Upload
others
View
3
Download
2
Embed Size (px)
Citation preview
Copyright Big Data Pr Serge Miranda, MBDS, Univ de Nice Sophia Antipolis (UCA)1
www.mbds-fr.org
MBDS course :
« From data bases to big data »
(7 lectures)
Professor Serge Miranda
Dept of Computer Science
University of Nice Sophia Antipolis (member of UCA)
Director of MBDS Master degree
(www.mbds-fr.org)
www.mbds-fr.org
Date’s manifesto on object-relational data model
(third manifesto on future data bases)
(Lecture 4)
Professor Serge Miranda
Departement of Computer Science
Université of Nice Sophia Antipolis (UCA)
Director of MBDS Master degree (www.mbds-fr.org)
Copyright Big Data Pr Serge Miranda, MBDS, Univ de Nice Sophia Antipolis (UCA)4
CONTENTS
1. Object paradigm and data bases
2. Object-oriented data model ?
➢Objects ?
➢Consensus on object properties
➢« RICE » properties of object systems
3. Object-relational data models
➢Two basic approachs
➢TABLE (ans SQL) centrics : Second Manifesto by M.Stonebraker
➢DOMAIN centrics : Third Manifesto by Chris date
➢ Examples : 2D DATA store and THESAURUS
4. TIPS Researchs
Short Seminar 1 : The second DB manifesto by M.Stonebraker
Short Seminar 2 : Object middleware and Universal access (Microsoft)
Copyright Big Data Pr Serge Miranda, MBDS, Univ de Nice Sophia Antipolis (UCA)5
DATA MODEL ?
➢Three families of DATA models :
➢Computing(IMS, DBTG…)
➢Mathematics (Relational/SQL, Category/ NO SQL, Linear Algebra. ML & DL)
➢Semantics(OBJECT, semantic nets)
DATA MODEL
DATA STRUCTURES
Structure-based operators
(algebra)
Integrity rules
Copyright Big Data Pr Serge Miranda, MBDS, Univ de Nice Sophia Antipolis (UCA)6
Codd’s data model and (relational data model)
➢« CODD’s relational data model »
➔ Prerequisite to SQL2
➢« DATE’s object-relational data
model ( 3rd manifesto) »
➔ Prerequisite to SQL3
(and ODMG)
➢DATA STRUCTURES of CODD’srelational model
➢« VALUES » (value paradigm)
➢« DOMAINS » (semantical data type)
➢ « RELATIONS »
➢Attributes/PK/FK
➢Dual formal definition (SET or PREDICATE)
➢Example of CODD’s relational data model on PILOT relation ?
Copyright Big Data Pr Serge Miranda, MBDS, Univ de Nice Sophia Antipolis (UCA)7
DATA BASE market & standards
(Stonebraker 96 & Gartner)
DATA
PROCESSING
SQL
No SQL
Simple Complex (graphs,..)
R-DBMS
SQL3Mobiquitous &
Big Data systems
ODMG CAD
OR-DBMS
OO-DBMSFile System
(Office automation)
(1) (2)
(3)
SQL2Production Decision
2010 2020
(1) 10 G dollars
20% of cont. Growing rate
(2) 2x (3) 2x (1)!
