Upload
darren-norman
View
218
Download
1
Embed Size (px)
Citation preview
1
Model Driven Developmentwith ORM 2 and NORMA
© 2007 T. Halpin & Neumont University
Terry HalpinNeumont University
2
• ORM Features, History, and Tool Support
• The Data Modeling Process
• ORM’s Graphical Language
• Comparison with ER and UML
• Case Study
• Relational Mapping and Tool Demos
3
• ORM Features, History, and Tool Support• The Data Modeling Process
• ORM’s Graphical Language
• Comparison with ER and UML
• Case Study
• Relational Mapping and Tool Demos
• Temporal aspects of Data Modeling
4
• A conceptual approach for modeling, querying, and transforming data
• Fact-oriented (attribute-free).
All facts are modeled as relationships (unary, binary, ternary …)1
• Semantic stability
(no remodeling or requerying to talk about an attribute)
• Facilitates validation by verbalization & population
• Richly expressive graphical constraint language
(compared with industrial ER, or UML class diagrams).
Object-Role Modeling (ORM)
1 The OMG’s SBVR approach is also fact-oriented.
5
Modeling Approach = Modeling Procedure(s) + Modeling Language(s)
ORM includes procedures for conceptual modelingand for mapping (transforming) ORM models and queries to attribute-based models(e.g. Relational, OO, ER, UML, XSD, OWL).
ORM includes graphical and textual modeling languagesand a textual query language.
ORM History and Tool Support
• Originated in 1970s in Europe
• Various flavors (NIAM2007, ORM 2, FCO-IM etc.)
• First ORM tools developed at Control Data labs in Brussels (IAST, RIDL*)
• USA tools: InfoDesigner, InfoModeler, ActiveQuery, ORM Source Model solution in Microsoft Visio for Enterprise Architects
• Current European tools: CaseTalk, Infagon, CogNIAM
• ORM 2 tool under development: NORMA
(open source plug-in to Visual Studio .NET)
This presentation focuses on ORM 2 (2nd generation ORM)
6
7
• ORM Features, History, and Tool Support
• The Data Modeling Process• ORM’s Graphical Language
• Comparison with ER and UML
• Case Study
• Relational Mapping and Tool Demos
• Temporal aspects of Data Modeling
Going from a Going from a process use process use
casecaseto a data to a data model model
Client(.nr)
uses
Account(.nr)
AccountType(.code)
has
holdsMoneyAmount
(USD)
Transaction(.nr)
is oninvolves
is of
TransactionType(.code)
ATM
Deposit
Withdraw
Transfer betweenaccounts
Get Balance
Client
8
• For data modeling, we need DATA use cases
(cases of data being used), e.g.– Sample reports– Sample input forms– Sample queries
• How to go from a data use case to a data model?– Have the domain expert verbalize the data– Rephrase this as unambiguous, elementary facts– Add and validate the business rules constraining the data
Data Use Cases
9
• The domain expert bestunderstands the business domain
• The modeler elicits andformalizes this understanding
• The modeler assists the domain expert to identify the business rules related to the data (constraints or derivation rules)
• The modeler validates the model with the client by– Verbalizing the model in natural language– Populating the model with positive/negative examples
Analysis is a Joint Activity
10
11
Patient# Temperature
571 100
Data (uninterpreted syntax)
The Patient with Patient# ‘571’has a Temperature of 100 oF
Fact (proposition taken to be true)
Information = data + semantics
571
12
An elementary fact is an assertion that
an object has a property *
or
one or more objects
participate in a relationship **
where the fact cannot be split
into simpler facts with the same objects (without info-
loss)
* plays a role by itself
** play different roles in the same association
Elementary Facts
Person
smokes
drivesCar
13
The Person named ‘Jack Smith’ smokes.
A unary fact
A binary factThe Executive named ‘William Portals’ climbedthe Mountain named ‘Mt Rainier’.
Facts may also be of higher arity(4 or more roles).
A ternary fact
The Person named ‘Don Bradman’ played the Sport named ‘Cricket’for the Country named ‘Australia’.
An Old but Fun Example
4 … 10 20
QLD Pot
NSW Middy, Ten Pint
WA Shetland Pony Middy, Ten Pot
… … … … …
QLD 10 Pot NSW 10 Middy ... ... ...
