Principles of

Programming
Languages
 CONTENTS

UNIT1.PRELIMINARYCONCEPTS
ReasonsforStudyingConceptsofProgrammingLanguages
ProgrammingDomains,LanguageEvaluationCriteria
InfluencesonLanguageDesign,LanguageCategories
ProgrammingParadigms-Imperative,FunctionalProgramminglanguage
LanguageImplementation-compilationandinterpretation
Programmingenvironments

SYNTAXANDSEMANTICS

TheGeneralProblemsofDescribingSyntaxandsemantics
BNF
EBNFforcommonprogramminglanguagefeatures
AmbiguousGrammar

UNIT2.DATATYPESANDVARIABLES
IntroductiontoPrimitiveDataTypes
User-DefinedDataTypesandArrayTypes
Record,UnionandPointerandReferencetypes
Names,Variables,TheConceptofBinding
TypeChecking,TypeCompatibility
StrongTypes,NamedConstants,variableinitialization

EXPRESSIONSANDSTATEMENTS&CONTROLSTRUCTURES

Expression:[Arithmetic,RelationalandBooleanExpressions]
Short-CircuitEvaluation
AssignmentStatements,Mixed-
ModeAssignment[ReferentialTransparency&FunctionalProgramming]
ControlStructures[statementLevelControlStructures,Compoundstat
ements]
SelectionStatements,IterativeStatements,UnconditionalBranching,Gu
ardedCommands

UNIT3.SUBPROGRAMSANDBLOCKS
FundamentalsofSubprograms
DesignIssuesforSubprograms
ParameterPassingMethods
ParametersSubprogramsasparameters
OverloadedSubprograms&Operations
GenericSubprograms,separatelycompiledmodules
Co-Routines

1
UNITIV.ABSTRACTDATATYPES
Abstract-datatypes[Abstraction&Encapsulation]
IntroductiontoDataAbstraction,DesignIssues
LanguageExamples
C++ParameterizedAbstractDataTypes
DataTypes
Object-OrientedProgramminginSmalltalk
Object-OrientedProgramminginC++
Object-OrientedProgramminginJava
Object-OrientedProgramminginC#
Object-OrientedProgramminginAda95

EXCEPTIONHANDLING&LOGICPROGRAMMING

ExceptionHandling:Exceptions,ExceptionPropagation
ExceptionHandlerinAda
C++andJava
Logic Programming Language : Introduction An Overview of Logic
Programming
TheBasicElementsofPROLOG
ApplicationsofLogicProgramming

UNIT-V.FUNCTIONALPROGRAMMINGLANGUAGES&SCRIPTINGLANGUAGE

FunctionalProgrammingLanguageIntroduction
FundamentalsofFunctionalProgrammingLanguages,LISPProgramming
FundamentalsofML,examples
FundamentalsofHaskell,functionsyntaxandexamples
Applications of Functional Programing language and Comparison
ofFunctionalandImperativeLanguages

2
 UNIT-1
PreliminaryConcepts

Background
 Frankly,wedidn‘thavethevaguestideahowthething[FORTRANlanguageandcompil
er] would work out in detail. …We struck out simply to optimize the
objectprogram,therunningtime,becausemostpeopleatthattimebelievedyoucould
n‘tdothatkindofthing.Theybelievedthatmachined-
codedprogramswouldbesoinefficientthatitwouldbeimpracticalformanyapplicati
ons.
 JohnBackus,unexpectedsuccessesarecommon–
thebrowserisanotherexampleofanunexpectedsuccess

ReasonsforStudyingConceptsofProgrammingLanguages-CO1
 Increasedabilitytoexpressideas
 Improvedbackgroundforchoosingappropriatelanguages
 Increasedabilitytolearnnewlanguages
 Betterunderstandingofsignificanceofimplementation
 Overalladvancementofcomputing

ProgrammingDomains–CO1
 Scientificapplications
– Largenumberoffloatingpointcomputations
– Fortran
 Businessapplications
– Producereports,usedecimalnumbersandcharacters
– COBOL
 Artificialintelligence
– Symbolsratherthannumbersmanipulated
– LISP
 Systemsprogramming
– Needefficiencybecauseofcontinuoususe
– C
 WebSoftware
– Eclecticcollectionoflanguages:markup(e.g.,XHTML),scripting(e.g.,PHP),gener
al-purpose(e.g.,Java)

3
 LanguageEvaluationCriteria–CO1,CO2

 Readability:theeasewithwhichprogramscanbereadandunderstood
 Writability :theeasewithwhichalanguagecanbeusedtocreateprograms
 Reliability :conformancetospecifications(i.e.,performstoitsspecifications)
 Cost :theultimatetotalcost

Readability
 Overallsimplicity
– Amanageablesetoffeaturesandconstructs
– Fewfeaturemultiplicity(meansofdoingthesameoperation)
– Minimaloperatoroverloading
 Orthogonality
– Arelativelysmallsetofprimitiveconstructscanbecombinedinarelativelysmallnumb
erofways
– Everypossiblecombinationislegal
 Controlstatements
– Thepresenceofwell-knowncontrolstructures(e.g.,whilestatement)
 Datatypesandstructures
– Thepresenceofadequatefacilitiesfordefiningdatastructures
 Syntaxconsiderations
– Identifierforms:flexiblecomposition
– Specialwordsandmethodsofformingcompoundstatements
– Formandmeaning:self-descriptiveconstructs,meaningfulkeywords
Writability
 SimplicityandOrthogonality
– Fewconstructs,asmallnumberofprimitives,asmallsetofrulesforcombiningthem
 Supportforabstraction
– Theabilitytodefineandusecomplexstructuresoroperationsinwaysthatallowdetails
tobeignored
 Expressivity
– Asetofrelativelyconvenientwaysofspecifyingoperations
– Example:theinclusionofforstatementinmanymodernlanguages
Reliability
 Typechecking
– Testingfortypeerrors
 Exceptionhandling
– Interceptrun-timeerrorsandtakecorrectivemeasures
 Aliasing
– Presenceoftwoormoredistinctreferencingmethodsforthesamememorylocation
 Readabilityandwritability
– Alanguagethatdoesnotsupport“natural”waysofexpressinganalgorithmwillnecess
arilyuse“unnatural”approaches,andhencereducedreliability
Cost
 Trainingprogrammerstouselanguage
 Writingprograms(closenesstoparticularapplications)
 Compilingprograms
 Executingprograms

4
  Languageimplementationsystem:availabilityoffreecompilers
 Reliability:poorreliabilityleadstohighcosts
 Maintainingprograms
Others
 Portability
– Theeasewithwhichprogramscanbemovedfromoneimplementationtoanother
 Generality
– Theapplicabilitytoawiderangeofapplications
 Well-definedness
– Thecompletenessandprecisionofthelanguage‘sofficialdefinition

5
 InfluencesonLanguageDesign-CO3
 ComputerArchitecture
– Languagesaredevelopedaroundtheprevalentcomputerarchitecture,knownast
hevonNeumannarchitecture
 ProgrammingMethodologies
– Newsoftwaredevelopmentmethodologies(e.g.,object-
orientedsoftwaredevelopment)ledtonewprogrammingparadigmsandbyextensio
n,newprogramminglanguages

ComputerArchitecture
 Well-knowncomputerarchitecture:VonNeumann
 Imperativelanguages,mostdominant,becauseofvonNeumanncomputers
– Dataandprogramsstoredinmemory
– MemoryisseparatefromCPU
– InstructionsanddataarepipedfrommemorytoCPU
– Basisforimperativelanguages
 Variablesmodelmemorycells
 Assignmentstatementsmodelpiping
 Iterationisefficient

Figure1.1ThevonNeumannComputerArchitecture

ProgrammingMethodologies
 1950sandearly1960s:Simpleapplications;worryaboutmachineefficiency
 Late1960s:Peopleefficiencybecameimportant;readability,bettercontrolstructures
– structuredprogramming
– top-downdesignandstep-wiserefinement
 Late1970s:Process-orientedtodata-oriented
– dataabstraction
 Middle1980s:Object-orientedprogramming
– Dataabstraction+inheritance+polymorphism

6
 LanguageCategories–CO1
 Imperative
– Centralfeaturesarevariables,assignmentstatements,anditeration
– Examples:C,Pascal
 Functional
– Mainmeansofmakingcomputationsisbyapplyingfunctionstogivenparameters
– Examples:LISP,Scheme
 Logic
– Rule-based(rulesarespecifiedinnoparticularorder)
– Example:Prolog
 Object-oriented
– Dataabstraction,inheritance,latebinding
– Examples:Java,C++
 Markup
– New;notaprogrammingperse,butusedtospecifythelayoutofinformationinWebdoc
uments
– Examples:XHTML,XML

LanguageDesignTrade-Offs
 Reliabilityvs.costofexecution
– Conflictingcriteria
– Example:Javademandsallreferencestoarrayelementsbecheckedforproperindexin
gbutthatleadstoincreasedexecutioncosts
 Readabilityvs.writability
– Anotherconflictingcriteria
– Example:APLprovidesmanypowerfuloperators(andalargenumberofnewsymbols
),allowingcomplexcomputationstobewritteninacompactprogrambutatthecostofp
oorreadability
 Writability(flexibility)vs.reliability
– Anotherconflictingcriteria
– Example:C++pointersarepowerfulandveryflexiblebutnotreliablyused

ImplementationMethods-CO2
 Compilation
– Programsaretranslatedintomachinelanguage
 PureInterpretation
– Programsareinterpretedbyanotherprogramknownasaninterpreter
 HybridImplementationSystems
– Acompromisebetweencompilersandpureinterpreters
Compilation
 Translatehigh-levelprogram(sourcelanguage)intomachinecode(machinelanguage)
 Slowtranslation,fastexecution
 Compilationprocesshasseveralphases:
– lexicalanalysis:convertscharactersinthesourceprogramintolexicalunits
– syntaxanalysis:transformslexicalunitsintoparsetreeswhichrepresentthesynta
cticstructureofprogram
– Semanticsanalysis:generateintermediatecode
– codegeneration:machinecodeisgenerated

7
Figure1.2LayeredViewofComputer:Theoperatingsys
temandlanguageimplementationarelayeredoverMac
hineinterfaceofacomputer Figure1.3TheCompilationProcess

AdditionalCompilationTerminologies
 Loadmodule(executableimage):theuserandsystemcodetogether
 Linkingandloading:theprocessofcollectingsystemprogramandlinkingthemtouserpro
gram
ExecutionofMachineCode
 Fetch-execute-cycle(onavonNeumannarchitecture)
initializetheprogramcounterrep
eatforever
fetchtheinstructionpointedbythecounterincrementt
hecounter
decodetheinstructionexecute
theinstruction
endrepeat
VonNeumannBottleneck
 Connectionspeedbetweenacomputer‘smemoryanditsprocessordeterminesthespeed
ofacomputer
 Programinstructionsoftencanbeexecutedalotfasterthantheaboveconnectionspee
d;theconnectionspeedthusresultsinabottleneck
 KnownasvonNeumannbottleneck;itistheprimarylimitingfactorinthespeedofcompute
rs

PureInterpretation
 Notranslation
 Easierimplementationofprograms(run-
timeerrorscaneasilyandimmediatelydisplayed)
 Slowerexecution(10to100timesslowerthancompiledprograms)
 Oftenrequiresmorespace
 Becomingrareonhigh-levellanguages

8
  Significantly comeback with some latest web scripting languages
(e.g.,JavaScript)

HybridImplementationSystems
 Acompromisebetweencompilersandpureinterpreters
 Ahigh-
levellanguageprogramistranslatedtoanintermediatelanguagethatallowseasyinterpr
etation
 Fasterthanpureinterpretation
 Examples
– Perlprogramsarepartiallycompiledtodetecterrorsbeforeinterpretation
– InitialimplementationsofJavawerehybrid;theintermediateform,bytecode,provide
sportabilitytoanymachinethathasabytecodeinterpreterandaruntimesystem(toge
ther,thesearecalledJavaVirtualMachine)

Figure1.4PureInterpreataion Figure1.5HybridImplementation

Just-in-TimeImplementationSystems
 Initiallytranslateprogramstoanintermediatelanguage
 Thencompileintermediatelanguageintomachinecode
 Machinecodeversioniskeptforsubsequentcalls
 JITsystemsarewidelyusedforJavaprograms
 .NETlanguagesareimplementedwithaJITsystem

Preprocessors-CO2
 Preprocessormacros(instructions)arecommonlyusedtospecifythatcodefromanother
fileistobeincluded
 Apreprocessorprocessesaprogramimmediatelybeforetheprogramiscompiledtoexpa
ndembeddedpreprocessormacros
 Awell-knownexample:Cpreprocessor
– expands#include,#define,andsimilarmacros

9
 SyntaxandSemantics

Introduction
 Syntax:theformorstructureoftheexpressions,statements,andprogramunits
 Semantics:themeaningoftheexpressions,statements,andprogramunits
 Syntaxandsemanticsprovidealanguage‘sdefinition
– Usersofalanguagedefinition
– Otherlanguagedesigners
– Implementers
– Programmers(theusersofthelanguage)

TheGeneralProblemofDescribingSyntax–CO1
 Asentenceisastringofcharactersoversomealphabet
 Alanguageisasetofsentences
 Alexemeisthelowestlevelsyntacticunitofalanguage(e.g.,*,sum,begin)
 Atokenisacategoryoflexemes(e.g.,identifier)
 LanguagesRecognizers
– Arecognitiondevicereadsinputstringsofthelanguageanddecideswhethertheinputs
tringsbelongtothelanguage
– Example:syntaxanalysispartofacompiler
 LanguagesGenerators
– Adevicethatgeneratessentencesofalanguage
– Onecandetermineifthesyntaxofaparticularsentenceiscorrectbycomparingittothes
tructureofthegenerator

FormalMethodsofDescribingSyntax–CO1,CO2
 Backus-NaurFormandContext-FreeGrammars
– Mostwidelyknownmethodfordescribingprogramminglanguagesyntax
 ExtendedBNF
– ImprovesreadabilityandwritabilityofBNF
 GrammarsandRecognizers
Backus-NaurFormandContext-FreeGrammars
 Context-FreeGrammars
 DevelopedbyNoamChomskyinthemid-1950s
 Languagegenerators,meanttodescribethesyntaxofnaturallanguages
 Defineaclassoflanguagescalledcontext-freelanguages
Backus-NaurForm(BNF)
 Backus-NaurForm(1959)
– InventedbyJohnBackustodescribeALGOL58
– BNFisequivalenttocontext-freegrammars
– BNFisametalanguageusedtodescribeanotherlanguage
– InBNF,abstractionsareusedtorepresentclassesofsyntacticstructures--
theyactlikesyntacticvariables(alsocallednonterminalsymbols)
BNFFundamentals
 Non-terminals:BNFabstractions
 Terminals:lexemesandtokens
 Grammar:acollectionofrules
– ExamplesofBNFrules:

10
 <ident_list>→identifier|identifer,<ident_list>
<if_stmt>→if<logic_expr>then<stmt>
BNFRules
 Arulehasaleft-handside(LHS)andaright-handside(RHS),andconsistsof
terminalandnonterminalsymbols
 Agrammarisafinitenonemptysetofrules
 Anabstraction(ornonterminalsymbol)canhavemorethanoneRHS
<stmt>→<single_stmt>
|begin<stmt_list>end
DescribingLists
 Syntacticlistsaredescribedusingrecursion
<ident_list>→ident
|ident,<ident_list>
 Aderivationisarepeatedapplicationofrules,startingwiththestartsymbolandendin
gwithasentence(allterminalsymbols)
AnExampleGrammar
<program>→<stmts>
<stmts>→<stmt>|<stmt>;<stmts>
<stmt> → <var> =<expr>
<var>→ a|b|c|d
<expr>→<term>+<term>|<term> -<term>
<term>→<var>|const

ParseTree
Ahierarchicalrepresentationofaderivation
Anexamplederivation Figure2.1ParseTree
<program><stmts>
 <stmt>
 <var>=<expr>
 a=<expr>
 a=<term>+<term>
 a=<var>+<term>
 a=b+<term>
 a=b+const

Derivation
 Everystringofsymbolsinthederivationisasententialform
 Asentenceisasententialformthathasonlyterminalsymbols
 Aleftmostderivationisoneinwhichtheleftmostnonterminalineachsententialformisthe
onethatisexpanded
 Aderivationmaybeneitherleftmostnorrightmost

AmbiguityinGrammars
 Agrammarisambiguousiffitgeneratesasententialformthathastwoormoredistinctpars
etrees
AnUnambiguousExpressionGrammar
Ifweusetheparsetreetoindicateprecedencelevelsoftheoperators,wecannothaveambig
uity
<expr>→<expr>-<term>|<term>
<term>→<term>/const|const

11
 Figure2.2AnAmbiguousExpressionGrammar Figure2.3AnUnambiguousExpressionGr
ammar

AssociativityofOperators Figure2.4ParseTreeforAs
sociativityoperator
Operatorassociativitycanalsobeindicatedbyagra
mmar
<expr>→<expr>+<expr>|const(ambiguous)
<expr>→<expr>+const|const(unambiguous)

ExtendedBackus-NaurForm(EBNF)–CO2
 Optionalpartsareplacedinbrackets([])
<proc_call>→ident[(<expr_list>)]
 AlternativepartsofRHSsareplacedinsideparenthesesandseparatedviaverticalbar
s
<term>→<term>(+|-)const
 Repetitions(0ormore)areplacedinsidebraces({})
<ident>→letter{letter|digit}
BNFandEBNF
 BNF
<expr> → <expr>+<term>
|<expr>-<term>
|<term>
<term>→<term>*<factor>
|<term>/<factor>
|<factor>
 EBNF
<expr>→<term>{(+|-)<term>}
<term>→<factor>{(*|/)<factor>}

AttributeGrammars–CO1,CO4
 Context-
freegrammars(CFGs)cannotdescribeallofthesyntaxofprogramminglanguages
 AdditionstoCFGstocarrysomesemanticinfoalongparsetrees
 Primaryvalueofattributegrammars(AGs):
– Staticsemanticsspecification
– Compilerdesign(staticsemanticschecking)

Definition
 Anattributegrammarisacontext-
freegrammarG=(S,N,T,P)withthefollowingadditions:
– ForeachgrammarsymbolxthereisasetA(x)ofattributevalues

12
 – Eachrulehasasetoffunctionsthatdefinecertainattributesofthenonterminalsint
herule
– Eachrulehasa(possiblyempty)setofpredicatestocheckforattributeconsistency
– LetX0X1...Xnbearule
– FunctionsoftheformS(X0)=f(A(X1),...,A(Xn))definesynthesizedattributes
– FunctionsoftheformI(Xj)=f(A(X0),...,A(Xn)),fori<=j<=n,define
inheritedattributes
– Initially,thereareintrinsicattributesontheleaves

Example
 Syntax
<assign>→<var>=<expr>
<expr>→<var>+<var>|<var>
<var>→A|B|C
 actual_type:synthesizedfor<var>and<expr>
 expected_type:inheritedfor<expr>
 Syntaxrule :<expr>→<var>[1]+<var>[2]
Semanticrules :<expr>.actual_type→<var>[1].actual_type
Predicate :<var>[1].actual_type==<var>[2].actual_type
<expr>.expected_type==<expr>.actual_type
 Syntaxrule :<var>→id
Semanticrule :<var>.actual_typelookup(<var>.string)
 Howareattributevaluescomputed?
– Ifallattributeswereinherited,thetreecouldbedecoratedintop-downorder.
– Ifallattributesweresynthesized,thetreecouldbedecoratedinbottom-uporder.
– Inmanycases,bothkindsofattributesareused,anditissomecombinationoftop-
downandbottom-upthatmustbeused.
<expr>.expected_typeinheritedfromparent
<var>[1].actual_typelookup(A)
<var>[2].actual_typelookup(B)
<var>[1].actual_type=?<var>[2].actual_type
<expr>.actual_type<var>[1].actual_type
<expr>.actual_type=?<expr>.expected_type

DescribingtheMeaningsofPrograms:DynamicSemantics
 Thereisnosinglewidelyacceptablenotationorformalismfordescribingsemantics
 OperationalSemantics
– Describethemeaningofaprogrambyexecutingitsstatementsonamachine,either
simulatedoractual.Thechangeinthestateofthemachine(memory,registers,etc.)
definesthemeaningofthestatement
 Touseoperationalsemanticsforahigh-levellanguage,avirtualmachineisneeded
 Ahardwarepureinterpreterwouldbetooexpensive
 Asoftwarepureinterpreteralsohasproblems:
– Thedetailedcharacteristicsoftheparticularcomputerwouldmakeactionsdifficu
lttounderstand
– Suchasemanticdefinitionwouldbemachine-dependent

13
OperationalSemantics
 Abetteralternative:Acompletecomputersimulation
 Theprocess:
– Buildatranslator(translatessourcecodetothemachinecodeofanidealizedcomputer
)
– Buildasimulatorfortheidealizedcomputer
 Evaluationofoperationalsemantics:
– Goodifusedinformally(languagemanuals,etc.)
– Extremelycomplexifusedformally(e.g.,VDL),itwasusedfordescribingsemanticsofP
L/I.
 AxiomaticSemantics
– Basedonformallogic(predicatecalculus)
– Originalpurpose:formalprogramverification
– Approach:Defineaxiomsorinferencerulesforeachstatementtypeinthelanguage(to
allowtransformationsofexpressionstootherexpressions)
– Theexpressionsarecalledassertions

AxiomaticSemantics
 Anassertionbeforeastatement(aprecondition)statestherelationshipsandconstrai
ntsamongvariablesthataretrueatthatpointinexecution
 Anassertionfollowingastatementisapostcondition
 Aweakestpreconditionistheleastrestrictivepreconditionthatwillguaranteethepostco
ndition
 Pre-postform:{P}statement{Q}
 Anexample:a=b+1{a>1}
 Onepossibleprecondition:{b>10}
 Weakestprecondition:{b>0}
 Programproofprocess:Thepostconditionforthewholeprogramisthedesiredresult.Wo
rkbackthroughtheprogramtothefirststatement.Ifthepreconditiononthefirststatemen
tisthesameastheprogramspec,theprogramiscorrect.
 Anaxiomforassignmentstatements(x=
E):
{Qx->E}x=E{Q}
 Aninferenceruleforsequences
– ForasequenceS1;S2:
– {P1}S1{P2}
– {P2}S2{P3}
 AninferenceruleforlogicalpretestloopsFort
heloopconstruct:
{P}whileBdoSend{Q}Characteristic
softheloopinvariantImustmeetthefoll
owingconditions:
– P=>I(theloopinvariantmustbetrueinitially)
– {I}B{I}(evaluationoftheBooleanmustnotchangethevalidityofI)
– {IandB}S{I}(Iisnotchangedbyexecutingthebodyoftheloop)
– (Iand(notB))=>Q(ifIistrueandBisfalse,Qisimplied)
– Theloopterminates(thiscanbedifficulttoprove)
 TheloopinvariantIisaweakenedversionofthelooppostcondition,anditisalsoaprec
ondition.

14
  Imustbeweakenoughtobesatisfiedpriortothebeginningoftheloop,butwhencombi
nedwiththeloopexitcondition,itmustbestrongenoughtoforcethetruthofthepostco
ndition.