(3) 1/100 x (1)
1/100 x (1)
Copyright Big Data Pr Serge Miranda, MBDS, Univ de Nice Sophia Antipolis (UCA)8
TIPS properties for STRUCTURED DATA stores
T
I
P
S
Transactions (with ACID properties)
No-procedural Interface (SQL)
Persistency (virtual paged memory)
Structuration (Schema)
Copyright Big Data Pr Serge Miranda, MBDS, Univ de Nice Sophia Antipolis (UCA)9
TOP DOWN approach of structureddata stores
➢2 Phases for DB management 1. SCHEMA DEFINITION (UML, Codd&Date, etc)
2. DB Creation with a DBMS Engine and SQL standard
➢TRANSACTION-oriented applications (OLTP : On-Line Transaction Processing)➢TIPS properties
➢ACID properties
➢Decision–support applications (OLCP : On-Line Complex Processing) with SQL extensions➢DATAWAREHOUSE
➢DATA MINING
Copyright Big Data Pr Serge Miranda, MBDS, Univ de Nice Sophia Antipolis (UCA)10
Structured-data store evolution
➢Huge data bases (cf BIG DATA) :
➢Petabytes DB (10**15 )
➢Exabytes(10**18),
➢Then Zetta bytes (10**21), Yotta bytes (10**24) and …Google(10**99)
➢SERVER scalability
➢SCALE UP (SMP, Cluster, MPP)
➢SCALE OUT
Copyright Big Data Pr Serge Miranda, MBDS, Univ de Nice Sophia Antipolis (UCA)11
DATA variety to be integrated
➢Stronger-structured data (schema)
➢Embedded objects
➢USER-DEFINED DATA TYPES/FUNCTIONS (extensible ADT)
➢Bottom up compatibility
➢Semi-structured and unstructured DATA
➢BLOBS – Binary Large Objects – (video,..) and CLOBS (Character LOB)
➢WEB data
➢Open DATA
➢Semantical links (hyper links)
➢New interfaces : multimedia-content based, neural nets (voice interface)
Copyright Big Data Pr Serge Miranda, MBDS, Univ de Nice Sophia Antipolis (UCA)12
Hybrid Object–Relational (OR) data model ?
➢OR DM extending… the relational data model
➢Formal unique relational basis
➢Multiple object data models (some consensus on properties)
➔ Enrich relational data model properties with specific
consensus object properties
« OBJECT »
Copyright Big Data Pr Serge Miranda, MBDS, Univ de Nice Sophia Antipolis (UCA)14
Object concepts
➢OBJECTS ?
➢ Object definitions
➢ Object integration : Hybrid Object-relational data model
Copyright Big Data Pr Serge Miranda, MBDS, Univ de Nice Sophia Antipolis (UCA)15
Weakness of relational data models
➢Operators separated from data
➢Stored procedures
➢Lack of hidden attributes
➢Atomic domains
➢Limitations to normalized relations (3NF)
➢Limitations to hierarchys and graphs representation (transitive closure)
➢Complex objects
➢Transaction focus with decision-support inadequacies
Copyright Big Data Pr Serge Miranda, MBDS, Univ de Nice Sophia Antipolis (UCA)16
Evolution of Computer Science and OBJECTS
PLOS
PL
AI
File systems DBMS
1965
1970
OS : Operating « SERVICES » ?
Copyright Big Data Pr Serge Miranda, MBDS, Univ de Nice Sophia Antipolis (UCA)17
Evolution of computing : OS and DBMS
1970
R-DBMSOS FS
OS services DBMS services
Physical resource managementSwapping algorithms (pagedmemory)
SECURITY Privacy protectionConcurrency Control(semaphors, Locking, deadlock,... )
MODELS(" process ") ...no programming language
(TIPS properties) :
(P) Persistance (Page) managementLRU ++
(T) SECURITY & Transaction support(ACID) DATA INTEGRITY
(S) DATA Structuration/Schema thru data models
(I) User Interface (SQL)
Copyright Big Data Pr Serge Miranda, MBDS, Univ de Nice Sophia Antipolis (UCA)18
Evolution of computing and OBJECTS
PL
AI OS
DBMS
1980-1990
OBJECTS
OS
DBMS
AI
PL
1990
Copyright Big Data Pr Serge Miranda, MBDS, Univ de Nice Sophia Antipolis (UCA)19
Computing future : Towards mobiquitous DATA spaces
? ? ?
2030
" It is hard to predict ..especially the future" N.Boehr
« SERVICES » ?