State(.code)
BeerServe(fl oz:)
… calls … by ...CommonName
14
15
• Feasibility study
• Requirements analysis
• Conceptual design (data, process)
• Logical design (data, process)
• External design (data, process)
• Prototyping
• Internal design and implementation
• Testing, validation, and maintenance
Large projects are often developed iteratively
Overall Procedure: Information systems life cycle
16
Conceptual analysis usually involves:
• high level service (essential business process) modeling
• information modeling
For large applications:
• divide the UoD into manageable sub-sections
• prioritize the order in which sub-sections will be modeled
• apply the Conceptual Schema Design Procedure (CSDP)
to each sub-section
• integrate the subschemas into a global conceptual schema
Many applications build on existing applications:
• reverse-engineer the existing model(s) to a conceptual model
• refine the conceptual model to fit the new business needs
17
Conceptual schema design procedure
1 Transform familiar examples into elementary facts
and apply quality checks
2 Draw the fact types
and apply a population check
3 Check for entity types that should be combined
and note any arithmetic derivations
4 Add uniqueness constraints
and check arity of fact types
5 Add mandatory role constraints
and check for logical derivations
6 Add value, set comparison, and subtyping constraints
7 Add other constraints and perform final checks
18
• ORM Features, History, and Tool Support
• The Data Modeling Process
• ORM’s Graphical Language• Comparison with ER and UML
• Case Study
• Relational Mapping and Tool Demos
ORM 2 Graphical Modeling Language
Object = Entity or Value
Entity = Object that is identified by a definite description. Entities typically change their state over time. Entities may be concrete or abstract.
e.g. The Country that has CountryCode ‘AU’.
The President named ‘Abraham Lincoln’.
The Course with course code ‘CS542’.
Entity types are depicted as named, soft rectangles.
As a configuration option, soft rectangles may bereplaced by hard rectangles or ellipses.
Country President
Country
Country President
President
19
Value = Lexical Constant (typically a character string or number). Values are literal and cannot change their state.
e.g. The CountryCode ‘AU’.
The PresidentName ‘Abraham Lincoln’.
The CourseCode ‘CS542’.
The RoomNumber ‘207’.
The SerialNumber 1090.
Value Types are depicted as named, dashed rectangles.Optionally, ellipses or hard rectangles may be used instead.
CountryCode
PresidentName
CourseCode
RoomNr
SerialNumber
20
21
Many entities are identified by their relationship to a simple value.If this is true for all instances of their entity type,the reference (identification) scheme for their entity typemay be displayed as a reference mode in parenthesis.
The reference mode may be popular, unit-based, or general.
A popular reference mode has a corresponding value typethat is used to identify entities of one type only,and is preceded by a dot.
e.g.
The value type name appends the reference mode name to theentity type name, with a user-definable format that may includea separatore.g. CountryCode CourseCode PresidentName etc.
Country_CodeCourse_Code President_Name etc.
Country(.code)
Course(.code)
President(.name)
Product(.name)
Building(.nr)
Employee(.nr)
22
A unit-based (or measurement) reference mode uses a unitbased on some unit dimension (whose display is often suppressed)1.A colon “:” is appended to the unit
e.g.
The value type name appends “Value” to the reference mode name(if the language is English) with a user-definable format e.g.
cmValue USDValuecm_Value USD_Value
If desired, the unit type may be displayedafter the colon, e.g.
Height(cm:)
Width(cm:)
Tax(USD:)
CostPrice(USD:)
MoneyAmount(USD:)
Height(cm: Length)
Tax(USD: Money)
1 Support for unit-based reference etc. is expected by end 2007
23
Different units based on the same unit dimensionare permitted in the same model,e.g.
General reference modes have the same name as their value type.The value type may be used to reference multiple entity typese.g.
Tax(USD: Money)
Fee(XEU: Money)
Height(cm: Length)
Distance(km: Length)
Book(ISBN)
Website(URL)
Link(URL)
24
An independent object type may have instancesthat exist in the model without participating in any other relationships.
Independent object types have a “!” placed after their namee.g.
If an object type shape is duplicated in the diagram(either on the same page or on different pages)this is shown by a shadowe.g.
