1 / 41

Centralized vs. Decentralized

Centralized vs. Decentralized. Interpreter Pattern Visitor Pattern. Defining a picture. Drawing. CompFig. CompFig. TextFig. TextFig. RectangleFig. RectangleFig. Defining a Language for Pictures. Figure - displayOn: is abstract method CompositeFigure Drawing RectangleFigure

kele
Download Presentation

Centralized vs. Decentralized

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Centralized vs. Decentralized Interpreter Pattern Visitor Pattern

  2. Defining a picture Drawing CompFig. CompFig. TextFig. TextFig. RectangleFig. RectangleFig.

  3. Defining a Language for Pictures • Figure - displayOn: is abstract method • CompositeFigure • Drawing • RectangleFigure • TextFigure

  4. Interpreter Pattern • Tree of objects is a kind of program • Interpret program by sending message to root, which is recursively sent to entire tree. • Not about parsing!!!

  5. Object-Oriented Interpreters • To write a little object-oriented interpreter for a language L: • 1) make a subclass of LParseNode for each rule in the grammar of L • 2) for each subclass, define an interpreter method that takes the current context as an argument.

  6. Object-Oriented Interpreters • 3) define protocol for making a tree of LParseNodes. • 4) define a program for L by building an abstract syntax tree.

  7. Example: ValueModels arguments ValueModel value BlockValue ValueHolder block value value value

  8. ValueModels BlockValue • | x y | • x := 3 asValue. • y := 7 asValue. • x + y • ValueModel>> + aValue • ^ BlockValue • block: [:a :b | a value * b value] • arguments: (Array with: self with: aValue asValue) ValueHolder ValueHolder

  9. Example: Regular Expression Checker • Grammar: exp ::= string | exp ‘+’ exp | exp ‘&’ exp | exp ‘repeat’ • Step 1: define RegExpNode, MatchRENode, AlternationRENode, SequenceRENode, RepeatRENode • Step 2: define a match: method for each class that takes an input state as an argument

  10. Regular Expression Classes RegExpNode Match MatchRENode RepeatRENode contents Match Match AlternationRENode SequenceRENode Match Match

  11. Regular Expression Objects 'this is a ' ('very ' *) ('big ’ + 'small ') ('dog' + 'cat' + 'mouse') SequenceRENode AlternationRENode MatchRENode 'this is a ' AlternationRENode RepeatRENode MatchRENode 'cat' MatchRENode 'small ' MatchRENode 'very ' MatchRENode 'big ' MatchRENode 'dog' MatchRENode 'mouse this is a very very very big mouse

  12. Matching • SequenceRENode>>match: inputState • ^components • inject: inputState • into: [:nextState :exp | • exp match: nextState]

  13. Matching • RepeatRENode>>match: inputState • | aState finalState | • aState := inputState. • finalState := inputState copy. • ...

  14. (continued) • [aState notEmpty] • whileTrue: • [aState := component match: aState. • finalState addAll: aState]. • ^finalState

  15. Matching • AlternationRENode>>match: inputState • | finalState | • finalState := REState new. • components do: [ :exp | finalState addAll: (exp match: inputState)] • ^finalState

  16. Matching • MatchRENode>>match: inputState • | finalState tStream | • finalState := REState new. • ... • REState is a collection of streams.

  17. (continued) • inputState • do: [:stream | • tStream := stream copy. • (tStream nextAvailable: • components size) = components • ifTrue: [finalState add: tStream]]. • ^finalState

  18. Protocol for Building Tree • Define ”+" and "repeat" and ”&" as methods on RegExpNodes and any literals that are to be treated as patterns, such as strings. • Then (('dog ' + 'cat ') repeat & 'weather') matches: 'dog dog cat weather' is true.

