Lecture 12: Proto-OCL, Modularisation & Design Patterns - PowerPoint PPT Presentation

Softwaretechnik / Software-Engineering Lecture 12: Proto-OCL, Modularisation & Design Patterns 2017-07-03 Prof. Dr. Andreas Podelski, Dr. Bernd Westphal Albert-Ludwigs-Universität Freiburg, Germany – 12 – 2017-07-03 – main –

Topic Area Architecture & Design: Content • Introduction and Vocabulary VL 10 • Software Modelling I . . (i) views and viewpoints, the 4+1 view . (ii) model-driven/-based software engineering (iii) Modelling structure VL 11 . a) (simplified) class diagrams . . b) (simplified) object diagrams VL 12 c) (simplified) object constraint logic (OCL) d) Unified Modelling Language (UML) • Principles of Design . (i) modularity, separation of concerns . . (ii) information hiding and data encapsulation (iii) abstract data types, object orientation (iv) Design Patterns VL 13 • Software Modelling II – 12 – 2017-07-03 – Sblockcontent – . (i) Modelling behaviour . . a) communicating finite automata b) Uppaal query language VL 14 c) basic state-machines . . . d) an outlook on hierarchical state-machines 2 /66

Content • Proto-OCL • syntax, semantics, • Proto-OCL vs. OCL. • Proto-OCL vs. Software • An outlook on UML • Principles of (Good) Design • modularity, separation of concerns • information hiding and data encapsulation • abstract data types, object orientation • ...by example • Architecture Patterns • Layered Architectures, Pipe-Filter, Model-View-Controller. • Design Patterns – 12 – 2017-07-03 – Scontent – • Strategy, Examples • Libraries and Frameworks 3 /66

Partial vs. Complete Object Diagrams • By now we discussed “ object diagram represents system state ”: { 1 C 7 � { p 7 � � , n 7 � { 5 C }} , 5 C : C p n 1 C : C 1 D : D 5 C 7 � { p 7 � � , n 7 � � } , � p = � p = � x = 23 1 D 7 � { p 7 � { 5 C } , x 7 � 23 }} n = � What about the other way round...? • Object diagrams can be partial , e.g. n 1 C : C 5 C : C 1 D : D or 1 C : C 5 C : C 1 D : D x = 23 � we may omit information. • Is the following object diagram partial or complete ? 5 C : C p n 1 C : C 1 D : D p = � p = � x = 23 n = � • If an object diagram – 11 – 2017-06-26 – Sod – • has values for all attributes of all objects in the diagram, and – 12 – 2017-07-03 – main – • if we say that it is meant to be complete then we can uniquely reconstruct a system state � . 39 /51 4 /66

Special Case: Anonymous Objects If the object diagram : C p n 1 C : C : D p = � p = � x = 23 n = � is considered as complete , then it denotes the set of all system states { 1 C 7 � { p 7 � � , n 7 � { c }}} , c 7 � { p 7 � � , n 7 � � } , d 7 � { p 7 � { c } , x 7 � 23 }} where c � D ( C ) , d � D ( D ) , c 6 = 1 C . Intuition : different boxes represent different objects. – 11 – 2017-06-26 – Sod – – 12 – 2017-07-03 – main – 40 /51 5 /66

Content • Proto-OCL • syntax, semantics, • Proto-OCL vs. OCL. • Proto-OCL vs. Software • An outlook on UML • Principles of (Good) Design • modularity, separation of concerns • information hiding and data encapsulation • abstract data types, object orientation • ...by example • Architecture Patterns • Layered Architectures, Pipe-Filter, Model-View-Controller. • Design Patterns – 12 – 2017-07-03 – Scontent – • Strategy, Examples • Libraries and Frameworks 6 /66

Towards Object Constraint Logic (OCL) — “Proto-OCL” — – 12 – 2017-07-03 – main – 7 /66