An external object type is defined in another model.The display notation “^” is tentativee.g.
Country !(.code)
PersonPerson StateCodeStateCode
Address^
[role1][role2]R R / S S R
A
25
Predicates (relationships) have one or more roles,each played by instances of a single object type.
Predicate readings may be shown in mixfix notation1
using … as an object placeholder,
e.g. … introduced … to …
For unary and binary predicates with no leading or trailing text,the placeholder may be omitted
e.g. smokes i.e. … smokeslikes i.e. … likes …
Roles may also be named.Duplicate predicate shapes are shadowed.
1 Mixfix allows natural verbalization of predicates of any arity, and non-infix predicates (common in many foreign languages).
26
Personsmokes
Personwas born in
Country
Person
reports to / manages
[manager]
Person
Sport
Country… played … for ...
Personemploys
Department[employee]
[employer]
For binary predicates, forward and inverse readings may be shown separated by “/”.Alternatively, arrow tips may be used.
Combining a predicate with its object type(s) forms an elementaryor compound fact type.
An elementary fact can’t be split into smaller factswith the same objects, without information losse.g.
27
A compound fact type includes two or more fact types,and if used in a model must be declared to be derivede.g.
An existential fact (or reference)simply asserts the existence of an objecte.g.
There exists a Country that has CountryCode ‘US’.
Existential fact types are displayedeither using a reference modeor an explicit relationship, e.g.
This includes constraints (see later).
Person
Car
Country
drives is imported from
… drives … imported from … *
Country(.code) Country
has / refers toCountryCode
28
e.g.
An elementary fact type may be objectified, resulting in another object type.
R“A”
Student Courseenrolled in
“Enrollment”
A fact type may be:
• asserted (base fact type)
• fully derived
• derived and stored
• semi-derived
R*
R**
R+
R
Uniqueness constraints require instances of their role or role sequenceto be unique in the role or role sequence population.
Internal uniqueness constraints apply to a single predicateand are depicted by bar(s) over the constrained role(s).
External uniqueness constraints apply to roles from different predicatesand are depicted by circled bars connected to the roles.
If the constraint applies to role(s) used to provide the preferredidentification scheme for an object type, a double-bar is used.
29
A simple mandatory role constraint requires its roleto be played by all instances of its object type’s populationand is shown by a solid dot at either end of the role-type connector.
An inclusive-or (or disjunctive mandatory role) constraintrequires at least one of its rolesto be played by all instances of its object type’s populationand is shown by a circled, solid dot connected to the roles.
A A
30
31
e.g.
Room
is inBuilding
(.nr)
hasRoomNr
Room(.nr)
HourSlot(dhCode)
Activity(.code)
… at … is booked for ...
20 Mon 9 a.m. ORC20 Tue 2 p.m. ORC33 Mon 9 a.m. XQC33 Fri 5 p.m. STP
ActivityName
has /refers to
ORC ORM class STM Staff Meeting STP Staff Party XQC XQuery class
20 Mon 9 a.m. XQC33 Mon 9 a.m. ORC
STP Staff PlanningSPY Staff Party
Consultant(.nr)
hasPassport
(.nr)
has
DriverLicense(.nr)
Language(.name)
is spoken by
32
A{a1, a2, a3} A
{a1 .. an}
{a ..}
Enumeration Range
Semi-bounded discrete range
Bounded continuous range
{(a1 .. a2)}
{[a1 .. a2]}
{[a1 .. a2)}
{(a1 .. a2]}
includes both end valuesexcludes both end valuesincludes first valueincludes last value
Object Value Constraints
Role Value Constraints A
{a1, a2}
Same patterns
as above
33
Subset Constraints 2
1
Simple:
Contiguous Role-pair:
1.1 1.2
2.1 2.2 Each object pair that playsthe role sequence 1.1, 1.2also playsthe role sequence 2.1, 2.2
Other cases:
Each object tuple that playsthe first role sequencealso playsthe second role sequence
ORM 2 also displays subsetconstraints over join paths
34
Equality Constraints
2 role-sequences (of 1 or more roles):
2
1 1.1 1.2
2.1 2.2
Populations of role-sequencesmust be equal
3 or more role-sequences: 1.1 1.2
2.1 2.2
3.1 3.2
e.g.
