29/7/2016

TheGNUAwkUsersGuide

TheGNUAwkUsersGuide
Next:Foreword3,Up:(dir)[Contents][Index]

GeneralIntroduction
Thisfiledocumentsawk,aprogramthatyoucanusetoselectparticularrecordsinafileandperformoperationsupon
them.
Copyright1989,1991,1992,1993,19962005,2007,20092015
FreeSoftwareFoundation,Inc.

ThisisEdition4.1ofGAWK:EffectiveAWKProgramming:AUsersGuideforGNUAwk,forthe4.1.2(orlater)
versionoftheGNUimplementationofAWK.
Permissionisgrantedtocopy,distributeand/ormodifythisdocumentunderthetermsoftheGNUFreeDocumentation
License,Version1.3oranylaterversionpublishedbytheFreeSoftwareFoundationwiththeInvariantSectionsbeing
GNUGeneralPublicLicense,withtheFrontCoverTextsbeingAGNUManual,andwiththeBackCoverTextsas
in(a)below.AcopyofthelicenseisincludedinthesectionentitledGNUFreeDocumentationLicense.
1.TheFSFsBackCoverTextis:YouhavethefreedomtocopyandmodifythisGNUmanual.
Foreword3:
Foreword4:
Preface:
GettingStarted:
InvokingGawk:
Regexp:
ReadingFiles:
Printing:
Expressions:
PatternsandActions:
Arrays:
Functions:
LibraryFunctions:
SamplePrograms:
AdvancedFeatures:
Internationalization:
Debugger:
ArbitraryPrecision
Arithmetic:
DynamicExtensions:
LanguageHistory:
Installation:
Notes:
BasicConcepts:
Glossary:

SomenicewordsaboutthisWebpage.
Morenicewords.
WhatthisWebpageisaboutbriefhistoryandacknowledgments.
Abasicintroductiontousingawk.Howtorunanawkprogram.Commandlinesyntax.
Howtorungawk.
Allaboutmatchingthingsusingregularexpressions.
Howtoreadfilesandmanipulatefields.
Howtoprintusingawk.Describestheprintandprintfstatements.Alsodescribes

redirectionofoutput.
Expressionsarethebasicbuildingblocksofstatements.
Overviewsofpatternsandactions.
Thedescriptionanduseofarrays.Alsoincludesarrayorientedcontrolstatements.
Builtinanduserdefinedfunctions.
ALibraryofawkFunctions.
Manyawkprogramswithcompleteexplanations.
Stuffforadvancedusers,specifictogawk.
Gettinggawktospeakyourlanguage.
Thegawkdebugger.
Arbitraryprecisionarithmeticwithgawk.

Addingnewbuiltinfunctionstogawk.
Theevolutionoftheawklanguage.
Installinggawkundervariousoperatingsystems.
Notesaboutaddingthingstogawkandpossiblefuturework.
Averyquickintroductiontoprogrammingconcepts.
Anexplanationofsomeunfamiliarterms.

https://www.gnu.org/software/gawk/manual/gawk.html

1/479

29/7/2016

TheGNUAwkUsersGuide

Copying:
GNUFreeDocumentation
License:
Index:

Yourrighttocopyanddistributegawk.
ThelicenseforthisWebpage.

History:
Names:
ThisManual:
Conventions:
ManualHistory:
HowToContribute:
Acknowledgments:
Runninggawk:
Oneshot:
ReadTerminal:
Long:
ExecutableScripts:
Comments:
Quoting:
DOSQuoting:
SampleDataFiles:
VerySimple:
TwoRules:
MoreComplex:
Statements/Lines:
OtherFeatures:
When:
IntroSummary:
CommandLine:
Options:
OtherArguments:
NamingStandardInput:
EnvironmentVariables:
AWKPATHVariable:
AWKLIBPATHVariable:
OtherEnvironment
Variables:
ExitStatus:
IncludeFiles:
LoadingSharedLibraries:
Obsolete:
Undocumented:
InvokingSummary:
RegexpUsage:
EscapeSequences:
RegexpOperators:
BracketExpressions:
LeftmostLongest:
ComputedRegexps:
GNURegexpOperators:
Casesensitivity:

Thehistoryofgawkandawk.
Whatnametousetofindawk.
UsingthisWebpage.Includessampleinputfilesthatyoucanuse.
TypographicalConventions.
BriefhistoryoftheGNUprojectandthisWebpage.
Helpingtosavetheworld.
Acknowledgments.
Howtorungawkprogramsincludescommandlinesyntax.
Runningashortthrowawayawkprogram.
Usingnoinputfiles(inputfromthekeyboardinstead).
Puttingpermanentawkprogramsinfiles.
Makingselfcontainedawkprograms.
Addingdocumentationtogawkprograms.
Morediscussionofshellquotingissues.
QuotinginWindowsBatchFiles.
SampledatafilesforuseintheawkprogramsillustratedinthisWebpage.
Averysimpleexample.
Alesssimpleonelineexampleusingtworules.
Amorecomplexexample.
Subdividingorcombiningstatementsintolines.
OtherFeaturesofawk.
Whentousegawkandwhentouseotherthings.
Summaryoftheintroduction.
Howtorunawk.
Commandlineoptionsandtheirmeanings.
Inputfilenamesandvariableassignments.
Howtospecifystandardinputwithotherfiles.
Theenvironmentvariablesgawkuses.
Searchingdirectoriesforawkprograms.
Searchingdirectoriesforawksharedlibraries.
Theenvironmentvariables.

RegexpSummary:

Regularexpressionssummary.

ConceptandVariableIndex.

gawksexitstatus.
Includingotherfilesintoyourprogram.
Loadingsharedlibrariesintoyourprogram.
ObsoleteOptionsand/orfeatures.
UndocumentedOptionsandFeatures.
Invocationsummary.
HowtoUseRegularExpressions.
Howtowritenonprintingcharacters.
RegularExpressionOperators.
Whatcangobetween[...].
Howmuchtextmatches.
UsingDynamicRegexps.
OperatorsspecifictoGNUsoftware.
Howtodocaseinsensitivematching.

https://www.gnu.org/software/gawk/manual/gawk.html

2/479

29/7/2016

TheGNUAwkUsersGuide

Records:
awksplitrecords:
gawksplitrecords:
Fields:
NonconstantFields:
ChangingFields:
FieldSeparators:
DefaultFieldSplitting:
RegexpFieldSplitting:
SingleCharacterFields:
CommandLineField
Separator:
FullLineFields:
FieldSplittingSummary:
ConstantSize:
SplittingByContent:
MultipleLine:
Getline:
PlainGetline:
Getline/Variable:
Getline/File:
Getline/Variable/File:
Getline/Pipe:
Getline/Variable/Pipe:
Getline/Coprocess:
Getline/Variable/Coprocess:
GetlineNotes:
GetlineSummary:
ReadTimeout:
Commandlinedirectories:
InputSummary:
InputExercises:
Print:
PrintExamples:
OutputSeparators:
OFMT:
Printf:
BasicPrintf:
ControlLetters:
FormatModifiers:
PrintfExamples:
Redirection:
SpecialFD:
SpecialFiles:
OtherInheritedFiles:
SpecialNetwork:
SpecialCaveats:
CloseFilesAndPipes:
OutputSummary:

Controllinghowdataissplitintorecords.
Howstandardawksplitsrecords.
Howgawksplitsrecords.
Anintroductiontofields.
NonconstantFieldNumbers.
ChangingtheContentsofaField.
Thefieldseparatorandhowtochangeit.
Howfieldsarenormallyseparated.
Usingregexpsasthefieldseparator.
Makingeachcharacteraseparatefield.
SettingFSfromthecommandline.

OutputExercises:
Values:

Exercises.
Constants,Variables,andRegularExpressions.

Makingthefulllinebeasinglefield.
Somefinalpointsandasummarytable.
Readingconstantwidthdata.
DefiningFieldsByContent
Readingmultilinerecords.
Readingfilesunderexplicitprogramcontrolusingthegetlinefunction.
Usinggetlinewithnoarguments.
Usinggetlineintoavariable.
Usinggetlinefromafile.
Usinggetlineintoavariablefromafile.
Usinggetlinefromapipe.
Usinggetlineintoavariablefromapipe.
Usinggetlinefromacoprocess.
Usinggetlineintoavariablefromacoprocess.
Importantthingstoknowaboutgetline.
SummaryofgetlineVariants.
Readinginputwithatimeout.
Whathappensifyouputadirectoryonthecommandline.
Inputsummary.
Exercises.
Theprintstatement.
Simpleexamplesofprintstatements.
Theoutputseparatorsandhowtochangethem.
ControllingNumericOutputWithprint.
Theprintfstatement.
Syntaxoftheprintfstatement.
Formatcontrolletters.
Formatspecificationmodifiers.
Severalexamples.
Howtoredirectoutputtomultiplefilesandpipes.
SpecialfilesforI/O.
Filenameinterpretationingawk.gawkallowsaccesstoinheritedfiledescriptors.
Accessingotheropenfileswithgawk.
Specialfilesfornetworkcommunications.
Thingstowatchoutfor.
ClosingInputandOutputFilesandPipes.
Outputsummary.

https://www.gnu.org/software/gawk/manual/gawk.html

3/479

29/7/2016

Constants:
ScalarConstants:
Nondecimalnumbers:
RegexpConstants:
UsingConstantRegexps:
Variables:
UsingVariables:
AssignmentOptions:
Conversion:
StringsAndNumbers:
Localeinfluences
conversions:
AllOperators:
ArithmeticOps:
Concatenation:
AssignmentOps:
IncrementOps:
TruthValuesand
Conditions:
TruthValues:
TypingandComparison:

TheGNUAwkUsersGuide

String,numericandregexpconstants.
Numericandstringconstants.
Whatareoctalandhexnumbers.
RegularExpressionconstants.
Whenandhowtousearegexpconstant.
Variablesgivenamestovaluesforlateruse.
Usingvariablesinyourprograms.
Settingvariablesonthecommandlineandasummaryofcommandlinesyntax.Thisis

anadvancedmethodofinput.
Theconversionofstringstonumbersandviceversa.
HowawkConvertsBetweenStringsAndNumbers.
Howthelocalemayaffectconversions.

gawksoperators.
Arithmeticoperations(+,,etc.)
Concatenatingstrings.
Changingthevalueofavariableorafield.
Incrementingthenumericvalueofavariable.
Testingfortrueandfalse.

Whatistrueandwhatisfalse.
Howvariablesacquiretypesandhowthisaffectscomparisonofnumbersandstrings

with<,etc.
VariableTyping:
Stringtypeversusnumerictype.
ComparisonOperators:
Thecomparisonoperators.
POSIXStringComparison: StringcomparisonwithPOSIXrules.
BooleanOps:
Combiningcomparisonexpressionsusingbooleanoperators||(or),&&(and)

and!(not).
ConditionalExp:
Conditionalexpressionsselectbetweentwosubexpressionsundercontrolofathird

subexpression.
FunctionCalls:
Afunctioncallisanexpression.
Precedence:
Howvariousoperatorsnest.
Locales:
Howthelocaleaffectsthings.
ExpressionsSummary:
Expressionssummary.
PatternOverview:
Whatgoesintoapattern.
RegexpPatterns:
Usingregexpsaspatterns.
ExpressionPatterns:
Anyexpressioncanbeusedasapattern.
Ranges:
Pairsofpatternsspecifyrecordranges.
BEGIN/END:
Specifyinginitializationandcleanuprules.
UsingBEGIN/END:
HowandwhytouseBEGIN/ENDrules.
I/OAndBEGIN/END:
I/OissuesinBEGIN/ENDrules.
BEGINFILE/ENDFILE:
Twospecialpatternsforadvancedcontrol.
Empty:
Theemptypattern,whichmatcheseveryrecord.
UsingShellVariables:
Howtouseshellvariableswithawk.
ActionOverview:
Whatgoesintoanaction.
Statements:
Describesthevariouscontrolstatementsindetail.
IfStatement:
Conditionallyexecutesomeawkstatements.
WhileStatement:
Loopuntilsomeconditionissatisfied.
DoStatement:
Dospecifiedactionwhileloopinguntilsomeconditionissatisfied.
ForStatement:
Anotherloopingstatement,thatprovidesinitializationandincrementclauses.
SwitchStatement:
Switch/caseevaluationforconditionalexecutionofstatementsbasedonavalue.
BreakStatement:
Immediatelyexittheinnermostenclosingloop.
https://www.gnu.org/software/gawk/manual/gawk.html

