Metamodeling. 16. April 2012 Real-Time Systems Lab Prof. Dr. Andy Schürr Dr. Gergely Varró 1

Transcription

1 Metamodeling The slides of this lecture are reused from the Model Engineering course at TU Vienna with the kind permission of Prof. Gerti Kappel (head of the Business Informatics Group) 16. April 2012 Real-Time Systems Lab Prof. Dr. Andy Schürr Dr. Gergely Varró 1

2 Contents Introduction Meta languages Metamodeling language: MOF 16. April 2012 Real-Time Systems Lab Prof. Dr. Andy Schürr Dr. Gergely Varró 2

3 Introduction Motivating example: a simple UML activity diagram Activity, ActivityTransition, InitialNode, FinalNode ad Course workflow Attend lecture Study content Write exam Question: Is this UML Activity diagram valid? Answer: Check the UML metamodel! Prefix meta : an operation is applied onto itself Further examples: meta discussion, meta learning, Aim of this lecture: What is meant by the term metamodel and how are metamodels defined? 16. April 2012 Real-Time Systems Lab Prof. Dr. Andy Schürr Dr. Gergely Varró 3

4 Anatomy of Formal Languages I. Although languages have, in general, divergent orientations and fields of application they still have a common language definition structure Formal languages Defines language elements/grammar Persistency and model exchange Semantics Abstract Syntax Meaning of the language elements Notation of the language elements Serialization Syntax Concrete Syntax 16. April 2012 Real-Time Systems Lab Prof. Dr. Andy Schürr Dr. Gergely Varró 4

5 Anatomy of Formal Languages II. Main components Abstract syntax: Defines the language concepts and how these concepts can be combined (~ grammar) However, it does not define the notation or meaning of the concepts Concrete syntax: Notation to illustrate the language concepts intuitively 2 ways: textual or graphical Semantics: Defines the meaning of the language concepts How are language concepts interpreted? Serialization syntax: For persistent storage and model exchange between tools XML, proprietary format, Additional components Extension of the language by new language concepts Domain- or technology-specific extensions Mapping to other languages, domains Transformation, semantic equivalence, abstraction, Examples: UML2Java, UML2SetTheory, PetriNet2BPEL, 16. April 2012 Real-Time Systems Lab Prof. Dr. Andy Schürr Dr. Gergely Varró 5

6 Contents Introduction Meta languages Metamodeling language: MOF 16. April 2012 Real-Time Systems Lab Prof. Dr. Andy Schürr Dr. Gergely Varró 6

7 Metamodeling Old Wine in New Bottles? Formal languages have a long tradition in computer science First attempts: Transition from machine code instructions to high-level programming languages (Algol60) Major successes Imperative, executable, high-level languages such as Java, C++, C#, Declarative languages such as XML Schema, DTD, RDF, OWL, SQL, Excursus How are programming languages and XML-based languages defined? What can be learned for modeling languages? 16. April 2012 Real-Time Systems Lab Prof. Dr. Andy Schürr Dr. Gergely Varró 7

8 Programming Languages I. John Backus and Peter Naur invented formal languages for the definition of languages so-called meta languages Examples for meta languages: BNF, EBNF, Used since 1960 for the definition of the syntax of programming languages Remark: The abstract and the concrete syntax are defined EBNF Example: Optional Java := [PackageDec] {ImportDec} ClassDec; PackageDec := package QualifiedIdentifier; ImportDec := import QualifiedIdentifier; ClassDec := Modifier class Identifier [ extends Identifier] [ implements IdentifierList] ClassBody; Production Rule Sequence 16. April 2012 Real-Time Systems Lab Prof. Dr. Andy Schürr Dr. Gergely Varró 8 Non-Terminal Terminal

9 Programming Languages II.: MiniJava Example Grammar Java := [PackageDec] {ImportDec} ClassDec; PackageDec := package QualifiedIdentifier; ImportDec := import QualifiedIdentifier; ClassDec := Modifier class Identifier [ extends Identifier] [ implements IdentifierList] ClassBody; Modifier := public private protected ; Identifier := { a z A Z 0 9 }; Program package de.tu-darmstadt.es; import java.util.*; public class Student extends Person { } Validation Does the program conform to the grammar? Compiler: javac, gcc, Interpreter: Ruby, Python, 16. April 2012 Real-Time Systems Lab Prof. Dr. Andy Schürr Dr. Gergely Varró 9