35
Exclusion Constraints n
1 1.1 1.2
n.1 n.2
Populations of 2 or more role-sequencesmust be mutually exclusive
: : :
Exclusive-Or Constraints
Each instance in A’s population playsexactly one of the n attached roles (n > 1)
A
1
n
: or A
1
n
:
Subtyping
36
A
B
B is a proper subtype ofA (its primary supertype) andC (a secondary supertype)
C
A
B C
A
B C
A
B C
Exclusive Total Partition
Frequency Constraints
37
1
f
1
f
The frequency specification fmay be any of the following
n exactly n (a positive integer)
n at least n
n at most nn..m at least n and at most m
1
f
2
2
Each instance that playsrole 1 does so f times
Each instance pair that playsroles 1, 2 does so f times
Each instance pair that playsroles 1, 2 does so f times
Ring Constraints
38
Irreflexive
Asymmetric
Intransitive
Antisymmetric
Acyclic
Asymmetric + Intransitive
Acyclic + Intransitive
Symmetric
Purely Reflexive
A
The previous constraints are allalethic(necessarily true for each state).
ORM 2 also supportsdeontic versionsof all these constraints
39
Uniqueness
Deontic constraints are colored bluerather than violet. Most include an “o”for “obligatory”. Deontic ring constraintsinstead use dashed lines.
Mandatory
Subset, Equality, Exclusion
Frequency f
Irreflexive Acyclic
Antisymmetric Symmetric
Intransitive Acyclic-Intrans
Asymmetric Asym-Intrans
Purely Reflexive
40
Uniqueness constraint on first role
+ve form: Each Moon orbits at most one Planet.
Illustrated by a satisfying fact population.
-ve form: It is impossible that the same Moon orbits more than one Planet.
Test with a counter-example.
Phobos Mars
Deimos
Mars
Io Jupiter
Io Mars }Counter-example
Model ValidationMoon
(.name)orbits / is orbited by
Planet(.name)
41
The absence of a uniqueness constraint on the second rolemay be verbalized using default form:
It is possible that the same Planet is orbited by more than one Moon.
Illustrated by a satisfying fact population.
Phobos Mars
Deimos
Mars
Io Jupiter
Moon(.name)
orbits / is orbited by
Planet(.name)
42
Sample screenshotshowing automatedverbalization (+veplus some default)for some selectedaspects.
Currently about 80%of constraints areverbalized.The rest shouldbe implementedin a few months.
43
Alethic:
It is possible that more than one Patient is a husband of the same Patient and that more than one Patient is a wife of the same Patient.
Each Patient, Patient combination occurs at most once in the population of Patient is a husband of Patient.
Deontic:
It is obligatory that each Patient is a husband of at most one Patient.
It is obligatory that each Patient is a wife of at most one Patient.
In ORM 2,rules may be assigneddifferent modalities
Patient(.nr)
is a husband of / is a wife of
44
• ORM Features, History, and Tool Support
• The Data Modeling Process
• ORM’s Graphical Language
• Comparison with ER and UML• Case Study
• Relational Mapping and Tool Demos
45
ER and UML class diagrams are attribute-based,leading to more compact diagramsthat are closer to implementation schemas.
UML also includes many other diagram types to deal withprocess modeling etc.
ORM’s attribute-free nature facilitatesvalidation by verbalization and populationand semantic stability.
ORM’s graphic language is far richer for data modelingthan that of ER and UML,and its textual languages are far easier for non-technical usersto understand than UML’s OCL.
ORM’s graphical language is also orthogonal and unambiguous(unlike UML).
46
UML’s multiplicity notation is fine for binaries but not for n-ariese.g.
Room(.nr)
HourSlot(dhCode)
Activity(.code)
… at … is booked for ...
ActivityNamehas / is of
can’t express as a multiplicity
*0..1 0..1
dhCode {P}
HourSlot
nr {P}
Room
code {P}name {U1}
Activity
Booking
Each activityhas a booking
47
Vehicle Company{xor}**
0..1
0..1
vehicleLeased lessor
vehicleSold seller
UML’s xor is defined between associations, not association roles,so this is ambiguous.