Copyright Big Data Pr Serge Miranda, MBDS, Univ de Nice Sophia Antipolis (UCA)20
Data computing and objects
DATA PROCESSING
Objects
Communication
SQL(Structured data management)
PL(Structured programming)
Copyright Big Data Pr Serge Miranda, MBDS, Univ de Nice Sophia Antipolis (UCA)21
Object approach and application development
INTERFACE
APPLICATION
DBMS
DB
Object integration
PB : IMPEDANCE MISMATCH
Copyright Big Data Pr Serge Miranda, MBDS, Univ de Nice Sophia Antipolis (UCA)22
Waves of distributed computing
CLOUD
CLIENT SERVER
File systems
•Distributed objects
•Web Services
•SQL server
•Groupware
•TP Monitor
Copyright Big Data Pr Serge Miranda, MBDS, Univ de Nice Sophia Antipolis (UCA)23
Software evolution towardsmiddleware for distributed computing
Middleware
DBMS
OS
• Tight coupling (Corba, RPC)
• Loose coupling(REST, web services)
• TIPS
• IP
Copyright Big Data Pr Serge Miranda, MBDS, Univ de Nice Sophia Antipolis (UCA)24
OBJECT duality
Encapsulated
OBJECT (PL)
-Specificoperators(Methods)
-Inheritance
BLACK BOX
Structured
OBJECT (DB)
-genericoperators(algebra)
-Persistence
WHITE BOX
Copyright Big Data Pr Serge Miranda, MBDS, Univ de Nice Sophia Antipolis (UCA)25
Object attractiveness
➢NATURALNESS :
➢Men deal with objects and computers with DATA
➢Normalized relations are (simple but) not natural
➢COMPLEXITY :
➢Att inventes C++ to deal with concentrators (200 000 lines of programming)
➢REUSE and flexibility :
➢Kit programming
➢ENCAPSULATION :
➢DATA along with associated processing
➢A fuzzy unifying concept !
Copyright Big Data Pr Serge Miranda, MBDS, Univ de Nice Sophia Antipolis (UCA)26
Unbearable Lightness of Beingan object !
OBJECT ?
« An identifiable thing which plays a role in regards of an operation..etc… » (X3-SPARC, SEPT.91, OODBTG, pp 3.6)
Copyright Big Data Pr Serge Miranda, MBDS, Univ de Nice Sophia Antipolis (UCA)27
Object ?
➢ (MIRA96) and (MIRA2002)
Un objet est une capsule logicielle (E) oblative identifiée (I)
avec un tropisme connatif incrémental (C) dont l’hétéronomie est la marque de la durée de l’éphémère
et la hoirie (R), la marque de la richesse.
« An OBJECT is an identified (I) oblative software capsule ( E) with an incremental (C) natural tropism whose heteronomy is its
ephemere–duration mark and hoirie (R), its wealth mark »
➢ (MIRA96) and (MIRA2002)
« An OBJECT is an identified (I) oblative software capsule ( E) with an incremental (C) natural tropism whose heteronomy is its
ephemere–duration mark and hoirie (R), its wealth mark »
Copyright Big Data Pr Serge Miranda, MBDS, Univ de Nice Sophia Antipolis (UCA)28
Object contributions to data management : RICE properties (Mira96)
R
I
C
E
Reusability (Inheritance or polymorphism)
Identification (OID : Object Identifier)
Complex Object construct
Encapsulation (Methods)
Copyright Big Data Pr Serge Miranda, MBDS, Univ de Nice Sophia Antipolis (UCA)29
OBJECTS and Object CLASS
➢OBJECT ?