4/479

29/7/2016

ContinueStatement:
NextStatement:
NextfileStatement:
ExitStatement:
BuiltinVariables:
Usermodified:
Autoset:
ARGCandARGV:
PatternActionSummary:
ArrayBasics:
ArrayIntro:
ReferencetoElements:
AssigningElements:
ArrayExample:
ScanninganArray:
ControllingScanning:
NumericArraySubscripts:
UninitializedSubscripts:
Delete:
Multidimensional:
Multiscanning:
ArraysofArrays:
ArraysSummary:
Builtin:
CallingBuiltin:
NumericFunctions:
StringFunctions:
GoryDetails:
I/OFunctions:
TimeFunctions:
BitwiseFunctions:
TypeFunctions:
I18NFunctions:
Userdefined:
DefinitionSyntax:
FunctionExample:
FunctionCaveats:
CallingAFunction:
VariableScope:
PassByValue/Reference:
ReturnStatement:
DynamicTyping:
IndirectCalls:
FunctionsSummary:
LibraryNames:
GeneralFunctions:
StrtonumFunction:
AssertFunction:
RoundFunction:
CliffRandomFunction:
OrdinalFunctions:

TheGNUAwkUsersGuide

Skiptotheendoftheinnermostenclosingloop.
Stopprocessingthecurrentinputrecord.
Stopprocessingthecurrentfile.
Stopexecutionofawk.
Summarizesthepredefinedvariables.
Builtinvariablesthatyouchangetocontrolawk.
Builtinvariableswhereawkgivesyouinformation.
WaystouseARGCandARGV.
PatternsandActionssummary.
Thebasicsofarrays.
IntroductiontoArrays
Howtoexamineoneelementofanarray.
Howtochangeanelementofanarray.
BasicExampleofanArray
Avariationoftheforstatement.Itloopsthroughtheindicesofanarraysexisting

elements.
Controllingtheorderinwhicharraysarescanned.
Howtousenumbersassubscriptsinawk.
UsingUninitializedvariablesassubscripts.
Thedeletestatementremovesanelementfromanarray.
Emulatingmultidimensionalarraysinawk.
Scanningmultidimensionalarrays.
Truemultidimensionalarrays.
Summaryofarrays.
Summarizesthebuiltinfunctions.
Howtocallbuiltinfunctions.
Functionsthatworkwithnumbers,includingint(),sin()andrand().
Functionsforstringmanipulation,suchassplit(),match()andsprintf().
Morethanyouwanttoknowabout\and&withsub(),gsub(),andgensub().
Functionsforfilesandshellcommands.
Functionsfordealingwithtimestamps.
Functionsforbitwiseoperations.
Functionsfortypeinformation.
Functionsforstringtranslation.
DescribesUserdefinedfunctionsindetail.
Howtowritedefinitionsandwhattheymean.
Anexamplefunctiondefinitionandwhatitdoes.
Thingstowatchoutfor.
Dontusespaces.
Controllingvariablescope.
Passingparameters.
Specifyingthevalueafunctionreturns.
Howvariabletypescanchangeatruntime.
Choosingthefunctiontocallatruntime.
Summaryoffunctions.
Howtobestnameprivateglobalvariablesinlibraryfunctions.
Functionsthatareofgeneraluse.
Areplacementforthebuiltinstrtonum()function.
Afunctionforassertionsinawkprograms.
Afunctionforroundingifsprintf()doesnotdoitcorrectly.
TheCliffRandomNumberGenerator.
Functionsforusingcharactersasnumbersandviceversa.

https://www.gnu.org/software/gawk/manual/gawk.html

5/479

29/7/2016

TheGNUAwkUsersGuide

JoinFunction:
Afunctiontojoinanarrayintoastring.
GetlocaltimeFunction:
Afunctiontogetformattedtimes.
ReadfileFunction:
Afunctiontoreadanentirefileatonce.
ShellQuoting:
Afunctiontoquotestringsfortheshell.
DataFileManagement:
Functionsformanagingcommandlinedatafiles.
FiletransFunction:
Afunctionforhandlingdatafiletransitions.
RewindFunction:
Afunctionforrereadingthecurrentfile.
FileChecking:
Checkingthatdatafilesarereadable.
EmptyFiles:
Checkingforzerolengthfiles.
IgnoringAssigns:
Treatingassignmentsasfilenames.
GetoptFunction:
Afunctionforprocessingcommandlinearguments.
PasswdFunctions:
Functionsforgettinguserinformation.
GroupFunctions:
Functionsforgettinggroupinformation.
WalkingArrays:
Afunctiontowalkarraysofarrays.
LibraryFunctionsSummary: Summaryoflibraryfunctions.
LibraryExercises:
Exercises.
RunningExamples:
Howtoruntheseexamples.
Clones:
Clonesofcommonutilities.
CutProgram:
Thecututility.
EgrepProgram:
Theegreputility.
IdProgram:
Theidutility.
SplitProgram:
Thesplitutility.
TeeProgram:
Theteeutility.
UniqProgram:
Theuniqutility.
WcProgram:
Thewcutility.
MiscellaneousPrograms:
Someinterestingawkprograms.
DupwordProgram:
Findingduplicatedwordsinadocument.
AlarmProgram:
Analarmclock.
TranslateProgram:
Aprogramsimilartothetrutility.
LabelsProgram:
Printingmailinglabels.
WordSorting:
Aprogramtoproduceawordusagecount.
HistorySorting:
Eliminatingduplicateentriesfromahistoryfile.
ExtractProgram:
PullingoutprogramsfromTexinfosourcefiles.
SimpleSed:
ASimpleStreamEditor.
IgawkProgram:
Awrapperforawkthatincludesfiles.
AnagramProgram:
Findinganagramsfromadictionary.
SignatureProgram:
Peopledoamazingthingswithtoomuchtimeontheirhands.
ProgramsSummary:
Summaryofprograms.
ProgramsExercises:
Exercises.
NondecimalData:
Allowingnondecimalinputdata.
ArraySorting:
Facilitiesforcontrollingarraytraversalandsortingarrays.
ControllingArrayTraversal: HowtousePROCINFO["sorted_in"].
ArraySortingFunctions:
Howtouseasort()andasorti().
TwowayI/O:
Twowaycommunicationswithanotherprocess.
TCP/IPNetworking:
Usinggawkfornetworkprogramming.
Profiling:
Profilingyourawkprograms.
AdvancedFeatures
Summaryofadvancedfeatures.

Summary:
I18NandL10N:
Explaininggettext:
Programmeri18n:

InternationalizationandLocalization.
HowGNUgettextworks.
Featuresfortheprogrammer.

https://www.gnu.org/software/gawk/manual/gawk.html

6/479

29/7/2016

TheGNUAwkUsersGuide

Translatori18n:
Featuresforthetranslator.
StringExtraction:
Extractingmarkedstrings.
PrintfOrdering:
Rearrangingprintfarguments.
I18NPortability:
awklevelportabilityissues.
I18NExample:
Asimplei18nexample.
GawkI18N:
gawkisalsointernationalized.
I18NSummary:
SummaryofI18Nstuff.
Debugging:
Introductiontogawkdebugger.
DebuggingConcepts:
DebugginginGeneral.
DebuggingTerms:
AdditionalDebuggingConcepts.
AwkDebugging:
AwkDebugging.
SampleDebuggingSession: Sampledebuggingsession.
DebuggerInvocation:
HowtoStarttheDebugger.
FindingTheBug:
FindingtheBug.
ListofDebugger
Maindebuggercommands.

Commands:
BreakpointControl:
ControlofBreakpoints.
DebuggerExecution
ControlofExecution.

Control:
ViewingAndChanging
ViewingandChangingData.

Data:
ExecutionStack:
DealingwiththeStack.
DebuggerInfo:
ObtainingInformationabouttheProgramandtheDebuggerState.
MiscellaneousDebugger
MiscellaneousCommands.

Commands:
ReadlineSupport:
Readlinesupport.
Limitations:
Limitationsandfutureplans.
DebuggingSummary:
Debuggingsummary.
ComputerArithmetic:
Aquickintrotocomputermath.
MathDefinitions:
Definingtermsused.
MPFRfeatures:
TheMPFRfeaturesingawk.
FPMathCaution:
Thingstoknow.
Inexactnessofcomputations: Floatingpointmathisnotexact.
Inexactrepresentation:
Numbersarenotexactlyrepresented.
ComparingFPValues:
Howtocomparefloatingpointvalues.
Errorsaccumulate:
Errorsgetbiggerastheygo.
GettingAccuracy:
Gettingmoreaccuracytakessomework.
TryToRound:
Adddigitsandround.
Settingprecision:
Howtosettheprecision.
Settingtheroundingmode: Howtosettheroundingmode.
ArbitraryPrecisionIntegers: ArbitraryPrecisionIntegerArithmeticwithgawk.
POSIXFloatingPoint
StandardsVersusExistingPractice.

Problems:
Floatingpointsummary:
Summaryoffloatingpointdiscussion.
ExtensionIntro:
Whatisanextension.
PluginLicense:
Anoteaboutlicensing.
ExtensionMechanism
Anoutlineofhowitworks.

Outline:
ExtensionAPIDescription: AfulldescriptionoftheAPI.
ExtensionAPIFunctions
IntroductiontotheAPIfunctions.

Introduction:
GeneralDataTypes:
Thedatatypes.
https://www.gnu.org/software/gawk/manual/gawk.html

7/479

29/7/2016

TheGNUAwkUsersGuide

MemoryAllocation
Functionsforallocatingmemory.
Functions:
ConstructorFunctions:
Functionsforcreatingvalues.
RegistrationFunctions:
Functionstoregisterthingswithgawk.
ExtensionFunctions:
Registeringextensionfunctions.
ExitCallbackFunctions:
Registeringanexitcallback.
ExtensionVersionString:
Registeringaversionstring.
InputParsers:
Registeringaninputparser.
OutputWrappers:
Registeringanoutputwrapper.
Twowayprocessors:
Registeringatwowayprocessor.
PrintingMessages:
Functionsforprintingmessages.
UpdatingERRNO:
FunctionsforupdatingERRNO.
RequestingValues:
Howtogetavalue.
AccessingParameters:
Functionsforaccessingparameters.
SymbolTableAccess:
Functionsforaccessingglobalvariables.
Symboltablebyname:
Accessingvariablesbyname.
Symboltablebycookie:
Accessingvariablesbycookie.
Cachedvalues:
Creatingandusingcachedvalues.
ArrayManipulation:
Functionsforworkingwitharrays.
ArrayDataTypes:
Datatypesforworkingwitharrays.
ArrayFunctions:
Functionsforworkingwitharrays.
FlatteningArrays:
Howtoflattenarrays.
CreatingArrays:
Howtocreateandpopulatearrays.
ExtensionAPIVariables:
VariablesprovidedbytheAPI.
ExtensionVersioning:
APIVersioninformation.
ExtensionAPIInformational Variablesprovidinginformationaboutgawksinvocation.

Variables:
ExtensionAPIBoilerplate: BoilerplatecodeforusingtheAPI.
FindingExtensions:
Howgawkfindscompiledextensions.
ExtensionExample:
ExampleCcodeforanextension.
InternalFileDescription:
Whatthenewfunctionswilldo.
InternalFileOps:
Thecodeforinternalfileoperations.
UsingInternalFileOps:
Howtouseanexternalextension.
ExtensionSamples:
Thesampleextensionsthatshipwithgawk.
ExtensionSampleFile
Thefilefunctionssample.