EvaluationofAxiomaticSemantics:
– Developingaxiomsorinferencerulesforallofthestatementsinalanguageisdifficult
– Itisagoodtoolforcorrectnessproofs,andanexcellentframeworkforreasoningabout
programs,butitisnotasusefulforlanguageusersandcompilerwriters
– Itsusefulnessindescribingthemeaningofaprogramminglanguageislimitedforlangu
ageusersorcompilerwriters

DenotationalSemantics
– Basedonrecursivefunctiontheory
– Themostabstractsemanticsdescriptionmethod
– OriginallydevelopedbyScottandStrachey(1970)
– Theprocessofbuildingadenotationalspecforalanguage(notnecessarilyeasy):
– Defineamathematicalobjectforeachlanguageentity
– Defineafunctionthatmapsinstancesofthelanguageentitiesontoinstancesofthecorr
espondingmathematicalobjects
– Themeaningoflanguageconstructsaredefinedbyonlythevaluesoftheprogram'svari
ables
– Thedifferencebetweendenotationalandoperationalsemantics:Inoperationalsema
ntics,thestatechangesaredefinedbycodedalgorithms;indenotationalsemantics,the
yaredefinedbyrigorousmathematicalfunctions
– Thestateofaprogramisthevaluesofallitscurrentvariables
s={<i1,v1>,<i2,v2>,…,<in,vn>}
– LetVARMAPbeafunctionthat,whengivenavariablenameandastate,returnsthecur
rentvalueofthevariable
VARMAP(ij,s)=vj
 DecimalNumbers
– Thefollowingdenotationalsemanticsdescriptionmapsdecimalnumbersasstringsof
symbolsintonumericvalues
<dec_num>→0|1|2|3|4|5|6|7|8|9
 |<dec_num>(0|1|2|3|4|5|6|7|8|9)
Mdec('0')=0,Mdec('1')=1,…,Mdec('9')=9
Mdec(<dec_num>'0')=10*Mdec(<dec_num>)Mdec(<d
ec_num>'1’)=10*Mdec(<dec_num>)+1
…
Mdec(<dec_num>'9')=10*Mdec(<dec_num>)+9

Expressions
 MapexpressionsontoZ{error}
 Weassumeexpressionsaredecimalnumbers,variables,orbinaryexpressionshavingon
earithmeticoperatorandtwooperands,eachofwhichcanbeanexpression
 AssignmentStatements
– Mapsstatesetstostatesets
 LogicalPretestLoops
– Mapsstatesetstostatesets

15
  Themeaningoftheloopisthevalueoftheprogramvariablesafterthestatementsintheloo
phavebeenexecutedtheprescribednumberoftimes,assumingtherehavebeennoerrors
 Inessence,theloophasbeenconvertedfromiterationtorecursion,wheretherecursiveco
ntrolismathematicallydefinedbyotherrecursivestatemappingfunctions
 Recursion,whencomparedtoiteration,iseasiertodescribewithmathematicalrigor
 Evaluationofdenotationalsemantics
– Canbeusedtoprovethecorrectnessofprograms
– Providesarigorouswaytothinkaboutprograms
– Canbeanaidtolanguagedesign
– Hasbeenusedincompilergenerationsystems
– Becauseofitscomplexity,theyareoflittleusetolanguageusers

Summary
 BNFandcontext-freegrammarsareequivalentmeta-languages
– Well-suitedfordescribingthesyntaxofprogramminglanguages
 Anattributegrammarisadescriptiveformalismthatcandescribeboththesynta
xandthesemanticsofalanguage
 Threeprimarymethodsofsemanticsdescription
– Operation,axiomatic,denotational

16
 UNIT-2
DataTypesandVariables

Introduction
 Adatatypedefinesacollectionofdataobjectsandasetofpredefinedoperationsonthoseob
jects
 Adescriptoristhecollectionoftheattributesofavariable
 Anobjectrepresentsaninstanceofauser-defined(abstractdata)type
 Onedesignissueforalldatatypes:Whatoperationsaredefinedandhowaretheyspecified
?

PrimitiveDataTypes–CO2
 Almostallprogramminglanguagesprovideasetofprimitivedatatypes
 Primitivedatatypes:Thosenotdefinedintermsofotherdatatypes
 Someprimitivedatatypesaremerelyreflectionsofthehardware
 Othersrequireonlyalittlenon-hardwaresupportfortheirimplementation
Integer
 Almostalwaysanexactreflectionofthehardwaresothemappingistrivial
 Theremaybeasmanyaseightdifferentintegertypesinalanguage
 Java‘ssignedintegersizes:byte,short,int,long
FloatingPoint
 Modelrealnumbers,butonlyasapproximations
 Languagesforscientificusesupportatleasttwofloating-
pointtypes(e.g.,floatanddouble;sometimesmore
 Usuallyexactlylikethehardware,butnotalways
 IEEEFloating-
PointStandard7
54
Complex
 Somelanguagessupportacomplextype,e.g.,FortranandPython
 Eachvalueconsistsoftwofloats,therealpartandtheimaginarypart
 Literalform(inPython):
(7+3j),where7istherealpartand3istheimaginarypart
Decimal
 Forbusinessapplications(money)
– EssentialtoCOBOL
– C#offersadecimaldatatype
 Storeafixednumberofdecimaldigits,incodedform(BCD)
 Advantage:accuracy
 Disadvantages:limitedrange,wastesmemory
Boolean
 Simplestofall
 Rangeofvalues:twoelements,onefor“true”andonefor“false”
 Couldbeimplementedasbits,butoftenasbytes
– Advantage:readability
Character
 Storedasnumericcoding
 Mostcommonlyusedcoding:ASCII
17
 Analternative,16-bitcoding:Unicode

18
 – Includescharactersfrommostnaturallanguages
– OriginallyusedinJava
– C#andJavaScriptalsosupportUnicode

CharacterStringTypes
 Valuesaresequencesofcharacters
 Designissues:
– Isitaprimitivetypeorjustaspecialkindofarray?
– Shouldthelengthofstringsbestaticordynamic?
Operations
 Typicaloperations:
– Assignmentandcopying
– Comparison(=,>,etc.)
– Catenation
– Substringreference
– Patternmatching
CharacterStringTypeinCertainLanguages
 CandC++
– Notprimitive
– Usechararraysandalibraryoffunctionsthatprovideoperations
 SNOBOL4(astringmanipulationlanguage)
 Primitive
 Manyoperations,includingelaboratepatternmatching
 FortranandPython
– Primitivetypewithassignmentandseveraloperations
 Java
– PrimitiveviatheStringclass
 Perl,JavaScript,Ruby,andPHP
– Providebuilt-inpatternmatching,usingregularexpressions
CharacterStringLengthOptions
 Static:COBOL,Java‘sStringclass
 LimitedDynamicLength:CandC++
– Intheselanguages,aspecialcharacterisusedtoindicatetheendofastring‘scharact
ers,ratherthanmaintainingthelength
 Dynamic(nomaximum):SNOBOL4,Perl,JavaScript
 Adasupportsallthreestringlengthoptions
Evaluation
 Aidtowritability
 Asaprimitivetypewithstaticlength,theyareinexpensivetoprovide—
whynothavethem?
 Dynamiclengthisnice,butisitworththeexpense?
Implementation
 Staticlength:compile-timedescriptor
 Limiteddynamiclength:mayneedarun-timedescriptorforlength(butnotinCandC++)
 Dynamiclength:needrun-timedescriptor;allocation/de-
allocationisthebiggestimplementationproblem

Figure3.1Compile-TimeDescriptor Figure3.2Run-TimeDescriptors

19
 User-DefinedOrdinalTypes-CO2
 Anordinaltypeisoneinwhichtherangeofpossiblevaluescanbeeasilyassociatedwith
thesetofpositiveintegers
 ExamplesofprimitiveordinaltypesinJava
– integer
– char
– boolean

EnumerationTypes
 Allpossiblevalues,whicharenamedconstants,areprovidedinthedefinition
 C#example
enumdays{mon,tue,wed,thu,fri,sat,sun};
 Designissues
– Isanenumerationconstantallowedtoappearinmorethanonetypedefinition,andifso,
howisthetypeofanoccurrenceofthatconstantchecked?
– Areenumerationvaluescoercedtointeger?
– Anyothertypecoercedtoanenumerationtype?

EvaluationofEnumeratedType
 Aidtoreadability,e.g.,noneedtocodeacolorasanumber
 Aidtoreliability,e.g.,compilercancheck:
– Operations(don‘tallowcolorstobeadded)
– Noenumerationvariablecanbeassignedavalueoutsideitsdefinedrange
– Ada,C#,andJava5.0providebettersupportforenumerationthanC++becauseenumer
ationtypevariablesintheselanguagesarenotcoercedintointegertypes

SubrangeTypes
 Anorderedcontiguoussubsequenceofanordinaltype
– Example:12..18isasubrangeofintegertype
 Ada‘sdesign
typeDaysis(mon,tue,wed,thu,fri,sat,sun);subty
peWeekdaysisDaysrangemon..fri;subtypeInde
xisIntegerrange1..100;

Day1:Days;Day
2:Weekday;Day
2:=Day1;

SubrangeEvaluation
 Aidtoreadability
– Makeitcleartothereadersthatvariablesofsubrangecanstoreonlycertainrangeofval
ues
 Reliability
– Assigningavaluetoasubrangevariablethatisoutsidethespecifiedrangeisdetecte
dasanerror

Implementation
 Enumerationtypesareimplementedasintegers
 Subrangetypesareimplementedliketheparenttypeswithcodeinserted(bythecompiler
)torestrictassignmentstosubrangevariables

20
ArrayTypes
 Anarrayisanaggregateofhomogeneousdataelementsinwhichanindividualelementisid
entifiedbyitspositionintheaggregate,relativetothefirstelement.
ArrayDesignIssues
 Whattypesarelegalforsubscripts?
 Aresubscriptingexpressionsinelementreferencesrangechecked?
 Whenaresubscriptrangesbound?
 Whendoesallocationtakeplace?
 Whatisthemaximumnumberofsubscripts?
 Canarrayobjectsbeinitialized?
 Areanykindofslicessupported?
ArrayIndexing
 Indexing(orsubscripting)isamappingfromindicestoelements
array_name(index_value_list)anelement
 IndexSyntax
– FORTRAN,PL/I,Adauseparentheses
 Adaexplicitlyusesparenthesestoshowuniformitybetweenarrayreferencesandfunctio
ncallsbecausebotharemappings
– Mostotherlanguagesusebrackets
ArraysIndex(Subscript)Types
 FORTRAN,C:integeronly
 Ada:integerorenumeration(includesBooleanandchar)
 Java:integertypesonly
 Indexrangechecking
– C,C++,Perl,andFortrandonotspecifyrangechecking
– Java,ML,C#specifyrangechecking
– InAda,thedefaultistorequirerangechecking,butitcanbeturnedoff
SubscriptBindingandArrayCategories
 Static:subscriptrangesarestaticallyboundandstorageallocationisstatic(beforerun-
time)
– Advantage:efficiency(nodynamicallocation)
 Fixedstack-
dynamic:subscriptrangesarestaticallybound,buttheallocationisdoneatdeclaratio
ntime
– Advantage:spaceefficiency
 Stack-dynamic:subscriptrangesare
dynamicallyboundandthestoragealloca
tionisdynamic(doneatrun-time)
– Advantage:flexibility(thesizeofanarrayneednotbeknownuntilthearrayistobeused
)
 Fixedheap-dynamic:similartofixedstack-
dynamic:storagebindingisdynamicbutfixedafterallocation(i.e.,bindingisdonewhenre
questedandstorageisallocatedfromheap,notstack)
 Heap-
dynamic:bindingofsubscriptrangesandstorageallocationisdynamicandcanchang
eanynumberoftimes
– Advantage:flexibility(arrayscangroworshrinkduringprogramexecution)
 CandC++arraysthatincludestaticmodifierarestatic
 CandC++arrayswithoutstaticmodifierarefixedstack-dynamic
 CandC++providefixedheap-dynamicarrays
 C#includesasecondarrayclassArrayListthatprovidesfixedheap-dynamic
21
 Perl,JavaScript,Python,andRubysupportheap-dynamicarrays

22
ArrayInitialization
 Somelanguageallowinitializationatthetimeofstorageallocation
– C,C++,Java,C#example
intlist[]={4,5,7,83}
– CharacterstringsinCandC++
charname[]=“freddie”;
– ArraysofstringsinCandC++
char*names[]={“Bob”,“Jake”,“Joe”};
– JavainitializationofStringobjects
String[]names={“Bob”,“Jake”,“Joe”};
HeterogeneousArrays
 Aheterogeneousarrayisoneinwhichtheelementsneednotbeofthesametype
 SupportedbyPerl,Python,JavaScript,andRuby
ArraysOperations
 APLprovidesthemostpowerfularrayprocessingoperationsforvectorsandmatrixesas
wellasunaryoperators(forexample,toreversecolumnelements)
 Adaallowsarrayassignmentbutalsocatenation
 Python‘sarrayassignments,buttheyareonlyreferencechanges.Pythonalsosupport
sarraycatenationandelementmembershipoperations
 Rubyalsoprovidesarraycatenation
 Fortranprovideselementaloperationsbecausetheyarebetweenpairsofarrayelements
– Forexample,+operatorbetweentwoarraysresultsinanarrayofthesumsoftheeleme
ntpairsofthetwoarrays
RectangularandJaggedArrays
 Arectangulararrayisamulti-
dimensionedarrayinwhichalloftherowshavethesamenumberofelementsandallcolum
nshavethesamenumberofelements
 Ajaggedmatrixhasrowswithvaryingnumberofelements
– Possiblewhenmulti-dimensionedarraysactuallyappearasarraysofarrays
 C,C++,andJavasupportjaggedarrays
 Fortran,Ada,andC#supportrectangulararrays(C#alsosupportsjaggedarrays)
Slices
 Asliceissomesubstructureofanarray;nothingmorethanareferencingmechanism
 Slicesareonlyusefulinlanguagesthathavearrayoperations

SliceExamples Figure3.3SlicesExamplesinFortran95
– Fortran95
Integer,Dimension(10)::VectorInteg
er,Dimension(3,3)::Mat
Integer,Dimension(3,3)::Cube

Vector(3:6)isafourelementarray

ImplementationofArrays
 Accessfunctionmapssubscriptexpressionstoanaddressinthearray
 Accessfunctionforsingle-dimensionedarrays:
address(list[k])=address(list[lower_bound])+((k-lower_bound)*element_size)

23
AccessingMulti-dimensionedArrays Figure3.4LocatinganElementina
Multi-dimensionedArray
 Twocommonways:
– Rowmajororder(byrows)–usedinmostlanguages
– Columnmajororder(bycolumns)–usedinFortran

Compile-TimeDescriptors

Figure3.5Single-DimensionedArray Figure3.6Multi-DimensionedArray

AssociativeArrays
 Anassociativearrayisanunorderedcollectionofdataelementsthatareindexedbyane
qualnumberofvaluescalledkeys
– User-definedkeysmustbestored
 Designissues:
– Whatistheformofreferencestoelements?
– Isthesizestaticordynamic?
AssociativeArraysinPerl
 Namesbeginwith%;literalsaredelimitedbyparentheses
%hi_temps=("Mon"=>77,"Tue"=>79,―Wed ‖=>65,…);
 Subscriptingisdoneusingbracesandkeys
$hi_temps{"Wed"}=83;
 Elementscanberemovedwithdelete
delete$hi_temps{"Tue"};

RecordTypes–CO2
 Arecordisapossiblyheterogeneousaggregateofdataelementsinwhichtheindividualele
mentsareidentifiedbynames
 Designissues:
– Whatisthesyntacticformofreferencestothefield?
– Areellipticalreferencesallowed?
DefinitionofRecordsinCOBOL
 COBOLuseslevelnumberstoshownestedrecords;othersuserecursivedefinition
01EMP-REC.
02EMP-NAME.
05FIRSTPICX(20).
05MIDPICX(10).
05LASTPICX(20).
02HOURLY-RATEPIC99V99.

24
DefinitionofRecordsinAda
 Recordstructuresareindicatedinanorthogonalway
typeEmp_Rec_TypeisrecordFirs
t:String(1..20);
Mid:String(1..10);
Last:String(1..20);Hourly_Ra
te:Float;
endrecord;
Emp_Rec:Emp_Rec_Type;
ReferencestoRecords
 Recordfieldreferences
– COBOL
field_nameOFrecord_name_1OF...OFrecord_name_n
– Others(dotnotation)
record_name_1.record_name_2 ....... record_name_n.field_name
 Fullyqualifiedreferencesmustincludeallrecordnames
 Ellipticalreferencesallowleavingoutrecordnamesaslongasthereferenceisunambi
guous,forexampleinCOBOL
FIRST, FIRST OF EMP-NAME, and FIRST of EMP-REC are
ellipticalreferencestotheemployee‘sfirstname
OperationsonRecords
 Assignmentisverycommonifthetypesareidentical
 Adaallowsrecordcomparison
 Adarecordscanbeinitializedwithaggregateliterals
 COBOLprovidesMOVECORRESPONDING
– Copiesafieldofthesourcerecordtothecorrespondingfieldinthetargetrecord
EvaluationandComparisontoArrays
 Recordsareusedwhencollectionofdatavaluesisheterogen
eous
 Accesstoarrayelementsismuchslowerthanaccesstorec
ordfields,becausesubscriptsaredynamic(fieldnamesar
estatic)
 Dynamicsubscriptscouldbeusedwithrecordfieldaccess
,butitwoulddisallowtypecheckinganditwouldbemuchs
lower
Figure3.7ImplementationofRecordType

UnionsTypes
 Aunionisatypewhosevariablesareallowedtostoredifferenttypevaluesatdifferenttime
sduringexecution
 Designissues
– Shouldtypecheckingberequired?
– Shouldunionsbeembeddedinrecords?

Discriminatedvs.FreeUnions
 Fortran,C,andC++provideunionconstructsinwhichthereisnolanguagesupportfort
ypechecking;theunionintheselanguagesiscalledfreeunion
 Typecheckingofunionsrequirethateachunionincludeatypeindicatorcalledadiscrimina
nt
– SupportedbyAda

25
 AdaUnionTypestypeS AdaUnionTypeIllustrated
hapeis(Circle,Triangle,Rectangle);typeColor
sis(Red,Green,Blue);
typeFigure(Form:Shape)isrecordFilled:Bo
olean;
Color:Colors;case
Formis
whenCircle=>Diameter:Float;
whenTriangle=>Leftside,Rightside:Integer;A
ngle:Float;
whenRectangle=>Side1,Side2:Integer;endca
se;
endrecord; Figure3.8ADiscriminatedUnionofThree
ShapeVariables
EvaluationofUnions
 Freeunionsareunsafe
– Donotallowtypechecking
 JavaandC#donotsupportunions
– Reflectiveofgrowingconcernsforsafetyinprogramminglanguage
 Ada‘sdescriminatedunionsaresafe

PointerandReferenceTypes
 Apointertypevariablehasarangeofvaluesthatconsistsofmemoryaddressesandasp
ecialvalue,nil
 Providethepowerofindirectaddressing
 Provideawaytomanagedynamicmemory
 Apointercanbeusedtoaccessalocationintheareawherestorageisdynamicallycreat
ed(usuallycalledaheap)
DesignIssuesofPointers
 Whatarethescopeofandlifetimeofapointervariable?
 Whatisthelifetimeofaheap-dynamicvariable?
 Arepointersrestrictedastothetypeofvaluetowhichtheycanpoint?
 Arepointersusedfordynamicstoragemanagement,indirectaddressing,orboth?
 Shouldthelanguagesupportpointertypes,referencetypes,orboth?
PointerOperations
 Twofundamentaloperations:assignmentanddereferencing
 Assignmentisusedtosetapointervariable‘svaluetosomeusefuladdress
 Dereferencingyieldsthevaluestoredatthelocationrepresentedbythepointer‘svalue
– Dereferencingcanbeexplicitorimplicit
– C++usesanexplicitoperationvia*
j=*ptr
setsjtothevaluelocatedatptr
PointerAssignmentIllustration

Figure3.9Theassignmentoperationj=*ptr

26
ProblemswithPointers
 Danglingpointers(dangerous)
– Apointerpointstoaheap-dynamicvariablethathasbeendeallocated
 Lostheap-dynamicvariable
– Anallocatedheap-
dynamicvariablethatisnolongeraccessibletotheuserprogram(oftencalledgarb
age)
 Pointerp1issettopointtoanewlycreatedheap-dynamicvariable
 Pointerp1islatersettopointtoanothernewlycreatedheap-dynamicvariable
 Theprocessoflosingheap-dynamicvariablesiscalledmemoryleakage
PointersinAda
 Somedanglingpointersaredisallowedbecausedynamicobjectscanbeautomatically
deallocatedattheendofpointer'stypescope
 Thelostheap-
dynamicvariableproblemisnoteliminatedbyAda(possiblewithUNCHECKED_DEAL
LOCATION)
PointersinCandC++
 Extremelyflexiblebutmustbeusedwithcare
 Pointerscanpointatanyvariableregardlessofwhenorwhereitwasallocated
 Usedfordynamicstoragemanagementandaddressing
 Pointerarithmeticispossible
 Explicitdereferencingandaddress-ofoperators
 Domaintypeneednotbefixed(void*)
void*canpointtoanytypeandcanbetypechecked(
cannotbede-referenced)
PointerArithmeticinCandC++
floatstuff[100];fl
oat*p;
p=stuff;
*(p+5)isequivalenttostuff[5]andp[5]
*(p+i)isequivalenttostuff[i]andp[i]
ReferenceTypes
 C++includesaspecialkindofpointertypecalledareferencetypethatisusedprimarilyforfo
rmalparameters
– Advantagesofbothpass-by-referenceandpass-by-value
 JavaextendsC++‘sreferencevariablesandallowsthemtoreplacepointersentirely
– Referencesarereferencestoobjects,ratherthanbeingaddresses
 C#includesboththereferencesofJavaandthepointersofC++
EvaluationofPointers
 Danglingpointersanddanglingobjectsareproblemsasisheapmanagement
 Pointersarelikegoto's--theywidentherangeofcellsthatcanbeaccessedbyavariable
 Pointersorreferencesarenecessaryfordynamicdatastructures--
sowecan'tdesignalanguagewithoutthem
RepresentationsofPointers
 Largecomputersusesinglevalues
 Intelmicroprocessorsusesegmentandoffset
DanglingPointerProblem
 Tombstone:extraheapcellthatisapointertotheheap-dynamicvariable
– Theactualpointervariablepointsonlyattombstones
– Whenheap-dynamicvariablede-allocated,tombstoneremainsbutsettonil
– Costlyintimeandspace
27
  Locks-and-keys:Pointervaluesarerepresentedas(key,address)pairs
– Heap-dynamicvariablesarerepresentedasvariablepluscellforintegerlockvalue
– Whenheap-
dynamicvariableallocated,lockvalueiscreatedandplacedinlockcellandkeycello
fpointer
HeapManagement
 Averycomplexrun-timeprocess
 Single-sizecellsvs.variable-sizecells
 Twoapproachestoreclaimgarbage
– Referencecounters(eagerapproach):reclamationisgradual
– Mark-
sweep(lazyapproach):reclamationoccurswhenthelistofvariablespacebecomesem
pty
ReferenceCounter
 Referencecounters:maintainacounterineverycellthatstorethenumberofpointerscurr
entlypointingatthecell
– Disadvantages:spacerequired,executiontimerequired,complicationsforcellsconne
ctedcircularly
– Advantage:itisintrinsicallyincremental,sosignificantdelaysintheapplicationex
ecutionareavoided
Mark-Sweep
 Therun-
timesystemallocatesstoragecellsasrequestedanddisconnectspointersfromcellsas
necessary;mark-sweepthenbegins
– Everyheapcellhasanextrabitusedbycollectionalgorithm
– Allcellsinitiallysettogarbage
– Allpointerstracedintoheap,andreachablecellsmarkedasnotgarbage
– Allgarbagecellsreturnedtolistofavailablecells
– Disadvantages:initsoriginalform,itwasdonetooinfrequently.Whendone,itcaus
edsignificantdelaysinapplicationexecution.Contemporarymarksweepalgorith
msavoidthisbydoingitmoreoften—calledincrementalmarksweep