➢OBJECT = (OID, VALUE) *➢OID : Object Identifier
➢➔ HASHING, b-TREE, …
➢An Object CLASS
➢IS (potential values)
➢HAS (real values)
a data type verifying the RICE properties
*cf (KEY, VALUE) in N.O.SQL data stores
SETconstruct
OBJECT
CLASS
Copyright Big Data Pr Serge Miranda, MBDS, Univ de Nice Sophia Antipolis (UCA)30
OR data model and properties
OR data model
(TIPS + RICE)
Object
Model
(RICE)
Relationaldata model
(TIPS)
Copyright Big Data Pr Serge Miranda, MBDS, Univ de Nice Sophia Antipolis (UCA)31
Object data model
➢NEW object data model : OO (Object-oriented)
➢First manifesto by Francois Bancilhon
➢ODMG standard
➢Enrichment of relational data model : OR (Object relational)
➢Second Manifesto by Mike Stonebraker
➢Third manifesto by Chris Date
Copyright Big Data Pr Serge Miranda, MBDS, Univ de Nice Sophia Antipolis (UCA)32
Three approaches for OBJECT Data models
•RICE properties• 1st Manifesto (F. Bancilhon)
• Based upon Object programming
VALUES
•RICE properties• 3rd Manifesto (C.Date)
• Based upon Codd’s relationaldata model
DOMAINS
•RICE properties• 2nd Manifesto
(M. Stonebraker)
• Based upon SQL
RELATIONS
SQL3
ODMG
Copyright Big Data Pr Serge Miranda, MBDS, Univ de Nice Sophia Antipolis (UCA)33
SQL3 and OR data model
➢SQL3 encompasses two manifestos
➔ 2 possibilities to create an object class
➢CREATE TYPE (cf « Date »)➢CREATE TABLE (cf « Stonebraker »)
Copyright Big Data Pr Serge Miranda, MBDS, Univ de Nice Sophia Antipolis (UCA)34
RICE properties in the relational DM
Relational data modelR
I
C
E
Copyright Big Data Pr Serge Miranda, MBDS, Univ de Nice Sophia Antipolis (UCA)35
Example : Object structural graph (cf IFO)
EMPLOYEE
HOSTESSPILOT
FLIGHT PLANE
Notation : « Inheritance»« SET OF »« TUPLE »« Class »
NUMBERH DC FLIGHT
STREET CITY NUMBER
E# ENAME ADDRESS
F# DC AC DT AT PLANE
PNAME CAP
Copyright Big Data Pr Serge Miranda, MBDS, Univ de Nice Sophia Antipolis (UCA)36
Exercice
➢Take the PILOT class and represent it in Codd’s relational data model
➢Take this example again at the end of the course and represent it in Date’s manifesto
Copyright Big Data Pr Serge Miranda, MBDS, Univ de Nice Sophia Antipolis (UCA)37
Back to the future withDate’s 3rd manifesto
C.Date, H. Darwen « Date base programming and design » January 1995 pp.25-34 http://thethirdmanifesto.com/
« Object features are orthogonal to the Relational DM and thereforeRDM needs no extension, no correction, no subsumption, no perversion in order for them to be accommodated »
Chris DATE
Copyright Big Data Pr Serge Miranda, MBDS, Univ de Nice Sophia Antipolis (UCA)38
OR data model : first possibility
➢First approach : RELATION with RICE properties
➢Example : UNISQL, POSTGRES,…EX : CREATE OBJECT CLASS PILOT PUBLIC PIL# NUMERIC, PILname CHAR, ADDR CHAR) ;
➢Advantage : unique concept for TABLE and CLASS
➢Problems :
➢PB1 : objects = tuples ➔ encapsulation and relational algebra ?