Functions:
ExtensionSampleFnmatch: Aninterfacetofnmatch().
ExtensionSampleFork:
Aninterfacetofork()andotherprocessfunctions.
ExtensionSampleInplace: Enablinginplacefileediting.
ExtensionSampleOrd:
Charactertovaluetocharacterconversions.
ExtensionSampleReaddir: Aninterfacetoreaddir().
ExtensionSampleRevout: Reversingoutputsampleoutputwrapper.
ExtensionSampleRev2way: Reversingdatasampletwowayprocessor.
ExtensionSampleRead
Serializinganarraytoafile.

writearray:
ExtensionSampleReadfile: Readinganentirefileintoastring.
ExtensionSampleTime:
Aninterfacetogettimeofday()andsleep().
ExtensionSampleAPITests: TestsfortheAPI.
gawkextlib:
Thegawkextlibproject.
Extensionsummary:
Extensionsummary.
ExtensionExercises:
Exercises.
V7/SVR3.1:
ThemajorchangesbetweenV7andSystemVRelease3.1.
https://www.gnu.org/software/gawk/manual/gawk.html

8/479

29/7/2016

TheGNUAwkUsersGuide

SVR4:
POSIX:
BTL:
POSIX/GNU:
FeatureHistory:
CommonExtensions:
RangesandLocales:
Contributors:
Historysummary:
GawkDistribution:
Getting:
Extracting:
Distributioncontents:
UnixInstallation:
QuickInstallation:
AdditionalConfiguration
Options:
ConfigurationPhilosophy:
NonUnixInstallation:
PCInstallation:
PCBinaryInstallation:
PCCompiling:
PCTesting:
PCUsing:
Cygwin:
MSYS:
VMSInstallation:
VMSCompilation:
VMSDynamicExtensions:
VMSInstallationDetails:
VMSRunning:
VMSGNV:
VMSOldGawk:
Bugs:
OtherVersions:
Installationsummary:
CompatibilityMode:

MinorchangesbetweenSystemVReleases3.1and4.
NewfeaturesfromthePOSIXstandard.
NewfeaturesfromBrianKernighansversionofawk.
TheextensionsingawknotinPOSIXawk.
Thehistoryofthefeaturesingawk.
CommonExtensionsSummary.
Howlocalesusedtoaffectregexpranges.
Themajorcontributorstogawk.
Historysummary.
Whatisinthegawkdistribution.
Howtogetthedistribution.
Howtoextractthedistribution.
Whatisinthedistribution.
InstallinggawkundervariousversionsofUnix.
CompilinggawkunderUnix.
Othercompiletimeoptions.

Additions:
AccessingTheSource:
AddingCode:
NewPorts:
DerivedFiles:
FutureExtensions:
ImplementationLimitations:
ExtensionDesign:
OldExtensionProblems:
ExtensionNewMechanism
Goals:
ExtensionOtherDesign
Decisions:
ExtensionFutureGrowth:
OldExtensionMechanism:

MakingAdditionsTogawk.
AccessingtheGitrepository.
Addingcodetothemainbodyofgawk.
Portinggawktoanewoperatingsystem.
WhyderivedfilesarekeptintheGitrepository.
Newfeaturesthatmaybeimplementedoneday.
Somelimitationsoftheimplementation.
DesignnotesabouttheextensionAPI.
Problemswiththeoldmechanism.
Goalsforthenewmechanism.

Howitsallsupposedtowork.
InstallationonOtherOperatingSystems.
InstallingandCompilinggawkonMSDOSandOS/2.
Installingaprepareddistribution.
CompilinggawkforMSDOS,Windows32,andOS/2.
TestinggawkonPCsystems.
RunninggawkonMSDOS,Windows32andOS/2.
BuildingandrunninggawkforCygwin.
UsinggawkInTheMSYSEnvironment.
InstallinggawkonVMS.
HowtocompilegawkunderVMS.
CompilinggawkdynamicextensionsonVMS.
HowtoinstallgawkunderVMS.
HowtorungawkunderVMS.
TheVMSGNVProject.
AnoldversioncomeswithsomeVMSsystems.
ReportingProblemsandBugs.
Otherfreelyavailableawkimplementations.
Summaryofinstallation.
Howtodisablecertaingawkextensions.

Someotherdesigndecisions.

Someroomforfuturegrowth.
Somecompatibilityforoldextensions.

https://www.gnu.org/software/gawk/manual/gawk.html

9/479

29/7/2016

TheGNUAwkUsersGuide

Notessummary:
BasicHighLevel:
BasicDataTyping:

Summaryofimplementationnotes.
Thehighlevelview.
Averyquickintrotodatatypes.

ShortTableofContents
ForewordtotheThirdEdition
ForewordtotheFourthEdition
Preface
PartI:
TheawkLanguage
1GettingStartedwithawk
2Runningawkandgawk
3RegularExpressions
4ReadingInputFiles
5PrintingOutput
6Expressions
7Patterns,Actions,andVariables
8Arraysinawk
9Functions
PartII:
ProblemSolvingwithawk
10ALibraryofawkFunctions
11PracticalawkPrograms
PartIII:
MovingBeyondStandardawkwithgawk
12AdvancedFeaturesofgawk
13Internationalizationwithgawk
14DebuggingawkPrograms
15ArithmeticandArbitraryPrecisionArithmeticwithgawk
16WritingExtensionsforgawk
PartIV:
Appendices
AppendixATheEvolutionoftheawkLanguage
AppendixBInstallinggawk
AppendixCImplementationNotes
AppendixDBasicProgrammingConcepts
Glossary
GNUGeneralPublicLicense
GNUFreeDocumentationLicense
Index

TableofContents
ForewordtotheThirdEdition
ForewordtotheFourthEdition
Preface
Historyofawkandgawk
ARosebyAnyOtherName
UsingThisBook
TypographicalConventions
DarkCorners
TheGNUProjectandThisBook
HowtoContribute
Acknowledgments
PartI:
https://www.gnu.org/software/gawk/manual/gawk.html

10/479

29/7/2016

TheGNUAwkUsersGuide

TheawkLanguage
1GettingStartedwithawk
1.1HowtoRunawkPrograms
1.1.1OneShotThrowawayawkPrograms
1.1.2RunningawkWithoutInputFiles
1.1.3RunningLongPrograms
1.1.4ExecutableawkPrograms
1.1.5CommentsinawkPrograms
1.1.6ShellQuotingIssues
1.1.6.1QuotinginMSWindowsBatchFiles
1.2DatafilesfortheExamples
1.3SomeSimpleExamples
1.4AnExamplewithTwoRules
1.5AMoreComplexExample
1.6awkStatementsVersusLines
1.7OtherFeaturesofawk
1.8WhentoUseawk
1.9Summary
2Runningawkandgawk
2.1Invokingawk
2.2CommandLineOptions
2.3OtherCommandLineArguments
2.4NamingStandardInput
2.5TheEnvironmentVariablesgawkUses
2.5.1TheAWKPATHEnvironmentVariable
2.5.2TheAWKLIBPATHEnvironmentVariable
2.5.3OtherEnvironmentVariables
2.6gawksExitStatus
2.7IncludingOtherFilesintoYourProgram
2.8LoadingDynamicExtensionsintoYourProgram
2.9ObsoleteOptionsand/orFeatures
2.10UndocumentedOptionsandFeatures
2.11Summary
3RegularExpressions
3.1HowtoUseRegularExpressions
3.2EscapeSequences
3.3RegularExpressionOperators
3.4UsingBracketExpressions
3.5HowMuchTextMatches?
3.6UsingDynamicRegexps
3.7gawkSpecificRegexpOperators
3.8CaseSensitivityinMatching
3.9Summary
4ReadingInputFiles
4.1HowInputIsSplitintoRecords
4.1.1RecordSplittingwithStandardawk
4.1.2RecordSplittingwithgawk
4.2ExaminingFields
4.3NonconstantFieldNumbers
4.4ChangingtheContentsofaField
4.5SpecifyingHowFieldsAreSeparated
4.5.1WhitespaceNormallySeparatesFields
4.5.2UsingRegularExpressionstoSeparateFields
4.5.3MakingEachCharacteraSeparateField
4.5.4SettingFSfromtheCommandLine
4.5.5MakingtheFullLineBeaSingleField
4.5.6FieldSplittingSummary
https://www.gnu.org/software/gawk/manual/gawk.html

11/479

29/7/2016

TheGNUAwkUsersGuide

4.6ReadingFixedWidthData
4.7DefiningFieldsbyContent
4.8MultipleLineRecords
4.9ExplicitInputwithgetline
4.9.1UsinggetlinewithNoArguments
4.9.2UsinggetlineintoaVariable
4.9.3UsinggetlinefromaFile
4.9.4UsinggetlineintoaVariablefromaFile
4.9.5UsinggetlinefromaPipe
4.9.6UsinggetlineintoaVariablefromaPipe
4.9.7UsinggetlinefromaCoprocess
4.9.8UsinggetlineintoaVariablefromaCoprocess
4.9.9PointstoRememberAboutgetline
4.9.10SummaryofgetlineVariants
4.10ReadingInputwithaTimeout
4.11DirectoriesontheCommandLine
4.12Summary
4.13Exercises
5PrintingOutput
5.1TheprintStatement
5.2printStatementExamples
5.3OutputSeparators
5.4ControllingNumericOutputwithprint
5.5UsingprintfStatementsforFancierPrinting
5.5.1IntroductiontotheprintfStatement
5.5.2FormatControlLetters
5.5.3ModifiersforprintfFormats
5.5.4ExamplesUsingprintf
5.6RedirectingOutputofprintandprintf
5.7SpecialFilesforStandardPreopenedDataStreams
5.8SpecialFilenamesingawk
5.8.1AccessingOtherOpenFileswithgawk
5.8.2SpecialFilesforNetworkCommunications
5.8.3SpecialFilenameCaveats
5.9ClosingInputandOutputRedirections
5.10Summary
5.11Exercises
6Expressions
6.1Constants,Variables,andConversions
6.1.1ConstantExpressions
6.1.1.1NumericandStringConstants
6.1.1.2OctalandHexadecimalNumbers
6.1.1.3RegularExpressionConstants
6.1.2UsingRegularExpressionConstants
6.1.3Variables
6.1.3.1UsingVariablesinaProgram
6.1.3.2AssigningVariablesontheCommandLine
6.1.4ConversionofStringsandNumbers
6.1.4.1HowawkConvertsBetweenStringsandNumbers
6.1.4.2LocalesCanInfluenceConversion
6.2Operators:DoingSomethingwithValues
6.2.1ArithmeticOperators
6.2.2StringConcatenation
6.2.3AssignmentExpressions
6.2.4IncrementandDecrementOperators
6.3TruthValuesandConditions
6.3.1TrueandFalseinawk
https://www.gnu.org/software/gawk/manual/gawk.html

12/479

29/7/2016

TheGNUAwkUsersGuide

6.3.2VariableTypingandComparisonExpressions
6.3.2.1StringTypeversusNumericType
6.3.2.2ComparisonOperators
6.3.2.3StringComparisonwithPOSIXRules
6.3.3BooleanExpressions
6.3.4ConditionalExpressions
6.4FunctionCalls
6.5OperatorPrecedence(HowOperatorsNest)
6.6WhereYouAreMakesaDifference
6.7Summary
7Patterns,Actions,andVariables
7.1PatternElements
7.1.1RegularExpressionsasPatterns
7.1.2ExpressionsasPatterns
7.1.3SpecifyingRecordRangeswithPatterns
7.1.4TheBEGINandENDSpecialPatterns
7.1.4.1StartupandCleanupActions
7.1.4.2Input/OutputfromBEGINandENDRules
7.1.5TheBEGINFILEandENDFILESpecialPatterns
7.1.6TheEmptyPattern
7.2UsingShellVariablesinPrograms
7.3Actions
7.4ControlStatementsinActions
7.4.1TheifelseStatement
7.4.2ThewhileStatement
7.4.3ThedowhileStatement
7.4.4TheforStatement
7.4.5TheswitchStatement
7.4.6ThebreakStatement
7.4.7ThecontinueStatement
7.4.8ThenextStatement
7.4.9ThenextfileStatement
7.4.10TheexitStatement
7.5PredefinedVariables
7.5.1BuiltinVariablesThatControlawk
7.5.2BuiltinVariablesThatConveyInformation
7.5.3UsingARGCandARGV
7.6Summary
8Arraysinawk
8.1TheBasicsofArrays
8.1.1IntroductiontoArrays
8.1.2ReferringtoanArrayElement
8.1.3AssigningArrayElements
8.1.4BasicArrayExample
8.1.5ScanningAllElementsofanArray
8.1.6UsingPredefinedArrayScanningOrderswithgawk
8.2UsingNumberstoSubscriptArrays
8.3UsingUninitializedVariablesasSubscripts
8.4ThedeleteStatement
8.5MultidimensionalArrays
8.5.1ScanningMultidimensionalArrays
8.6ArraysofArrays
8.7Summary
9Functions
9.1BuiltinFunctions
9.1.1CallingBuiltinFunctions
9.1.2NumericFunctions
https://www.gnu.org/software/gawk/manual/gawk.html