  19. Interface method • RegExpNode>>matches: anArg • | inputState | • inputState := (streamOrCollection isKindOf: Collection) • ifTrue: [REState with: (ReadStream on: streamOrCollection)] • ifFalse: [REState with: streamOrCollection]. • (self match: inputState) do: [:stream | stream atEnd ifTrue: [^true]]. • ^false

  20. Interface methods • Define +, &, repeat, and asRENode in RegExpNode and String • + anArg • ^AlternationRENode new • components: (Array with: self with: anArg asRENode)

  21. Other examples of Interpreter pattern: • • producing Postscript for a document • • figuring out the value of an insurance policy • • spreadsheet • • compiling a program • C program would use case statement

  22. Replacing Cases with Subclasses • Advantages • • instead of modifying case statements, add a new subclass • • easier to parameterize • • can use inheritance to make new options

  23. Replacing Cases with Subclasses • Disadvantages • • program is spread out, • + harder to understand • + harder to replace algorithm • • state of object can change, but class can not

  24. How To Centralize Algorithms • Define isAlternative, isRepeat, isSequence, isMatch, childrenDo: • fix: anRENode • anRENode isMatch • ifTrue: [anRENode contents: (anRENode contents capitalize)] • ifFalse: [anRENode childrenDo: [:child | self fix: child]]

  25. When to Centralize Algorithm • Use centralized algorithm when you need to • • change entire algorithm at once • • look at entire algorithm at once • • change algorithm, but not add new classes of components

  26. Visitor pattern • Visitor lets you centralize algorithm, lets you create a family of algorithms by inheritance, and makes it easy to create new algorithms. • Major problem is that adding a new kind of parse node requires adding a new method to each visitor.

  27. Visitor pattern • • two kinds of classes: nodes and node visitor • • nodes have accept: method with visitor as an argument • • accept: method sends visit message • SequenceRENode>>accept: aVisitor • aVisitor visitSequence: self

  28. Visitor pattern • • each node class sends a different visit message to visitor • visitSequence:, visitAlternation:, visitRepetition:, visitMatch: • • visitor defines a visit method for each class of parse tree node • • uses double dispatching

  29. REVisitor and MatchVisitor • An REVisitor implements visitSequence:, visitAlternation:, visitRepetition:, visitMatch: • MatchVisitor is an REVisitor with one instance variable, state, and methods to access it.

  30. MatchVisitor>>visitSequence: sequence • ^state := sequence components • inject: state • into: [:nextState :exp | • visitor := MatchVisitor with: nextState. • exp match: visitor. • visitor state]

  31. MatchVisitor>>visitAlternation: aNode • | initialState finalState | • initialState := state. • finalState := REState new. • components do: [ :eachComponent | • state := initialState. • eachComponent accept: self. • finalState addAll: state] • state := finalState

  32. MatchVisitor>>visitRepeat: aNode • | finalState | • finalState := state copy. • [state notEmpty] • whileTrue: • [component accept: self. • finalState addAll: state]. • ^finalState

  33. FixVisitor • FixVisitor>>visitMatch: aMatch • aMatch contents: (aMatch contents capitalize) • FixVisitor>>visitSequence: aSequence • aSequence • childrenDo: [:child | child accept: self]]

  34. Visitor as a refactoring • Convert an “interpret:” method to Visitor. • Precondition: interpret: must not return anything. • Create a visitor class V. • Find common superclass of the classes that define interpret:, and define • interpret: aContext • self accept: (V newWith: aContext)

  35. Visitor as a refactoring • Add an accept: method to every class X with interpret:, and have it send visitX: to argument. • Move code for interpret: in class X to visitX: of V. Add accessing methods to X. Change sends of interpret: to accept:

  36. Comments • Design pattern is a complex refactoring • Visitor illustrates that location of code is arbitrary

  37. Visitor Pattern • Strengths • Can add a new algorithm to operate on a class hierarchy without changing existing classes. • Weakness • Adding a new node class requires changing visitors.

  38. ParseTreeEnumerator • Smalltalk compiler parse tree hierachy rarely changes. • Many programs need to traverse tree: • • consistency checkers • • program transformation • • metrics • Keep them out of the parse tree.

  39. VisualWorks GUI Specs • A tree of Specs describes the components of a window. • You build a window for an application by asking the Spec for that window to do it. • UIPolicy is a Visitor that converts a tree of Specs into a tree of VisualComponents.

  40. dispatchTo: policy with: builder • ActionButtonSpec • policy actionButton: self into: builder • CompositeSpec • policy composite: self into: builder

  41. Summary • Two related patterns: • Interpreter - distribute code over class hierarchy • Visitor - centralize code in single class

More Related