Vehicle
is leased from
was purchased from
Company
ORM correctly defines the constraint between roles and treats it as a combination of exclusion and inclusive-or.
Vehicle
is leased from
was purchased from
Company
48
ORM models are immune to changes that reshape attributesas entity types or relationships.
The meaning of a query is not changed if we change a constraint or add a new fact type.
ORM queries respect this principleand hence facilitate schema evolution.
ER and OO queries do not: such changes can cause attributes to be remodeled; hence, existing queries need to be reformulated.
Semantic stability
49
List titled people and their gender
Person --- has Title --- is of Gender
SSN {P}gendertitle [0..1]
Person
select SSN, genderfrom Personwhere title is not null
Person(SSN)
is of
Gender(.code)
hasTitle
50
List titled people and their gender
Person --- has Title --- is of Gender
select Person.SSN, genderfrom Person join PersonTitle on Person.SSN = PersonTitle.SSN
SSN {P}gender
Person
name {P}
Title* *
has
precedes
* *
Person(SSN)
is of
Gender(.code)
hasTitle
precedes
51
Have your cake and eat it too
by using ORM for conceptual analysis
and mapping it to ER or UML views as desired.
It is expected that the NORMA tool will provide
automatic, live generation of both ER and UML views
by the end of 2008.
52
• ORM Features, History, and Tool Support
• The Data Modeling Process
• ORM’s Graphical Language
• Comparison with ER and UML
• Case Study• Relational Mapping and Tool Demos
Specify an ORM schema for this report from a book publisher.
Sales ISBN Title Published
Translation of Year Nr Total
Best Seller?
1-33456-012-3 Mizu no Kokoro
2002 2003 2004 2005
5000 6000 5000
16000
Y 2-55860-123-6 Mind Like
Water 2004 1-33456-012-3 2004
2005 3000 3000
6000
N
3-540-25432-2 Informatics 2005 2005 2000 2000 N 4-567-12345-3 Informatics 2006 5-123-45678-5 Semantics
Year(CE)
Year(CE)
Year(CE)
Year(CE)
Book(ISBN)
BookTitle
has
was published in
PublishedBook
is translated from
is a best seller*[copiesSoldInYear]
[totalCopiesSold]
Each PublishedBook is a Book that was published in some Year.* For each PublishedBook, totalCopiesSold= sum(copiesSoldInYear).* PublishedBook is a best seller iff PublishedBook sold total NrCopies >= 10000.
NrCopies… in … sold ...
sold total- *
Model this report from the same business domain.
PNr Name Title Gender Books authored 1 2 3 4 5 6 7 8 9
John Smith Don Bradchap Sue Yakamoto Yoko Ohyes Isaac Seldon Ann Gables John Smith Ann Jones Selena Moore
Mr Sir Mrs Dr Dr Ms Mr Ms Mrs
M M F F M F M F F
1-33456-012-3 2-55860-123-6 3-540-25432-2, 5-123-45678-5 4-567-12345-3 5-123-45678-5
Person(.nr)
PersonName
has/is of
Gender(.code)
is of
{‘M’, ‘F’}
has
PersonTitle
is restricted to
Book(ISBN)
authored
Model this final report from the same business domain.
Review Assignment ISBN Title PNr Name Result
1-33456-012-3 2-55860-123-6 3-540-25432-2 4-567-12345-3
Mizu no Kokoro Mind Like Water Informatics Informatics
1 4 2 5 6 1 7 1 5
John Smith Yoko Ohyes Don Bradchap Isaac Seldon Ann Gables John Smith John Smith John Smith Isaac Seldon
4 5 5 5 4 4 5
Book(ISBN)
is authored by /authored
Person(.nr)
is assigned for review by
“ReviewAssignment !”
resulted in
Grade(.nr)
{1..5}
2
BookTitle
has
Book(ISBN)
is authored by
Person(.nr)
is assigned for review by“ReviewAssignment !”
PersonName
has/is of
Gender(.code)
is of
{‘M’, ‘F’}
has
PersonTitle
is restricted to
resulted in
Grade(.nr) {1..5}
BookTitle
has
Year(CE)
was published in
PublishedBook
is translated from
… in … sold ...