13/479

29/7/2016

TheGNUAwkUsersGuide

9.1.3StringManipulationFunctions
9.1.3.1Moreabout\and&withsub(),gsub(),andgensub()
9.1.4Input/OutputFunctions
9.1.5TimeFunctions
9.1.6BitManipulationFunctions
9.1.7GettingTypeInformation
9.1.8StringTranslationFunctions
9.2UserDefinedFunctions
9.2.1FunctionDefinitionSyntax
9.2.2FunctionDefinitionExamples
9.2.3CallingUserDefinedFunctions
9.2.3.1WritingaFunctionCall
9.2.3.2ControllingVariableScope
9.2.3.3PassingFunctionArgumentsbyValueOrbyReference
9.2.4ThereturnStatement
9.2.5FunctionsandTheirEffectsonVariableTyping
9.3IndirectFunctionCalls
9.4Summary
PartII:
ProblemSolvingwithawk
10ALibraryofawkFunctions
10.1NamingLibraryFunctionGlobalVariables
10.2GeneralProgramming
10.2.1ConvertingStringstoNumbers
10.2.2Assertions
10.2.3RoundingNumbers
10.2.4TheCliffRandomNumberGenerator
10.2.5TranslatingBetweenCharactersandNumbers
10.2.6MerginganArrayintoaString
10.2.7ManagingtheTimeofDay
10.2.8ReadingaWholeFileatOnce
10.2.9QuotingStringstoPasstotheShell
10.3DatafileManagement
10.3.1NotingDatafileBoundaries
10.3.2RereadingtheCurrentFile
10.3.3CheckingforReadableDatafiles
10.3.4CheckingforZeroLengthFiles
10.3.5TreatingAssignmentsasFilenames
10.4ProcessingCommandLineOptions
10.5ReadingtheUserDatabase
10.6ReadingtheGroupDatabase
10.7TraversingArraysofArrays
10.8Summary
10.9Exercises
11PracticalawkPrograms
11.1RunningtheExamplePrograms
11.2ReinventingWheelsforFunandProfit
11.2.1CuttingOutFieldsandColumns
11.2.2SearchingforRegularExpressionsinFiles
11.2.3PrintingOutUserInformation
11.2.4SplittingaLargeFileintoPieces
11.2.5DuplicatingOutputintoMultipleFiles
11.2.6PrintingNonduplicatedLinesofText
11.2.7CountingThings
11.3AGrabBagofawkPrograms
11.3.1FindingDuplicatedWordsinaDocument
11.3.2AnAlarmClockProgram
https://www.gnu.org/software/gawk/manual/gawk.html

14/479

29/7/2016

TheGNUAwkUsersGuide

11.3.3TransliteratingCharacters
11.3.4PrintingMailingLabels
11.3.5GeneratingWordUsageCounts
11.3.6RemovingDuplicatesfromUnsortedText
11.3.7ExtractingProgramsfromTexinfoSourceFiles
11.3.8ASimpleStreamEditor
11.3.9AnEasyWaytoUseLibraryFunctions
11.3.10FindingAnagramsfromaDictionary
11.3.11AndNowforSomethingCompletelyDifferent
11.4Summary
11.5Exercises
PartIII:
MovingBeyondStandardawkwithgawk
12AdvancedFeaturesofgawk
12.1AllowingNondecimalInputData
12.2ControllingArrayTraversalandArraySorting
12.2.1ControllingArrayTraversal
12.2.2SortingArrayValuesandIndiceswithgawk
12.3TwoWayCommunicationswithAnotherProcess
12.4UsinggawkforNetworkProgramming
12.5ProfilingYourawkPrograms
12.6Summary
13Internationalizationwithgawk
13.1InternationalizationandLocalization
13.2GNUgettext
13.3InternationalizingawkPrograms
13.4TranslatingawkPrograms
13.4.1ExtractingMarkedStrings
13.4.2RearrangingprintfArguments
13.4.3awkPortabilityIssues
13.5ASimpleInternationalizationExample
13.6gawkCanSpeakYourLanguage
13.7Summary
14DebuggingawkPrograms
14.1IntroductiontothegawkDebugger
14.1.1DebugginginGeneral
14.1.2DebuggingConcepts
14.1.3awkDebugging
14.2SamplegawkDebuggingSession
14.2.1HowtoStarttheDebugger
14.2.2FindingtheBug
14.3MainDebuggerCommands
14.3.1ControlofBreakpoints
14.3.2ControlofExecution
14.3.3ViewingandChangingData
14.3.4WorkingwiththeStack
14.3.5ObtainingInformationAbouttheProgramandtheDebuggerState
14.3.6MiscellaneousCommands
14.4ReadlineSupport
14.5Limitations
14.6Summary
15ArithmeticandArbitraryPrecisionArithmeticwithgawk
15.1AGeneralDescriptionofComputerArithmetic
15.2OtherStufftoKnow
15.3ArbitraryPrecisionArithmeticFeaturesingawk
15.4FloatingPointArithmetic:CaveatEmptor!
15.4.1FloatingPointArithmeticIsNotExact
https://www.gnu.org/software/gawk/manual/gawk.html

15/479

29/7/2016

TheGNUAwkUsersGuide

15.4.1.1ManyNumbersCannotBeRepresentedExactly
15.4.1.2BeCarefulComparingValues
15.4.1.3ErrorsAccumulate
15.4.2GettingtheAccuracyYouNeed
15.4.3TryaFewExtraBitsofPrecisionandRounding
15.4.4SettingthePrecision
15.4.5SettingtheRoundingMode
15.5ArbitraryPrecisionIntegerArithmeticwithgawk
15.6StandardsVersusExistingPractice
15.7Summary
16WritingExtensionsforgawk
16.1Introduction
16.2ExtensionLicensing
16.3HowItWorksataHighLevel
16.4APIDescription
16.4.1Introduction
16.4.2GeneralPurposeDataTypes
16.4.3MemoryAllocationFunctionsandConvenienceMacros
16.4.4ConstructorFunctions
16.4.5RegistrationFunctions
16.4.5.1RegisteringAnExtensionFunction
16.4.5.2RegisteringAnExitCallbackFunction
16.4.5.3RegisteringAnExtensionVersionString
16.4.5.4CustomizedInputParsers
16.4.5.5CustomizedOutputWrappers
16.4.5.6CustomizedTwowayProcessors
16.4.6PrintingMessages
16.4.7UpdatingERRNO
16.4.8RequestingValues
16.4.9AccessingandUpdatingParameters
16.4.10SymbolTableAccess
16.4.10.1VariableAccessandUpdatebyName
16.4.10.2VariableAccessandUpdatebyCookie
16.4.10.3CreatingandUsingCachedValues
16.4.11ArrayManipulation
16.4.11.1ArrayDataTypes
16.4.11.2ArrayFunctions
16.4.11.3WorkingWithAllTheElementsofanArray
16.4.11.4HowToCreateandPopulateArrays
16.4.12APIVariables
16.4.12.1APIVersionConstantsandVariables
16.4.12.2InformationalVariables
16.4.13BoilerplateCode
16.5HowgawkFindsExtensions
16.6Example:SomeFileFunctions
16.6.1Usingchdir()andstat()
16.6.2CCodeforchdir()andstat()
16.6.3IntegratingtheExtensions
16.7TheSampleExtensionsinthegawkDistribution
16.7.1FileRelatedFunctions
16.7.2Interfacetofnmatch()
16.7.3Interfacetofork(),wait(),andwaitpid()
16.7.4EnablingInPlaceFileEditing
16.7.5CharacterandNumericvalues:ord()andchr()
16.7.6ReadingDirectories
16.7.7ReversingOutput
16.7.8TwoWayI/OExample
https://www.gnu.org/software/gawk/manual/gawk.html

16/479

29/7/2016

TheGNUAwkUsersGuide

16.7.9DumpingandRestoringanArray
16.7.10ReadinganEntireFile
16.7.11ExtensionTimeFunctions
16.7.12APITests
16.8ThegawkextlibProject
16.9Summary
16.10Exercises
PartIV:
Appendices
AppendixATheEvolutionoftheawkLanguage
A.1MajorChangesBetweenV7andSVR3.1
A.2ChangesBetweenSVR3.1andSVR4
A.3ChangesBetweenSVR4andPOSIXawk
A.4ExtensionsinBrianKernighansawk
A.5ExtensionsingawkNotinPOSIXawk
A.6HistoryofgawkFeatures
A.7CommonExtensionsSummary
A.8RegexpRangesandLocales:ALongSadStory
A.9MajorContributorstogawk
A.10Summary
AppendixBInstallinggawk
B.1ThegawkDistribution
B.1.1GettingthegawkDistribution
B.1.2ExtractingtheDistribution
B.1.3ContentsofthegawkDistribution
B.2CompilingandInstallinggawkonUnixLikeSystems
B.2.1CompilinggawkforUnixLikeSystems
B.2.2AdditionalConfigurationOptions
B.2.3TheConfigurationProcess
B.3InstallationonOtherOperatingSystems
B.3.1InstallationonPCOperatingSystems
B.3.1.1InstallingaPreparedDistributionforPCSystems
B.3.1.2CompilinggawkforPCOperatingSystems
B.3.1.3TestinggawkonPCOperatingSystems
B.3.1.4UsinggawkonPCOperatingSystems
B.3.1.5UsinggawkInTheCygwinEnvironment
B.3.1.6UsinggawkInTheMSYSEnvironment
B.3.2CompilingandInstallinggawkonVax/VMSandOpenVMS
B.3.2.1CompilinggawkonVMS
B.3.2.2CompilinggawkDynamicExtensionsonVMS
B.3.2.3InstallinggawkonVMS
B.3.2.4RunninggawkonVMS
B.3.2.5TheVMSGNVProject
B.3.2.6SomeVMSSystemsHaveAnOldVersionofgawk
B.4ReportingProblemsandBugs
B.5OtherFreelyAvailableawkImplementations
B.6Summary
AppendixCImplementationNotes
C.1DownwardCompatibilityandDebugging
C.2MakingAdditionstogawk
C.2.1AccessingThegawkGitRepository
C.2.2AddingNewFeatures
C.2.3PortinggawktoaNewOperatingSystem
C.2.4WhyGeneratedFilesAreKeptInGit
C.3ProbableFutureExtensions
C.4SomeLimitationsoftheImplementation
C.5ExtensionAPIDesign
https://www.gnu.org/software/gawk/manual/gawk.html

17/479

29/7/2016

TheGNUAwkUsersGuide

C.5.1ProblemsWithTheOldMechanism
C.5.2GoalsForANewMechanism
C.5.3OtherDesignDecisions
C.5.4RoomForFutureGrowth
C.6CompatibilityForOldExtensions
C.7Summary
AppendixDBasicProgrammingConcepts
D.1WhataProgramDoes
D.2DataValuesinaComputer
Glossary
GNUGeneralPublicLicense
GNUFreeDocumentationLicense
ADDENDUM:HowtousethisLicenseforyourdocuments
Index
Next:Foreword4,Previous:Top,Up:Top[Contents][Index]