10 Programming Languages III. Four layer architecture Grammar := {Rule}; Rule := Terminal NonTerminal ; Definition of EBNF in EBNF - reflexive M3-Layer Java := [PackageDec] {ImportDec} ClassDec; PackageDec := package QualifiedIdentifier; Definition of Java in EBNF grammar M2-Layer package de.tu-darmstadt.es; public class Student extends Person { } Program Sentence conform to the grammar M1-Layer Execution of the program M0-Layer 16. April 2012 Real-Time Systems Lab Prof. Dr. Andy Schürr Dr. Gergely Varró 10

11 XML-Based Languages I. XML files require definite structures to allow for a standardized and automatic processing Examples for XML meta languages: DTD, XML Schema Characteristics of XML files Well-formed (character level) vs. valid (grammar level) DTD Example: Element Containment 1..* <!ELEMENT cookbook (title, meal+)> <!ELEMENT title (#PCDATA)> <!ELEMENT meal (ingredient+)> <!ELEMENT ingredient> <!ATTLIST ingredient name CDATA #REQUIRED amount CDATA #IMPLIED Attributes unit CDATA #IMPLIED> April 2012 Real-Time Systems Lab Prof. Dr. Andy Schürr Dr. Gergely Varró 11

12 XML-Based Languages II. DTD <!ELEMENT cookbook (title, meal+)> <!ELEMENT title (#PCDATA)> <!ELEMENT meal (ingredient+)> <!ELEMENT ingredient> <!ATTLIST ingredient name CDATA #REQUIRED amount CDATA #IMPLIED unit CDATA #IMPLIED> XML <cookbook> <title>how to cook!</title> <meal name="spaghetti"> <ingredient name="tomato" amount="300" unit="gramm"/> <ingredient name="meat" amount="200" unit="gramm"/> </meal> </cookbook> Validation XML Parser: Xerces, 16. April 2012 Real-Time Systems Lab Prof. Dr. Andy Schürr Dr. Gergely Varró 12

13 XML-Based Languages III. Five layer architecture (revised with XML-Schema) Grammar := {Rule}; Rule := Terminal NonTerminal ; Element := <!ELEMENT Identifier > AttList; AttList := <!ATTLIST Identifier ; <!ELEMENT javaprog (packagedec*, importdec*, classdec)> <!ELEMENT packagedec (#PCDATA)> Definition of EBNF in EBNF Definition of DTD in EBNF Definition of Java in DTD Grammar M4-Layer M3-Layer M2-Layer <javaprog> <packagedec>de.tu-darmstadt.es</packagedec> <classdec name="student" extends="person"/> </javaprog> XML conform to the DTD M1-Layer Objects (e.g., of type Student) M0-Layer 16. April 2012 Real-Time Systems Lab Prof. Dr. Andy Schürr Dr. Gergely Varró 13

14 Contents Introduction Meta languages Metamodeling language: MOF 16. April 2012 Real-Time Systems Lab Prof. Dr. Andy Schürr Dr. Gergely Varró 14

15 Spirit and Purpose of Metamodeling I. Metamodels define language concepts and their grammar for the specification of models The Greek prefix»meta«means»about«- hence a metamodel states something»about«other models Differentiation between model elements and metamodel elements Metamodel Model Model Excerpt of reality View View (transitive) View 16. April 2012 Real-Time Systems Lab Prof. Dr. Andy Schürr Dr. Gergely Varró 15

16 Spirit and Purpose of Metamodeling II. Advantages of metamodels Concise and precise definition of the language concepts Standardized exchange format Checking correctness of models Management of models in repositories Extensibility of the language Generalization on a higher level of abstraction by means of the meta-metamodel Language concepts for the definition of a metamodel MOF is considered as a universally accepted meta-metamodel 16. April 2012 Real-Time Systems Lab Prof. Dr. Andy Schürr Dr. Gergely Varró 16