NrCopiessold total- * is a best seller*
Each PublishedBook is a Book that was published in some Year.* For each PublishedBook, totalCopiesSold= sum(copiesSoldInYear).* PublishedBook is a best seller iff PublishedBook sold total NrCopies >= 10000.
[copiesSoldInYear]
[totalCopiesSold]
2
The full schema
60
• ORM Features, History, and Tool Support
• The Data Modeling Process
• ORM’s Graphical Language
• Comparison with ER and UML
• Case Study
• Relational Mapping and Tool Demos
61
Conceptual Schema Relational Schema
Allergy
PK,FK1 patientNrPK drugName
Patient
PK patientNr
patientNamesmokes
Rmap procedure generates 5th normal form by default.
Patient(.nr)
PatientName
has
is allergic toDrug
(.name)
smokes
[allergy]
Relational Mapping
62
Patient
1025* PatientNr:
* Name:
Allergies:
Ann Jones
OK
PenicillinCodeine
Smokes
(a) (b)Patient
1056* PatientNr:
* Name:
Allergies:
John B. Smith
OK
Smokes
Patient Allergy
patientNr patientName smokes
1025
1056
Ann Jones
John B. Smith
true
false
patientNr
drugName
1025
1025
Penicillin
Codeine
External (forms) and Logical (relational) schemas provide different structures for grouping elementary (conceptual) facts.
63
Tool Demos
(1) Microsoft Visio ORM Source Model Solution
(2) NORMA
64
Microsoft Visio for Enterprise Architects supports:
• Entry of ORM 1 schemas
• Forward engineering of ORM schemas to relational schemas
• Forward engineering of ORM updates to relational updates
• Direct entry of relational schemas
• Multiple styles for relational schemas (pure relational, IDEF1X, …)
• Reverse engineering of relational schemas to ORM schemas
• Report generation
• NORMA1 is the first tool to support ORM 2
• Coded in C#, XML and XSLT
• Open source plug-in to Microsoft Visual Studio .NET 2005, utilizing Microsoft’s Domain Specific Language (DSL) toolkit.
• Supports entry of ORM 2 models
• Automated live error checking and verbalization
• Automatic transformation to implementation artifacts
• Can import ORM models from Visio (via free Orthogonal Toolbox)
• Currently pre-beta. A usable version for industry is expected in 2008
1 Public version downloadable from http://sourceforge.net/projects/orm. A newer build will soon be provided. See NORMA Labs 1-5 to start.
NORMA (Neumont ORM Architect)
65
66
• NORMA supports mappings to various implementation artifacts
SQL: 2003
IBM DB2
Oracle
PostgreSQL
MySQL
MS SQL Server
DDILDCILOIALn-aryORM
BinaryORM
WSDL PLiX
C# VB.NET PHP
OIAL ORM Intermediate Abstraction LanguageDCIL Database Conceptual Intermediate LanguageDDIL Data Definition Intermediate LanguagePLiX Programming Language in XML
OWL DSL
.NETTiers
EDM
Java
DTD
early development
XSD
mid-stage development
67
Main current limitations of NORMA
• Forward engineering to new schemas only (generation of incremental schema updates is coming)
• Verbalization of only about 75% of constraints (full verbalization expected early 2008)
• No reverse engineering (we have a basic prototype. RevEng expected in a 2008 release)
• Constraint code generation only for basics (ma, unique, value) (full code generation of graphic constraints expected late 2008) (code generation for formal textual constraints expected later)
Future plans include• complete application generation (including forms)• ER, UML etc. views (editable)• multi-model support (including model component reuse)• integration with conceptual process modeling
68
www.orm.net -- my ORM websitewww.ORMFoundation.net -- ORM Foundation websitewww.inConcept.com -- Journal of Conceptual Modeling, …www.ORMcentral.com -- COM API details for VEA, …www.objectrolemodeling.com/ -- Orthogonal Toolbox, …www.brcommunity.com/ -- Business Rules Journal articles
Halpin, T. 2001, Information Modeling and Relational DatabaseDesign, Morgan Kaufmann.
Halpin, T. et al. 2003, Database Modeling with Microsoft Visiofor Enterprise Architects, Morgan Kaufmann.
Further resources