➢PB2 : Generic and specific operators within query expressions
➢PB3: Closure
Copyright Big Data Pr Serge Miranda, MBDS, Univ de Nice Sophia Antipolis (UCA)39
OR data model : second possibility
➢Second approach : DOMAINS with RICE properties
➢Advantages :
➢Object duality➔ Dual Object structures
➢Domains are naturally extendable to classes
➢Problem : structural inheritance
Copyright Big Data Pr Serge Miranda, MBDS, Univ de Nice Sophia Antipolis (UCA)40
CLASS and DOMAIN consistency
➢OBJECT CLASS = « DATA TYPE with RICE properties »
➢DOMAIN of Codd’s relational data model :
« Semantical data type »
➢DOMAIN in Date’s manifesto and OR data model :
« Semantical data type with RICE properties »
Copyright Big Data Pr Serge Miranda, MBDS, Univ de Nice Sophia Antipolis (UCA)41
Double inheritance in DATA BASES
➢DB INHERITANCE is twofold :
➢SI : STRUCTURAL inheritance (attributes)
➢OI : OPERATIONAL inheritance (methods)
➢In the 2nd manifesto : SI + OI on TABLES
➢In the 3rd manifesto :
➢OI at DOMAIN level
➢SI (at table level) thru a dedicated inheritance operator noted « ➔ » involving primary domains
Copyright Big Data Pr Serge Miranda, MBDS, Univ de Nice Sophia Antipolis (UCA)42
DATE’s manifesto : Inheritance at primary-domain level
➢Declarative approach for primary domains
➢Example :
Create Domain PILNO UNDER EMPNO
Create Domain HOSTESSNO UNDER EMPNO 100 103
100 103
EMPNO
PILNO
HostessNO
➢Inheritance graph among primary domains
Copyright Big Data Pr Serge Miranda, MBDS, Univ de Nice Sophia Antipolis (UCA)43
Structural inheritance
➢2 possibilities :
➢Declarative approach : DATA STRUCTURE ➢Example : UNDER Clause among TABLES (2nd manifesto)
➢Manipulation approach : OPERATOR➢Example : ‘➔’ operator on primary keys (3rd manifesto)
PK➔ {ATTi} <ATTi being any attribute in the inheritance graph>
➢DOMAIN index and ‘➔’ operator
➢Example : value 100 for EMPNO and access to any tuple having the ‘100’ value for PK or FK
Copyright Big Data Pr Serge Miranda, MBDS, Univ de Nice Sophia Antipolis (UCA)44
Structural inhéritance in DATE’smanifesto «PK➔ {ATTi} » ?
➢TABLEs➢EMPLOYEE (E# : EMPNO, Ename, Salary, ADDR)➢PILOT (PIL# : PILNO, NBHF)➢HOSTESS (H# : HOSTESSNO, RANK)
Q : SELECT PIL#➔ Ename, Salary,NBHF
FROM PILOT
WHERE (PIL#➔) NBHF >
(SELECT MAX NBHF FROM PILOT
WHERE PIL#➔ ADDR = ‘Nice’);
EMPNO
PILNO
HostessNO
100 103
100 103
Copyright Big Data Pr Serge Miranda, MBDS, Univ de Nice Sophia Antipolis (UCA)45
Exercice
➢Give another form of the query using E# ➔
Copyright Big Data Pr Serge Miranda, MBDS, Univ de Nice Sophia Antipolis (UCA)46
Solution with E#
SELECT E# ➔ Ename, Salary,NBHF
FROM EMPLOYEE
WHERE E# ➔ NBHF >
( SELECT MAX NBHF FROM PILOT
WHERE E#➔ ADDR = ‘Nice ’);
Copyright Big Data Pr Serge Miranda, MBDS, Univ de Nice Sophia Antipolis (UCA)47
D language
(« D » for « Domain » ?... « C++ » ?... « DATE » ?)
➢Object class attached to relational DOMAINS (RICE)
➢« Encapsulated semantical data types whose values could be complex withassociated functions »
➢and (R)ICE properties at the domain level
➢UNDER Clause among primary domains :
« SUB-DOMAIN » and « SUPER-DOMAIN »
Example : Create Domain Hostessno UNDER Empno
➢C.Date, « user-defined data types and functions including inheritance are orthogonal to the relational data model »
Copyright Big Data Pr Serge Miranda, MBDS, Univ de Nice Sophia Antipolis (UCA)48
D language (SQL-D)
➢Examples illustrating ( R ) ICE at domain level(3rd manifesto) with the D language :
➢2D
➢Thesaurus
➢Exercice : Inheritance with primary domains and ‘➔’ operator
Copyright Big Data Pr Serge Miranda, MBDS, Univ de Nice Sophia Antipolis (UCA)49
Summary of dual Date’s enrichment of Codd’smodel
➢PRIMARY DOMAIN inheritance➢EX : PILNO under EMPNO (➔ operational inheritance)
➢ INHERITANCE operator➢Ex : PIL#➔ EMPADDR, NBHF (➔ structural inheritance)
➢< PL#, PK definied on the Primary domain PILNO>
Copyright Big Data Pr Serge Miranda, MBDS, Univ de Nice Sophia Antipolis (UCA)50
Example 1 : 2D data (M. Stonebraker and C.Date)
CREATE TABLE RECTANGLES
(RECTID, x1, x2, y1,y2)
PRIMARY KEY (RECTID)
UNIQUE (x1, x2, y1, y2)
QUESTION (in SQL2) :What are the rectangles overlapping the square (0, 1, 0, 1) ?