17 Spirit and Purpose of Metamodeling III. Metamodeling stack Language definitions Examples Meta- Metamodel defines Meta language MOF, Ecore represents conformsto conformsto Metamodel defines Language UML, ER, Model System Used terms Intra-level Inter-level Author instanceof conformsto [Bézivin01] instof meta [Favre04] ontological instof linguistic instof [Atkinson02] 16. April 2012 Real-Time Systems Lab Prof. Dr. Andy Schürr Dr. Gergely Varró 17

18 Meta Object Facility I. OMG standard for the definition of metamodels Defines set of language concepts for the modeling of object-oriented structures Language concepts based on object-orientation Objects are described by classes which can be generalized Intrinsic properties of objects Relationships (associations) between objects Packages to group classes MOF itself is defined by MOF (reflexive) and divided into EMOF (Essential MOF) Simple language for the definition of metamodels Interesting for metamodelers CMOF (Complete MOF) Extends EMOF Supports management of meta data via enhanced services (e.g., reflection) Interesting for tool manufacturers 16. April 2012 Real-Time Systems Lab Prof. Dr. Andy Schürr Dr. Gergely Varró 18

19 Meta Object Facility II. Offers modeling infrastructure for MDA MDA dictates MOF as meta-metamodel Metamodels are defined by means of MOF language concepts MOF is used as a metamodeling language and at the same time it is used for the storage of models UML, CWM and further OMG standards are conform to MOF In addition, there are mapping rules for various technical platforms defined for MOF XML: XML Metadata Interchange (XMI) Java: Java Metadata Interfaces (JMI) CORBA: Interface Definition Language (IDL) 16. April 2012 Real-Time Systems Lab Prof. Dr. Andy Schürr Dr. Gergely Varró 19

20 Meta Object Facility III. OMG language definition stack MOF Model M3-Layer Meta-Metamodel UML Metamodel IDL Metamodel CWM Metamodel M2-Layer Metamodel Models Models UML Models Models Models IDL Interfaces CWM Models Models Models M1-Layer Model M0-Layer Instances 16. April 2012 Real-Time Systems Lab Prof. Dr. Andy Schürr Dr. Gergely Varró 20

21 MOF 2.0 Language Architecture I. MOF 2.0 Minimal OO language range Extended reflection and OO concepts EMOF CMOF Extension Extensionmechanism Reflection Selfreflection Identity Unambiguous identification 16. April 2012 Real-Time Systems Lab Prof. Dr. Andy Schürr Dr. Gergely Varró 21

22 MOF 2.0 Language Architecture II. Core of EMOF Based on object-orientation Classes, attributes, associations, operations and parameter Type TypedElement MultiplicityElement Property Class isabstract: Boolean ownedattribute 0..1 * isreadonly: Boolean = false default: String [0..1] iscomposite: Boolean = false isderived: Boolean = false 0..1 opposite superclass * 0..1 TypedElement * Operation MultiplicityElement TypedElement ownedparameter * Parameter MultiplicityElement ownedoperation raisedexception * Type 16. April 2012 Real-Time Systems Lab Prof. Dr. Andy Schürr Dr. Gergely Varró 22

23 MOF 2.0 Language Architecture III. Abstract classes of EMOF Definition of general properties NamedElement Name TypedElement Type MultiplicityElement Multiplicities Order Set semantics Type isinstance(element:element): Boolean 0..1 type Taxonomy of the abstract classes Object Element NamedElement name:string TypedElement MultiplicityElement isordered: Boolean = false isunique: Boolean = true lower: Integer upper: UnlimitedNatural 16. April 2012 Real-Time Systems Lab Prof. Dr. Andy Schürr Dr. Gergely Varró 23

24 Classes A class specifies structure and behavior of a set of objects An unlimited number of instances (objects) of a class can be created A class has a unique name At least in the local namespace Abstract classes: cannot be instantiated! only useful in inheritance hierarchies used for»highlighting«of common features of a set of subclasses Concrete classes: can be instantiated directly! MOF Example Transition Class name : String isabstract : boolea name: String isabstract : boolean Activity Event 16. April 2012 Real-Time Systems Lab Prof. Dr. Andy Schürr Dr. Gergely Varró 24

