,X), q(X,Y).
∆Tk-1
Tk-1
…
Tk-1
Qu…
∆Tk-1 Tk-2 … Tk-2 Q u
∆Tk-1 ∆Tk-1 … Tk-2 Q u
∆Tk-1 Tk-2 … ∆Tk-1 Q u
∆Tk-1 ∆Tk-1 … ∆Tk-1 Q u

24. 
Auxiliary-table Approach

ID-based Approach

25. Auxiliary-table
Approach
DLVDB
Strategy
DLVDB-like
Strategy

26. P
P
P_d1
P_d
∆ at k-1
∆ at k

27. P
P
P_new
P_old
∆ at k
Tk-1
P_temp

28. P
P

29. P
50
30
10

35. 
a RESTful web service for DLV (aka web API)

a remote workspace for Datalog Programs

a simple way to use DLV/DLVDB functionalities

developed in Java

37. <HTTP VERB, URI>
RESTdlv
WS
JSON
XML
Network
Bean
RESTdlv
Bean
Client

38. URI
RESTdlv WS
Resource
Servlet
HTTP
Verb
•POST
•GET
•PUT
•DELETE
Method
•C
•R
•U
•D

39. RESTdlv
Bean
WS
Bean
Network
RESTdlv
Client

40. RESTdlv
WS
Datalog Programs

41. DLV Wrapper
use
RESTdlv
Client
use
RESTdlv
Client
(Java)
Third-Party SW
RESTdlv WS
N
E
T
W
O
R
K
(Java)
Third-Party SW
(any)
Web App
(HTML + Javascript)
Mobile App
(Android, iOS, ecc)

42. 
A DEMO of the Sudoku game

A proof of concept for RESTdlv WS

45. Bean
WS
Network
Bean
Client

47. 
RelGrounder

Three implementations of the Semi-Naïve method

RESTdlv WS + SudokuLogic

RESTdlv Client (integrated in DLVWrapper)

48. Thank you all.