ForewordtotheThirdEdition
ArnoldRobbinsandIaregoodfriends.Wewereintroducedin1990bycircumstancesandourfavoriteprogramming
language,AWK.Thecircumstancesstartedacoupleofyearsearlier.Iwasworkingatanewjobandnoticedan
unpluggedUnixcomputersittinginthecorner.Nooneknewhowtouseit,andneitherdidI.However,acoupleofdays
later,itwasrunning,andIwasrootandtheoneandonlyuser.Thatday,IbeganthetransitionfromstatisticiantoUnix
programmer.
OnoneofmanytripstothelibraryorbookstoreinsearchofbooksonUnix,IfoundthegrayAWKbook,a.k.a.Alfred
V.Aho,BrianW.Kernighan,andPeterJ.WeinbergersTheAWKProgrammingLanguage(AddisonWesley,1988).
awkssimpleprogrammingparadigmfindapatternintheinputandthenperformanactionoftenreducedcomplexor
tediousdatamanipulationstoafewlinesofcode.IwasexcitedtotrymyhandatprogramminginAWK.
Alas,theawkonmycomputerwasalimitedversionofthelanguagedescribedinthegraybook.Idiscoveredthatmy
computerhadoldawkandthebookdescribednewawk.Ilearnedthatthiswastypicaltheoldversionrefusedtostep
asideorrelinquishitsname.Ifasystemhadanewawk,itwasinvariablycallednawk,andfewsystemshadit.Thebest
waytogetanewawkwastoftpthesourcecodeforgawkfromprep.ai.mit.edu.gawkwasaversionofnewawkwritten
byDavidTruemanandArnold,andavailableundertheGNUGeneralPublicLicense.
(Incidentally,itsnolongerdifficulttofindanewawk.gawkshipswithGNU/Linux,andyoucandownloadbinariesor
sourcecodeforalmostanysystemmywifeusesgawkonherVMSbox.)
MyUnixsystemstartedoutunpluggedfromthewallitcertainlywasnotpluggedintoanetwork.So,oblivioustothe
existenceofgawkandtheUnixcommunityingeneral,anddesiringanewawk,Iwrotemyown,calledmawk.BeforeI
wasfinished,Iknewaboutgawk,butitwastoolatetostop,soIeventuallypostedtoacomp.sourcesnewsgroup.
Afewdaysaftermyposting,IgotafriendlyemailfromArnoldintroducinghimself.Hesuggestedwesharedesignand
algorithmsandattachedadraftofthePOSIXstandardsothatIcouldupdatemawktosupportlanguageextensionsadded
afterpublicationofTheAWKProgrammingLanguage.
Frankly,ifourroleshadbeenreversed,Iwouldnothavebeensoopenandweprobablywouldhavenevermet.Imglad
wedidmeet.HeisanAWKexpertsAWKexpertandagenuinelyniceperson.Arnoldcontributessignificantamounts
ofhisexpertiseandtimetotheFreeSoftwareFoundation.
Thisbookisthegawkreferencemanual,butatitscoreitisabookaboutAWKprogrammingthatwillappealtoawide
audience.ItisadefinitivereferencetotheAWKlanguageasdefinedbythe1987BellLaboratoriesreleaseandcodified
inthe1992POSIXUtilitiesstandard.
Ontheotherhand,thenoviceAWKprogrammercanstudyawealthofpracticalprogramsthatemphasizethepowerof
AWKsbasicidioms:datadrivencontrolflow,patternmatchingwithregularexpressions,andassociativearrays.Those
https://www.gnu.org/software/gawk/manual/gawk.html

18/479

29/7/2016

TheGNUAwkUsersGuide

lookingforsomethingnewcantryoutgawksinterfacetonetworkprotocolsviaspecial/inetfiles.
TheprogramsinthisbookmakeclearthatanAWKprogramistypicallymuchsmallerandfastertodevelopthana
counterpartwritteninC.Consequently,thereisoftenapayofftoprototypinganalgorithmordesigninAWKtogetit
runningquicklyandexposeproblemsearly.Often,theinterpretedperformanceisadequateandtheAWKprototype
becomestheproduct.
Thenewpgawk(profilinggawk),producesprogramexecutioncounts.Irecentlyexperimentedwithanalgorithmthatfor
nlinesofinput,exhibited~Cn^2performance,whiletheorypredicted~Cnlognbehavior.Afewminutesporingover
theawkprof.outprofilepinpointedtheproblemtoasinglelineofcode.pgawkisawelcomeadditiontomy
programmerstoolbox.
ArnoldhasdistilledoveradecadeofexperiencewritingandusingAWKprograms,anddevelopinggawk,intothisbook.
IfyouuseAWKorwanttolearnhow,thenreadthisbook.
MichaelBrennan
Authorofmawk
March2001
Next:Preface,Previous:Foreword3,Up:Top[Contents][Index]

ForewordtotheFourthEdition
Somethingsdontchange.ThirteenyearsagoIwrote:IfyouuseAWKorwanttolearnhow,thenreadthisbook.
Truethen,andstilltruetoday.
Learningtouseaprogramminglanguageisaboutmorethanmasteringthesyntax.Oneneedstoacquirean
understandingofhowtousethefeaturesofthelanguagetosolvepracticalprogrammingproblems.Afocusofthisbook
ismanyexamplesthatshowhowtouseAWK.
Somethingsdochange.Ourcomputersaremuchfasterandhavemorememory.Consequently,speedandstorage
inefficienciesofahighlevellanguagematterless.PrototypinginAWKandthenrewritinginCforperformancereasons
happensless,becausemoreoftentheprototypeisfastenough.
Ofcourse,therearecomputingoperationsthatarebestdoneinCorC++.Withgawk4.1andlater,youdonothaveto
choosebetweenwritingyourprograminAWKorinC/C++.YoucanwritemostofyourprograminAWKandthe
aspectsthatrequireC/C++capabilitiescanbewritteninC/C++,andthenthepiecesgluedtogetherwhenthegawk
moduleloadstheC/C++moduleasadynamicplugin.DynamicExtensions,hasallthedetails,and,asexpected,many
examplestohelpyoulearntheinsandouts.
IenjoyprogramminginAWKandhadfun(re)readingthisbook.Ithinkyouwilltoo.
MichaelBrennan
Authorofmawk
October2014
Next:GettingStarted,Previous:Foreword4,Up:Top[Contents][Index]

Preface
Severalkindsoftasksoccurrepeatedlywhenworkingwithtextfiles.Youmightwanttoextractcertainlinesanddiscard
therest.Oryoumayneedtomakechangeswherevercertainpatternsappear,butleavetherestofthefilealone.Such
jobsareofteneasywithawk.Theawkutilityinterpretsaspecialpurposeprogramminglanguagethatmakesiteasyto
handlesimpledatareformattingjobs.
https://www.gnu.org/software/gawk/manual/gawk.html

19/479

29/7/2016

TheGNUAwkUsersGuide

TheGNUimplementationofawkiscalledgawkifyouinvokeitwiththeproperoptionsorenvironmentvariables,itis
fullycompatiblewiththePOSIX1specificationoftheawklanguageandwiththeUnixversionofawkmaintainedby
BrianKernighan.Thismeansthatallproperlywrittenawkprogramsshouldworkwithgawk.Somostofthetime,we
dontdistinguishbetweengawkandotherawkimplementations.
Usingawkyoucan:
Managesmall,personaldatabases
Generatereports
Validatedata
Produceindexesandperformotherdocumentpreparationtasks
Experimentwithalgorithmsthatyoucanadaptlatertoothercomputerlanguages
Inaddition,gawkprovidesfacilitiesthatmakeiteasyto:
Extractbitsandpiecesofdataforprocessing
Sortdata
Performsimplenetworkcommunications
Profileanddebugawkprograms
ExtendthelanguagewithfunctionswritteninCorC++
ThisWebpageteachesyouabouttheawklanguageandhowyoucanuseiteffectively.Youshouldalreadybefamiliar
withbasicsystemcommands,suchascatandls,2aswellasbasicshellfacilities,suchasinput/output(I/O)redirection
andpipes.
Implementationsoftheawklanguageareavailableformanydifferentcomputingenvironments.ThisWebpage,while
describingtheawklanguageingeneral,alsodescribestheparticularimplementationofawkcalledgawk(whichstandsfor
GNUawk).gawkrunsonabroadrangeofUnixsystems,rangingfromIntelarchitecturePCbasedcomputersup
throughlargescalesystems.gawkhasalsobeenportedtoMacOSX,MicrosoftWindows(allversions)andOS/2PCs,
andOpenVMS.3
History:
Thehistoryofgawkandawk.
Names:
Whatnametousetofindawk.
ThisManual:
UsingthisWebpage.Includessampleinputfilesthatyoucanuse.
Conventions:
TypographicalConventions.
ManualHistory:
BriefhistoryoftheGNUprojectandthisWebpage.
HowToContribute: Helpingtosavetheworld.
Acknowledgments: Acknowledgments.
Next:Names,Up:Preface[Contents][Index]

Historyofawkandgawk
RecipeforaProgrammingLanguage
1partegrep 1partsnobol
2partsed 3partsC
Blendallpartswellusinglexandyacc.Documentminimallyandrelease.
Aftereightyears,addanotherpartegrepandtwomorepartsC.Documentverywellandrelease.
Thenameawkcomesfromtheinitialsofitsdesigners:AlfredV.Aho,PeterJ.Weinberger,andBrianW.Kernighan.The
originalversionofawkwaswrittenin1977atAT&TBellLaboratories.In1985,anewversionmadetheprogramming
languagemorepowerful,introducinguserdefinedfunctions,multipleinputstreams,andcomputedregularexpressions.
https://www.gnu.org/software/gawk/manual/gawk.html

20/479

29/7/2016

TheGNUAwkUsersGuide

ThisnewversionbecamewidelyavailablewithUnixSystemVRelease3.1(1987).TheversioninSystemVRelease4
(1989)addedsomenewfeaturesandcleanedupthebehaviorinsomeofthedarkcornersofthelanguage.The
specificationforawkinthePOSIXCommandLanguageandUtilitiesstandardfurtherclarifiedthelanguage.Boththe
gawkdesignersandtheoriginalawkdesignersatBellLaboratoriesprovidedfeedbackforthePOSIXspecification.
PaulRubinwrotegawkin1986.JayFenlasoncompletedit,withadvicefromRichardStallman.JohnWoodscontributed
partsofthecodeaswell.In1988and1989,DavidTrueman,withhelpfromme,thoroughlyreworkedgawkfor
compatibilitywiththenewerawk.Circa1994,Ibecametheprimarymaintainer.Currentdevelopmentfocusesonbug
fixes,performanceimprovements,standardscompliance,and,occasionally,newfeatures.
InMay1997,JrgenKahrsfelttheneedfornetworkaccessfromawk,andwithalittlehelpfromme,setaboutadding
featurestodothisforgawk.Atthattime,healsowrotethebulkofTCP/IPInternetworkingwithgawk(aseparate
document,availableaspartofthegawkdistribution).Hiscodefinallybecamepartofthemaingawkdistributionwith
gawkversion3.1.
JohnHaquerewrotethegawkinternals,intheprocessprovidinganawkleveldebugger.Thisversionbecameavailableas
gawkversion4.0in2011.
SeeContributors,forafulllistofthosewhohavemadeimportantcontributionstogawk.
Next:ThisManual,Previous:History,Up:Preface[Contents][Index]

ARosebyAnyOtherName
Theawklanguagehasevolvedovertheyears.FulldetailsareprovidedinLanguageHistory.Thelanguagedescribedin
thisWebpageisoftenreferredtoasnewawk.Byanalogy,theoriginalversionofawkisreferredtoasoldawk.
Onmostcurrentsystems,whenyouruntheawkutilityyougetsomeversionofnewawk.4Ifyoursystemsstandardawk
istheoldone,youwillseesomethinglikethisifyoutrythetestprogram:
$awk1/dev/null
errorawk:syntaxerrornearline1
errorawk:bailingoutnearline1

Inthiscase,youshouldfindaversionofnewawk,orjustinstallgawk!
ThroughoutthisWebpage,wheneverwerefertoalanguagefeaturethatshouldbeavailableinanycomplete
implementationofPOSIXawk,wesimplyusethetermawk.WhenreferringtoafeaturethatisspecifictotheGNU
implementation,weusethetermgawk.
Next:Conventions,Previous:Names,Up:Preface[Contents][Index]

UsingThisBook
Thetermawkreferstoaparticularprogramaswellastothelanguageyouusetotellthisprogramwhattodo.Whenwe
needtobecareful,wecallthelanguagetheawklanguage,andtheprogramtheawkutility.ThisWebpageexplains
bothhowtowriteprogramsintheawklanguageandhowtoruntheawkutility.Thetermawkprogramreferstoa
programwrittenbyyouintheawkprogramminglanguage.
Primarily,thisWebpageexplainsthefeaturesofawkasdefinedinthePOSIXstandard.Itdoessointhecontextofthe
gawkimplementation.Whiledoingso,italsoattemptstodescribeimportantdifferencesbetweengawkandotherawk
implementations.5Finally,itnotesanygawkfeaturesthatarenotinthePOSIXstandardforawk.
ThisWebpagehasthedifficulttaskofbeingbothatutorialandareference.Ifyouareanovice,feelfreetoskipover
detailsthatseemtoocomplex.Youshouldalsoignorethemanycrossreferencestheyarefortheexpertuserandforthe
InfoandHTMLversionsoftheWebpage.
https://www.gnu.org/software/gawk/manual/gawk.html