(1,1)
1
1
X1 X2
Y2Y1
(0,0)
Copyright Big Data Pr Serge Miranda, MBDS, Univ de Nice Sophia Antipolis (UCA)51
SOLUTION : 2D query in SQL2
➢Q : What are the rectangles overlapping the square (0, 1, 0, 1) ?
<With 10 conditions>
SELECT * From RECTANGLES Where
(x1>=0 AND x1<=1 AND y1>=0 AND y2<=1)
OR (x2>=0 AND x2<=1...)...OR...
OR (x1<=0 AND x2<=1 AND y2<=0 and y2<=1)…
<with some trick >
SELECT * FROM rectangle
WHERE x1 BETWEEN 0 AND 1) INTERSECT (SELECT * FROM rectangle WHERE y1 BETWEEN 0 AND 1)
UNION
(SELECT * FROM rectangle
WHERE x2 BETWEEN 0 AND 1) INTERSECT (SELECT * FROM rectangle WHERE y2 BETWEEN 0 AND 1);
--
Copyright Big Data Pr Serge Miranda, MBDS, Univ de Nice Sophia Antipolis (UCA)52
2D Example (C.Date) : Domain creation
CREATE DOMAIN RECTANGLE ( ... RTREE ... )CREATE FUNCTION MAKE-RECT
(A FLOAT, B FLOAT, C FLOAT, D FLOAT)RETURNS ( RECTANGLE )AS BEGIN
DECLARE R RECTANGLER.x1 = A; R.x2 = B; R.y1 = C; R.y2 = D;RETURN (R)
END;CREATE FUNCTION OVERLAP( R1 RECTANGLE, R2 RECTANGLE )RETURNS ( BOOLEAN )AS BEGIN ... END;
Copyright Big Data Pr Serge Miranda, MBDS, Univ de Nice Sophia Antipolis (UCA)53
2D Example : relation creation
CREATE RELATION RECTANGLES
( RECTID : RECTNO,
AR : RECTANGLE, ...)
PRIMARY KEY ( RECTID )
UNIQUE ( R ) ;
Copyright Big Data Pr Serge Miranda, MBDS, Univ de Nice Sophia Antipolis (UCA)54
2D Example : Query in D language (Date)
Q :
SELECT *
FROM RECTANGLES R
WHERE
OVERLAP (R, MAKE_RECT(0,1,0,1))= TRUE;
Copyright Big Data Pr Serge Miranda, MBDS, Univ de Nice Sophia Antipolis (UCA)55
DB research on object-relational waves
Copyright Big Data Pr Serge Miranda, MBDS, Univ de Nice Sophia Antipolis (UCA)56
DB research (TIPS)
T
I
P
Long-duration and bursty transactions
Multimedia and neural voice interfaces
Big data and very-light DB ; data lakes ; virtual memory management ; main memory DB ; access methos
Copyright Big Data Pr Serge Miranda, MBDS, Univ de Nice Sophia Antipolis (UCA)57
DB research (TIPS)
S
Object-relational DBMS (bridges between ODMG and SQL3)
Semi-structured data (bridges SaprQL and SQL)
SQL extension to handle ML, DL and NO SQL
Query optimization
Copyright Big Data Pr Serge Miranda, MBDS, Univ de Nice Sophia Antipolis (UCA)58
Exercice : THESAURUS
➢THESAURUS : a DOMAIN i.e. a set of key words (values) with strongsemantical links among them
➢Note : a thesaurus is a semantical net
➢ THESAURUS role
➢Uniform indexing
➢Extension or reduction of the query outcome
Copyright Big Data Pr Serge Miranda, MBDS, Univ de Nice Sophia Antipolis (UCA)59
Thesaurus (ontology) : semantical links among KEY WORDS (values)
Computer Science
Software
/LOGICIEL
Operating systems
/OS
Data base systems
/DBMS
Hardware
/MATERIEL
Data base machines
/MPP
3 semantical links :- SYNONYMY (/)- CLOSENESS ( )- HIERARCHY
Copyright Big Data Pr Serge Miranda, MBDS, Univ de Nice Sophia Antipolis (UCA)60
Exercice
1.Give the following entity in a normalized relational format ?