Figure3.10MarkingAlgorithm

Variable-SizeCells
 Allthedifficultiesofsingle-sizecellsplusmore
 Requiredbymostprogramminglanguages
 Ifmark-sweepisused,additionalproblemsoccur
– Theinitialsettingoftheindicatorsofallcellsintheheapisdifficult
– Themarkingprocessinnontrivial
– Maintainingthelistofavailablespaceisanothersourceofoverhead

28
 Names–CO2
 Designissuesfornames:
– Maximumlength?
– Areconnectorcharactersallowed?
– Arenamescasesensitive?
– Arespecialwordsreservedwordsorkeywords?
 Length
– Iftooshort,theycannotbeconnotative
– Languageexamples:
o FORTRANI:maximum6
o COBOL:maximum30
o FORTRAN90andANSIC:maximum31
o AdaandJava:nolimit,andallaresignificant
o C++:nolimit,butimplementorsoftenimposeone
 Connectors
– Pascal,Modula-2,andFORTRAN77don'tallow
– Othersdo
 Casesensitivity
– Disadvantage:readability(namesthatlookalikearedifferent)
o worseinC++andJavabecausepredefinednamesaremixedcase(e.g.,IndexOut
OfBoundsException)
– C,C++,andJavanamesarecasesensitive
– Thenamesinotherlanguagesarenot
 Specialwords
– Anaidtoreadability;usedtodelimitorseparatestatementclauses
– Def:Akeywordisawordthatisspecialonlyincertaincontexts
o i.e.inFortran:
– RealVarName(Realisdatatypefollowedwithaname,thereforeRealisakeyword
)
– Real=3.4(Realisavariable)
– Disadvantage:poorreadability
– Def:Areservedwordisaspecialwordthatcannotbeusedasauser-definedname

Variables–CO2
 Avariableisanabstractionofamemorycell
 Variablescanbecharacterizedasasextupleofattributes:(na
me,address,value,type,lifetime,andscope)
 Name-notallvariableshavethem(anonymous)
 Address-thememoryaddresswithwhichitisassociated(alsocalledl-value)
– Avariablemayhavedifferentaddressesatdifferenttimesduringexecution
– Avariablemayhavedifferentaddressesatdifferentplacesinaprogram
– Iftwovariablenamescanbeusedtoaccessthesamememorylocation,theyarecalle
daliases
– Aliasesareharmfultoreadability(programreadersmustrememberallofthem)
 Howaliasescanbecreated:
– Pointers,referencevariables,CandC++unions
– Someoftheoriginaljustificationsforaliasesarenolongervalid;e.g.,memoryreuse
inFORTRAN
– Replacethemwithdynamicallocation

29
  Type-
determinestherangeofvaluesofvariablesandthesetofoperationsthataredefinedforval
uesofthattype;inthecaseoffloatingpoint,typealsodeterminesthePrecision
 Value-thecontentsofthelocationwithwhichthevariableisassociated
 Abstractmemorycell-thephysicalcellorcollectionofcellsassociatedwithavariable

TheConceptofBinding
 Thel-valueofavariableisitsaddress
 Ther-valueofavariableisitsvalue
 Def:Abindingisanassociation,suchasbetweenanattributeandanentity,orbetweena
noperationandasymbol
 Def:Bindingtimeisthetimeatwhichabindingtakesplace.
 Possiblebindingtimes:
– Languagedesigntime--e.g.,bindoperatorsymbolstooperations
– Language implementation time--e.g., bind floating point type to
arepresentation
– Compiletime--e.g.,bindavariabletoatypeinCorJava
– Loadtime--e.g.,bindaFORTRAN77variabletoamemorycell(oraCstatic
variable)
– Runtime--e.g.,bindanonstaticlocalvariabletoamemorycell
 Def:Abindingisstaticifitfirstoccursbeforeruntimeandremainsunchangedthrough
outprogramexecution.
 Def:Abindingisdynamicifitfirstoccursduringexecutionorcanchangeduringexecutiono
ftheprogram.
 TypeBindings
– Howisatypespecified?
– Whendoesthebindingtakeplace?
– Ifstatic,thetypemaybespecifiedbyeitheranexplicitoranimplicitdeclaration
 Def:Anexplicitdeclarationisaprogramstatementusedfordeclaringthetypesofvariables
 Def:Animplicitdeclarationisadefaultmechanismforspecifyingtypesofvariables(thefir
stappearanceofthevariableintheprogram)
 FORTRAN,PL/I,BASIC,andPerlprovideimplicitdeclarations
– Advantage:writability
– Disadvantage:reliability(lesstroublewithPerl)
 DynamicTypeBinding(JavaScriptandPHP)
 Specifiedthroughanassignmentstatemente.g.,JavaScript
list=[2,4.33,6,8];
list=17.3;
– Advantage:flexibility(genericprogramunits)
– Disadvantages:
o Highcost(dynamictypecheckingandinterpretation)
o Typeerrordetectionbythecompilerisdifficult
 TypeInferencing(ML,Miranda,andHaskell)
– Ratherthanbyassignmentstatement,typesaredeterminedfromthecontextoftheref
erence
 StorageBindings&Lifetime
– Allocation-gettingacellfromsomepoolofavailablecells
– Deallocation-puttingacellbackintothepool

30
 Def:Thelifetimeofavariableisthetimeduringwhichitisboundtoaparticularmemory
cell
 Categoriesofvariablesbylifetimes
– Static--
boundtomemorycellsbeforeexecutionbeginsandremainsboundtothesameme
morycellthroughoutexecution.
e.g.,allFORTRAN77variables,Cstaticvariables
– Advantages: efficiency (direct addressing), history-sensitive
subprogramsupport
– Disadvantage:lackofflexibility(norecursion)
 Categoriesofvariablesbylifetimes
– Stack-dynamic--Storage bindings are created for variables when
theirdeclarationstatementsareelaborated.
– Ifscalar,allattributesexceptaddressarestaticallybound
e.g.,localvariablesinCsubprogramsandJavamethods
– Advantage:allowsrecursion;conservesstorage
– Disadvantages:
o Overheadofallocationanddeallocation
o Subprogramscannotbehistorysensitive
o Inefficientreferences(indirectaddressing)
 Categoriesofvariablesbylifetimes
– Explicit heap-dynamic--Allocated and deallocated by explicit
directives,specifiedbytheprogrammer,whichtakeeffectduringexecution
– Referencedonlythroughpointersorreferences
e.g.,dynamicobjectsinC++(vianewanddelete)allobjectsinJava
– Advantage:providesfordynamicstoragemanagement
– Disadvantage:inefficientandunreliable
 Categoriesofvariablesbylifetimes
– Implicitheapdynamic--
Allocationanddeallocationcausedbyassignmentstatements
e.g.,allvariablesinAPL;allstringsandarraysinPerlandJavaScript
– Advantage:flexibility
– Disadvantages:
o Inefficient,becauseallattributesaredynamic
o Lossoferrordetection

TypeChecking-CO3
 Generalizetheconceptofoperandsandoperatorstoincludesubprogramsandassign
ments
 Typecheckingistheactivityofensuringthattheoperandsofanoperatorareofcompatible
types
 Acompatibletypeisonethatiseitherlegalfortheoperator,orisallowedunderlanguagerul
estobeimplicitlyconverted,bycompiler-
generatedcode,toalegaltype.Thisautomaticconversioniscalledascoercion.
 Atypeerroristheapplicationofanoperatortoanoperandofaninappropriatetype
 Ifalltypebindingsarestatic,nearlyalltypecheckingcanbestatic
 Iftypebindingsaredynamic,typecheckingmustbedynamic
 Def:Aprogramminglanguageisstronglytypediftypeerrorsarealwaysdetected

31
TypeCompatibility

 Ourconcernisprimarilyforstructuredtypes
 Def:Nametypecompatibilitymeansthetwovariableshavecompatibletypesiftheyareine
itherthesamedeclarationorindeclarationsthatusethesametypename
 Easytoimplementbuthighlyrestrictive:
– Subrangesofintegertypesarenotcompatiblewithintegertypes
– Formalparametersmustbethesametypeastheircorrespondingactualparameters(P
ascal)
 Structuretypecompatibilitymeansthattwovariableshavecompatibletypesiftheirtypes
haveidenticalstructures
 Moreflexible,buthardertoimplement
 Considertheproblemoftwostructuredtypes:
– Aretworecordtypescompatibleiftheyarestructurallythesamebutusedifferentfieldna
mes?
– Aretwoarraytypescompatibleiftheyarethesameexceptthatthesubscriptsaredifferen
t?
(e.g.,[1..10]and[0..9])
– Are two enumeration types compatible if their components are
spelleddifferently?
– Withstructuraltypecompatibility,youcannotdifferentiatebetweentypesofthesamestr
ucture(e.g.,differentunitsofspeed,bothfloat)
 Languageexamples:
– Pascal:usuallystructure,butinsomecasesnameisused(formalparameters)
– C:structure,exceptforrecords
– Ada:restrictedformofname
o Derivedtypesallowtypeswiththesamestructuretobedifferent
o Anonymoustypesareallunique,evenin:
A,B:array(1..10)ofINTEGER:

StrongTyping–CO2

 Advantageofstrongtyping:allowsthedetectionofthemisusesofvariablesthatresultinty
peerrors
 Languageexamples:
– FORTRAN77isnot:parameters,EQUIVALENCE
– Pascalisnot:variantrecords
– CandC++arenot:parametertypecheckingcanbeavoided;unionsarenottypechecked
– Adais,almost(UNCHECKEDCONVERSIONisloophole)(Ja
vaissimilar)
 Coercionrulesstronglyaffectstrongtyping--
theycanweakenitconsiderably(C++versusAda)
 AlthoughJavahasjusthalftheassignmentcoercionsofC++,itsstrongtypingisstillfarle
sseffectivethanthatofAda

32
NamedConstants
 Def:Anamedconstantisavariablethatisboundtoavalueonlywhenitisboundtostora
ge
 Advantages:readabilityandmodifiability
 Usedtoparameterizeprograms
 Thebindingofvaluestonamedconstantscanbeeitherstatic(calledmanifestconstant
s)ordynamic
 Languages:
– Pascal:literalsonly
– FORTRAN90:constant-valuedexpressions
– Ada,C++,andJava:expressionsofanykind

VariableInitialization
 Def:Thebindingofavariabletoavalueatthetimeitisboundtostorageiscalledinitializatio
n
 Initializationisoftendoneonthedeclarationstatemente.g.,Javaintsum=0

Summary
 Thedatatypesofalanguagearealargepartofwhatdeterminesthatlanguage‘sstylean
dusefulness
 Theprimitivedatatypesofmostimperativelanguagesincludenumeric,charact
er,andBooleantypes
 Theuser-
definedenumerationandsubrangetypesareconvenientandaddtothereadabilityan
dreliabilityofprograms
 Arraysandrecordsareincludedinmostlanguages
 Pointersareusedforaddressingflexibilityandtocontroldynamicstoragemanage
ment
 Casesensitivityandtherelationshipofnamestospecialwordsrepresentdesignissuesof
names
 Variablesarecharacterizedbythesextuples:name,address,value,type,lifeti
me,scope
 Bindingistheassociationofattributeswithprogramentities
 Scalarvariablesarecategorizedas:static,stackdynamic,explicitheapdyna
mic,implicitheapdynamic
 Strongtypingmeansdetectingalltypeerrors

33
 ExpressionsandStatements&ControlStructures

Introduction
 Expressions are the fundamental means of specifying computations in
aprogramminglanguage.
 Tounderstandexpressionevaluation,needtobefamiliarwiththeordersofoperatorandoper
andevaluation.
 EssenceofimperativelanguagesisdominantroleofassignmentstatementsArithmeticExpr
essions.
 Arithmeticevaluationwasoneofthemotivationsforthedevelopmentofthefirstprogrammi
nglanguages.

ArithmeticExpressions–CO2,CO3
Arithmetic Expressions consist of operators, operands, parentheses
andfunctioncalls.
DesignIssues
Designissuesforarithmeticexpressions
 Operatorprecedencerules?
 Operatorassociativityrules?
 Orderofoperandevaluation?
 Operandevaluationsideeffects?
 Operatoroverloading?
 Typemixinginexpressions?
Operators
 Aunaryoperatorhasoneoperand.
 Abinaryoperatorhastwooperands.
 Aternaryoperatorhasthreeoperands.
OperatorPrecedenceRules
Theoperatorprecedencerulesforexpressionevaluationdefinetheorderinwhichadjacent
operatorsofdifferentprecedencelevelsareevaluated.

Typicalprecedencelevels:
 parentheses
 unaryoperators
 **(ifthelanguagesupportsit)
 *,/
 +,-
OperatorAssociativityRule
Theoperatorassociativityrulesforexpressionevaluationdefinetheorderinwhichadjacent
operatorswiththesameprecedencelevelareevaluated.

Typicalassociativityrules:
 Lefttoright,except**(RubyandFortran),whichisrighttoleft
 Sometimesunaryoperatorsassociaterighttoleft(e.g.,inFORTRAN)
– APLisdifferent;alloperatorshaveequalprecedenceandalloperatorsassociaterightt
oleft.
– Precedenceandassociativityrulescanbeoverriddenwithparentheses.

ConditionalExpressions
34
ConditionalExpressions(ternaryoperator?:)availableinC-
basedlanguages.Anexample:(C,C++)
average=(count==0)?0:sum/count
Evaluatesasifwrittenlike
if(count==0)
average=0
else
average=sum/count
OperandEvaluationOrder
Operandevaluationorderasfollows.
 Variables:fetchthevaluefrommemory.
 Constants:sometimesafetchfrommemory;sometimestheconstantisinthemachinel
anguageinstruction.
 Parenthesizedexpressions:evaluatealloperandsandoperatorsfirst.
 Themostinterestingcaseiswhenanoperandisafunctioncall.

PotentialsforSideEffectsFunctionalsideeffects:whenafunctionchangesatwo-
wayparameteroranon-localvariable
Problemwithfunctionalsideeffects:Whenafunctionreferencedinanexpressionaltersano
theroperandoftheexpression;
e.g.,foraparameterchange:
a=10;
/*assumethatfunchangesitsparameter*/
b=a+fun(a);
Twopossiblesolutionstothefunctionalsideeffectsproblem:Writethelanguagedefinition
todisallowfunctionalsideeffects.
 Notwo-wayparametersinfunctions
 Nonon-localreferencesinfunctions

Advantage:itworks!
Disadvantages:inflexibilityofone-wayparametersandlackofnon-localreferences
– Writethelanguagedefinitiontodemandthatoperandevaluationorderbefixed.
Disadvantage:limitssomecompileroptimizations
– Javarequiresthatoperandsappeartobeevaluatedinleft-to-rightorder.

OverloadedOperators

Useofanoperatorformorethanonepurposeiscalledoperatoroverloading.
 Somearecommon(e.g.,+forintandfloat)
 Somearepotentialtrouble(e.g.,*inCandC++)
Problems:
 Lossofcompilererrordetection(omissionofanoperandshouldbeadetectableerror)
 Somelossofreadability
 Canbeavoidedbyintroductionofnewsymbols(e.g.,Pascal‘sdivforintegerdivision)
 C++,Ada,Fortran95,andC#allowuser-definedoverloadedoperators
Potentialproblems:
 Userscandefinenonsenseoperations.
 Readabilitymaysuffer,evenwhentheoperatorsmakesense.

35
TypeConversions
Anarrowingconversionisonethatconvertsanobjecttoatypethatcannotincludeallofth
evaluesoftheoriginaltype.
e.g.,floattoint
Awideningconversionisoneinwhichanobjectisconvertedtoatypethatcanincludeatleasta
pproximationstoallofthevaluesoftheoriginaltype.
e.g.,inttofloat
MixedMode
Amixed-
modeexpressionisonethathasoperandsofdifferenttypesAcoercionisanim
plicittypeconversion
Disadvantageofcoercions:
– Theydecreaseinthetypeerrordetectionabilityofthecompiler
 Inmostlanguages,allnumerictypesarecoercedinexpressions,usingwideningconve
rsions.
 InAda,therearevirtuallynocoercionsinexpressionsExplicitT
ypeConversionscalledascastinginC-basedlanguages.Examples:-
C:(int)angle,Ada:Float(Sum)
– NotethatAda’ssyntaxissimilartothatoffunctioncalls
ErrorsinExpressionscauses
 Inherentlimitationsofarithmetice.g.,divisionbyzero
 Limitationsofcomputerarithmetice.g.,overflow
 Oftenignoredbytherun-timesystem

RelationalandBooleanExpressions
RelationalExpressions:
 Userelationaloperatorsandoperandsofvarioustypes
 EvaluatetosomeBooleanrepresentation
 Operatorsymbolsusedvarysomewhatamonglanguages(!=,/=,.NE.,<>,#)
 JavaScriptandPHPhavetwoadditionalrelationaloperator,===and!==
 Similartotheircousins,==and!=,exceptthattheydonotcoercetheiroperands
BooleanExpressions:
OperandsareBooleanandtheresultisBooleanExampleop
erators
FORTRAN77 FORTRAN90 C Ada
.AND. and && and
.OR. or || or
.NOT. not ! not
xor --- --- ----
 NoBooleanTypeinC
 C89hasnoBooleantype--itusesinttypewith0forfalseandnonzerofortrue
 OneoddcharacteristicofC‘sexpressions:a<b<cisalegalexpression,buttheresultisn
otwhatyoumightexpect:
 Leftoperatorisevaluated,producing0or1
 Theevaluationresultisthencomparedwiththethirdoperand(i.e.,c)

ShortCircuitEvaluation-CO3
Anexpressioninwhichtheresultisdeterminedwithoutevaluatingalloftheoperandsand/
oroperators
Example:(13*a)*(b/13–1)
If‘a’iszero,thereisnoneedtoevaluate(b/13-1)

36
Problemwithnon-short-circuitevaluation
index=1;
while(index<=length)&&(LIST[index]!=value)index++;
 Whenindex=length,LIST[index]willcauseanindexingproblem(assumingLISThasl
ength-1elements)
 C,C++andJava:useshort-
circuitevaluationfortheusualBooleanoperators(&&and||),butalsoprovidebitwise
Booleanoperatorsthatarenotshortcircuit(&and|)
 Ada:programmercanspecifyeither(short-circuitisspecifiedwithandthenandorelse)
 Short-
circuitevaluationexposesthepotentialproblemofsideeffectsinexpressionse.g.,(a>
b)||(b++/3)

AssignmentStatementsCO3

Thegeneralsyntax:<target_var><assign_operator><expression>Thea
ssignmentoperator
 ‘=’inFORTRAN,BASIC,theC-basedlanguages
 ‘:=’inALGOL,Pascal,Ada
–
Equal‘=’canbebadwhenitisoverloadedfortherelationaloperatorforequality(that‘swh
ytheC-basedlanguagesuse==astherelationaloperator)
ConditionalTargets(Perl)
($flag?$total:$subtotal)=0
Whichisequivalentto
if($flag){$total=0}else
{$subtotal=0}

CompoundAssignmentOperators
 Ashorthandmethodofspecifyingacommonlyneededformofassignment.
 IntroducedinALGOL;adoptedbyCEx
:‘a=a+b’ iswrittenas‘a+=b’

UnaryAssignmentOperators
UnaryassignmentoperatorsinC-
basedlanguagescombineincrementanddecrementoperationswithassignment.Forexam
ple
sum+=++count(countincremented,addedtosum)sum+
=count++(countincremented,addedtosum)count++(co
untincremented)
-count++(countincrementedthennegated)

AssignmentasanExpression
InC,C++,andJava,theassignmentstatementproducesaresultandcanbeusedasopera
nds.
while((ch=getchar())!=EOF){…}
ch=getchar()iscarriedout;theresult(assignedtoch)isusedasaconditionalvalueforthewh
ilestatement
ListAssignments
PerlandRubysupportlistassignments
e.g.,($first,$second,$third)=(20,30,40);

37
Mixed-ModeAssignment
Assignmentstatementscanalsobemixed-mode,forexample
inta,b;float
c;
c=a/b;
– InFortran,C,andC++,anynumerictypevaluecanbeassignedtoanynumerictypevaria
ble.
– InJava,onlywideningassignmentcoercionsaredone.
– InAda,thereisnoassignmentcoercion.

ControlStructures-CO3
Acontrolstructureisacontrolstatementandthestatementswhoseexecutionitcontrols
.

LevelsofControlFlow
 Withinexpressions
 Amongprogramunits
 Amongprogramstatements

ControlStatements:Evolution
 FORTRANIcontrolstatementswerebaseddirectlyonIBM704hardware
 Muchresearchandargumentinthe1960sabouttheissue
Oneimportantresult:Itwasproventhatallalgorithmsrepresentedbyflowchartscanbecod
edwithonlytwo-wayselectionandpretestlogicalloops

SelectionStatements–CO3
Aselectionstatementprovidesthemeansofchoosingbetweentwoormorepathsofexecutio
n.Twogeneralcategories:
 Two-wayselectors
 Multiple-
wayselectorsTwo-
WaySelectionStatementsGenera
l form as follows…
ifcontrol_expressionth
enclause
elseclause
DesignIssues:
 Whatistheformandtypeofthecontrolexpression?
 Howarethethenandelseclausesspecified?
 Howshouldthemeaningofnestedselectorsbespecified?
TheControlExpression
 Ifthe‘then’reservedwordorsomeothersyntacticmarkerisnotusedtointroducethe‘then
’clause,thecontrolexpressionisplacedinparentheses.
 InC89,C99,Python,andC++,thecontrolexpressioncanbearithmetic.
 InlanguagessuchasAda,Java,Ruby,andC#,thecontrolexpressionmustbeBoolean.
ClauseForm
 Inmanycontemporarylanguages,thethenandelseclausescanbesinglestatementsorco
mpoundstatements
 InPerl,allclausesmustbedelimitedbybraces(theymustbecompound)
 InFortran95,Ada,andRuby,clausesarestatementsequences
 Pythonusesindentationtodefineclauses
38
 ifx>y:
x=y
print"case1"
NestingSelectors:Javaexample
if(sum==0)if(
count==0)res
ult=0;
elseresult=1;
 Whichifgetstheelse?
 Java'sstaticsemanticsrule:elsematcheswiththenearestifNestingSelectors
 Toforceanalternativesemantics,compoundstatementsmaybeused:
if(sum==0){if(
count==0)resu
lt=0;}elseresul
t=1;
 TheabovesolutionisusedinC,C++,andC#
 Perlrequiresthatallthenandelseclausestobecompound
 Statementsequencesasclauses:Ruby
ifsum==0thenifco
unt==0thenresult
=0
elseresult
=1end
end
-Python
ifsum==0:
ifcount==0:
result=0els
e:
result=1

Multiple-WaySelectionStatements
AllowtheselectionofoneofanynumberofstatementsorstatementgroupsDesignIssues:
 Whatistheformandtypeofthecontrolexpression?
 Howaretheselectablesegmentsspecified?
 Isexecutionflowthroughthestructurerestrictedtoincludejustasingleselectablesegme
nt?
 Howarecasevaluesspecified?
 Whatisdoneaboutunrepresentedexpressionvalues?