Motivation C c 0,1 a D A 0,1 • How do I precisely, formally tell my developers that All D -instances having a link to the same C object should have links to the same A . • That is, the following system state is forbidden in the software: a c c a : A : D : C : D : A Note: formally, it is a proper system state . – 12 – 2017-07-03 – Socl – • Use (Proto-)OCL : “Dear developers, please only use system states which satisfy:” ∀ d 1 ∈ allInstances C • ∀ d 2 ∈ allInstances C • c ( d 1 ) = c ( d 2 ) = ⇒ a ( d 1 ) = a ( d 2 ) 8 /66

Constraints on System States C x : Int • Example : for all C -instances, x should never have the value 27 . ∀ c ∈ allInstances C • x ( c ) � = 27 • Proto-OCL Syntax wrt. signature ( T , C , V, atr , F, mth ) , c is a logical variable , C ∈ C : F ::= c : τ C : 2 τ C | allInstances C | v ( F ) : τ C → τ ⊥ , if v : τ ∈ atr ( C ) | v ( F ) : τ C → τ D , if v : D 0 , 1 ∈ atr ( C ) : τ C → 2 τ D , | v ( F ) if v : D ∗ ∈ atr ( C ) | f ( F 1 , . . . , F n ) : τ 1 × · · · × τ n → τ, if f : τ 1 × · · · × τ n → τ : τ C × 2 τ C × B ⊥ → B ⊥ | ∀ c ∈ F 1 • F 2 – 12 – 2017-07-03 – Socl – • The formula above in prefix normal form : ∀ c ∈ allInstances C • � = ( x ( c ) , 27) 9 /66

Semantics • Proto-OCL Types: I � 2 τ C � = D ( C ∗ ) ˙ • I � τ C � = D ( C ) ˙ I � τ ⊥ � = D ( τ ) ˙ ∪ {⊥} , ∪ {⊥} , ∪ {⊥} • I � B ⊥ � = { true , false } ˙ I � Z ⊥ � = Z ˙ ∪ {⊥} , ∪ {⊥} • Functions: • We assume f I given for each function symbol f ( → in a minute). • Proto-OCL Semantics (interpretation function): • I � c � ( σ, β ) = β ( c ) (assuming β is a type-consistent valuation of the logical variables), • I � allInstances C � ( σ, β ) = dom( σ ) ∩ D ( C ) , � σ ( I � F � ( σ, β )) ( v ) , if I � F � ( σ, β ) ∈ dom( σ ) • I � v ( F ) � ( σ, β ) = (if not v : C 0 , 1 ) ⊥ , otherwise � σ ( u ′ )( v ) , if I � F � ( σ, β ) = { u ′ } ⊆ dom( σ ) • I � v ( F ) � ( σ, β ) = (if v : C 0 , 1 ) ⊥ , otherwise • I � f ( F 1 , . . . , F n ) � ( σ, β ) = f I ( I � F 1 � ( σ, β ) , . . . , I � F n � ( σ, β )) , – 12 – 2017-07-03 – Socl –  true , if I � F 2 � ( σ, β [ c := u ]) = true for all u ∈ I � F 1 � ( σ, β )   • I � ∀ c ∈ F 1 • F 2 � ( σ, β ) = false , if I � F 2 � ( σ, β [ c := u ]) = false for some u ∈ I � F 1 � ( σ, β )  ⊥ , otherwise  10 /66

Semantics Cont’d • Proto-OCL is a three-valued logic: a formula evaluates to true , false , or ⊥ . • Example : ∧ I ( · , · ) : { true , false , ⊥} × { true , false , ⊥} → { true , false , ⊥} is defined as follows: x 1 true true true false false false ⊥ ⊥ ⊥ x 2 true false ⊥ true false ⊥ true false ⊥ ∧ I ( x 1 , x 2 ) true false ⊥ false false false ⊥ false ⊥ We assume common logical connectives ¬ , ∧ , ∨ , . . . with canonical 3-valued interpretation. • Example : + I ( · , · ) : ( Z ˙ ∪ {⊥} ) × ( Z ˙ ∪ {⊥} ) → Z ˙ ∪ {⊥} � x 1 + x 2 , if x 1 � = ⊥ and x 2 � = ⊥ + I ( x 1 , x 2 ) = ⊥ , otherwise We assume common arithmetic operations − , /, ∗ , . . . and relation symbols >, <, ≤ , . . . with monotone 3-valued interpretation. • And we assume the special unary function symbol isUndefined : – 12 – 2017-07-03 – Socl – � true , if x = ⊥ , isUndefined I ( x ) = false , otherwise isUndefined I is definite : it never yields ⊥ . 11 /66