DOCUMENT entity :
(DOC#, title, editor, PAGES, {authors}, {keywords})
2.Give a relational schema to represent a thesaurus itself
(with binary relations)
3.Give a SQL2 form for the following query :
What are the documents concerning software (using
thesaurus semantical links to extend the query) whose title
starts with « concepts » ?
Copyright Big Data Pr Serge Miranda, MBDS, Univ de Nice Sophia Antipolis (UCA)61
Exercice
4.Give Dates’s object-relational schema and query for this thesaurus
example (in D language)
5.Introduce a super class : « Document » [LOAN function] with 2 sub
classes : « BOOK » and « HANDOUT » [COPY Function]
➢Buid primary domain hierarchy
➢Use inheritance operator to process the following query
What is the number of pages of books published by Dunod which are on loan ?
Copyright Big Data Pr Serge Miranda, MBDS, Univ de Nice Sophia Antipolis (UCA)62
Solution : Thesaurus
➢Documentary DB in the relational schema :
➢3 normalized relations :
➢DOCUMENT (DOC#, title, editor, PAGES)
➢DOC-author (DOC#, author)
➢DOC-keyword (DOC#, keyword)
Copyright Big Data Pr Serge Miranda, MBDS, Univ de Nice Sophia Antipolis (UCA)63
Thesaurus (Cont’)
Thesaurus (type..)
1. With one relation :
➢THESAURUS (keyword1, keyword2, TYPE)
DB Data Base synonymy
Software Data Base hierarchy...
2. OR 3 binary relations :
➢SYNONYMY*(keyword1, keyword2)
Data Base DB...
➢HIERARCHY( keyword-GEN, keyword-SPEC)
Computing Software...
➢CLOSENESS*(keyword1, keyword2)
DB DB Machine
* and REFLEXIVITy !
Copyright Big Data Pr Serge Miranda, MBDS, Univ de Nice Sophia Antipolis (UCA)64
SQL2 for the query
➢Q - What are the documents concerning software (using thesaurus semanticallinks to extend the query) whose title starts with « concepts » ?
SELECT *FROM DOCUMENT, DOC-author, DOC-keyword
WHERE DOCUMENT.DOC#=DOC-author.DOC#
and DOCUMENT.DOC#=DOC-keyword.DOC#
and TITLE=“CONCEPT%”
and (DOC-keyword.keyword = “Software”
or DOC-keyword.keyword IN
(SELECT keyword2 FROM SYNONYMY
WHERE keyword1 = “Software”) <+ INVERSE>
or DOC-keyword.keyword IN
(SELECT keyword-SPEC FROM HIERARCHY
WHERE keyword-GEN = “Software”)
or DOC-keyword.keyword IN
(SELECT keyword2 FROM CLOSENESS
WHERE keyword1 = “Software” ); <+ INVERSE>)
Copyright Big Data Pr Serge Miranda, MBDS, Univ de Nice Sophia Antipolis (UCA)65
OR schema (Chris DATE)
With encapsulated domains
CREATE DOMAIN THESAURUS/KEYWORD ( ... )
CREATE FUNCTION SYNONYMY(keyword THESAURUS) RETURNS SET-OF (THESAURUS)
CREATE FUNCTION HIERARCHY(keyword THESAURUS)RETURNS SET-OF (THESAURUS)
CREATE FUNCTION CLOSENESS(keyword THESAURUS)RETURNS SET-OF (THESAURUS)
Copyright Big Data Pr Serge Miranda, MBDS, Univ de Nice Sophia Antipolis (UCA)66
Date’s model (cont’)
CREATE FUNCTION CONCERN(keyword THESAURUS)RETURNS SET-OF (THESAURUS)AS BEGIN < composition of synonymy/ hierarchy /closeness functions> END
Copyright Big Data Pr Serge Miranda, MBDS, Univ de Nice Sophia Antipolis (UCA)67
Date’s model (D language)
Let us create a unique relation (« as in SQL ») with multivaluedattributes with the SET-OF construct
CREATE RELATION DOCUMENT
(… Setof key word : THESAURUS)…
Q : SELECT *
FROM DOCUMENTS
WHERE title = “CONCEPT%”
and keyword IN {SOFTWARE, CONCERN (SOFTWARE)}
Copyright Big Data Pr Serge Miranda, MBDS, Univ de Nice Sophia Antipolis (UCA)68
Complete example with Date’s model
The 3 following Classes with :
(i) the super class DOCUMENTS (DNO, title, PAGES, {keywords})
(ii) two sub classes BOOKS ({authors}, editor) and HANDOUTS (author)
1. complete Object-relational schema (Date’s) ?