Multiple-WaySelection:Examples
 C,C++,andJava
switch(expression){
caseconst_expr_1:stmt_1;
…
caseconst_expr_n:stmt_n;[defau
lt:stmt_n+1]}
 DesignchoicesforC‘sswitchstatement
 Controlexpressioncanbeonlyanintegertype
 Selectable segments can be statement sequences, blocks, or
compoundstatements
 Anynumberofsegmentscanbeexecutedinoneexecutionoftheconstruct(thereisnoimpli
citbranchattheendofselectablesegments)

39
  defaultclauseisforunrepresentedvalues(ifthereisnodefault,thewholestatementdoes
nothing)
 C#
– DiffersfromCinthatithasastaticsemanticsrulethatdisallowstheimplicitexecutio
nofmorethanonesegment
– Eachselectablesegmentmustendwithanunconditionalbranch(gotoorbreak)
 Ada
caseexpressionis
whenchoicelist=>stmt_sequence;
…
whenchoicelist=>stmt_sequence;w
henothers=>stmt_sequence;]endca
se;
MorereliablethanC‘sswitch(onceastmt_sequenceexecutioniscompleted,controlispassedt
othefirststatementafterthecasestatement
 Adadesignchoices:
1. Expressioncanbeanyordinaltype
2. Segmentscanbesingleorcompound
3. Onlyonesegmentcanbeexecutedperexecutionoftheconstruct
4. Unrepresentedvaluesarenotallowed
 ConstantListForms:
1. Alistofconstants
2. Caninclude:
- Subranges
- BooleanORoperators(|)

Multiple-WaySelectionUsingif
MultipleSelectorscanappearasdirectextensionstotwo-wayselectors,usingelse-
ifclauses,forexampleinPython:
ifcount<10:
bag1=Truee
lsifcount<100:
bag2=Trueels
eifcount<1000:
bag3=True

IterativeStatements
 Therepeatedexecutionofastatementorcompoundstatementisaccomplishedeitherbyit
erationorrecursion
 Generaldesignissuesforiterationcontrolstatements:
1. Howisiterationcontrolled?
2. Whereisthecontrolmechanismintheloop?
Counter-ControlledLoops
Acountingiterativestatementhasaloopvariable,andameansofspecifyingtheinitialan
dterminal,andstepsizevalues
DesignIssues:
 Whatarethetypeandscopeoftheloopvariable?
 Whatisthevalueoftheloopvariableatlooptermination?
 Shoulditbelegalfortheloopvariableorloopparameterstobechangedintheloopbody,an
difso,doesthechangeaffectloopcontrol?
 Shouldtheloopparametersbeevaluatedonlyonce,oronceforeveryiteration?

40
IterativeStatements:Examples
FORTRAN95syntax
DOlabelvar=start,finish[,stepsize]Stepsiz
ecanbeanyvaluebutzeroParameterscanbe
expressions
Designchoices:
1. LoopvariablemustbeINTEGER
2. Loopvariablealwayshasitslastvalue
3. Theloopvariablecannotbechangedintheloop,buttheparameterscan;becausetheyareeval
uatedonlyonce,itdoesnotaffectloopcontrol
4. Loopparametersareevaluatedonlyonce
 FORTRAN95:asecondform:
[name:]Dovariable=initial,terminal[,stepsize]
…
EndDo[name]
– CannotbranchintoeitherofFortran‘sDostatements
 Ada
forvarin[reverse]discrete_rangeloop...endl
oop
 Designchoices:
– Typeoftheloopvariableisthatofthediscreterange(Adiscreterangeisasub-
rangeofanintegerorenumerationtype).
– Loopvariabledoesnotexistoutsidetheloop
– Theloopvariablecannotbechangedintheloop,butthediscreterangecan;itdoesnotaff
ectloopcontrol
– Thediscreterangeisevaluatedjustonce
– Cannotbranchintotheloopbody
 C-basedlanguages
for([expr_1];[expr_2];[expr_3])statement
– Theexpressionscanbewholestatements,orevenstatementsequences,withthest
atementsseparatedbycommas
– Thevalueofamultiple-
statementexpressionisthevalueofthelaststatementintheexpression
– Ifthesecondexpressionisabsent,itisaninfiniteloop
 Designchoices:
– Thereisnoexplicitloopvariable
– Everythingcanbechangedintheloop
– Thefirstexpressionisevaluatedonce,buttheothertwoareevaluatedwitheachiterati
on
 C++differsfromCintwoways:
– ThecontrolexpressioncanalsobeBoolean
– Theinitialexpressioncanincludevariabledefinitions(scopeisfromthedefinitiontoth
eendoftheloopbody)
 JavaandC#
– DiffersfromC++inthatthecontrolexpressionmustbeBoolean
– IterativeStatements:Logically-ControlledLoops
– RepetitioncontrolisbasedonaBooleanexpression
 Designissues:
– Pretestorposttest?
– Shouldthelogicallycontrolledloopbeaspecialcaseofthecountingloopstatement
oraseparatestatement?

41
  IterativeStatements:Logically-ControlledLoops:Examples
CandC++havebothpretestandposttestforms,inwhichthecontrolexpressioncanbear
ithmetic:
while(ctrl_expr)doloop
bodyloopbodywhile(ctrl
_expr)
 JavaislikeCandC++,exceptthecontrolexpressionmustbeBoolean(andthebodycano
nlybeenteredatthebeginning--Javahasnogoto
 IterativeStatements:Logically-ControlledLoops:Examples
 Adahasapretestversion,butnoposttest
 FORTRAN95hasneither
 PerlandRubyhavetwopretestlogicalloops,whileanduntil.Perlalsohastwoposttestloop
s

UnconditionalBranching:User-LocatedLoopControlMechanisms
 Sometimesitisconvenientfortheprogrammerstodecidealocationforloopcontrol(other
thantoporbottomoftheloop)
 Simpledesignforsingleloops(e.g.,break)
 Designissuesfornestedloops
– Shouldtheconditionalbepartoftheexit?
– Shouldcontrolbetransferableoutofmorethanoneloop?User-
LocatedLoopControlMechanismsbreakandcontinue
 C,C++,Python,Ruby,andC#haveunconditionalunlabeledexits(break)
 JavaandPerlhaveunconditionallabeledexits(breakinJava,lastinPerl)
 C,C++,andPythonhaveanunlabeledcontrolstatement,continue,thatskipstheremai
nderofthecurrentiteration,butdoesnotexittheloop
 JavaandPerlhavelabeledversionsofcontinue
IterativeStatements:IterationBasedonDataStructures
 Numberofelementsofinadatastructurecontrolloopiteration
 Controlmechanismisacalltoaniteratorfunctionthatreturnsthenextelementinsomecho
senorder,ifthereisone;elseloopisterminate
 C'sforcanbeusedtobuildauser-definediterator:
for(p=root;p==NULL;traverse(p)){
}
 C#‘sforeachstatementiteratesontheelementsofarraysandothercollections:
Strings[]=strList={"Bob","Carol","Ted"};foreach(St
ringsnameinstrList)Console.WriteLine("Name:{0}"
,name);
Thenotation{0}indicatesthepositioninthestringtobedisplayed
 Perlhasabuilt-initeratorforarraysandhashes,foreachUnconditionalBranching
 Transfersexecutioncontroltoaspecifiedplaceintheprogram
 Representedoneofthemostheateddebatesin1960‘sand1970‘s
 Well-knownmechanism:gotostatement
 Majorconcern:Readability
 Somelanguagesdonotsupportgotostatement(e.g.,Java)
 C#offersgotostatement(canbeusedinswitchstatements)
 Loopexitstatementsarerestrictedandsomewhatcamouflagedgoto‘s

42
GuardedCommands
 DesignedbyDijkstra
 Purpose: to support a new programming methodology that
supportedverification(correctness)duringdevelopment
 Basisfortwolinguisticmechanismsforconcurrentprogramming(inCSPandAda)
 BasicIdea:iftheorderofevaluationisnotimportant,theprogramshouldnotspecifyone
SelectionGuardedCommand
• Form
if<Booleanexp>-
><statement>[]<Booleanexp>-
><statement>
...
[]<Booleanexp>-><statement>fi
Semantics:whenconstructisreached,
 EvaluateallBooleanexpressions
 Ifmorethanonearetrue,chooseonenon-deterministically
 Ifnonearetrue,itisaruntimeerror

LoopGuardedCommand
• Form
do<Boolean>-
><statement>[]<Boolean>-
><statement>
...
[]<Boolean>-><statement>Od

Semantics:foreachiteration
 EvaluateallBooleanexpressions
 Ifmorethanonearetrue,chooseonenon-deterministically;thenstartloopagain
 Ifnonearetrue,exitloop

GuardedCommands:Rationale
 Connectionbetweencontrolstatementsandprogramverificationisintimate
 Verificationisimpossiblewithgotostatements
 Verificationispossiblewithonlyselectionandlogicalpretestloops
 Verificationisrelativelysimplewithonlyguardedcommands

Summary
 Expressions
 Operatorprecedenceandassociativity
 Operatoroverloading
 Mixed-typeexpressions
 Variousformsofassignment
 Varietyofstatement-levelstructures
 Choiceofcontrolstatementsbeyondselectionandlogicalpretestloopsisatrade-
offbetweenlanguagesizeandwritability
 Functional and logic programming languages are quite different
controlstructures

43
 UNIT-3
SubprogramsandBlocks

Introduction
 Twofundamentalabstractionfacilities
– Processabstraction
 Emphasizedfromearlydays
– Dataabstraction
 Emphasizedinthe1980s

FundamentalsofSubprograms–CO4
 Eachsubprogramhasasingleentrypoint
 Thecallingprogramissuspendedduringexecutionofthecalledsubprogram
 Controlalwaysreturnstothecallerwhenthecalledsubprogram‘sexecutionterminates
BasicDefinitions
 Asubprogramdefinitiondescribestheinterfacetoandtheactionsofthesubprogramabstr
action
 Asubprogramcallisanexplicitrequestthatthesubprogrambeexecuted
 Asubprogramheaderisthefirstpartofthedefinition,includingthename,thekindofsu
bprogram,andtheformalparameters
 Theparameterprofile(akasignature)ofasubprogramisthenumber,order,andtypesofits
parameters
 Theprotocolisasubprogram‘sparameterprofileand,ifitisafunction,itsreturntype
 FunctiondeclarationsinCandC++areoftencalledprototypes
 Asubprogramdeclarationprovidestheprotocol,butnotthebody,ofthesubprogram
 Aformalparameterisadummyvariablelistedinthesubprogramheaderandusedinth
esubprogram
 Anactualparameterrepresentsavalueoraddressusedinthesubprogramcallstatement
Actual/FormalParameterCorrespondence
 Positional
– Thebindingofactualparameterstoformalparametersisbyposition:thefirstactualpa
rameterisboundtothefirstformalparameterandsoforth
– Safeandeffective
 Keyword
– Thenameoftheformalparametertowhichanactualparameteristobeboundisspecifie
dwiththeactualparameter
– Parameterscanappearinanyorder
FormalParameterDefaultValues
 Incertainlanguages(e.g.,C++,Ada),formalparameterscanhavedefaultvalues(ifnotactu
alparameterispassed)
– InC++,defaultparametersmustappearlastbecauseparametersarepositionallya
ssociated
 C#methodscanacceptavariablenumberofparametersaslongastheyareofthesametype

44
ProceduresandFunctions
 Therearetwocategoriesofsubprograms
– Procedures are collection of statements that define
parameterizedcomputations
– Functionsstructurallyresembleproceduresbutaresemanticallymodeledonmat
hematicalfunctions
 Theyareexpectedtoproducenosideeffects
 Inpractice,programfunctionshavesideeffects

DesignIssuesforSubprograms-CO3
 Whatparameterpassingmethodsareprovided?
 Areparametertypeschecked?
 Arelocalvariablesstaticordynamic?
 Cansubprogramdefinitionsappearinothersubprogramdefinitions?
 Cansubprogramsbeoverloaded?
 Cansubprogrambegeneric?

ScopeandLifetime
 Thescopeofavariableistherangeofstatementsoverwhichitisvisible
 Thenonlocalvariablesofaprogramunitarethosethatarevisiblebutnotdeclaredthere
 Thescoperulesofalanguagedeterminehowreferencestonamesareassociatedwithvaria
bles
 Scopeandlifetimearesometimescloselyrelated,butaredifferentconcepts
 ConsiderastaticvariableinaCorC++function
Staticscope
– Basedonprogramtext
– Toconnectanamereferencetoavariable,you(orthecompiler)mustfindthedeclaratio
n
– Searchprocess:searchdeclarations,firstlocally,theninincreasinglylargerenclos
ingscopes,untiloneisfoundforthegivenname
– Enclosingstaticscopes(toaspecificscope)arecalleditsstaticancestors;theneares
tstaticancestoriscalledastaticparent
 Variablescanbehiddenfromaunitbyhavinga"closer"variablewiththesamename
 C++andAdaallowaccesstothese"hidden"variables
– InAda:unit.name
– InC++:class_name::name
 Blocks
– Amethodofcreatingstaticscopesinsideprogramunits--fromALGOL60
– Examples:
CandC++:for(...)
{intindex;
...
}
Ada:declareLCL:FLOAT;begin
...
end
 EvaluationofStaticScoping
 Considertheexample:
AssumeMAINcallsAandB

45
 AcallsCandD
BcallsAandE
 SupposethespecischangedsothatDmustnowaccesssomedatainB
 Solutions:
– PutDinB(butthenCcannolongercallitandDcannotaccessA'svariables)
– MovethedatafromBthatDneedstoMAIN(butthenallprocedurescanaccessthem)
 Sameproblemforprocedureaccess
 Overall:staticscopingoftenencouragesmanyglobals
DynamicScope
– Basedoncallingsequencesofprogramunits,nottheirtextuallayout(temporalver
susspatial)
– Referencestovariablesareconnectedtodeclarationsbysearchingbackthroughth
echainofsubprogramcallsthatforcedexecutiontothispoint

ScopeExample
MAIN
-
declarationofx
SUB1
-declarationofx-
...
callSUB2
...SU
B2
...
-referencetox-
...
...
callSUB1
…
 Staticscoping
– ReferencetoxistoMAIN'sx
 Dynamicscoping
– ReferencetoxistoSUB1'sx
 EvaluationofDynamicScoping:
– Advantage:convenience
– Disadvantage:poorreadability

LocalReferencingEnvironments
 Def:Thereferencingenvironmentofastatementisthecollectionofallnamesthatarevisibl
einthestatement
 Inastatic-
scopedlanguage,itisthelocalvariablesplusallofthevisiblevariablesinalloftheenclos
ingscopes
 Asubprogramisactiveifitsexecutionhasbegunbuthasnotyetterminated
 Inadynamic-
scopedlanguage,thereferencingenvironmentisthelocalvariablesplusallvisiblevari
ablesinallactivesubprograms
 Localvariablescanbestack-dynamic(boundtostorage)
– Advantages
 Supportforrecursion
 Storageforlocalsissharedamongsomesubprograms
– Disadvantages

46
 Allocation/de-allocation,initializationtime

47
  Indirectaddressing
 Subprogramscannotbehistorysensitive
 Localvariablescanbestatic
– Moreefficient(noindirection)
– Norun-timeoverhead

ParameterPassingMethods–CO3
 Waysinwhichparametersaretransmittedtoand/orfromcalledsubprograms
– Pass-by-value
– Pass-by-result
– Pass-by-value-result
– Pass-by-reference
– Pass-by-name

Figure5.1ModelsofParameterPassing

 Thevalueoftheactualparameterisusedtoinitializethecorrespondingformalparameter
– Normallyimplementedbycopying
– Canbeimplementedbytransmittinganaccesspathbutnotrecommended(enforcing
writeprotectionisnoteasy)
– Whencopiesareused,additionalstorageisrequired
– Storageandcopyoperationscanbecostly

Pass-by-Result(OutMode)
 Whenaparameterispassedbyresult,novalueistransmittedtothesubprogram;thecorre
spondingformalparameteractsasalocalvariable;itsvalueistransmittedtocaller‘sactual
parameterwhencontrolisreturnedtothecaller
– Requireextrastoragelocationandcopyoperation
 Potentialproblem:sub(p1,p1);whicheverformalparameteriscopiedbackwillrepresen
tthecurrentvalueofp1
Pass-by-Value-Result(InoutMode)
 Acombinationofpass-by-valueandpass-by-result
 Sometimescalledpass-by-copy
 Formalparametershavelocalstorage
 Disadvantages:
– Thoseofpass-by-result
– Thoseofpass-by-value

48
Pass-by-Reference(InoutMode)
 Passanaccesspath
 Alsocalledpass-by-sharing
 Passingprocessisefficient(nocopyingandnoduplicatedstorage)
 Disadvantages
– Sloweraccesses(comparedtopass-by-value)toformalparameters
– Potentialsforun-wantedsideeffects
– Un-wantedaliases(accessbroadened)
Pass-by-Name(InoutMode)
 Bytextualsubstitution
 Formalsareboundtoanaccessmethodatthetimeofthecall,butactualbindingtoavalu
eoraddresstakesplaceatthetimeofareferenceorassignment
 Allowsflexibilityinlatebinding
ImplementingParameter-PassingMethods
 Inmostlanguageparametercommunicationtakesplacethrutherun-timestack
 Pass-by-referencearethesimplesttoimplement;onlyanaddressisplacedinthestack
 Asubtlebutfatalerrorcanoccurwithpass-by-referenceandpass-by-
valueresult:aformalparametercorrespondingtoaconstantcanmistakenlybechanged
ParameterPassingMethodsofMajorLanguages
 Fortran
– Alwaysusedtheinoutsemanticsmodel
– BeforeFortran77:pass-by-reference
– Fortran77andlater:scalarvariablesareoftenpassedbyvalue-result
 C
– Pass-by-value
– Pass-by-referenceisachievedbyusingpointersasparameters
 C++
– Aspecialpointertypecalledreferencetypeforpass-by-reference
 Java
– Allparametersarepassedarepassedbyvalue
– Objectparametersarepassedbyreference
 Ada
– Threesemanticsmodesofparametertransmission:in,out,inout;inisthedefaultm
ode
– Formalparametersdeclaredoutcanbeassignedbutnotreferenced;thosedeclaredinc
anbereferencedbutnotassigned;inoutparameterscanbereferencedandassigned
 C#
– Defaultmethod:pass-by-value
– Pass-by-
referenceisspecifiedbyprecedingbothaformalparameteranditsactualparameterw
ithref
 PHP:verysimilartoC#
 Perl:allactualparametersareimplicitlyplacedinapredefinedarraynamed@_
TypeCheckingParameters
 Consideredveryimportantforreliability
 FORTRAN77andoriginalC:none
 Pascal,FORTRAN90,Java,andAda:itisalwaysrequired

49
  ANSICandC++:choiceismadebytheuser
– Prototypes
 RelativelynewlanguagesPerl,JavaScript,andPHPdonotrequiretypechecking
MultidimensionalArraysasParameters
 Ifamultidimensionalarrayispassedtoasubprogramandthesubprogramisseparatel
ycompiled,thecompilerneedstoknowthedeclaredsizeofthatarraytobuildthestora
gemappingfunction
MultidimensionalArraysasParameters:CandC++
 Programmerisrequiredtoincludethedeclaredsizesofallbutthefirstsubscriptintheactu
alparameter
 Disallowswritingflexiblesubprograms
 Solution:passapointertothearrayandthesizesofthedimensionsasotherparameters;th
eusermustincludethestoragemappingfunctionintermsofthesizeparameters
MultidimensionalArraysasParameters:PascalandAda
 Pascal
– Notaproblem;declaredsizeispartofthearray‘stype
 Ada
– Constrainedarrays-likePascal
– Unconstrainedarrays-declaredsizeispartoftheobjectdeclaration
MultidimensionalArraysasParameters:Fortran
 Formalparameterthatarearrayshaveadeclarationaftertheheader
– Forsingle-dimensionarrays,thesubscriptisirrelevant
– Formulti-dimensionalarrays,thesubscriptsallowthestorage-mappingfunction
MultidimensionalArraysasParameters:JavaandC#
 SimilartoAda
 Arraysareobjects;theyareallsingle-dimensioned,buttheelementscanbearrays
 Eacharrayinheritsanamedconstant(lengthinJava,LengthinC#)thatissettothelengt
hofthearraywhenthearrayobjectiscreated

DesignConsiderationsforParameterPassing
 Twoimportantconsiderations
– Efficiency
– One-wayortwo-waydatatransfer
 Buttheaboveconsiderationsareinconflict
– Goodprogrammingsuggestlimitedaccesstovariables,whichmeansonewaywhenev
erpossible
– Butpass-by-referenceismoreefficienttopassstructuresofsignificantsize

ParametersSubprogramsasparameters–CO3
 Itissometimesconvenienttopasssubprogramnamesasparameters
 Issues:
– Areparametertypeschecked?
– Whatisthecorrectreferencingenvironmentforasubprogramthatwassentasaparam
eter?
ParametersthatareSubprogramNames:ParameterTypeChecking
 CandC++:functionscannotbepassedasparametersbutpointerstofunctionscanbepasse
d;parameterscanbetypechecked
 FORTRAN95typechecks

50
  LaterversionsofPascaland
 Adadoesnotallowsubprogramparameters;asimilaralternativeisprovidedviaAda‘sgen
ericfacility
ParametersthatareSubprogramNames:ReferencingEnvironment
 Shallowbinding:Theenvironmentofthecallstatementthatenactsthepassedsubprogra
m
 Deepbinding:Theenvironmentofthedefinitionofthepassedsubprogram
 Adhocbinding:Theenvironmentofthecallstatementthatpassedthesubprogram
OverloadedSubprograms–CO3
 Anoverloadedsubprogramisonethathasthesamenameasanothersubprograminthe
samereferencingenvironment
– Everyversionofanoverloadedsubprogramhasauniqueprotocol
 C++,Java,C#,andAdaincludepredefinedoverloadedsubprograms
 InAda,thereturntypeofanoverloadedfunctioncanbeusedtodisambiguatecalls(thu
stwooverloadedfunctionscanhavethesameparameters)
 Ada,Java,C++,andC#allowuserstowritemultipleversionsofsubprogramswiththes
amename
GenericSubprograms–CO3
 Agenericorpolymorphicsubprogramtakesparametersofdifferenttypesondifferentacti
vations
 Overloadedsubprogramsprovideadhocpolymorphism
 Asubprogramthattakesagenericparameterthatisusedinatypeexpressionthatdesc
ribesthetypeoftheparametersofthesubprogramprovidesparametricpolymorphis
m
Examplesofparametricpolymorphism:C++
template<classType>
Typemax(Typefirst,Typesecond){retu
rnfirst>second?first:second;
}
 Theabovetemplatecanbeinstantiatedforanytypeforwhichoperator>isdefined
intmax(intfirst,intsecond){returnfirs
t>second?first:second;
}
DesignIssuesforFunctions
 Aresideeffectsallowed?
– Parametersshouldalwaysbein-modetoreducesideeffect(likeAda)
 Whattypesofreturnvaluesareallowed?
– Mostimperativelanguagesrestrictthereturntypes
– Callowsanytypeexceptarraysandfunctions
– C++islikeCbutalsoallowsuser-definedtypes
– Adaallowsanytype
– JavaandC#donothavefunctionsbutmethodscanhaveanytype
User-DefinedOverloadedOperators
 OperatorscanbeoverloadedinAdaandC++
 AnAdaexample
Function“*”(A,B:inVec_Type):returnIntegerisSum
:Integer:=0;
begin
forIndexinA‘rangeloop
Sum:=Sum+A(Index)*B(Index)endlo
op

51
 returnsum;
end“*”;
…
c=a*b;--a,b,andcareoftypeVec_Type
Co-Routines–CO3
 Acoroutineisasubprogramthathasmultipleentriesandcontrolsthemitself
 Alsocalledsymmetriccontrol:callerandcalledcoroutinesareonamoreequalbasis
 Acoroutinecallisnamedaresume
 Thefirstresumeofacoroutineistoitsbeginning,butsubsequentcallsenteratthepoint
justafterthelastexecutedstatementinthecoroutine
 Coroutinesrepeatedlyresumeeachother,possiblyforever
 Coroutinesprovidequasi-
concurrentexecutionofprogramunits(thecoroutines);theirexecutionisinterleaved,but
notoverlapped

Figure5.2PossibleExecutionControls Figure5.3PossibleExecutionControls

Figure5.4PossibleExecutionControlswithLoops
Summary
 Asubprogramdefinitiondescribestheactionsrepresentedbythesubprogram
 Subprogramscanbeeitherfunctionsorprocedures
 Localvariablesinsubprogramscanbestack-dynamicorstatic
 Threemodelsofparameterpassing:inmode,outmode,andinoutmode
 Somelanguagesallowoperatoroverloading
 Subprogramscanbegeneric
 Aco-routineisaspecialsubprogramwithmultipleentries

52
 UNIT-4
AbstractDataTypes

TheConceptofAbstraction–CO3
 Anabstractionisavieworrepresentationofanentitythatincludesonlythemostsignifican
tattributes
 Theconceptofabstractionisfundamentalinprogramming(andcomputerscience)
 Nearly all programming languages support process abstraction
withsubprograms
 Nearlyallprogramminglanguagesdesignedsince1980supportdataabstraction

IntroductiontoDataAbstraction–CO3
 AnAbstractDataTypeisauser-
defineddatatypethatsatisfiesthefollowingtwoconditions:
– Therepresentationof,andoperationson,objectsofthetypearedefinedinasinglesynta
cticunit
– Therepresentationofobjectsofthetypeishiddenfromtheprogramunitsthatusethes
eobjects,sotheonlyoperationspossiblearethoseprovidedinthetype'sdefinition

AdvantagesofDataAbstraction
 Advantageofthefirstcondition
– Programorganization,modifiability(everythingassociatedwithadatastructurei
stogether),andseparatecompilation
 Advantagethesecondcondition
– Reliability--
byhidingthedatarepresentations,usercodecannotdirectlyaccessobjectsofthetype
ordependontherepresentation,allowingtherepresentationtobechangedwithoutaf
fectingusercode
LanguageRequirementsforADTs:
 Asyntacticunitinwhichtoencapsulatethetypedefinition
 Amethodofmakingtypenamesandsubprogramheadersvisibletoclients,whilehidingac
tualdefinitions
 SomeprimitiveoperationsmustbebuiltintothelanguageprocessorDesignI
ssues:
 Canabstracttypesbeparameterized?
 Whataccesscontrolsareprovided?