Example: Evaluate Formula for System State C 1 C : C σ : x : Int x = 13 ∀ c ∈ allInstances C • x ( c ) � = 27 • Recall prefix notation : ∀ c ∈ allInstances C • � =( x ( c ) , 27) Note : � = is a binary function symbol, 27 is a 0 -ary function symbol. • Example : I � ∀ c ∈ allInstances C • � =( x ( c ) , 27) � ( σ, ∅ ) = true , because... I � � =( x ( c ) , 27) � ( σ, β ) , β := ∅ [ c := 1 C ] = { c �→ 1 C } = � = I ( I � x ( c ) � ( σ, β ) , I � 27 � ( σ, β ) ) = � = I ( σ ( I � c � ( σ, β ) )( x ) , 27 I ) = � = I ( σ ( β ( c ) )( x ) , 27 I ) – 12 – 2017-07-03 – Socl – = � = I ( σ ( 1 C )( x ) , 27 I ) = � = I ( 13 , 27 ) = true ...and 1 C is the only C -object in σ : I � allInstances C � ( σ, ∅ ) = { 1 C } . 12 /66

More Interesting Example C n 1 C : C | σ : x : Int n x = 13 0 .. 1 ∀ c : allInstances C • x ( n ( c )) � = 27 • Similar to the previous slide, we need the value of I � x ( n ( c )) � ( σ, β ) , β = { c �→ 1 C } • I � c � ( σ, β ) = β ( c ) = 1 C • I � n ( c ) � ( σ, β ) = ⊥ since σ ( I � c � ( σ, β ) )( n ) = ∅ � = { u ′ } by rule � u ′ , if I � F � ( σ, β ) ∈ dom( σ ) and σ ( I � F � ( σ, β ))( v ) = { u ′ } I � v ( F ) � ( σ, β ) = (if v : C 0 , 1 ) ⊥ , otherwise • I � x ( n ( c )) � ( σ, β ) = ⊥ since I � n ( c ) � ( σ, β ) = ⊥ by rule – 12 – 2017-07-03 – Socl – � σ ( I � F � ( σ, β )) ( v ) , if I � F � ( σ, β ) ∈ dom( σ ) I � v ( F ) � ( σ, β ) = (if not v : C 0 , 1 ) ⊥ , otherwise 13 /66

More Interesting Example C n 1 C : C | σ : x : Int n x = 13 0 .. 1 ∀ c : C • x ( n ( c )) � = 27 • Similar to the previous slide, we need the value of σ ( σ ( I � c � ( σ, β ) )( n ) ) ( x ) • I � c � ( σ, β ) = β ( c ) = 1 C • σ ( I � c � ( σ, β ) )( n ) = σ ( 1 C )( n ) = ∅ • σ ( σ ( I � c � ( σ, β ) )( n ) ) ( x ) = ⊥ by the following rule: – 12 – 2017-07-03 – Socl – � σ ( u ′ )( v ) , if I � F � ( σ, β ) = { u ′ } ⊆ dom( σ ) I � v ( F ) � ( σ, β ) = (if v : C 0 , 1 ) ⊥ , otherwise 13 /66

Object Constraint Language (OCL) OCL is the same — just with less readable (?) syntax. Literature: (OMG, 2006; Warmer and Kleppe, 1999). – 12 – 2017-07-03 – Socl – 14 /66