2. 2 following (SQL-D) queries using ‘➔’ :
➢Q1 What is the number of pages of DUNOD books on loan ?
➢Q2 What is the number of pages of DUNOD books on loan whose title starts with CONCEPT and which concerns SOFTWARE ?
Copyright Big Data Pr Serge Miranda, MBDS, Univ de Nice Sophia Antipolis (UCA)69
Example : Complete OR schema (Date’s)
➢PHASE 1 : DOMAIN definition (with inheritance among primary domains)
Create Domain DOCNO primaryFunction LOAN < boolean value>
Create domain BNO UNDER DocNO primary
Create Domain HandoutNO UNDER DocNO primaryFunction COPY
Create Domain Dtitle Character (12)
Create Domain Dauthor Character (12)
Create Domain Dpage INT
Create Domain Deditor Character (12)
Copyright Big Data Pr Serge Miranda, MBDS, Univ de Nice Sophia Antipolis (UCA)70
Date’s schema (cont’)
Create Domain THESAURUS <already done>
Function SYN…
Function HIER…
Function CLOSE…
Function CONCERN,…
➢ Phase 2 : Relation creation
Create RELATION Document (
DOC# : DOCNO, Primay Key
title : Dtitle
NBREPAGE : Dpage
key word SET-OF: THESAURUS)
Copyright Big Data Pr Serge Miranda, MBDS, Univ de Nice Sophia Antipolis (UCA)71
Thesaurus schema
Create Relation BOOK(B# : BNO Primary keyauthor SET-OF: Dauthoreditor : Deditor)
Create Relation HANDOUT(H# :HANDOUTNO Primary Keyauthor : Dauthor)
Copyright Big Data Pr Serge Miranda, MBDS, Univ de Nice Sophia Antipolis (UCA)72
Q1 query (Thesaurus)
Select B# ➔ NBREPAGE <structural inheritance with ➔ operator>
From Book
Where editor = « Dunod » and LOAN = « True » ;
Copyright Big Data Pr Serge Miranda, MBDS, Univ de Nice Sophia Antipolis (UCA)73
Q2 query ☺
Select B#➔ NBREPAGE
From Book
Where
editor = « Dunod » and LOAN = « True » and B#➔ title = « Concept% »
and B#➔ keyword in { Software, CONCERN (Software)}
Copyright Big Data Pr Serge Miranda, MBDS, Univ de Nice Sophia Antipolis (UCA)74
Object-relational DBMS (Books and web site)
➢http://thethirdmanifesto.com/
➢C. J. Date and Hugh Darwen « Databases, Types, and the Relational Model : The Third Manifesto », Addison-Wesley, 2006 (for the 3rd edition)
➢M.Stonebraker« Object-relational DBMS (The next great wave) »Morgan Kaufmann, 1996
Copyright Big Data Pr Serge Miranda, MBDS, Univ de Nice Sophia Antipolis (UCA)75
Seminars : Second DB manifesto by M. Stonebraker and DCOM from Microsoft
« Everything (concerning data store interface) is
and will be SQL »