LanguageExamplesCO2,CO3,CO4
LanguageExamples:Ada
 Theencapsulationconstructiscalledapackage
– Specificationpackage(theinterface)
– Bodypackage(implementationoftheentitiesnamedinthespecification)
 InformationHiding
– Thespecpackagehastwoparts,publicandprivate
– Thenameoftheabstracttypeappearsinthepublicpartofthespecificationpackage.Thi
spartmayalsoincluderepresentationsofunhiddentypes

53
 – Therepresentationoftheabstracttypeappearsinapartofthespecificationcalledthep
rivatepart
 Morerestrictedformwithlimitedprivatetypes
– Privatetypeshavebuilt-inoperationsforassignmentandcomparison
– LimitedprivatetypeshaveNObuilt-inoperations
 Reasonsforthepublic/privatespecpackage:
1. Thecompilermustbeabletoseetherepresentationafterseeingonlythespecpacka
ge(itcannotseetheprivatepart)
2. Clientsmustseethetypename,butnottherepresentation(theyalsocannotseethe
privatepart)
 Havingpartoftheimplementationdetails(therepresentation)inthespecpackageandpa
rt(themethodbodies)inthebodypackageisnotgood

Onesolution:makeallADTspointers
Problemswiththis:
1. Difficultieswithpointers
2. Objectcomparisons
3. ControlofobjectallocationislostAnE
xampleinAda
packageStack_Packis
typestack_typeislimitedprivate;max_size:
constant:=100;
functionempty(stk:instack_type)returnBoolean;procedurepu
sh(stk:inoutstack_type;elem:inInteger);procedurepop(stk:in
outstack_type);
functiontop(stk:instack_type)returnInteger;private--
hiddenfromclients
typelist_typeisarray(1..max_size)ofInteger;typesta
ck_typeisrecord
list:list_type;
topsub:Integerrange0..max_size):=0;end
record;
endStack_Pack

LanguageExamples:C++
 BasedonCstructtypeandSimula67classes
 Theclassistheencapsulationdevice
 Alloftheclassinstancesofaclassshareasinglecopyofthememberfunctions
 Eachinstanceofaclasshasitsowncopyoftheclassdatamembers
 Instancescanbestatic,stackdynamic,orheapdynamic
 InformationHiding
– Privateclauseforhiddenentities
– Publicclauseforinterfaceentities
– Protectedclauseforinheritance(Chapter12)
 Constructors:
– Functionstoinitializethedatamembersofinstances(theydonotcreatetheobjects)
– Mayalsoallocatestorageifpartoftheobjectisheap-dynamic
– Canincludeparameterstoprovideparameterizationoftheobjects
– Implicitlycalledwhenaninstanceiscreated
– Canbeexplicitlycalled

54
 – Nameisthesameastheclassname
 Destructors
– Functionstocleanupafteraninstanceisdestroyed;usuallyjusttoreclaimheapstor
age
– Implicitlycalledwhentheobject‘slifetimeends
– Canbeexplicitlycalled
– Nameistheclassname,precededbyatilde(~)AnExa
mpleinC++
classstack{
private:
int*stackPtr,maxLen,topPtr;pu
blic:
stack(){//aconstructorstackPt
r=newint[100];maxLen=99
;
topPtr=-1;
};
~stack(){delete[]stackPtr;};voi
dpush(intnum){…};voidpop(){
…};
inttop(){…};
intempty(){…};
}
EvaluationofADTsinC++andAda
 C++supportforADTsissimilartoexpressivepowerofAda
 Bothprovideeffectivemechanismsforencapsulationandinformationhiding
 Adapackagesaremoregeneralencapsulations;classesaretypes
 Friendfunctionsorclasses-
toprovideaccesstoprivatememberstosomeunrelatedunitsorfunctions
– NecessaryinC++
LanguageExamples:Java
 SimilartoC++,except:
– Alluser-definedtypesareclasses
– Allobjectsareallocatedfromtheheapandaccessedthroughreferencevariables
– Individualentitiesinclasseshaveaccesscontrolmodifiers(privateorpublic),rathert
hanclauses
– Javahasasecondscopingmechanism,packagescope,whichcanbeusedinplaceoff
riends
 Allentitiesinallclassesinapackagethatdonothaveaccesscontrolmodifiersarevisibl
ethroughoutthepackage
AnExampleinJava
classStackClass{priva
te:
privateint[]*stackRef;
privateint[]maxLen,topIndex;publicSt
ackClass(){//aconstructor
stackRef=newint[100];ma
xLen=99;
topPtr=-1;
};
public void push (int num)
{…};public void pop () {…};
publicinttop(){…};
publicbooleanempty(){…};
}

55
LanguageExamples:C#
 BasedonC++andJava
 Addstwoaccessmodifiers,internalandprotectedinternal
 Allclassinstancesareheapdynamic
 Defaultconstructorsareavailableforallclasses
 Garbagecollectionisusedformostheapobjects,sodestructorsarerarelyused
 structsarelightweightclassesthatdonotsupportinheritance
 Commonsolutiontoneedforaccesstodatamembers:accessormethods(getterandsetter
)
 C#providespropertiesasawayofimplementinggettersandsetterswithoutrequiringexp
licitmethodcalls

C#PropertyExample
publicclassWeather{
publicintDegreeDays{//**DegreeDaysisapropertyget{ret
urndegreeDays;}
set{
if(value<0||value>30)
Console.WriteLine("Valueisoutofrange:{0}",value);elsede
greeDays=value;}
}
privateintdegreeDays;
...
}
...
Weatherw=newWeather();
intdegreeDaysToday,oldDegreeDays;
...
w.DegreeDays=degreeDaysToday;
...
oldDegreeDays=w.DegreeDays;

ParameterizedAbstractDataTypes-CO4
 ParameterizedADTsallowdesigninganADTthatcanstoreanytypeelements(amongoth
erthings)
 Alsoknownasgenericclasses
 C++,Ada,Java5.0,andC#2005providesupportforparameterizedADTs
ParameterizedADTsinAda
 AdaGenericPackages
– Makethestacktypemoreflexiblebymakingtheelementtypeandthesizeofthestackge
neric
generic
Max_Size:Positive;
typeElem_Typeisprivate;packa
geGeneric_Stackis
TypeStack_Typeislimitedprivate;
functionTop(Stk:inoutStackType)returnElem_type;
…
endGeneric_Stack;
PackageInteger_StackisnewGeneric_Stack(100,Integer);Package
Float_StackisnewGeneric_Stack(100,Float);
ParameterizedADTsinC++
 Classes can be somewhat generic by writing parameterized
constructorfunctions
classstack{
…
56
 stack(intsize){stk_ptr=n
ewint[size];max_len=siz
e-1;
top=-1;
};
…
}
stackstk(100);
 Thestackelementtypecanbeparameterizedbymakingtheclassatemplatedclass
template<classType>cla
ssstack{
private:
Type*stackPtr;const
intmaxLen;inttopPt
r;
public:
stack(){
stackPtr=newType[100];m
axLen=99;
topPtr=-1;
}
…
}
ParameterizedClassesinJava5.0
 Genericparametersmustbeclasses
 Mostcommongenerictypesarethecollectiontypes,suchasLinkedListandArrayList
 Eliminatetheneedtocastobjectsthatareremoved
 Eliminatetheproblemofhavingmultipletypesinastructure
ParameterizedClassesinC#2005
 SimilartothoseofJava5.0
 Elementsofparameterizedstructurescanbeaccessedthroughindexing
SummaryofADT
 TheconceptofADTsandtheiruseinprogramdesignwasamilestoneinthedevelopmentof
languages
 TwoprimaryfeaturesofADTsarethepackagingofdatawiththeirassociatedoperationsa
ndinformationhiding
 AdaprovidespackagesthatsimulateADTs
 C++dataabstractionisprovidedbyclasses
 Java‘sdataabstractionissimilartoC++
 Ada,C++,Java5.0,andC#2005supportparameterizedADTs

Object-OrientedProgramming–CO4
 Abstractdatatypes
 Inheritance
– InheritanceisthecentralthemeinOOPandlanguagesthatsupportit
 Polymorphism
Inheritance
 Productivityincreasescancomefromreuse
– ADTsaredifficulttoreuse—alwaysneedchanges
– AllADTsareindependentandatthesamelevel
 Inheritanceallowsnewclassesdefinedintermsofexistingones,i.e.,byallowingthemtoin
heritcommonparts

57
  Inheritanceaddressesbothoftheaboveconcerns--
reuseADTsafterminorchangesanddefineclassesinahierarchy
Object-OrientedConcepts
 ADTsareusuallycalledclasses
 Classinstancesarecalledobjects
 Aclassthatinheritsisaderivedclassorasubclass
 Theclassfromwhichanotherclassinheritsisaparentclassorsuperclass
 Subprogramsthatdefineoperationsonobjectsarecalledmethods
 Callstomethodsarecalledmessages
 Theentirecollectionofmethodsofanobjectiscalleditsmessageprotocolor
messageinterface
 Messageshavetwoparts--amethodnameandthedestinationobject
 Inthesimplestcase,aclassinheritsalloftheentitiesofitsparent
 Inheritancecanbecomplicatedbyaccesscontrolstoencapsulatedentities
– Aclasscanhideentitiesfromitssubclasses
– Aclasscanhideentitiesfromitsclients
– Aclasscanalsohideentitiesforitsclientswhileallowingitssubclassestoseethem
 Besidesinheritingmethodsasis,aclasscanmodifyaninheritedmethod
– Thenewoneoverridestheinheritedone
– Themethodintheparentisoverriden
 Therearetwokindsofvariablesinaclass:
– Classvariables-one/class
– Instancevariables-one/object
 Therearetwokindsofmethodsinaclass:
– Classmethods–acceptmessagestotheclass
– Instancemethods–acceptmessagestoobjects
 Singlevs.MultipleInheritance
 Onedisadvantageofinheritanceforreuse:
– Createsinterdependenciesamongclassesthatcomplicatemaintenance
DynamicBinding
 Apolymorphicvariablecanbedefinedinaclassthatisabletoreference(orpointto)objects
oftheclassandobjectsofanyofitsdescendants
 Whenaclasshierarchyincludesclassesthatoverridemethodsandsuchmethodsarecalle
dthroughapolymorphicvariable,thebindingtothecorrectmethodwillbedynamic
 Allowssoftwaresystemstobemoreeasilyextendedduringbothdevelopmentandmainte
nance
DynamicBindingConcepts
 Anabstractmethodisonethatdoesnotincludeadefinition(itonlydefinesaprotocol)
 Anabstractclassisonethatincludesatleastonevirtualmethod
 Anabstractclasscannotbeinstantiated

DesignIssuesforOOPLanguages–CO4
 TheExclusivityofObjects
 AreSubclassesSubtypes?
 TypeCheckingandPolymorphism
 SingleandMultipleInheritance
 ObjectAllocationandDeAllocation

58
  DynamicandStaticBinding
 NestedClasses
TheExclusivityofObjects
 Everythingisanobject
– Advantage-eleganceandpurity
– Disadvantage-slowoperationsonsimpleobjects
 Addobjectstoacompletetypingsystem
– Advantage-fastoperationsonsimpleobjects
– Disadvantage-resultsinaconfusingtypesystem(twokindsofentities)
 Includeanimperative-styletypingsystemforprimitivesbutmakeeverythingelseobjects
– Advantage-fastoperationsonsimpleobjectsandarelativelysmalltypingsystem
– Disadvantage-stillsomeconfusionbecauseofthetwotypesystems
AreSubclassesSubtypes?
 Doesan“is-
a”relationshipholdbetweenaparentclassobjectandanobjectofthesubclass?
– Ifaderivedclassis-
aparentclass,thenobjectsofthederivedclassmustbehavethesameastheparentclass
object
 Aderivedclassisasubtypeifithasanis-arelationshipwithitsparentclass
– Subclasscanonlyaddvariablesandmethodsandoverrideinheritedmethodsin
“compatible”ways
TypeCheckingandPolymorphism
 Polymorphismmayrequiredynamictypecheckingofparametersandthereturnvalue
– Dynamictypecheckingiscostlyanddelayserrordetection
 Ifoverridingmethodsarerestrictedtohavingthesameparametertypesandreturntype,t
hecheckingcanbestatic
SingleandMultipleInheritance
 Multipleinheritanceallowsanewclasstoinheritfromtwoormoreclasses
 Disadvantagesofmultipleinheritance:
– Languageandimplementationcomplexity(inpartduetonamecollisions)
– Potentialinefficiency-
dynamicbindingcostsmorewithmultipleinheritance(butnotmuch)
 Advantage:
– Sometimesitisquiteconvenientandvaluable
AllocationandDeAllocationofObjects
 Fromwhereareobjectsallocated?
– IftheybehavelinetheADTs,theycanbeallocatedfromanywhere
 Allocatedfromtherun-timestack
 Explicitlycreateontheheap(vianew)
– Iftheyareallheap-
dynamic,referencescanbeuniformthruapointerorreferencevariable
 Simplifiesassignment-dereferencingcanbeimplicit
– Ifobjectsarestackdynamic,thereisaproblemwithregardtosubtypes
 Isdeallocationexplicitorimplicit?
DynamicandStaticBinding
 Shouldallbindingofmessagestomethodsbedynamic?
– Ifnoneare,youlosetheadvantagesofdynamicbinding
– Ifallare,itisinefficient
 Allowtheusertospecify

59
NestedClasses
 Ifanewclassisneededbyonlyoneclass,thereisnoreasontodefinesoitcanbeseenbyother
classes
– Canthenewclassbenestedinsidetheclassthatusesit?
– Insomecases,thenewclassisnestedinsideasubprogramratherthandirectlyinan
otherclass
 Otherissues:
– Whichfacilitiesofthenestingclassshouldbevisibletothenestedclassandviceversa

SupportforOOPinSmalltalk–CO4
 SmalltalkisapureOOPlanguage
– Everythingisanobject
– Allobjectshavelocalmemory
– Allcomputationisthroughobjectssendingmessagestoobjects
– Noneoftheappearancesofimperativelanguages
– Allobjectedareallocatedfromtheheap
– Alldeallocationisimplicit
 TypeCheckingandPolymorphism
– Allbindingofmessagestomethodsisdynamic
 Theprocessistosearchtheobjecttowhichthemessageissentforthemethod;ifnotfou
nd,searchthesuperclass,etc.uptothesystemclasswhichhasnosuperclass
– TheonlytypecheckinginSmalltalkisdynamicandtheonlytypeerroroccurswhen
amessageissenttoanobjectthathasnomatchingmethod
 Inheritance
– ASmalltalksubclassinheritsalloftheinstancevariables,instancemethods,andcla
ssmethodsofitssuperclass
– Allsubclassesaresubtypes(nothingcanbehidden)
– Allinheritanceisimplementationinheritance
– Nomultipleinheritance
 EvaluationofSmalltalk
– Thesyntaxofthelanguageissimpleandregular
– Goodexampleofpowerprovidedbyasmalllanguage
– Slowcomparedwithconventionalcompiledimperativelanguages
– Dynamicbindingallowstypeerrorstogoundetecteduntilruntime
– Introducedthegraphicaluserinterface
– Greatestimpact:advancementofOOP

SupportforOOPinC++-CO4
 GeneralCharacteristics:
– EvolvedfromCandSIMULA67
– AmongthemostwidelyusedOOPlanguages
– Mixedtypingsystem
– Constructorsanddestructors
– Elaborateaccesscontrolstoclassentities
 Inheritance
– Aclassneednotbethesubclassofanyclass
– Accesscontrolsformembersare
– Private(visibleonlyintheclassandfriends)(disallowssubclassesfrombeingsubtype
s)

60
 – Public(visibleinsubclassesandclients)
– Protected(visibleintheclassandinsubclasses,butnotclients)
 Inaddition,thesubclassingprocesscanbedeclaredwithaccesscontrols(privateorpu
blic),whichdefinepotentialchangesinaccessbysubclasses
– Privatederivation-
inheritedpublicandprotectedmembersareprivateinthesubclasses
– Public derivation public and protected members are also public
andprotectedinsubclasses
InheritanceExampleinC++
classbase_class{priva
te:
inta;float
x;
protected:
intb;floa
ty;public
:intc;floa
tz;
};
classsubclass_1:publicbase_class{…};
//Inthisone,bandyareprotectedand
//candzarepublic
classsubclass_2:privatebase_class{…};
//Inthisone,b,y,c,andzareprivate,
//andnoderivedclasshasaccesstoany
//memberofbase_class
 Amemberthatisnotaccessibleinasubclass(becauseofprivatederivation)canbedeclare
dtobevisiblethereusingthescoperesolutionoperator(::),e.g.,
classsubclass_3:privatebase_class{base_cl
ass::c;
…
}
 Onemotivationforusingprivatederivation
– Aclassprovidesmembersthatmustbevisible,sotheyaredefinedtobepublicmembers
;aderivedclassaddssomenewmembers,butdoesnotwantitsclientstoseethemembe
rsoftheparentclass,eventhoughtheyhadtobepublicintheparentclassdefinition
 Multipleinheritanceissupported
– Iftherearetwoinheritedmemberswiththesamename,theycanbothbereferencedusi
ngthescoperesolutionoperator
 DynamicBinding
– Amethodcanbedefinedtobevirtual,whichmeansthattheycanbecalledthroughp
olymorphicvariablesanddynamicallyboundtomessages
– Apurevirtualfunctionhasnodefinitionatall
– Aclassthathasatleastonepurevirtualfunctionisanabstractclass
 Evaluation
– C++providesextensiveaccesscontrols(unlikeSmalltalk)
– C++providesmultipleinheritance
– InC++,theprogrammermustdecideatdesigntimewhichmethodswillbestatically
boundandwhichmustbedynamicallybound
 Staticbindingisfaster!
– Smalltalktypecheckingisdynamic(flexible,butsomewhatunsafe)
– Becauseofinterpretationanddynamicbinding,Smalltalkis~10timesslowerthanC+
+

61
SupportforOOPinJava–CO4
 BecauseofitscloserelationshiptoC++,focusisonthedifferencesfromthatlanguage
 GeneralCharacteristics
– Alldataareobjectsexcepttheprimitivetypes
– Allprimitivetypeshavewrapperclassesthatstoreonedatavalue
– Allobjectsareheap-
dynamic,arereferencedthroughreferencevariables,andmostareallocatedwithnew
– Afinalizemethodisimplicitlycalledwhenthegarbagecollectorisabouttoreclaimthes
torageoccupiedbytheobject
 Inheritance
– Singleinheritancesupportedonly,butthereisanabstractclasscategorythatprovi
dessomeofthebenefitsofmultipleinheritance(interface)
– Aninterfacecanincludeonlymethoddeclarationsandnamedconstants,
e.g.,
publicinterfaceComparable{
publicintcomparedTo(Objectb);}
– Methodscanbefinal(cannotbeoverriden)
 DynamicBinding
– InJava,allmessagesaredynamicallyboundtomethods,unlessthemethodisfinal(i.
e.,itcannotbeoverriden,thereforedynamicbindingservesnopurpose)
– Staticbindingisalsousedifthemethodsisstaticorprivatebothofwhichdisallowoverri
ding
 Severalvarietiesofnestedclasses
 Allarehiddenfromallclassesintheirpackage,exceptforthenestingclass
 Nestedclassescanbeanonymous
 Alocalnestedclassisdefinedinamethodofitsnestingclass
– Noaccessspecifierisused
 Evaluation
– DesigndecisionstosupportOOParesimilartoC++
– Nosupportforproceduralprogramming
– Noparentlessclasses
– Dynamicbindingisusedas“normal”waytobindmethodcallstomethoddefinitions
– Usesinterfacestoprovideasimpleformofsupportformultipleinheritance

SupportforOOPinC#-CO4
 Generalcharacteristics
– SupportforOOPsimilartoJava
– Includesbothclassesandstructs
– ClassesaresimilartoJava‘sclasses
– structsarelesspowerfulstack-dynamicconstructs(e.g.,noinheritance)
 Inheritance
– UsesthesyntaxofC++fordefiningclasses
– Amethodinheritedfromparentclasscanbereplacedinthederivedclassbymarkingits
definitionwithnew
– Theparentclassversioncanstillbecalledexplicitlywiththeprefixbase:base.Draw()
 Dynamicbinding
– Toallowdynamicbindingofmethodcallstomethods:

62
 Thebaseclassmethodismarkedvirtual
 Thecorrespondingmethodsinderivedclassesaremarkedoverride
– Abstractmethodsaremarkedabstractandmustbeimplementedinallsubclasses
– AllC#classesareultimatelyderivedfromasinglerootclass,Object
 NestedClasses
– AC#classthatisdirectlynestedinanestingclassbehaveslikeaJavastaticnestedcla
ss
– C#doesnotsupportnestedclassesthatbehavelikethenon-staticclassesofJava
 Evaluation
– C#isthemostrecentlydesignedC-basedOOlanguage
– ThedifferencesbetweenC#‘sandJava‘ssupportforOOParerelativelyminor

SupportforOOPinAda95–CO4
 GeneralCharacteristics
– OOPwasoneofthemostimportantextensionstoAda83
– Encapsulationcontainerisapackagethatdefinesataggedtype
– Ataggedtypeisoneinwhicheveryobjectincludesatagtoindicateduringexecutionitst
ype(thetagsareinternal)
– Taggedtypescanbeeitherprivatetypesorrecords
– Noconstructorsordestructorsareimplicitlycalled
 Inheritance
– Subclassescanbederivedfromtaggedtypes
– Newentitiesareaddedtotheinheritedentitiesbyplacingtheminarecorddefinition
– Allsubclassesaresubtypes
– Nosupportformultipleinheritance
 AcomparableeffectcanbeachievedusinggenericclassesExamp
leofaTaggedType
PackagePerson_Pkgis
typePersonistaggedprivate;procedure
Display(P:inoutPerson);
private
typePersonistaggedreco
rd
Name:String(1..30);Addr
ess:String(1..30);Age:Int
eger;
endrecord;en
dPerson_Pkg;
withPerson_Pkg;usePerson_Pkg;packag
eStudent_Pkgis
typeStudentisnewPersonwithreco
rd
Grade_Point_Average:Float;G
rade_Level:Integer;
endrecord;
procedureDisplay(St:inStudent);e
ndStudent_Pkg;
//Note:DisplayisbeingoverriddenfromPerson_Pkg
 DynamicBinding
– Dynamicbindingisdoneusingpolymorphicvariablescalledclasswidetypes
 ForthetaggedtypePrtdon,theclasswidetypeisPerson‘class
– Otherbindingsarestatic

63
 – Anymethodmaybedynamicallybound
– PurelyabstractbasetypescanbedefinedinAda95byincludingthereservedworda
bstract
 Evaluation
– AdaofferscompletesupportforOOP
– C++offersbetterformofinheritancethanAda
– Adaincludesnoinitializationofobjects(e.g.,constructors)
– DynamicbindinginC-
basedOOPlanguagesisrestrictedtopointersand/orreferencestoobjects;Adahasnos
uchrestrictionandisthusmoreorthogonal

ImplementingOOPsConstructs
 Twointerestingandchallengingparts
– Storagestructuresforinstancevariables
– Dynamicbindingofmessagestomethods
InstanceDataStorage
 Classinstancerecords(CIRs)storethestateofanobject
– Static(builtatcompiletime)
 Ifaclasshasaparent,thesubclassinstancevariablesareaddedtotheparentCIR
 BecauseCIRisstatic,accesstoallinstancevariablesisdoneasitisinrecords
– Efficient
DynamicBindingofMethodsCalls
 MethodsinaclassthatarestaticallyboundneednotbeinvolvedintheCIR;methodstha
twillbedynamicallyboundmusthaveentriesintheCIR
– Callstodynamicallyboundmethodscanbeconnectedtothecorrespondingcodethrua
pointerintheCIR
– Thestoragestructureissometimescalledvirtualmethodtables(vtable)
– Methodcallscanberepresentedasoffsetsfromthebeginningofthevtable
SummaryofOOPs
 OOprogramminginvolvesthreefundamentalconcepts:ADTs,inheritance,dynamic
binding
 Majordesignissues:exclusivityofobjects,subclassesandsubtypes,typecheckingandpol
ymorphism,singleandmultipleinheritance,dynamicbinding,explicitandimplicitde-
allocationofobjects,andnestedclasses
 SmalltalkisapureOOL
 C++hastwodistincttypesystem(hybrid)
 JavaisnotahybridlanguagelikeC++;itsupportsonlyOOprogramming
 C#isbasedonC++andJava
 ImplementingOOPinvolvessomenewdatastructures

Concurrency
 Concurrencycanoccuratfourlevels:
– Machineinstructionlevel
– High-levellanguagestatementlevel
– Unitlevel
– Programlevel
 Becausetherearenolanguageissuesininstruction-andprogram-
levelconcurrency,theyarenotaddressedhere

64
MultiprocessorArchitectures
 Late1950s-onegeneral-
purposeprocessorandoneormorespecialpurposeprocessorsforinputandoutputopera
tions
 Early1960s-multiplecompleteprocessors,usedforprogram-levelconcurrency
 Mid-1960s-multiplepartialprocessors,usedforinstruction-levelconcurrency
 Single-InstructionMultiple-Data(SIMD)machines
 Multiple-InstructionMultiple-Data(MIMD)machines
– Independentprocessorsthatcanbesynchronized(unit-levelconcurrency)
CategoriesofConcurrency
 Athreadofcontrolinaprogramisthesequenceofprogrampointsreachedascontrolflowst
hroughtheprogram
 CategoriesofConcurrency:
– Physicalconcurrency-Multipleindependentprocessors(multiplethreadsofcontrol)
– Logicalconcurrency-Theappearanceofphysicalconcurrencyispresentedbytime-
sharingoneprocessor(softwarecanbedesignedasifthereweremultiplethreadsofco
ntrol)
 Coroutines(quasi-concurrency)haveasinglethreadofcontrol
MotivationsforStudyingConcurrency
 Involvesadifferentwayofdesigningsoftwarethatcanbeveryuseful—manyreal-
worldsituationsinvolveconcurrency
 Multiprocessorcomputerscapableofphysicalconcurrencyarenowwidelyused

Subprogram-LevelConcurrency
 Ataskorprocessisaprogramunitthatcanbeinconcurrentexecutionwithotherprogra
munits
 Tasksdifferfromordinarysubprogramsinthat:
– Ataskmaybeimplicitlystarted
– Whenaprogramunitstartstheexecutionofatask,itisnotnecessarilysuspended
– Whenatask‘sexecutioniscompleted,controlmaynotreturntothecaller
 Tasksusuallyworktogether
TwoGeneralCategoriesofTasks
 Heavyweighttasksexecuteintheirownaddressspace
 Lightweighttasksallruninthesameaddressspace
 Ataskisdisjointifitdoesnotcommunicatewithoraffecttheexecutionofanyothertaskinth
eprograminanyway
TaskSynchronization
 Amechanismthatcontrolstheorderinwhichtasksexecute
 Twokindsofsynchronization
– Cooperationsynchronization
– Competitionsynchronization
 Taskcommunicationisnecessaryforsynchronization,providedby:
– Sharednonlocalvariables
– Parameters
– Messagepassing
Kindsofsynchronization
 Cooperation:TaskAmustwaitfortaskBtocompletesomespecificactivitybeforetaskAca
ncontinueitsexecution,e.g.,theproducer-consumerproblem

65
  Competition:Twoormoretasksmustusesomeresourcethatcannotbesimultaneousl
yused,e.g.,asharedcounter
– Competitionisusuallyprovidedbymutuallyexclusiveaccess(approachesarediscuss
edlater)
–

Figure6.1NeedforCompetitionSynchronization
Scheduler
 Providingsynchronizationrequiresamechanismfordelayingtaskexecution
 Taskexecutioncontrolismaintainedbyaprogramcalledthescheduler,whichmapsta
skexecutionontoavailableprocessors
TaskExecutionStates
 New-createdbutnotyetstarted
 Ready-readytorunbutnotcurrentlyrunning(noavailableprocessor)
 Running
 Blocked-
hasbeenrunning,butcannotnowcontinue(usuallywaitingforsomeeventtooccur)
 Dead-nolongeractiveinanysense
LivenessandDeadlock
 Livenessisacharacteristicthataprogramunitmayormaynothave
– Insequentialcode,itmeanstheunitwilleventuallycompleteitsexecution
 Inaconcurrentenvironment,ataskcaneasilyloseitsliveness
 Ifalltasksinaconcurrentenvironmentlosetheirliveness,itiscalleddeadlock
DesignIssuesforConcurrency
 Competitionandcooperationsynchronization
 Controllingtaskscheduling
 Howandwhentasksstartandendexecution
 Howandwhenaretaskscreated
MethodsofProvidingSynchronization
 Semaphores
 Monitors
 MessagePassing

Semaphores
 Dijkstra-1965
 Asemaphoreisadatastructureconsistingofacounterandaqueueforstoringtaskdesc
riptors
 Semaphorescanbeusedtoimplementguardsonthecodethataccessesshareddatastr
uctures
 Semaphoreshaveonlytwooperations,waitandrelease(originallycalledPand
VbyDijkstra)

66
  Semaphores can be used to provide both competition and
cooperationsynchronization
CooperationSynchronizationwithSemaphores
 Example:Asharedbuffer
 ThebufferisimplementedasanADTwiththeoperationsDEPOSITandFETCHastheonlyw
aystoaccessthebuffer
 Usetwosemaphoresforcooperation:emptyspotsandfullspots
 Thesemaphorecountersareusedtostorethenumbersofemptyspotsandfullspotsintheb
uffer
 DEPOSITmustfirstcheckemptyspotstoseeifthereisroominthebuffer
 Ifthereisroom,thecounterofemptyspotsisdecrementedandthevalueisinserted
 Ifthereisnoroom,thecallerisstoredinthequeueofemptyspots
 WhenDEPOSITisfinished,itmustincrementthecounteroffullspots
 FETCHmustfirstcheckfullspotstoseeifthereisavalue
– Ifthereisafullspot,thecounteroffullspotsisdecrementedandthevalueisremoved
– Iftherearenovaluesinthebuffer,thecallermustbeplacedinthequeueoffullspots
– WhenFETCHisfinished,itincrementsthecounterofemptyspots
 TheoperationsofFETCHandDEPOSITonthesemaphoresareaccomplishedthrought
wosemaphoreoperationsnamedwaitandrelease

Semaphores:WaitOperation
wait(aSemaphore)
ifaSemaphore‘scounter>0thendecrement
aSemaphore‘scounter
else
putthecallerinaSemaphore‘squeueattemp
ttotransfercontroltoareadytask
--ifthetaskreadyqueueisempty,
--deadlockoccurs
end

Semaphores:ReleaseOperation
release(aSemaphore)
ifaSemaphore‘squeueisemptythenincrementa
Semaphore‘scounter
else
putthecallingtaskinthetaskreadyqueuetransfercont
roltoataskfromaSemaphore‘squeue
end

ProducerConsumerCode
semaphorefullspots,emptyspots;fullst
ops.count=0;emptyspots.coun
t=BUFLEN;taskproducer;
loop
--produceVALUE–-
wait(emptyspots);{waitforspace}DEPO
SIT(VALUE);
release(fullspots);{increasefilled}e
ndloop;
endproducer;

67
ProducerConsumerCode
taskconsumer;
loop
wait(fullspots);{waittillnotempty}}FET
CH(VALUE);
release(emptyspots);{increaseempty}
--consumeVALUE–-
endloop;
endconsumer;
CompetitionSynchronizationwithSemaphores
 Athirdsemaphore,namedaccess,isusedtocontrolaccess(competitionsynchronization)
– Thecounterofaccesswillonlyhavethevalues0and1
– Suchasemaphoreiscalledabinarysemaphore
 Notethatwaitandreleasemustbeatomic!
ProducerConsumerCode
semaphoreaccess,fullspots,emptyspots;access.count
=0;
fullstops.count=0;emptyspots.
count=BUFLEN;taskproducer;
loop
--produceVALUE–-
wait(emptyspots);{waitforspace}wait(
access);{waitforaccess)DEPOSIT(VAL
UE);
release(access);{relinquishaccess}relea
se(fullspots);{increasefilled}
endloop;
endproducer;

ProducerConsumerCode
taskconsumer;
loop
wait(fullspots);{waittillnotempty}wait(
access);{waitforaccess}FETCH(VALUE);
release(access);{relinquishaccess}rele
ase(emptyspots);{increaseempty}
--consumeVALUE–-
endloop;
endconsumer;
EvaluationofSemaphores
 Misuseofsemaphorescancausefailuresincooperationsynchronization,
e.g.,thebufferwilloverflowifthewaitoffullspotsisleftout
 Misuseofsemaphorescancausefailuresincompetitionsynchronization,
e.g.,theprogramwilldeadlockifthereleaseofaccessisleftout

Monitors
 Ada,Java,C#
 Theidea:encapsulatetheshareddataanditsoperationstorestrictaccess
 Amonitorisanabstractdatatypeforshareddata
CompetitionSynchronization
 Shareddataisresidentinthemonitor(ratherthanintheclientunits)
 Allaccessresidentinthemonitor

68
 – Monitorimplementationguaranteesynchronizedaccessbyallowingonlyoneaccessa
tatime
– Callstomonitorproceduresareimplicitlyqueuedifthemonitorisbusyatthetimeofthe
call
CooperationSynchronization
 Cooperationbetweenprocessesisstillaprogrammingtask
– Programmermustguaranteethatasharedbufferdoesnotexperienceunderfloworov
erflow

Figure6.2CooperationSynchronization
EvaluationofMonitors
 Abetterwaytoprovidecompetitionsynchronizationthanaresemaphores
 Semaphorescanbeusedtoimplementmonitors
 Monitorscanbeusedtoimplementsemaphores
 Supportforcooperationsynchronizationisverysimilaraswithsemaphores,soithast
hesameproblems

MessagePassing
 Messagepassingisageneralmodelforconcurrency
– Itcanmodelbothsemaphoresandmonitors
– Itisnotjustforcompetitionsynchronization
 Centralidea:taskcommunicationislikeseeingadoctor--
mostofthetimeshewaitsforyouoryouwaitforher,butwhenyouarebothready,yougetto
gether,orrendezvous
MessagePassingRendezvous
 Tosupportconcurrenttaskswithmessagepassing,alanguageneeds:
– Amechanismtoallowatasktoindicatewhenitiswillingtoacceptmessages
– Awaytorememberwhoiswaitingtohaveitsmessageacceptedandsome
“fair”wayofchoosingthenextmessage
 Whenasendertask‘smessageisacceptedbyareceivertask,theactualmessagetransm
issioniscalledarendezvous

AdaSupportforConcurrency
 TheAda83Message-PassingModel
– Adataskshavespecificationandbodyparts,likepackages;thespechastheinterfac
e,whichisthecollectionofentrypoints:
taskTask_Exampleis
entryENTRY_1(Item:inInteger);endT
ask_Example;

69
TaskBody
 Thebodytaskdescribestheactionthattakesplacewhenarendezvousoccurs
 Ataskthatsendsamessageissuspendedwhilewaitingforthemessagetobeacceptedandd
uringtherendezvous
 Entrypointsinthespecaredescribedwithacceptclausesinthebodyaccept
entry_name(formalparameters)do
…
endentry_name

ExampleofaTaskBody
taskbodyTask_Exampleisbegi
n
loop
acceptEntry_1(Item:inFloat)do
...
endEntry_1;
endloop;en
dTask_Example;

AdaMessagePassingSemantics
 Thetaskexecutestothetopoftheacceptclauseandwaitsforamessage
 Duringexecutionoftheacceptclause,thesenderissuspended
 acceptparameterscantransmitinformationineitherorbothdirections
 Everyacceptclausehasanassociatedqueuetostorewaitingmessages

Figure6.3RendezvousTimeLines

MessagePassing:Server/ActorTasks
 Ataskthathasacceptclauses,butnoothercodeiscalledaservertask(theexampleabov
eisaservertask)
 Ataskwithoutacceptclausesiscalledanactortask
– Anactortaskcansendmessagestoothertasks
– Note:Asendermustknowtheentrynameofthereceiver,butnotviceversa(asymmetri
c)

70
 Figure6.4GraphicalRepresentationofaRendezvous

Example:ActorTask
taskWater_Monitor;--specificationtaskbodybodyWater_Monitoris--bodybegin
loop
ifWater_Level>Max_Levelthe
nSound_Alarm;
endif;
delay1.0;--Nofurtherexecution
--foratleast1second
endloop;end
Water_Monitor;
MultipleEntryPoints
 Taskscanhavemorethanoneentrypoint
– Thespecificationtaskhasanentryclauseforeach
– Thetaskbodyhasanacceptclauseforeachentryclause,placedinaselectclause,whi
chisinaloop
ATaskwithMultipleEntries
taskbodyTellerisloop
select
acceptDrive_Up(formalparams)do
...
endDrive_Up;
...
or
acceptWalk_Up(formalparams)do
...
endWalk_Up;
...
endselect;en
dloop;
endTeller;

SemanticsofTaskswithMultipleacceptClauses
 Ifexactlyoneentryqueueisnonempty,chooseamessagefromit
 Ifmorethanoneentryqueueisnonempty,chooseone,nondeterministically,fromwhi
chtoacceptamessage
 Ifallareempty,wait
 Theconstructisoftencalledaselectivewait
 Extendedacceptclause-codefollowingtheclause,butbeforethenextclause
– Executedconcurrentlywiththecaller

71
CooperationSynchronizationwithMessagePassing
 ProvidedbyGuardedacceptclauses
whennotFull(Buffer)=>acceptD
eposit(New_Value)do
 Anacceptclausewithawithawhenclauseiseitheropenorclosed
– Aclausewhoseguardistrueiscalledopen
– Aclausewhoseguardisfalseiscalledclosed
– Aclausewithoutaguardisalwaysopen

SemanticsofselectwithGuardedacceptClauses:
 selectfirstcheckstheguardsonallclauses
 Ifexactlyoneisopen,itsqueueischeckedformessages
 Ifmorethanoneareopen,non-
deterministicallychooseaqueueamongthemtocheckformessages
 Ifallareclosed,itisaruntimeerror
 Aselectclausecanincludeanelseclausetoavoidtheerror
– Whentheelseclausecompletes,thelooprepeats

ExampleofaTaskwithGuardedacceptClauses
 Note:Thestationmaybeoutofgasandtheremayormaynotbeapositionavailableinthegar
age
taskGas_Station_Attendantis
entryService_Island(Car:Car_Type);entr
yGarage(Car:Car_Type);
endGas_Station_Attendant;

ExampleofaTaskwithGuardedacceptClauses
taskbodyGas_Station_Attendantisbegi
n
loop
select
whenGas_Available=>
acceptService_Island(Car:Car_Type)doFill_With_Gas(C
ar);
endService_Island;or
whenGarage_Available=>
acceptGarage(Car:Car_Type)doFix(C
ar);
endGarage;else
Sleep;
endselect;
endloop;
endGas_Station_Attendant;
CompetitionSynchronizationwithMessagePassing
 Modelingmutuallyexclusiveaccesstoshareddata
 Example--asharedbuffer
 Encapsulatethebufferanditsoperationsinatask
 Competitionsynchronizationisimplicitinthesemanticsofacceptclauses
– Onlyoneacceptclauseinataskcanbeactiveatanygiventime

72
TaskTermination
 Theexecutionofataskiscompletedifcontrolhasreachedtheendofitscodebody
 Ifataskhascreatednodependenttasksandiscompleted,itisterminated
 Ifataskhascreateddependenttasksandiscompleted,itisnotterminateduntilallitsdepen
denttasksareterminated
TheterminateClause
 Aterminateclauseinaselectisjustaterminatestatement
 Aterminateclauseisselectedwhennoacceptclauseisopen
 Whenaterminateisselectedinatask,thetaskisterminatedonlywhenitsmasterandalloft
hedependentsofitsmasterareeithercompletedorarewaitingataterminate
 Ablockorsubprogramisnotleftuntilallofitsdependenttasksareterminated
MessagePassingPriorities
 ThepriorityofanytaskcanbesetwiththepragmaprioritypragmaPriority(expression);
 Thepriorityofataskappliestoitonlywhenitisinthetaskreadyqueue

BinarySemaphores
 ForsituationswherethedatatowhichaccessistobecontrolledisNOTencapsulatedin
atask
taskBinary_Semaphoreisent
ryWait;
entryrelease;end
Binary_Semaphore;
taskbodyBinary_Semaphoreisbegin
loop
acceptWait;acce
ptRelease;endloo
p;
endBinary_Semaphore;

ConcurrencyinAda95
 Ada95includesAda83featuresforconcurrency,plustwonewfeatures
– Protectedobjects:Amoreefficientwayofimplementingshareddatatoallowaccesstoash
areddatastructuretobedonewithoutrendezvous
– Asynchronouscommunication

Ada95:ProtectedObjects
 Aprotectedobjectissimilartoanabstractdatatype
 Accesstoaprotectedobjectiseitherthroughmessagespassedtoentries,aswithatask,
orthroughprotectedsubprograms
 A protected procedure provides mutually exclusive read-write access
toprotectedobjects
 Aprotectedfunctionprovidesconcurrentread-onlyaccesstoprotectedobjects

AsynchronousCommunication
 Providedthroughasynchronousselectstructures
 Anasynchronousselecthastwotriggeringalternatives,anentryclauseoradelay
– Theentryclauseistriggeredwhensentamessage
– Thedelayclauseistriggeredwhenitstimelimitisreached

73
EvaluationoftheAda
 Messagepassingmodelofconcurrencyispowerfulandgeneral
 Protectedobjectsareabetterwaytoprovidesynchronizedshareddata
 Intheabsenceofdistributedprocessors,thechoicebetweenmonitorsandtaskswithmess
agepassingissomewhatamatteroftaste
 Fordistributedsystems,messagepassingisabettermodelforconcurrency

JavaThreads
 TheconcurrentunitsinJavaaremethodsnamedrun
– Arunmethodcodecanbeinconcurrentexecutionwithothersuchmethods
– Theprocessinwhichtherunmethodsexecuteiscalledathread
ClassmyThreadextendsThread{
publicvoidrun(){…}
}
…
ThreadmyTh=newMyThread();myTh.st
art();

ControllingThreadExecution
 TheThreadclasshasseveralmethodstocontroltheexecutionofthreads
– Theyieldisarequestfromtherunningthreadtovoluntarilysurrendertheprocessor
– Thesleepmethodcanbeusedbythecallerofthemethodtoblockthethread
– Thejoinmethodisusedtoforceamethodtodelayitsexecutionuntiltherunmethodofa
notherthreadhascompleteditsexecution

ThreadPriorities
 Athread‘sdefaultpriorityisthesameasthethreadthatcreateit
– Ifmaincreatesathread,itsdefaultpriorityisNORM_PRIORITY
 Threads defined two other priority constants, MAX_PRIORITY
andMIN_PRIORITY
 ThepriorityofathreadcanbechangedwiththemethodssetPriority

CompetitionSynchronizationwithJavaThreads
 Amethodthatincludesthesynchronizedmodifierdisallowsanyothermethodfromrunni
ngontheobjectwhileitisinexecution
…
public synchronized void deposit( int i) {…}
public synchronized int fetch() {…}
…
 Theabovetwomethodsaresynchronizedwhichpreventsthemfrominterferingwitheac
hother
 Ifonlyapartofamethodmustberunwithoutinterference,itcanbesynchronizedthroughs
ynchronizedstatement
synchronized(expression)
statement
CooperationSynchronizationwithJavaThreads
 CooperationsynchronizationinJavaisachievedviawait,notify,andnotifyAllmethod
s
– AllmethodsaredefinedinObject,whichistherootclassinJava,soallobjectsinheritthe
m
 Thewaitmethodmustbecalledinaloop