25 Generalization Generalization: taxonomic relation between a specialized class (subclass) and a general class (superclass) Subclass inherits properties of the superclass and appends further properties Discriminator: virtual attribute used for the classification Disjoint (non-overlapping) generalization Multiple inheritance but no multiple classification (Reason: see previous bullet) MOF Example TimeEvent Class 0..* superclasses Event TimeOrCall Event CallEvent 16. April 2012 Real-Time Systems Lab Prof. Dr. Andy Schürr Dr. Gergely Varró 25 not possible in MOF!

26 Attributes Attributes describe inherent characteristics of concepts (classes) Consist of a name and a range of values Initial value - default Derived (calculated) attributes Multiplicity: how many values can be stored in an attribute Interval: upper and lower limit are natural numbers * asterisk - also possible for upper limit (Semantics: unlimited number) 0..1 (optional): null values are allowed MOF Example id: Integer [1..1] TimeEvent kinds: String [1..*] Class * ownedattribute Property isreadonly: Boolean default: String[0..1] iscomposite: Boolean isderived: Boolean Event CallEvent synchron: boolean = true [0..1] 16. April 2012 Real-Time Systems Lab Prof. Dr. Andy Schürr Dr. Gergely Varró 26

27 Associations An association describes the common structure of a set of relationships between objects MOF only allows binary associations, i.e., an association is always defined between two classes Associations consist of: Role name Objects can play multiple roles Objects can dynamically take on/off roles Multiplicity limits the number of partner objects of an object Navigation navigable end getter and setter methods non-navigable end no access Composition part-whole relationship (also part-of relationship) One part can always be at most part of one composed object Asymmetric and transitive Multiplicity: 1 or April 2012 Real-Time Systems Lab Prof. Dr. Andy Schürr Dr. Gergely Varró 27

28 Association Example Association multiplicity A 1..* association * role name B Composition non-navigable role navigable role Example 1 Example 2 Example 3 A 0..1 * B A 1 C B 1 * * Syntax Semantics A 0..1 {xor} C B * * Syntax Semantics April 2012 Real-Time Systems Lab Prof. Dr. Andy Schürr Dr. Gergely Varró 28

29 Packages Packages serve as a grouping mechanism Grouping of related type declarations Partitioning criteria Functional cohesion Information cohesion Packages form own namespace Usage of identical names in different parts of a metamodel Packages can be nested Hierarchical grouping Model elements can be assigned to at most one package abstract: elements derived from Type concrete: classes, enumerations MOF NamedElement name:string Package Example X Y A C B * +nestingpackage nestedpackage 0..* Z A X Type B 16. April 2012 Real-Time Systems Lab Prof. Dr. Andy Schürr Dr. Gergely Varró 29

30 Types I. Primitive data types: Predefined types for integers, character strings and Boolean values Enumerations: Enumeration types consisting of named constants Allowed values are defined in the course of the declaration Example: enum Color {red, blue, green} Enumeration types can be used as data types for attributes Type isinstance() DataType NamedElement name:string PrimitiveType Enumeration Enumeration Literal enumeration ownedliteral 0..1 {ordered} 0..* <<primitive>> Integer <<primitive>> Boolean <<primitive>> String <<primitive>> UnlimitedNatural 16. April 2012 Real-Time Systems Lab Prof. Dr. Andy Schürr Dr. Gergely Varró 30

31 Types II. Differentiation between Value types and Reference types Value types: contain a direct value (e.g., 123 or x ) Reference types: contain a reference to an object Types Value types Reference types Primitive types Enumerations Classes Boolean Integer String user-defined types Primitive types Enumerations Reference types Car color: String Car color: CarColor «enumeration» CarColor red green blue 16. April 2012 Real-Time Systems Lab Prof. Dr. Andy Schürr Dr. Gergely Varró 31 Car owner: Person Person