21/479

29/7/2016

TheGNUAwkUsersGuide

TherearesidebarsscatteredthroughouttheWebpage.Theyaddamorecompleteexplanationofpointsthatarerelevant,
butnotlikelytobeofinterestonfirstreading.Allappearintheindex,undertheheadingsidebar.
Mostofthetime,theexamplesusecompleteawkprograms.Someofthemoreadvancedsectionsshowonlythepartof
theawkprogramthatillustratestheconceptbeingdescribed.
AlthoughthisWebpageisaimedprincipallyatpeoplewhohavenotbeenexposedtoawk,thereisalotofinformation
herethateventheawkexpertshouldfinduseful.Inparticular,thedescriptionofPOSIXawkandtheexampleprograms
inLibraryFunctions,andinSamplePrograms,shouldbeofinterest.
ThisWebpageissplitintoseveralparts,asfollows:
PartIdescribestheawklanguageandthegawkprogramindetail.Itstartswiththebasics,andcontinuesthroughall
ofthefeaturesofawk.Itcontainsthefollowingchapters:
GettingStarted,providestheessentialsyouneedtoknowtobeginusingawk.
InvokingGawk,describeshowtorungawk,themeaningofitscommandlineoptions,andhowitfindsawk
programsourcefiles.
Regexp,introducesregularexpressionsingeneral,andinparticulartheflavorssupportedbyPOSIXawk
andgawk.
ReadingFiles,describeshowawkreadsyourdata.Itintroducestheconceptsofrecordsandfields,aswell
asthegetlinecommand.I/Oredirectionisfirstdescribedhere.NetworkI/Oisalsobrieflyintroducedhere.
Printing,describeshowawkprogramscanproduceoutputwithprintandprintf.
Expressions,describesexpressions,whicharethebasicbuildingblocksforgettingmostthingsdoneina
program.
PatternsandActions,describeshowtowritepatternsformatchingrecords,actionsfordoingsomething
whenarecordismatched,andthepredefinedvariablesawkandgawkuse.
Arrays,coversawksoneandonlydatastructure:theassociativearray.Deletingarrayelementsandwhole
arraysisdescribed,aswellassortingarraysingawk.Thechapteralsodescribeshowgawkprovidesarraysof
arrays.
Functions,describesthebuiltinfunctionsawkandgawkprovide,aswellashowtodefineyourown
functions.Italsodiscusseshowgawkletsyoucallfunctionsindirectly.
PartIIshowshowtouseawkandgawkforproblemsolving.Thereislotsofcodehereforyoutoreadandlearn
from.Thispartcontainsthefollowingchapters:
LibraryFunctions,providesanumberoffunctionsmeanttobeusedfrommainawkprograms.
SamplePrograms,providesmanysampleawkprograms.
Readingthesetwochaptersallowsyoutoseeawksolvingrealproblems.
PartIIIfocusesonfeaturesspecifictogawk.Itcontainsthefollowingchapters:
AdvancedFeatures,describesanumberofadvancedfeatures.Ofparticularnotearetheabilitiestocontrol
theorderofarraytraversal,havetwowaycommunicationswithanotherprocess,performTCP/IP
networking,andprofileyourawkprograms.
Internationalization,describesspecialfeaturesfortranslatingprogrammessagesintodifferentlanguages
atruntime.
Debugger,describesthegawkdebugger.
ArbitraryPrecisionArithmetic,describesadvancedarithmeticfacilities.
DynamicExtensions,describeshowtoaddnewvariablesandfunctionstogawkbywritingextensionsinC
orC++.
PartIVprovidestheappendices,theGlossary,andtwolicensesthatcoverthegawksourcecodeandthisWeb
page,respectively.Itcontainsthefollowingappendices:
LanguageHistory,describeshowtheawklanguagehasevolvedsinceitsfirstreleasetothepresent.Italso
describeshowgawkhasacquiredfeaturesovertime.
Installation,describeshowtogetgawk,howtocompileitonPOSIXcompatiblesystems,andhowto
compileanduseitondifferentnonPOSIXsystems.Italsodescribeshowtoreportbugsingawkandwhere
togetotherfreelyavailableawkimplementations.
Notes,describeshowtodisablegawksextensions,aswellashowtocontributenewcodetogawk,and
somepossiblefuturedirectionsforgawkdevelopment.
BasicConcepts,providessomeverycursorybackgroundmaterialforthosewhoarecompletelyunfamiliar
https://www.gnu.org/software/gawk/manual/gawk.html

22/479

29/7/2016

TheGNUAwkUsersGuide

withcomputerprogramming.
TheGlossary,definesmost,ifnotall,ofthesignificanttermsusedthroughouttheWebpage.Ifyoufind
termsthatyouarentfamiliarwith,trylookingthemuphere.
Copying,andGNUFreeDocumentationLicense,presentthelicensesthatcoverthegawksourcecodeand
thisWebpage,respectively.
Next:ManualHistory,Previous:ThisManual,Up:Preface[Contents][Index]

TypographicalConventions
ThisWebpageiswritteninTexinfo,theGNUdocumentationformattinglanguage.AsingleTexinfosourcefileisused
toproduceboththeprintedandonlineversionsofthedocumentation.Becauseofthis,thetypographicalconventionsare
slightlydifferentthaninotherbooksyoumayhaveread.
Examplesyouwouldtypeatthecommandlineareprecededbythecommonshellprimaryandsecondaryprompts,$
and>.Inputthatyoutypeisshownlikethis.Outputfromthecommandisprecededbytheglyph|.Thistypically
representsthecommandsstandardoutput.Errormessagesandotheroutputonthecommandsstandarderrorare
precededbytheglypherror.Forexample:
$echohionstdout
|hionstdout
$echohelloonstderr1>&2
errorhelloonstderr

Inthetext,almostanythingrelatedtoprogramming,suchascommandnames,variableandfunctionnames,andstring,
numericandregexpconstantsappearinthisfont.Codefragmentsappearinthesamefontandquoted,likethis.
Thingsthatarereplacedbytheuserorprogrammerappearinthisfont.Optionslooklikethis:f.Filenamesare
indicatedlikethis:/path/to/ourfile.Somethingsareemphasizedlikethis,andifapointneedstobemadestrongly,it
isdonelikethis.Thefirstoccurrenceofanewtermisusuallyitsdefinitionandappearsinthesamefontastheprevious
occurrenceofdefinitioninthissentence.
Charactersthatyoutypeatthekeyboardlooklikethis.Inparticular,therearespecialcharacterscalledcontrol
characters.ThesearecharactersthatyoutypebyholdingdownboththeCONTROLkeyandanotherkey,atthesametime.
Forexample,aCtrldistypedbyfirstpressingandholdingtheCONTROLkey,nextpressingthedkey,andfinally
releasingbothkeys.
Forthesakeofbrevity,throughoutthisWebpage,werefertoBrianKernighansversionofawkasBWKawk.(See
OtherVersions,forinformationonhisandotherversions.)
DarkCorners
Darkcornersarebasicallyfractalnomatterhowmuchyouilluminate,theresalwaysasmallerbut
darkerone.
BrianKernighan
UntilthePOSIXstandard(andGAWK:EffectiveAWKProgramming),manyfeaturesofawkwereeitherpoorly
documentedornotdocumentedatall.Descriptionsofsuchfeatures(oftencalleddarkcorners)arenotedinthisWeb
pagewith(d.c.).Theyalsoappearintheindexundertheheadingdarkcorner.
But,asnotedbytheopeningquote,anycoverageofdarkcornersisbydefinitionincomplete.
Extensionstothestandardawklanguagethataresupportedbymorethanoneawkimplementationaremarked(c.e.),
andlistedintheindexundercommonextensionsandextensions,common.

https://www.gnu.org/software/gawk/manual/gawk.html

23/479

29/7/2016

TheGNUAwkUsersGuide

Next:HowToContribute,Previous:Conventions,Up:Preface[Contents][Index]

TheGNUProjectandThisBook
TheFreeSoftwareFoundation(FSF)isanonprofitorganizationdedicatedtotheproductionanddistributionoffreely
distributablesoftware.ItwasfoundedbyRichardM.Stallman,theauthoroftheoriginalEmacseditor.GNUEmacsis
themostwidelyusedversionofEmacstoday.
TheGNU6ProjectisanongoingeffortonthepartoftheFreeSoftwareFoundationtocreateacomplete,freely
distributable,POSIXcompliantcomputingenvironment.TheFSFusestheGNUGeneralPublicLicense(GPL)to
ensurethatitssoftwaressourcecodeisalwaysavailabletotheenduser.AcopyoftheGPLisincludedinthisWeb
pageforyourreference(seeCopying).TheGPLappliestotheClanguagesourcecodeforgawk.Tofindoutmoreabout
theFSFandtheGNUProjectonline,seetheGNUProjectshomepage.ThisWebpagemayalsobereadfromGNUs
website.
Ashell,aneditor(Emacs),highlyportableoptimizingC,C++,andObjectiveCcompilers,asymbolicdebuggerand
dozensoflargeandsmallutilities(suchasgawk),haveallbeencompletedandarefreelyavailable.TheGNUoperating
systemkernel(theHURD),hasbeenreleasedbutremainsinanearlystageofdevelopment.
UntiltheGNUoperatingsystemismorefullydeveloped,youshouldconsiderusingGNU/Linux,afreelydistributable,
UnixlikeoperatingsystemforIntel,PowerArchitecture,SunSPARC,IBMS/390,andothersystems.7Many
GNU/LinuxdistributionsareavailablefordownloadfromtheInternet.
TheWebpageyouarereadingisactuallyfreeatleast,theinformationinitisfreetoanyone.Themachinereadable
sourcecodefortheWebpagecomeswithgawk.(TakeamomenttochecktheFreeDocumentationLicenseinGNUFree
DocumentationLicense.)
TheWebpageitselfhasgonethroughmultiplepreviouseditions.PaulRubinwrotetheveryfirstdraftofTheGAWK
Manualitwasaround40pageslong.DianeCloseandRichardStallmanimprovedit,yieldingaversionthatwas
around90pagesandbarelydescribedtheoriginal,oldversionofawk.
Istartedworkingwiththatversioninthefallof1988.Asworkonitprogressed,theFSFpublishedseveralpreliminary
versions(numbered0.x).In1996,edition1.0wasreleasedwithgawk3.0.0.TheFSFpublishedthefirsttwoeditions
underthetitleTheGNUAwkUsersGuide.
Thiseditionmaintainsthebasicstructureofthepreviouseditions.ForFSFedition4.0,thecontentwasthoroughly
reviewedandupdated.Allreferencestogawkversionspriorto4.0wereremoved.Ofsignificantnoteforthateditionwas
theadditionofDebugger.
ForFSFedition4.1,thecontenthasbeenreorganizedintoparts,andthemajornewadditionsareArbitraryPrecision
Arithmetic,andDynamicExtensions.
ThisWebpagewillundoubtedlycontinuetoevolve.IfyoufindanerrorintheWebpage,pleasereportit!SeeBugs,for
informationonsubmittingproblemreportselectronically.
Next:Acknowledgments,Previous:ManualHistory,Up:Preface[Contents][Index]

HowtoContribute
AsthemaintainerofGNUawk,IoncethoughtthatIwouldbeabletomanageacollectionofpubliclyavailableawk
programsandIevensolicitedcontributions.MakingthingsavailableontheInternethelpskeepthegawkdistribution
downtomanageablesize.
Theinitialcollectionofmaterial,suchasitis,isstillavailableatftp://ftp.freefriends.org/arnold/Awkstuff.Inthehopes
ofdoingsomethingmorebroad,Iacquiredtheawk.infodomain.
However,IfoundthatIcouldnotdedicateenoughtimetomanagingcontributedcode:thearchivedidnotgrowandthe
https://www.gnu.org/software/gawk/manual/gawk.html