74
  Thenotifymethodiscalledtotellonewaitingthreadthattheeventitwaswaitinghashappe
ned
 ThenotifyAllmethodawakensallofthethreadsontheobject‘swaitlist

Java’sThreadEvaluation
 Java‘ssupportforconcurrencyisrelativelysimplebuteffective
 NotaspowerfulasAda‘stasks

C#Threads
 LooselybasedonJavabuttherearesignificantdifferences
 Basicthreadoperations
– Anymethodcanruninitsownthread
– AthreadiscreatedbycreatingaThreadobject
– Creatingathreaddoesnotstartitsconcurrentexecution;itmustberequestedthro
ughtheStartmethod
– AthreadcanbemadetowaitforanotherthreadtofinishwithJoin
– AthreadcanbesuspendedwithSleep
– AthreadcanbeterminatedwithAbort

SynchronizingThreads
 ThreewaystosynchronizeC#threads
– TheInterlockedclass
 Usedwhentheonlyoperationsthatneedtobesynchronizedareincrementingordecreme
ntingofaninteger
– Thelockstatement
 Usedtomarkacriticalsectionofcodeinathreadlock(expression){…}
– TheMonitorclass
 Provides four methods that can be used to provide more
sophisticatedsynchronization
C#’sConcurrencyEvaluation
 AnadvanceoverJavathreads,e.g.,anymethodcanrunitsownthread
 ThreadterminationiscleanerthaninJava
 Synchronizationismoresophisticated

Statement-LevelConcurrency
 Objective:Provideamechanismthattheprogrammercanusetoinformcompilerofwaysit
canmaptheprogramontomultiprocessorarchitecture
 Minimizecommunicationamongprocessorsandthememoriesoftheotherprocessor
s
High-PerformanceFortran
 Acollectionofextensionsthatallowtheprogrammertoprovideinformationtothecompil
ertohelpitoptimizecodeformultiprocessorcomputers
 Specifythenumberofprocessors,thedistributionofdataoverthememoriesofthoseproc
essors,andthealignmentofdata
PrimaryHPFSpecifications
 Numberofprocessors
!HPF$PROCESSORSprocs(n)
 Distributionofdata
!HPF$DISTRIBUTE(kind)ONTOprocs::identifier_list
– kindcanbeBLOCK(distributedatatoprocessorsinblocks)orCYCLIC(
distributedatatoprocessorsoneelementatatime)
75
  Relatethedistributionofonearraywiththatofanother
ALIGNarray1_elementWITHarray2_element
Statement-
LevelConcurrencyExampleRE
ALlist_1(1000),list_2(1000)INTEGERlis
t_3(500),list_4(501)
!HPF$PROCESSORSproc(10)
!HPF$DISTRIBUTE(BLOCK)ONTOprocs::
list_1,list_2
!HPF$ALIGNlist_1(index)WITHlist_4
(index+1)
…
list_1(index)=list_2(index)list_3(i
ndex)=list_4(index+1)
 FORALLstatementisusedtospecifyalistofstatementsthatmaybeexecutedconcurrently
FORALL(index=1:1000)
list_1(index)=list_2(index)
 Specifiesthatall1,000RHSsoftheassignmentscanbeevaluatedbeforeanyassignmen
ttakesplace

Summary
 Concurrentexecutioncanbeattheinstruction,statement,orsubprogramlevel
 Physicalconcurrency:whenmultipleprocessorsareusedtoexecuteconcurrentunits
 Logicalconcurrency:concurrentunitedareexecutedonasingleprocessor
 Two primary facilities to support subprogram concurrency:
competitionsynchronizationandcooperationsynchronization
 Mechanisms:semaphores,monitors,rendezvous,threads
 High-
PerformanceFortranprovidesstatementsforspecifyinghowdataistobedistributedove
rthememoryunitsconnectedtomultipleprocessors

76
 ExceptionHandling&LogicProgrammingLanguage

IntroductiontoExceptionHandling
 Inalanguagewithoutexceptionhandling
– Whenanexceptionoccurs,controlgoestotheoperatingsystem,whereamessageisdis
playedandtheprogramisterminated
 Inalanguagewithexceptionhandling
– Programsareallowedtotrapsomeexceptions,therebyprovidingthepossibilityoffixi
ngtheproblemandcontinuing
BasicConcepts–CO3
 Manylanguagesallowprogramstotrapinput/outputerrors(includingEOF)
 Anexceptionisanyunusualevent,eithererroneousornot,detectablebyeitherhardwareo
rsoftware,thatmayrequirespecialprocessing
 Thespecialprocessingthatmayberequiredafterdetectionofanexceptioniscalledexcepti
onhandling
 Theexceptionhandlingcodeunitiscalledanexceptionhandler

ExceptionHandlingAlternatives
 Anexceptionisraisedwhenitsassociatedeventoccurs
 Alanguagethatdoesnothaveexceptionhandlingcapabilitiescanstilldefine,detect,ra
ise,andhandleexceptions(userdefined,softwaredetected)
 Alternatives:
– Sendanauxiliaryparameterorusethereturnvaluetoindicatethereturnstatusofasub
program
– Passanexceptionhandlingsubprogramtoallsubprograms

AdvantagesofBuilt-inExceptionHandling
 Errordetectioncodeistedioustowriteanditclutterstheprogram
 Exception handling encourages programmers to consider many
differentpossibleerrors
 Exceptionpropagationallowsahighlevelofreuseofexceptionhandlingcode

DesignIssues
 Howareuser-definedexceptionsspecified?
 Shouldtherebedefaultexceptionhandlersforprogramsthatdonotprovidetheirown?
 Canbuilt-inexceptionsbeexplicitlyraised?
 Arehardware-detectableerrorstreatedasexceptionsthatcanbehandled?
 Arethereanybuilt-inexceptions?
 Howcanexceptionsbedisabled,ifatall?
 Howandwhereexceptionhandlersspecifiedandwhataretheirscope?
 Howisanexceptionoccurrenceboundtoanexceptionhandler?
 Caninformationabouttheexceptionbepassedtothehandler?
 Wheredoesexecutioncontinue,ifatall,afteranexceptionhandlercompletesitsexecution
?(continuationvs.resumption)
 Issomeformoffinalizationprovided?

77
 Figure7.1ExceptionHandlingControlFlow

ExceptionHandlinginAda–CO3
 TheframeofanexceptionhandlerinAdaiseitherasubprogrambody,apackagebody,a
task,orablock
 Becauseexceptionhandlersareusuallylocaltothecodeinwhichtheexceptioncanberaise
d,theydonothaveparameters
AdaExceptionHandlers
 Handlerform:
whenexception_choice{|exception_choice}=>statement_sequence
...
[whenothers=>statement_sequ
ence]
exception_choiceform:excep
tion_name|others
 Handlersareplacedattheendoftheblockorunitinwhichtheyoccur
BindingExceptionstoHandlers
 Iftheblockorunitinwhichanexceptionisraiseddoesnothaveahandlerforthatexception,
theexceptionispropagatedelsewheretobehandled
– Procedures-propagateittothecaller
– Blocks-propagateittothescopeinwhichitappears
– Packagebody-
propagateittothedeclarationpartoftheunitthatdeclaredthepackage(ifitisalibraryu
nit,theprogramisterminated)
– Task-nopropagation;ifithasahandler,executeit;ineithercase,markit"completed"
Continuation
 Theblockorunitthatraisesanexceptionbutdoesnothandleitisalwaysterminated(alsoa
nyblockorunittowhichitispropagatedthatdoesnothandleit)
OtherDesignChoices
 User-definedExceptionsform:
exception_name_list:exception;
 RaisingExceptionsform:
raise[exception_name]
– (theexceptionnameisnotrequiredifitisinahandler--
inthiscase,itpropagatesthesameexception)
 Exceptionconditionscanbedisabledwith:
pragmaSUPPRESS(exception_list)
PredefinedExceptions
 CONSTRAINT_ERROR-indexconstraints,rangeconstraints,etc.
 NUMERIC_ERROR-
numericoperationcannotreturnacorrectvalue(overflow,divisionbyzero,etc.)

78
  PROGRAM_ERROR-calltoasubprogramwhosebodyhasnotbeenelaborated
 STORAGE_ERROR-systemrunsoutofheap
 TASKING_ERROR-anerrorassociatedwithtasks
Evaluation
 TheAdadesignforexceptionhandlingembodiesthestate-of-the-
artinlanguagedesignin1980
 AsignificantadvanceoverPL/I
 AdawastheonlywidelyusedlanguagewithexceptionhandlinguntilitwasaddedtoC++

ExceptionHandlinginC++-CO3
 AddedtoC++in1990
 DesignisbasedonthatofCLU,Ada,andML
C++ExceptionHandlers
 ExceptionHandlersForm:
try{
--codethatisexpectedtoraiseanexception
}
catch(formalparameter){
--handlercode
}
...
catch(formalparameter){
--handlercode
}
ThecatchFunction
 catchisthenameofallhandlers--
itisanoverloadedname,sotheformalparameterofeachmustbeunique
 Theformalparameterneednothaveavariable
– Itcanbesimplyatypenametodistinguishthehandleritisinfromothers
 Theformalparametercanbeusedtotransferinformationtothehandler
 Theformalparametercanbeanellipsis,inwhichcaseithandlesallexceptionsnotyeth
andled

ThrowingExceptions
 Exceptionsareallraisedexplicitlybythestatement:throw[expression];
 Thebracketsaremetasymbols
 Athrowwithoutanoperandcanonlyappearinahandler;whenitappears,itsimplyre-
raisestheexception,whichisthenhandledelsewhere
 Thetypeoftheexpressiondisambiguatestheintendedhandler

UnhandledExceptions
 Anunhandledexceptionispropagatedtothecallerofthefunctioninwhichitisraised
 Thispropagationcontinuestothemainfunction

Continuation
 Afterahandlercompletesitsexecution,controlflowstothefirststatementafterthelastha
ndlerinthesequenceofhandlersofwhichitisanelement
 Otherdesignchoices

79
 – Allexceptionsareuser-defined
– Exceptionsareneitherspecifiednordeclared
– Functionscanlisttheexceptionstheymayraise
– Withoutaspecification,afunctioncanraiseanyexception(thethrowclause)
Evaluation
 Itisoddthatexceptionsarenotnamedandthathardware-andsystemsoftware-
detectableexceptionscannotbehandled
 Bindingexceptionstohandlersthroughthetypeoftheparametercertainlydoesnotprom
otereadability

ExceptionHandlinginJava–CO3
 BasedonthatofC++,butmoreinlinewithOOPphilosophy
 AllexceptionsareobjectsofclassesthataredescendantsoftheThrowableclass
ClassesofExceptions
 TheJavalibraryincludestwosubclassesofThrowable:
– Error
o ThrownbytheJavainterpreterforeventssuchasheapoverflow
o Neverhandledbyuserprograms
– Exception
o User-definedexceptionsareusuallysubclassesofthis
o Hastwopredefinedsubclasses,IOExceptionandRuntimeException
e.g.,ArrayIndexOutOfBoundsExceptionandNullPointerException
JavaExceptionHandlers
 LikethoseofC++,excepteverycatchrequiresanamedparameterandallparametersmust
bedescendantsofThrowable
 SyntaxoftryclauseisexactlythatofC++
 Exceptionsarethrownwiththrow,asinC++,butoftenthethrowincludesthenewoperator
tocreatetheobject,asin:thrownewMyException();
BindingExceptionstoHandlers
 BindinganexceptiontoahandlerissimplerinJavathanitisinC++
– Anexceptionisboundtothefirsthandlerwithaparameteristhesameclassasthethrow
nobjectoranancestorofit
 Anexceptioncanbehandledandrethrownbyincludingathrowinthehandler(ahandlerco
uldalsothrowadifferentexception)
Continuation
 Ifnohandlerisfoundinthemethod,theexceptionispropagatedtothemethod‘scaller
 Ifnohandlerisfound(allthewaytomain),theprogramisterminated
 Toensurethatallexceptionsarecaught,ahandlercanbeincludedinanytryconstructt
hatcatchesallexceptions
– SimplyuseanExceptionclassparameter
– Ofcourse,itmustbethelastinthetryconstruct
CheckedandUncheckedExceptions
 TheJavathrowsclauseisquitedifferentfromthethrowclauseofC++
 ExceptionsofclassErrorandRunTimeExceptionandalloftheirdescendantsarecalle
duncheckedexceptions;allotherexceptionsarecalledcheckedexceptions
 Checkedexceptionsthatmaybethrownbyamethodmustbeeither:
– Listedinthethrowsclause,or
– Handledinthemethod

80
OtherDesignChoices
 Amethodcannotdeclaremoreexceptionsinitsthrowsclausethanthemethoditoverrides
 Amethodthatcallsamethodthatlistsaparticularcheckedexceptioninitsthrowsclauseha
sthreealternativesfordealingwiththatexception:
– Catchandhandletheexception
– Catchtheexceptionandthrowanexceptionthatislistedinitsownthrowsclause
– Declareitinitsthrowsclauseanddonothandleit
ThefinallyClause
 Canappearattheendofatryconstruct
 Form:
finally{..}
 Purpose:Tospecifycodethatistobeexecuted,regardlessofwhathappensinthetryconstr
uct
Example
 Atryconstructwithafinallyclausecanbeusedoutsideexceptionhandling
try{
for(index=0;index<100;index++){
…
if(…){
return;
}//**endofif
}//**endoftryclausefinally{
…
}//**endoftryconstruct
Assertions
 Statementsintheprogramdeclaringabooleanexpressionregardingthecurrentstate
ofthecomputation
 Whenevaluatedtotruenothinghappens
 WhenevaluatedtofalseanAssertionErrorexceptionisthrown
 Canbedisabledduringruntimewithoutprogrammodificationorrecompilation
 Twoforms
– assertcondition;
– assertcondition:expression;
Evaluation
 ThetypesofexceptionsmakesmoresensethaninthecaseofC++
 ThethrowsclauseisbetterthanthatofC++(ThethrowclauseinC++sayslittletotheprogra
mmer)
 Thefinallyclauseisoftenuseful
 TheJavainterpreterthrowsavarietyofexceptionsthatcanbehandledbyuserprograms
SummaryofExceptionHandling
 Adaprovidesextensiveexception-handlingfacilitieswithacomprehensivesetofbuilt-
inexceptions.
 C++includesnopredefinedexceptions.Exceptionsareboundtohandlersbyconnectingt
hetypeofexpressioninthethrowstatementtothatoftheformalparameterofthecatchfun
ction
 JavaexceptionsaresimilartoC++exceptionsexceptthataJavaexceptionmustbeades
cendantoftheThrowableclass.AdditionallyJavaincludesafinallyclause

81
 LogicProgrammingIntroductionCO4
 Logic programming languages, sometimes called declarative
programminglanguages
 Expressprogramsinaformofsymboliclogic
 Usealogicalinferencingprocesstoproduceresults
 Declarativeratherthatprocedural:
– Onlyspecificationofresultsarestated(notdetailedproceduresforproducingthem)
Proposition
 Alogicalstatementthatmayormaynotbetrue
– Consistsofobjectsandrelationshipsofobjectstoeachother

SymbolicLogic
 Logicwhichcanbeusedforthebasicneedsofformallogic:
– Expresspropositions
– Expressrelationshipsbetweenpropositions
– Describehownewpropositionscanbeinferredfromotherpropositions
 Particularformofsymboliclogicusedforlogicprogrammingcalledpredicatecalculus

ObjectRepresentation
 Objectsinpropositionsarerepresentedbysimpleterms:eitherconstantsorvariables
 Constant:asymbolthatrepresentsanobject
 Variable:asymbolthatcanrepresentdifferentobjectsatdifferenttimes
– Differentfromvariablesinimperativelanguages

CompoundTerms
 Atomicpropositionsconsistofcompoundterms
 Compoundterm:oneelementofamathematicalrelation,writtenlikeamathematicalf
unction
– Mathematicalfunctionisamapping
– Canbewrittenasatable

PartsofaCompoundTerm
 Compoundtermcomposedoftwoparts
– Functor:functionsymbolthatnamestherelationship
– Orderedlistofparameters(tuple)
 Examples:
student(jon)like(set
h,OSX)like(nick,win
dows)like(jim,linux)

FormsofaProposition
 Propositionscanbestatedintwoforms:
– Fact:propositionisassumedtobetrue
– Query:truthofpropositionistobedetermined
 Compoundproposition:
– Havetwoormoreatomicpropositions
– Propositionsareconnectedbyoperators

82
 LogicalOperators
Name Symbol Example Meaning
Negation  a anotb
Conjunction  ab aandb
Disjunction  ab aorb
Equivalence  ab aisequivalenttob
 aba aimpliesb
Implication
 b bimpliesa

Quantifiers
Name Example Meaning
universal X.P ForallX,Pistrue
existential X.P ThereexistsavalueofXsuchthatPistrue

ClausalForm
 Toomanywaystostatethesamething
 Useastandardformforpropositions
 Clausalform:
– B1B2…BnA1A2…Am
– meansifalltheAsaretrue,thenatleastoneBistrue
 Antecedent:rightside
 Consequent:leftside
PredicateCalculusandProvingTheorems
 Auseofpropositionsistodiscovernewtheoremsthatcanbeinferredfromknownaxiomsa
ndtheorems
 Resolution:aninferenceprinciplethatallowsinferredpropositionstobecomputedfromg
ivenpropositions
Resolution
 Unification:findingvaluesforvariablesinpropositionsthatallowsmatchingprocesstosu
cceed
 Instantiation:assigningtemporaryvaluestovariablestoallowunificationtosucceed
 Afterinstantiatingavariablewithavalue,ifmatchingfails,mayneedto
backtrackandinstantiatewithadifferentvalue

TheoremProving
 Basisforlogicprogramming
 Whenpropositionsusedforresolution,onlyrestrictedformcanbeused
 Hornclause-canhaveonlytwoforms
– Headed:singleatomicpropositiononleftside
– Headless:emptyleftside(usedtostatefacts)
 MostpropositionscanbestatedasHornclauses

AnOverviewofLogicProgramming
 Declarativesemantics
– Thereisasimplewaytodeterminethemeaningofeachstatement
– Simplerthanthesemanticsofimperativelanguages
 Programmingisnonprocedural
– Programsdonotstatenowaresultistobecomputed,butrathertheformoftheresult

83
TheOriginsofProlog
 UniversityofAix-Marseille
– Naturallanguageprocessing
 UniversityofEdinburgh
– Automatedtheoremproving

TheBasicElementsofProLogCO4
 EdinburghSyntax
 Term:aconstant,variable,orstructure
 Constant:anatomoraninteger
 Atom:symbolicvalueofProlog
 Atomconsistsofeither:
– astringofletters,digits,andunderscoresbeginningwithalowercaseletter
– astringofprintableASCIIcharactersdelimitedbyapostrophes
Terms:VariablesandStructures
 Variable:anystringofletters,digits,andunderscoresbeginningwithanuppercaselett
er
 Instantiation:bindingofavariabletoavalue
– Lastsonlyaslongasittakestosatisfyonecompletegoal
 Structure:representsatomicproposition
functor(parameterlist)
FactStatements
 Usedforthehypotheses
 HeadlessHornclauses
female(shelley).m
ale(bill).father(bi
ll,jake).
RuleStatements
 Usedforthehypotheses
 HeadedHornclause
 Rightside:antecedent(ifpart)
– Maybesingletermorconjunction
 Leftside:consequent(thenpart)
– Mustbesingleterm
 Conjunction:multipletermsseparatedbylogicalANDoperations(implied)
ExampleRules
ancestor(mary,shelley):-mother(mary,shelley).
 Canusevariables(universalobjects)togeneralizemeaning:
parent(X,Y):-mother(X,Y).
parent(X,Y):-father(X,Y).
grandparent(X,Z):-parent(X,Y),parent(Y,Z).
sibling(X,Y):-mother(M,X),mother(M,Y),
father(F,X),father(F,Y).
GoalStatements
 Fortheoremproving,theoremisinformofpropositionthatwewantsystemtoproveordis
prove– goalstatement
 SameformatasheadlessHorn
man(fred)
 Conjunctivepropositionsandpropositionswithvariablesalsolegalgoals
father(X,mike)
InferencingProcessofProlog
 Queriesarecalledgoals

84
  Ifagoalisacompoundproposition,eachofthefactsisasubgoal
 Toproveagoalistrue,mustfindachainofinferencerulesand/orfacts.ForgoalQ:
B:-A
C:-B
…
Q:-P
 Processofprovingasubgoalcalledmatching,satisfying,orresolution
Approaches
 Bottom-upresolution,forwardchaining
– Beginwithfactsandrulesofdatabaseandattempttofindsequencethatleadstogoal
– Workswellwithalargesetofpossiblycorrectanswers
 Top-downresolution,backwardchaining
– Beginwithgoalandattempttofindsequencethatleadstosetoffactsindatabase
– Workswellwithasmallsetofpossiblycorrectanswers
 Prologimplementationsusebackwardchaining
SubgoalStrategies
 Whengoalhasmorethanonesubgoal,canuseeither
– Depth-firstsearch:findacompleteproofforthefirstsubgoalbeforeworkingonothers
– Breadth-firstsearch:workonallsubgoalsinparallel
 Prologusesdepth-firstsearch
– Canbedonewithfewercomputerresources
Backtracking
 Withagoalwithmultiplesubgoals,iffailtoshowtruthofoneofsubgoals,reconsiderprevio
ussubgoaltofindanalternativesolution:backtracking
 Beginsearchwhereprevioussearchleftoff
 Cantakelotsoftimeandspacebecausemayfindallpossibleproofstoeverysubgoal
SimpleArithmetic
 Prologsupportsintegervariablesandintegerarithmetic
 isoperator:takesanarithmeticexpressionasrightoperandandvariableasleftoperand
AisB/17+C
 Notthesameasanassignmentstatement!
Example
speed(ford,100).speed(ch
evy,105).speed(dodge,95)
.speed(volvo,80).time(for
d,20).time(chevy,21).time
(dodge,24).time(volvo,24
).
distance(X,Y):- speed(X,Speed),
time(X,Time),
YisSpeed*Time.
Trace
• Built-instructurethatdisplaysinstantiationsateachstep
• Tracingmodelofexecution-fourevents:
– Call(beginningofattempttosatisfygoal)
– Exit(whenagoalhasbeensatisfied)

85
 – Redo(whenbacktrackoccurs)
– Fail(whengoalfails)
Example
likes(jake,chocolate).likes(jake,apric
ots).likes(darcie,licorice).likes(darci
e,apricots).

trace.likes(jak
e,X),likes(darci
e,X).