32 Example I. Activity diagram example Activity, ActivityTransition, InitialNode, FinalNode ad Course workflow Attend lecture Study content Write exam How does the metamodel of this language look like? Metamodel development process 1. Identification of the modeling concepts 2. Determining the properties of the modeling concepts 3. Object-oriented design of the language 16. April 2012 Real-Time Systems Lab Prof. Dr. Andy Schürr Dr. Gergely Varró 32

33 Example II.: Identifying Modeling Concepts ad Course workflow Attend lecture Study content Write exam Notation table Syntax ad name ActivityDiagram Concept FinalNode InitialNode name Activity Transition 16. April 2012 Real-Time Systems Lab Prof. Dr. Andy Schürr Dr. Gergely Varró 33

34 Example III.: Determining Properties ad Course workflow Attend lecture Study content Write exam Concept properties Concept Intrinsic properties Relationships ActivityDiagram Name 1 InitialNode 1 FinalNode Unlimited number of Activities and Transitions FinalNode - Incoming Transition InitialNode - Outgoing Transition Activity Name Incoming and outgoing Transitions Transition - Source node and target node Nodes: InitialNode, FinalNode, Activity 16. April 2012 Real-Time Systems Lab Prof. Dr. Andy Schürr Dr. Gergely Varró 34

35 Example IV.: Object-Oriented Design MOF Class Property Property Concept Intrinsic properties Relationships ActivityDiagram Name 1 InitialNode 1 FinalNode Unlimited number of Activities and Transitions FinalNode - Incoming Transition InitialNode - Outgoing Transition Activity Name Incoming and outgoing Transitions Transition - Source node and target node Nodes: InitialNode, FinalNode, Activity ActivityDiagram name:string 1 InitialNode Transition 16. April 2012 Real-Time Systems Lab Prof. Dr. Andy Schürr Dr. Gergely Varró 35 * in out src 1 out 1 in trg src 0..1 trg Activity 0..1 name:string 0..1 FinalNode 1

36 Model Metamodel Example V.: Overview ActivityDiagram name:string 1 InitialNode * in 1 Transition out src 1 out 1 in trg src 0..1 trg * Activity 0..1 name:string 0..1 FinalNode 1 2:InitialNode 1:ActivityDiagram 3:FinalNode 7:Transition name="course workflow" 10:Transition 4:Activity 8:Transition 5:Activity 9:Transition 6:Activity name="attend lecture" name="study content" name="write exam" Abstract syntax Concrete syntax ad Course workflow Attend lecture Study content Write exam 16. April 2012 Real-Time Systems Lab Prof. Dr. Andy Schürr Dr. Gergely Varró 36

37 Correspondence between EBNF and MOF Mapping table (excerpt): EBNF MOF Production Composition Non-Terminal Class Sequence Multiplicity 0..* Example: Grammar Model ::= {Class} Class ::= Name {Attribute} {Method} C Metamodel Model * Class 1 Name * Attribute * Method 16. April 2012 Real-Time Systems Lab Prof. Dr. Andy Schürr Dr. Gergely Varró 37

38 Correspondence between DTD and MOF Mapping table (excerpt): DTD MOF Item Composition Element Class Cardinality * Multiplicity 0..* Example: DTD <!ELEMENT Model (Class*)> <!ELEMENT Class (Name, Attribute*, Method*)> C Metamodel Model * Class 1 Name * Attribute * Method 16. April 2012 Real-Time Systems Lab Prof. Dr. Andy Schürr Dr. Gergely Varró 38

39 MOF vs. UML MOF only a subset of UML? MOF very similar to the UML class diagram, but much more limited No n-ary or qualifying associations, no association classes No overlapping inheritance No interfaces, dependencies, Main differences result from the field of application UML Domain: object-oriented modeling Comprehensive modeling language for various software systems Structural and behavioral modeling Conceptual and implementation modeling MOF Domain: metamodeling Simple conceptual structure modeling language Conclusion MOF is a specialized DSML for metamodeling What is included in UML? Core of UML and MOF (almost) identical What is not included in UML? Concepts which can be used on the programming level to realize repositories, e.g., reflection 16. April 2012 Real-Time Systems Lab Prof. Dr. Andy Schürr Dr. Gergely Varró 39