24/479

29/7/2016

TheGNUAwkUsersGuide

domainwentunusedforseveralyears.
Latein2008,avolunteertookonthetaskofsettingupanawkrelatedwebsitehttp://awk.infoanddidaverynice
job.
Ifyouhavewrittenaninterestingawkprogram,orhavewrittenagawkextensionthatyouwouldliketosharewiththe
restoftheworld,pleaseseehttp://awk.info/?contributeforhowtocontributeittothewebsite.
Previous:HowToContribute,Up:Preface[Contents][Index]

Acknowledgments
TheinitialdraftofTheGAWKManualhadthefollowingacknowledgments:
Manypeopleneedtobethankedfortheirassistanceinproducingthismanual.JayFenlasoncontributed
manyideasandsampleprograms.RichardMlynarikandRobertChassellgavehelpfulcommentsondrafts
ofthismanual.ThepaperASupplementalDocumentforAWKbyJohnW.PierceoftheChemistry
DepartmentatUCSanDiego,pinpointedseveralissuesrelevantbothtoawkimplementationandtothis
manual,thatwouldotherwisehaveescapedus.
IwouldliketoacknowledgeRichardM.Stallman,forhisvisionofabetterworldandforhiscourageinfoundingthe
FSFandstartingtheGNUProject.
EarliereditionsofthisWebpagehadthefollowingacknowledgements:
Thefollowingpeople(inalphabeticalorder)providedhelpfulcommentsonvariousversionsofthisbook:
RickAdams,Dr.NelsonH.F.Beebe,KarlBerry,Dr.MichaelBrennan,RichBurridge,ClaireCloutier,
DianeClose,ScottDeifik,Christopher(Topher)Eliot,JeffreyFriedl,Dr.DarrelHankerson,Michal
Jaegermann,Dr.RichardJ.LeBlanc,MichaelLijewski,PatRankin,MiriamRobbins,MarySheehan,and
ChuckToporek.
RobertJ.ChassellprovidedmuchvaluableadviceontheuseofTexinfo.Healsodeservesspecialthanks
forconvincingmenottotitlethisWebpageHowtoGawkPolitely.KarlBerryhelpedsignificantlywiththe
TeXpartofTexinfo.
IwouldliketothankMarshallandElaineHartholzofSeattleandDr.BertandRitaSchreiberofDetroitfor
largeamountsofquietvacationtimeintheirhomes,whichallowedmetomakesignificantprogressonthis
Webpageandongawkitself.
PhilHughesofSSCcontributedinaveryimportantwaybyloaningmehislaptopGNU/Linuxsystem,not
once,buttwice,whichallowedmetodoalotofworkwhileawayfromhome.
DavidTruemandeservesspecialcredithehasdoneayeomanjobofevolvinggawksothatitperformswell
andwithoutbugs.Althoughheisnolongerinvolvedwithgawk,workingwithhimonthisprojectwasa
significantpleasure.
TheintrepidmembersoftheGNITSmailinglist,andmostnotablyUlrichDrepper,providedinvaluable
helpandfeedbackforthedesignoftheinternationalizationfeatures.
ChuckToporek,MarySheehan,andClaireCloutierofOReilly&Associatescontributedsignificant
editorialhelpforthisWebpageforthe3.1releaseofgawk.
Dr.NelsonBeebe,AndreasBuening,Dr.ManuelCollado,AntonioColombo,StephenDavies,ScottDeifik,Akim
Demaille,DarrelHankerson,MichalJaegermann,JrgenKahrs,StepanKasal,JohnMalmberg,DavePitts,Chet
Ramey,PatRankin,AndrewSchorr,CorinnaVinschen,andEliZaretskii(inalphabeticalorder)makeupthecurrent
gawkcrackportabilityteam.Withouttheirhardworkandhelp,gawkwouldnotbenearlytherobust,portableprogram
itistoday.Ithasbeenandcontinuestobeapleasureworkingwiththisteamoffinepeople.
https://www.gnu.org/software/gawk/manual/gawk.html

25/479

29/7/2016

TheGNUAwkUsersGuide

Notablecodeanddocumentationcontributionsweremadebyanumberofpeople.SeeContributors,forthefulllist.
ThankstoMichaelBrennanfortheForewords.
ThankstoPatriceDumasforthenewmakeinfoprogram.ThankstoKarlBerry,whocontinuestoworktokeepthe
Texinfomarkuplanguagesane.
RobertP.J.Day,MichaelBrennan,andBrianKernighankindlyactedasreviewersforthe2015editionofthisWebpage.
Theirfeedbackhelpedimprovethefinalwork.
IwouldalsoliketothankBrianKernighanforhisinvaluableassistanceduringthetestinganddebuggingofgawk,and
forhisongoinghelpandadviceinclarifyingnumerouspointsaboutthelanguage.Wecouldnothavedonenearlyas
goodajoboneithergawkoritsdocumentationwithouthishelp.
Brianisinaclassbyhimselfasaprogrammerandtechnicalauthor.Ihavetothankhim(yetagain)forhisongoing
friendshipandforbeingarolemodeltomeforcloseto30years!Havinghimasareviewerisanexcitingprivilege.It
hasalsobeenextremelyhumbling...
Imustthankmywonderfulwife,Miriam,forherpatiencethroughthemanyversionsofthisproject,forher
proofreading,andforsharingmewiththecomputer.Iwouldliketothankmyparentsfortheirlove,andforthegrace
withwhichtheyraisedandeducatedme.Finally,IalsomustacknowledgemygratitudetoGd,forthemany
opportunitiesHehassentmyway,aswellasforthegiftsHehasgivenmewithwhichtotakeadvantageofthose
opportunities.

ArnoldRobbins
NofAyalon
Israel
February2015
Next:InvokingGawk,Previous:Preface,Up:Top[Contents][Index]

PartI:
TheawkLanguage
1GettingStartedwithawk
Thebasicfunctionofawkistosearchfilesforlines(orotherunitsoftext)thatcontaincertainpatterns.Whenaline
matchesoneofthepatterns,awkperformsspecifiedactionsonthatline.awkcontinuestoprocessinputlinesinthisway
untilitreachestheendoftheinputfiles.
Programsinawkaredifferentfromprogramsinmostotherlanguages,becauseawkprogramsaredatadriven(i.e.,you
describethedatayouwanttoworkwithandthenwhattodowhenyoufindit).Mostotherlanguagesareprocedural
youhavetodescribe,ingreatdetail,everysteptheprogramshouldtake.Whenworkingwithprocedurallanguages,itis
usuallymuchhardertoclearlydescribethedatayourprogramwillprocess.Forthisreason,awkprogramsareoften
refreshinglyeasytoreadandwrite.
Whenyourunawk,youspecifyanawkprogramthattellsawkwhattodo.Theprogramconsistsofaseriesofrules(it
mayalsocontainfunctiondefinitions,anadvancedfeaturethatwewillignorefornowseeUserdefined).Eachrule
specifiesonepatterntosearchforandoneactiontoperformuponfindingthepattern.
https://www.gnu.org/software/gawk/manual/gawk.html

26/479

29/7/2016

TheGNUAwkUsersGuide

Syntactically,aruleconsistsofapatternfollowedbyanaction.Theactionisenclosedinbracestoseparateitfromthe
pattern.Newlinesusuallyseparaterules.Therefore,anawkprogramlookslikethis:
pattern{action}
pattern{action}

Runninggawk:
Howtorungawkprogramsincludescommandlinesyntax.
SampleDataFiles: SampledatafilesforuseintheawkprogramsillustratedinthisWebpage.
VerySimple:
Averysimpleexample.
TwoRules:
Alesssimpleonelineexampleusingtworules.
MoreComplex: Amorecomplexexample.
Statements/Lines: Subdividingorcombiningstatementsintolines.
OtherFeatures:
OtherFeaturesofawk.
When:
Whentousegawkandwhentouseotherthings.
IntroSummary: Summaryoftheintroduction.
Next:SampleDataFiles,Up:GettingStarted[Contents][Index]

1.1HowtoRunawkPrograms
Thereareseveralwaystorunanawkprogram.Iftheprogramisshort,itiseasiesttoincludeitinthecommandthatruns
awk,likethis:
awk'program'inputfile1inputfile2

Whentheprogramislong,itisusuallymoreconvenienttoputitinafileandrunitwithacommandlikethis:
awkfprogramfileinputfile1inputfile2

Thissectiondiscussesbothmechanisms,alongwithseveralvariationsofeach.
Oneshot:
Runningashortthrowawayawkprogram.
ReadTerminal:
Usingnoinputfiles(inputfromthekeyboardinstead).
Long:
Puttingpermanentawkprogramsinfiles.
ExecutableScripts: Makingselfcontainedawkprograms.
Comments:
Addingdocumentationtogawkprograms.
Quoting:
Morediscussionofshellquotingissues.
Next:ReadTerminal,Up:Runninggawk[Contents][Index]
1.1.1OneShotThrowawayawkPrograms
Onceyouarefamiliarwithawk,youwilloftentypeinsimpleprogramsthemomentyouwanttousethem.Thenyoucan
writetheprogramasthefirstargumentoftheawkcommand,likethis:
awk'program'inputfile1inputfile2

whereprogramconsistsofaseriesofpatternsandactions,asdescribedearlier.
Thiscommandformatinstructstheshell,orcommandinterpreter,tostartawkandusetheprogramtoprocessrecordsin
theinputfile(s).Therearesinglequotesaroundprogramsotheshellwontinterpretanyawkcharactersasspecialshell
characters.Thequotesalsocausetheshelltotreatallofprogramasasingleargumentforawk,andallowprogramtobe
morethanonelinelong.
Thisformatisalsousefulforrunningshortormediumsizedawkprogramsfromshellscripts,becauseitavoidstheneed
foraseparatefilefortheawkprogram.Aselfcontainedshellscriptismorereliablebecausetherearenootherfilesto
https://www.gnu.org/software/gawk/manual/gawk.html

27/479

29/7/2016

TheGNUAwkUsersGuide

misplace.
Laterinthischapter,inVerySimple,wellseeexamplesofseveralshort,selfcontainedprograms.
Next:Long,Previous:Oneshot,Up:Runninggawk[Contents][Index]
1.1.2RunningawkWithoutInputFiles
Youcanalsorunawkwithoutanyinputfiles.Ifyoutypethefollowingcommandline:
awk'program'
awkappliestheprogramtothestandardinput,whichusuallymeanswhateveryoutypeonthekeyboard.Thiscontinues
untilyouindicateendoffilebytypingCtrld.(OnnonPOSIXoperatingsystems,theendoffilecharactermaybe
different.Forexample,onOS/2,itisCtrlz.)

Asanexample,thefollowingprogramprintsafriendlypieceofadvice(fromDouglasAdamssTheHitchhikersGuide
totheGalaxy),tokeepyoufromworryingaboutthecomplexitiesofcomputerprogramming:
$awk'BEGIN{print"Don\47tPanic!"}'
|Don'tPanic!
awkexecutesstatementsassociatedwithBEGINbeforereadinganyinput.Iftherearenootherstatementsinyour
program,asisthecasehere,awkjuststops,insteadoftryingtoreadinputitdoesntknowhowtoprocess.The\47isa

magicway(explainedlater)ofgettingasinglequoteintotheprogram,withouthavingtoengageinuglyshellquoting
tricks.
NOTE:IfyouuseBashasyourshell,youshouldexecutethecommandset+Hbeforerunningthis
programinteractively,todisabletheCshellstylecommandhistory,whichtreats!asaspecialcharacter.
Werecommendputtingthiscommandintoyourpersonalstartupfile.
Thisnextsimpleawkprogramemulatesthecatutilityitcopieswhateveryoutypeonthekeyboardtoitsstandard
output(whythisworksisexplainedshortly):
$awk'{print}'
Nowisthetimeforallgoodmen
|Nowisthetimeforallgoodmen
tocometotheaidoftheircountry.
|tocometotheaidoftheircountry.
Fourscoreandsevenyearsago,...
|Fourscoreandsevenyearsago,...
What,meworry?
|What,meworry?
Ctrld

Next:ExecutableScripts,Previous:ReadTerminal,Up:Runninggawk[Contents][Index]
1.1.3RunningLongPrograms
Sometimesawkprogramsareverylong.Inthesecases,itismoreconvenienttoputtheprogramintoaseparatefile.In
ordertotellawktousethatfileforitsprogram,youtype:
awkfsourcefileinputfile1inputfile2

Thefinstructstheawkutilitytogettheawkprogramfromthefilesourcefile(seeOptions).Anyfilenamecanbeused
forsourcefile.Forexample,youcouldputtheprogram:
BEGIN{print"Don'tPanic!"}
https://www.gnu.org/software/gawk/manual/gawk.html

28/479

29/7/2016

TheGNUAwkUsersGuide

intothefileadvice.Thenthiscommand:
awkfadvice

doesthesamethingasthisone:
awk'BEGIN{print"Don\47tPanic!"}'

Thiswasexplainedearlier(seeReadTerminal).Notethatyoudontusuallyneedsinglequotesaroundthefilenamethat
youspecifywithf,becausemostfilenamesdontcontainanyoftheshellsspecialcharacters.Noticethatinadvice,
theawkprogramdidnothavesinglequotesaroundit.Thequotesareonlyneededforprogramsthatareprovidedonthe
awkcommandline.(Also,placingtheprograminafileallowsustousealiteralsinglequoteintheprogramtext,instead
ofthemagic\47.)
Ifyouwanttoclearlyidentifyanawkprogramfileassuch,youcanaddtheextension.awktothefilename.Thisdoesnt
affecttheexecutionoftheawkprogrambutitdoesmakehousekeepingeasier.
Next:Comments,Previous:Long,Up:Runninggawk[Contents][Index]
1.1.4ExecutableawkPrograms
Onceyouhavelearnedawk,youmaywanttowriteselfcontainedawkscripts,usingthe#!scriptmechanism.Youcan
dothisonmanysystems.8Forexample,youcouldupdatethefileadvicetolooklikethis:
#!/bin/awkf
BEGIN{print"Don'tPanic!"}

Aftermakingthisfileexecutable(withthechmodutility),simplytypeadviceattheshellandthesystemarrangesto
runawkasifyouhadtypedawkfadvice:
$chmod+xadvice
$advice
|Don'tPanic!