ListStructures
 Otherbasicdatastructure(besidesatomicpropositionswehavealreadyseen):list
 Listisasequenceofanynumberofelements
 Elementscanbeatoms,atomicpropositions,orotherterms(includingotherlists)
[apple,prune,grape,kumquat][](e
mptylist)
[X|Y](headXandtailY)
AppendExample
append([],List,List).
append([Head|List_1],List_2,[Head|List_3]):-
append(List_1,List_2,List_3).
ReverseExample
reverse([],[]).
reverse([Head|Tail],List):-
reverse(Tail,Result),append(
Result,[Head],List).
DeficienciesofProlog
 Resolutionordercontrol
 Theclosed-worldassumption
 Thenegationproblem
 Intrinsiclimitations

ApplicationsofLogicProgramming–CO4
 Relationaldatabasemanagementsystems
 Expertsystems
 Naturallanguageprocessing
SummaryofLogicProgramming
 Symboliclogicprovidesbasisforlogicprogramming
 Logicprogramsshouldbenonprocedural
 Prologstatementsarefacts,rules,orgoals
 ResolutionistheprimaryactivityofaProloginterpreter
 Althoughthereareanumberofdrawbackswiththecurrentstateoflogicprogrammingith
asbeenusedinanumberofareas

86
 UNIT-5
FunctionalProgrammingLanguages&ScriptingLanguage

FunctionalProgrammingLanguageIntroduction–CO5
 ThedesignoftheimperativelanguagesisbaseddirectlyonthevonNeumannarchitecture
– Efficiencyistheprimaryconcern,ratherthanthesuitabilityofthelanguageforsoftwar
edevelopment
 Thedesignofthefunctionallanguagesisbasedonmathematicalfunctions
– Asolidtheoreticalbasisthatisalsoclosertotheuser,butrelativelyunconcernedwithth
earchitectureofthemachinesonwhichprogramswillrun

MathematicalFunctions
 Amathematicalfunctionisamappingofmembersofoneset,calledthedomainset,toan
otherset,calledtherangeset
 Alambdaexpressionspecifiestheparameter(s)andthemappingofafunctioninthefoll
owingform
(x)x*x*x
forthefunctioncube(x)=x*x*x
LambdaExpressions
 Lambdaexpressionsdescribenamelessfunctions
 Lambdaexpressionsareappliedtoparameter(s)byplacingtheparameter(s)aftertheexp
ression
e.g.,((x)x*x*x)(2)whi
chevaluatesto8
FunctionalForms
 Ahigher-
orderfunction,orfunctionalform,isonethateithertakesfunctionsasparametersoryields
afunctionasitsresult,orboth
FunctionComposition
 Afunctionalformthattakestwofunctionsasparametersandyieldsafunctionwhosev
alueisthefirstactualparameterfunctionappliedtotheapplicationofthesecond
Form:hf°g
whichmeansh(x)f(g(x))
Forf(x)x+2andg(x)3*x,hf°gy
ields(3*x)+2
Apply-to-all
 Afunctionalformthattakesasinglefunctionasaparameterandyieldsalistofvaluesobtain
edbyapplyingthegivenfunctiontoeachelementofalistofparameters
Form:
Forh(x)x*x
(h,(2,3,4))yields(4,9,16)

FundamentalsofFunctionalProgrammingLanguages-CO5
 TheobjectiveofthedesignofaFPListomimicmathematicalfunctionstothegreatestex
tentpossible
 ThebasicprocessofcomputationisfundamentallydifferentinaFPLthaninanimperat
ivelanguage

87
 – Inanimperativelanguage,operationsaredoneandtheresultsarestoredinvariablesfo
rlateruse
– Managementofvariablesisaconstantconcernandsourceofcomplexityforimperativ
eprogramming
 InanFPL,variablesarenotnecessary,asisthecaseinmathematics
ReferentialTransparency
 InanFPL,theevaluationofafunctionalwaysproducesthesameresultgiventhesamepara
meters

TheFirstFunctionalProgrammingLanguage:LISP–CO5

LISPDataTypesandStructures
 Dataobjecttypes:originallyonlyatomsandlists
 Listform:parenthesizedcollectionsofsublistsand/oratoms
e.g.,(AB(CD)E)
 Originally,LISPwasatypelesslanguage
 LISPlistsarestoredinternallyassingle-linkedlists
LISPInterpretation
 Lambdanotationisusedtospecifyfunctionsandfunctiondefinitions.Func
tionapplicationsanddatahavethesameform.
e.g.,Ifthelist(ABC)isinterpretedasdataitisasimplelistofthreeatoms,A,BandC
Ifitisinterpretedasafunctionapplication,itmeansthatthefunctionnamedAisappli
edtothetwoparameters,BandC
 ThefirstLISPinterpreterappearedonlyasademonstrationoftheuniversalityofthecomp
utationalcapabilitiesofthenotation

ML–CO5
 Astatic-scopedfunctionallanguagewithsyntaxthatisclosertoPascalthantoLISP
 Usestypedeclarations,butalsodoestypeinferencingtodeterminethetypesofundecla
redvariables
 Itisstronglytyped(whereasSchemeisessentiallytypeless)andhasnotypecoercions
 Includesexceptionhandlingandamodulefacilityforimplementingabstractdatatypes
 Includeslistsandlistoperations
MLSpecifics
 Functiondeclarationform:
funname(parameters)=body;
e.g.,funcube(x:int)=x*x*x;
– Thetypecouldbeattachedtoreturnvalue,asin
funcube(x):int=x*x*x;
– Withnotypespecified,itwoulddefaultto
int(thedefaultfornumericvalues)
– User-
definedoverloadedfunctionsarenotallowed,soifwewantedacubefunctionforrealpa
rameters,itwouldneedtohaveadifferentname
– TherearenotypecoercionsinML
 MLselection
ifexpressionthenthen_expression
elseelse_expression

88
 wherethefirstexpressionmustevaluatetoaBooleanvalue
 Patternmatchingisusedtoallowafunctiontooperateondifferentparameterforms
funfact(0)=1
|fact(n:int):int=n*fact(n–1)
 Lists
Literallistsarespecifiedinbrackets[3,5,
7]
[]istheemptylist
CONSisthebinaryinfixoperator,::
4::[3,5,7],whichevaluatesto[4,3,5,7]CARistheu
naryoperatorhd
CDRistheunaryoperatortlfunl
ength([])=0
|length(h::t)=1+length(t);funa
ppend([],lis2)=lis2
|append(h::t,lis2)=h::append(t,lis2);
 Thevalstatementbindsanametoavalue(similartoDEFINEinScheme)valdist
ance=time*speed;
– AsisthecasewithDEFINE,valisnothinglikeanassignmentstatementinanimperat
ivelanguage

Haskell–CO5
 SimilartoML(syntax,staticscoped,stronglytyped,typeinferencing,patternmatchin
g)
 DifferentfromML(andmostotherfunctionallanguages)inthatitispurelyfunctional(
e.g.,novariables,noassignmentstatements,andnosideeffectsofanykind)
SyntaxdifferencesfromML
fact0=1
factn=n*fact(n– 1)fib0=1
fib1=1
fib(n+2)=fib(n+1)+fibn
FunctionDefinitionswithDifferentParameterRanges
factn
|n==0=1
|n>0=n*fact(n–1)

subn
|n<10=0
|n>100=2
|otherwise=1s
quarex=x*x
- Worksforanynumerictypeofx
Lists
 Listnotation:Putelementsinbrackets
e.g.,directions=["north","south","east","west"]
 Length:#
e.g.,#directionsis4
 Arithmeticserieswiththe..operator
e.g.,[2,4..10]is[2,4,6,8,10]
 Catenationiswith++
e.g.,[1,3]++[5,7]resultsin[1,3,5,7]

89
  CONS,CAR,CDRviathecolonoperator(asinProlog)
e.g.,1:[3,5,7]resultsin[1,3,5,7]
FactorialRevisited
product[]=1
product(a:x)=a*productxfactn=
product[1..n]
ListComprehension
 Setnotation
 Listofthesquaresofthefirst20positiveintegers:[n*n|n←[1..20]]
 Allofthefactorsofitsgivenparameter:
factorsn=[i|i←[1..ndiv2],
nmodi==0]
Quicksort
sort[]=[]sort
(a:x)=
sort[b|b←x;b<=a]++
[a]++
sort [b|b←x;b>a]
LazyEvaluation
 Alanguageisstrictifitrequiresallactualparameterstobefullyevaluated
 Alanguageisnonstrictifitdoesnothavethestrictrequirement
 Nonstrictlanguagesaremoreefficientandallowsomeinterestingcapabilities
– infinitelists
 Lazyevaluation-Onlycomputethosevaluesthatarenecessary
 Positivenumbers
positives=[0..]
 Determiningif16isasquarenumber
member[]b=False
member(a:x)b=(a==b)||memberxb
squares=[n * n|n← [0..]]
membersquares16
MemberRevisited
 Thememberfunctioncouldbewrittenas:me
mber[]b=False
member(a:x)b=(a==b)||memberxb
 However,thiswouldonlyworkiftheparametertosquareswasaperfectsquare;ifnot,itwil
lkeepgeneratingthemforever.Thefollowingversionwillalwayswork:
member2(m:x)n
|m<n=member2xn
|m==n=True
|otherwise=False

ApplicationsofFunctionalLanguages–CO5
 APLisusedforthrow-awayprograms
 LISPisusedforartificialintelligence
– Knowledgerepresentation
– Machinelearning
– Naturallanguageprocessing
– Modelingofspeechandvision
 Schemeisusedtoteachintroductoryprogrammingatsomeuniversities

90
ComparingFunctionalandImperativeLanguages
 ImperativeLanguages:
– Efficientexecution
– Complexsemantics
– Complexsyntax
– Concurrencyisprogrammerdesigned
 FunctionalLanguages:
– Simplesemantics
– Simplesyntax
– Inefficientexecution
– Programscanautomaticallybemadeconcurrent

SummaryofFunctionalProgrammingLanguages
 Functionalprogramminglanguagesusefunctionapplication,conditionalexpression
s,recursion,andfunctionalformstocontrolprogramexecutioninsteadofimperativef
eaturessuchasvariablesandassignments
 LISPbeganasapurelyfunctionallanguageandlaterincludedimperativefeatures
 SchemeisarelativelysimpledialectofLISPthatusesstaticscopingexclusively
 COMMONLISPisalargeLISP-basedlanguage
 MLisastatic-
scopedandstronglytypedfunctionallanguagewhichincludestypeinference,excepti
onhandling,andavarietyofdatastructuresandabstractdatatypes
 Haskellisalazyfunctionallanguagesupportinginfinitelistsandsetcomprehension.
 Purelyfunctionallanguageshaveadvantagesoverimperativealternatives,buttheirl
owerefficiencyonexistingmachinearchitectureshaspreventedthemfromenjoying
widespreaduse

Pragmatics
Asoftwaresystemoftenconsistsofanumberofsubsystemscontrolledorconnectedbyascript.
Scriptingisaparadigmcharacterizedby:
 Useofscriptstogluesubsystemstogether.
 Rapiddevelopmentandevolutionofscripts.
 Modestefficiencyrequirements.
 Veryhigh-levelfunctionalityinapplication-specificareas.

KeyConcepts
Thefollowingconceptsarecharacteristicofscriptinglanguages:
 Veryhigh-levelstringprocessing.
 Veryhigh-levelgraphicaluserinterfacesupport.
 Dynamictyping.

CaseStudy:PYTHON
 PYTHONwasdesignedintheearly1990sbyGuidovanRossum.
 IthasbeenusedtohelpimplementthesuccessfulWebsearchengineGOOGLE,andinav
arietyofotherapplicationareasrangingfromsciencefiction(visualeffectsfortheStar
Warsseries)torealscience(computer-aideddesigninNASA).

91
ValuesandTypes
 PYTHONhasalimitedrepertoireofprimitivetypes:integer,real,andcomplexnumbers.
 Ithasnospecificcharactertype;single-
characterstringsareusedinstead.Itsbooleanvalues(namedFalseandTrue)arejusts
mallintegers.
 PYTHONhasarichrepertoireofcompositetypes:tuples,strings,lists,dictionariesandobj
ects.APYTHONlistisaheterogeneoussequenceofvalues.
 Adictionary(sometimescalledanassociativearray)isaheterogeneousmappingfromkey
stovalues,wherethekeysaredistinctimmutablevalues.
 Thefollowingcodeillustratestupleconstruction:
date=1998,"Nov",19
Nowdate[0]yields1998,date[1]yields‘‘Nov’’,anddate[2]yields19.

 The following code illustrates two list constructions, which construct

ahomogeneouslistandaheterogeneouslist,respectively:
primes=[2,3,5,7,11]
years=["unknown",1314,1707,date[0]]
Nowprimes[0]yields2,years[1]yields1314,years[3]yields1998,‘‘years[0]=843’’up
datesthefirstcomponentofyears,andsoon.Also,‘‘years.append(1999)’’
adds1999attheendofyears.

 Thefollowingcodeillustratesdictionaryconstruction:
phones={"David":6742,"Carol":6742,"Ali":6046}
Nowphones["Carol"]yields6742,phones["Ali"]yields6046,‘‘phones["Ali"]=1234’’
updatesthecomponentofphoneswhosekeyis‘‘Ali’’,andsoon.Also,‘‘David”inphonesret
urnsTrue,and‘‘phones.keys()’’returnsalistcontaining‘‘Ali’’,‘‘Carol’’,and‘‘David’’
(innoparticularorder).

Variables,StorageandControl
 PYTHONsupportsglobalandlocalvariables.
 Variablesarenotexplicitlydeclared,simplyinitializedbyassignment.Afterinitializat
ion,avariablemaylaterbeassignedanyvalueofanytype.
 PYTHON’srepertoireofcommandsincludeassignments,procedurecalls,conditional(if-
butnotcase-)commands,iterative(while-andfor-)commandsandexception-
handlingcommands.
 However,PYTHONdiffersfromCinnotallowinganassignmenttobeusedasanexpress
ion.
 PYTHONadditionallysupportssimultaneousassignment.
 Forexample:
y,m,d=date
assignsthethreecomponentsofthetupledatetothreeseparatevariables.Also:
m,n=n,m
conciselyswapsthevaluesoftwovariablesmandn.(Actually,itfirstconstructsapair,the
nassignsthetwocomponentsofthepairtothetwoleft-sidevariables)
 PYTHONif-andwhile-commandsareconventional.
 PYTHONfor-commandssupportdefiniteiteration.
 Wecaneasilyachievetheconventionaliterationoverasequenceofnumbersbyusingtheli
braryprocedurerange(m,n),whichreturnsalistofintegersfrommthroughn−1.

92
  PYTHONsupportsbreak,continue,andreturnsequencers.Italsosupportsexception
s,whichareobjectsofasubclassofException,andwhichcancarryvalues.

 ThefollowingcodecomputestheGreatestCommonDivisoroftwointegers,mandn:
p,q=m,n
whilep%q!=0:
p,q=q,p%qgcd=q
 Notetheeleganceofsimultaneousassignment.
 Notealsothatindentationisrequiredtoindicatetheextentoftheloopbody.
 Thefollowingcodesumsthenumericcomponentsofalistrow,ignoringanynonnumericc
omponents:
sum=0.0
forxinrow:
ifisinstance(x,(int,float)):su
m+=x

PYTHONExceptions
 Thefollowingcodepromptstheusertoenteranumericliteral,andstoresthecorrespondi
ngrealnumberinnum:
whileTrue:
try:
response=raw_input("Enteranumericliteral:")nu
m=float(response)
break
exceptValueError:
print"Yourresponsewasill-formed."
Thiswhile-commandkeepspromptinguntiltheuserentersawell-
formednumericliteral.Thelibraryprocedureraw_input(...)displaysthegivenpromp
tandreturnstheuser’sresponseasastring.Thetypeconversion‘‘float(response)’’
attemptstoconverttheresponsetoarealnumber.Ifthistypeconversionispossible,th
efollowingbreaksequencerterminatestheloop.Ifnot,thetypeconversionthrowsaV
alueErrorexception,controlistransferredtotheValueErrorexceptionhandler,whic
hdisplaysawarningmessage,andfinallytheloopisiteratedagain.

BindingsandScope
 APYTHONprogramconsistsofanumberofmodules,whichmaybegroupedintopackages.
 Withinamodulewemayinitializevariables,defineprocedures,anddeclareclasses.Withi
naprocedurewemayinitializelocalvariablesanddefinelocalprocedures.Withinaclassw
emayinitializevariablecomponentsanddefineprocedures(methods).
 DuringaPYTHONsession,wemayinteractivelyissuedeclarations,commands,andex
pressionsfromthekeyboard.
 Theseareallacteduponimmediately.Wheneverweissueanexpression,itsvalueisdispla
yedonthescreen.Wemayalsoimportanamedmodule(orselectedcomponentsofit)atan
ytime.
 PYTHONwasoriginallyadynamically-scopedlanguage,butitisnowstaticallyscoped.

93
ProceduralAbstraction
 PYTHONsupportsfunctionproceduresandproperprocedures.
 Theonlydifferenceisthatafunctionprocedurereturnsavalue,whileaproperprocedurer
eturnsnothing.
 SincePYTHONisdynamicallytyped,aproceduredefinitionstatesthenamebutnotthetyp
eofeachformalparameter.Thecorrespondingargumentmaybeofdifferenttypesondiffe
rentcallstotheprocedure.

PYTHONProcedures
 Thefollowingfunctionprocedurereturnsthegreatestcommondivisorofitstwoargumen
ts:
defgcd(m,n):
p,q=m,n
whilep%q!=0:p,q=
q,p%q
returnq
Herepandqarelocalvariables.
 ThefollowingproperproceduretakesadaterepresentedbyatripleandprintsthatdateinI
SOformat(e.g.,‘‘2000-01-01’’):
defprint_date(date):y
,m,d=date
ifm="Jan":m=
1
elifm="Feb":m=
2
...
elifm="Dec":m=
12
print"%04d-%02d-%02d"%(y,m,d)
Herey,m,anddarelocalvariables.

PYTHONprocedurewithdynamictyping
 Thefollowingfunctionprocedureillustratestheflexibilityofdynamictyping.Itreturnsth
eminimumandmaximumcomponentofagivensequence:
defminimax(vals):
min=max=vals[0]
forvalinvals:
ifval<min:
min=val
elifval>max:max=
val
returnmin,max
 Inacalltothisprocedure,theargumentmaybeeitheratupleoralist.
 Ineffectithastworesults,whichwecaneasilyseparateusingsimultaneousassignmen
t:
readings=[...]
low,high=minimax(readings)
 SomeolderlanguagessuchasChavelibraryprocedureswithvariablenumbersofargu
ments.
 PYTHONisalmostuniqueinallowingsuchprocedurestobedefinedbyprogrammers.
 Thisisachievedbythesimpleexpedientofallowingasingleformalparametertorefertoaw
holetuple(ordictionary)ofarguments.

94
PYTHONprocedurewithavariablenumberofarguments
 Thefollowingproperprocedureacceptsanynumberofarguments,andprintsthemonepe
rline:
defprintall(*args):
forarginargs:
printarg
 Thenotation‘‘*args’’declaresthatargswillrefertoatupleofarguments.
 Allofthefollowingprocedurecallsworksuccessfully:
printall(name)printall(na
me,address)
printall(name,address,zipcode)

DataAbstraction
 PYTHONhasthreedifferentconstructsrelevanttodataabstraction:packages,modul
es,andclasses.
 Modulesandclassessupportencapsulation,usinganamingconventiontodistinguish
betweenpublicandprivatecomponents.
 Apackageissimplyagroupofmodules.Amoduleisagroupofcomponentsthatmaybev
ariables,procedures,andclasses.
 Thesecomponentsmaybeimportedforusebyanyothermodule.Allcomponentsofamod
ulearepublic,exceptthosewhoseidentifiersstartwith‘‘_’’ whichareprivate.
 Aclassisagroupofcomponentsthatmaybeclassvariables,classmethods,andinstanc
emethods.Aproceduredefinedinaclassdeclarationactsasaninstancemethodifitsfir
stformalparameterisnamedselfandreferstoanobjectoftheclassbeingdeclared.Oth
erwisetheprocedureactsasaclassmethod.
 Toachievetheeffectofaconstructor,weusuallyequipeachclasswithaninitializationmet
hodnamed‘‘init’’;thismethodisautomaticallycalledwhenanobjectoftheclassisconstruc
ted.Instancevariablesarenamedusingtheusual‘‘.’’Notation(asinself.attr),andtheymay
beinitializedbytheinitializationmethodorbyanyothermethod.Allcomponentsofaclass
arepublic,exceptthosewhoseidentifiersstartwith‘‘’’,whichareprivate.

PYTHONClass
 Considerthefollowingclass:
classPerson:
definit(self,sname,fname,gender,birth):self.sur
name=sname
self.forename=fname
self.female=(gender=="F"orgender=="f")self.birt
h=birth
defget_surname(self):
returnself.surname
defchange_surname(self,sname):sel
f.surname=sname
defprint_details(self):
printself.forename+""+self.surname
 Thisclassisequippedwithaninitializationmethodandthreeotherinstancemethods,eac
hofwhichhasaselfparameterandperhapssomeotherparameters.Inthefollowingcode:
dw=Person("Watt","David","M",1946)

95
 theobjectconstructionontherightfirstcreatesanobjectofclassPerson;itthenpassesthea
bovearguments,togetherwithareferencetothenewlycreatedobject,totheinitialization
method.Thelatterinitializestheobject’sinstancevariables,whicharenamedsurname,fo
rename,female,andbirth(andthusareallprivate).
 PYTHONsupportsmultipleinheritance:aclassmaydesignateanynumberofsupercla
sses.Ambiguousreferencestoclasscomponentsareresolvedbysearchingthesuperc
lassesintheorderinwhichtheyarenamedintheclassdeclaration.
 PYTHON’ssupportforobject-
orientedprogrammingisdevelopingbutisnotyetmature.Theuseofthe‘‘’’
namingconventiontoindicateprivacyisclumsyanderror-
prone;classcomponentsarepublicbydefault.Stillmoreseriously,variablecomponentsc
anbecreated(anddeleted)atanytime,byanymethodandevenbyapplicationcode.

SeparateCompilation
 PYTHONmodulesarecompiledseparately.Eachmodulemustexplicitlyimportevery
othermoduleonwhichitdepends.Eachmodule’ssourcecodeisstoredinatextfile.
 Forexample,amodulenamedwidgetisstoredinafilenamedwidget.py.Whenthatmo
duleisfirstimported,itiscompiledanditsobjectcodeisstoredinafilenamedwidget.p
yc.
 Wheneverthemoduleissubsequentlyimported,itisrecompiledonlyifthesourcecod
ehasbeeneditedinthemeantime.Compilationiscompletelyautomatic.
 ThePYTHONcompilerdoesnotrejectcodethatreferstoundeclaredidentifiers.Suchcode
simplyfailsifandwhenitisexecuted.
ModuleLibrary
 PYTHONisequippedwithaveryrichmodulelibrary,whichsupportsstringhandling,mar
kup,mathematicsandcryptography,multimedia,GUIs,operatingsystemservices,Intern
etservices,compilation,andsoon.
 Unlikeolderscriptinglanguages,PYTHONdoesnothavebuilt-inhigh-
levelstringprocessingorGUIsupport.Instead,thePYTHONmodulelibraryprovidess
uchfunctionality.Forexample,therelibrarymoduleprovidespowerfulstringmatchi
ngfacilitiesusingregularexpressions.

SummaryofScriptingLanguages
 Thepragmaticissuesthatinfluencethedesignofscriptinglanguages:gluing,rapiddev
elopmentandevolution,modestefficiencyrequirements,andveryhigh-
levelfunctionalityinrelevantareas.
 Theconceptscommontomostscriptinglanguages:veryhigh-
levelsupportforstringprocessing,veryhigh-
levelsupportforGUIs,anddynamictyping.
 ThedesignofamajorscriptinglanguagePYTHON,resemblesaconventionalprogrammin
glanguage,exceptthatitisdynamicallytyped,andthatitderivesmuchofitsexpressivenes
sfromarichmodulelibrary.

96