(Weassumeyouhavethecurrentdirectoryinyourshellssearchpathvariable[typically$PATH].Ifnot,youmayneedto
type./adviceattheshell.)
Selfcontainedawkscriptsareusefulwhenyouwanttowriteaprogramthatuserscaninvokewithouttheirhavingto
knowthattheprogramiswritteninawk.
Understanding#!
awkisaninterpretedlanguage.Thismeansthattheawkutilityreadsyourprogramandthenprocessesyourdata

accordingtotheinstructionsinyourprogram.(ThisisdifferentfromacompiledlanguagesuchasC,whereyour
programisfirstcompiledintomachinecodethatisexecuteddirectlybyyoursystemsprocessor.)Theawkutilityis
thustermedaninterpreter.Manymodernlanguagesareinterpreted.
Thelinebeginningwith#!liststhefullfilenameofaninterpretertorunandasingleoptionalinitialcommandline
argumenttopasstothatinterpreter.Theoperatingsystemthenrunstheinterpreterwiththegivenargumentandthefull
argumentlistoftheexecutedprogram.Thefirstargumentinthelististhefullfilenameoftheawkprogram.Therest
oftheargumentlistcontainseitheroptionstoawk,ordatafiles,orboth.(Notethatonmanysystemsawkmaybefound
in/usr/bininsteadofin/bin.)
Somesystemslimitthelengthoftheinterpreternameto32characters.Often,thiscanbedealtwithbyusinga
symboliclink.
Youshouldnotputmorethanoneargumentonthe#!lineafterthepathto awk.Itdoesnotwork.Theoperating
systemtreatstherestofthelineasasingleargumentandpassesittoawk.Doingthisleadstoconfusingbehavior
https://www.gnu.org/software/gawk/manual/gawk.html

29/479

29/7/2016

TheGNUAwkUsersGuide

mostlikelyausagediagnosticofsomesortfromawk.
Finally,thevalueofARGV[0](seeBuiltinVariables)variesdependinguponyouroperatingsystem.Somesystemsput
awkthere,someputthefullpathnameofawk(suchas/bin/awk),andsomeputthenameofyourscript(advice).
(d.c.)DontrelyonthevalueofARGV[0]toprovideyourscriptname.

Next:Quoting,Previous:ExecutableScripts,Up:Runninggawk[Contents][Index]
1.1.5CommentsinawkPrograms
Acommentissometextthatisincludedinaprogramforthesakeofhumanreadersitisnotreallyanexecutablepartof
theprogram.Commentscanexplainwhattheprogramdoesandhowitworks.Nearlyallprogramminglanguageshave
provisionsforcomments,asprogramsaretypicallyhardtounderstandwithoutthem.
Intheawklanguage,acommentstartswiththenumbersigncharacter(#)andcontinuestotheendoftheline.The#
doesnothavetobethefirstcharacterontheline.Theawklanguageignorestherestofalinefollowinganumbersign.
Forexample,wecouldhaveputthefollowingintoadvice:
#Thisprogramprintsanice,friendlymessage.Ithelps
#keepnoviceusersfrombeingafraidofthecomputer.
BEGIN{print"Don'tPanic!"}

Youcanputcommentlinesintokeyboardcomposedthrowawayawkprograms,butthisusuallyisntveryusefulthe
purposeofacommentistohelpyouoranotherpersonunderstandtheprogramwhenreadingitatalatertime.
CAUTION:AsmentionedinOneshot,youcanencloseshorttomediumsizedprogramsinsinglequotes,
inordertokeepyourshellscriptsselfcontained.Whendoingso,dontputanapostrophe(i.e.,asingle
quote)intoacomment(oranywhereelseinyourprogram).Theshellinterpretsthequoteastheclosing
quotefortheentireprogram.Asaresult,usuallytheshellprintsamessageaboutmismatchedquotes,andif
awkactuallyruns,itwillprobablyprintstrangemessagesaboutsyntaxerrors.Forexample,lookatthe
following:
$awk'BEGIN{print"hello"}#let'sbecute'
>

Theshellseesthatthefirsttwoquotesmatch,andthatanewquotedobjectbeginsattheendofthe
commandline.Itthereforepromptswiththesecondaryprompt,waitingformoreinput.WithUnixawk,
closingthequotedstringproducesthisresult:
$awk'{print"hello"}#let'sbecute'
>'
errorawk:can'topenfilebe
errorsourcelinenumber1

Puttingabackslashbeforethesinglequoteinlet'swouldnthelp,becausebackslashesarenotspecial
insidesinglequotes.Thenextsubsectiondescribestheshellsquotingrules.
Previous:Comments,Up:Runninggawk[Contents][Index]
1.1.6ShellQuotingIssues
DOSQuoting: QuotinginWindowsBatchFiles.
Forshorttomediumlengthawkprograms,itismostconvenienttoentertheprogramontheawkcommandline.Thisis
bestdonebyenclosingtheentireprograminsinglequotes.Thisistruewhetheryouareenteringtheprogram
interactivelyattheshellprompt,orwritingitaspartofalargershellscript:
https://www.gnu.org/software/gawk/manual/gawk.html

30/479

29/7/2016

TheGNUAwkUsersGuide

awk'programtext'inputfile1inputfile2

Onceyouareworkingwiththeshell,itishelpfultohaveabasicknowledgeofshellquotingrules.Thefollowingrules
applyonlytoPOSIXcompliant,Bournestyleshells(suchasBash,theGNUBourneAgainShell).IfyouusetheC
shell,youreonyourown.
Beforedivingintotherules,weintroduceaconceptthatappearsthroughoutthisWebpage,whichisthatofthenull,or
empty,string.
Thenullstringischaracterdatathathasnovalue.Inotherwords,itisempty.Itiswritteninawkprogramslikethis:"".
Intheshell,itcanbewrittenusingsingleordoublequotes:""or''.Althoughthenullstringhasnocharactersinit,it
doesexist.Forexample,considerthiscommand:
$echo""

Here,theechoutilityreceivesasingleargument,eventhoughthatargumenthasnocharactersinit.Intherestofthis
Webpage,weusethetermsnullstringandemptystringinterchangeably.Now,ontothequotingrules:
Quoteditemscanbeconcatenatedwithnonquoteditemsaswellaswithotherquoteditems.Theshellturns
everythingintooneargumentforthecommand.
Precedinganysinglecharacterwithabackslash(\)quotesthatcharacter.Theshellremovesthebackslashand
passesthequotedcharacterontothecommand.
Singlequotesprotecteverythingbetweentheopeningandclosingquotes.Theshelldoesnointerpretationofthe
quotedtext,passingitonverbatimtothecommand.Itisimpossibletoembedasinglequoteinsidesinglequoted
text.ReferbacktoComments,foranexampleofwhathappensifyoutry.
Doublequotesprotectmostthingsbetweentheopeningandclosingquotes.Theshelldoesatleastvariableand
commandsubstitutiononthequotedtext.Differentshellsmaydoadditionalkindsofprocessingondouble
quotedtext.
Becausecertaincharacterswithindoublequotedtextareprocessedbytheshell,theymustbeescapedwithinthe
text.Ofnotearethecharacters$,`,\,and",allofwhichmustbeprecededbyabackslashwithindouble
quotedtextiftheyaretobepassedonliterallytotheprogram.(Theleadingbackslashisstrippedfirst.)Thus,the
exampleseenpreviouslyinReadTerminal:
awk'BEGIN{print"Don\47tPanic!"}'

couldinsteadbewrittenthisway:
$awk"BEGIN{print\"Don'tPanic!\"}"
|Don'tPanic!

Notethatthesinglequoteisnotspecialwithindoublequotes.
Nullstringsareremovedwhentheyoccuraspartofanonnullcommandlineargument,whileexplicitnull
objectsarekept.Forexample,tospecifythatthefieldseparatorFSshouldbesettothenullstring,use:
awkF""'program'files#correct

Dontusethis:
awkF""'program'files#wrong!

Inthesecondcase,awkattemptstousethetextoftheprogramasthevalueofFS,andthefirstfilenameasthetext
oftheprogram!Thisresultsinsyntaxerrorsatbest,andconfusingbehavioratworst.
Mixingsingleanddoublequotesisdifficult.Youhavetoresorttoshellquotingtricks,likethis:
$awk'BEGIN{print"Hereisasinglequote<'"'"'>"}'
|Hereisasinglequote<'>

Thisprogramconsistsofthreeconcatenatedquotedstrings.Thefirstandthethirdaresinglequoted,andthesecondis
https://www.gnu.org/software/gawk/manual/gawk.html

31/479

29/7/2016

TheGNUAwkUsersGuide

doublequoted.
Thiscanbesimplifiedto:
$awk'BEGIN{print"Hereisasinglequote<'\''>"}'
|Hereisasinglequote<'>

Judgeforyourselfwhichofthesetwoisthemorereadable.
Anotheroptionistousedoublequotes,escapingtheembedded,awkleveldoublequotes:
$awk"BEGIN{print\"Hereisasinglequote<'>\"}"
|Hereisasinglequote<'>

Thisoptionisalsopainful,becausedoublequotes,backslashes,anddollarsignsareverycommoninmoreadvancedawk
programs.
Athirdoptionistousetheoctalescapesequenceequivalents(seeEscapeSequences)forthesingleanddoublequote
characters,likeso:
$awk'BEGIN{print"Hereisasinglequote<\47>"}'
|Hereisasinglequote<'>
$awk'BEGIN{print"Hereisadoublequote<\42>"}'
|Hereisadoublequote<">

Thisworksnicely,butyoushouldcommentclearlywhattheescapesmean.
Afourthoptionistousecommandlinevariableassignment,likethis:
$awkvsq="'"'BEGIN{print"Hereisasinglequote<"sq">"}'
|Hereisasinglequote<'>

(Here,thetwostringconstantsandthevalueofsqareconcatenatedintoasinglestringthatisprintedbyprint.)
Ifyoureallyneedbothsingleanddoublequotesinyourawkprogram,itisprobablybesttomoveitintoaseparatefile,
wheretheshellwontbepartofthepictureandyoucansaywhatyoumean.
Up:Quoting[Contents][Index]
1.1.6.1QuotinginMSWindowsBatchFiles
AlthoughthisWebpagegenerallyonlyworriesaboutPOSIXsystemsandthePOSIXshell,thefollowingissuearises
oftenenoughformanyusersthatitisworthaddressing.
TheshellsonMicrosoftWindowssystemsusethedoublequotecharacterforquoting,andmakeitdifficultor
impossibletoincludeanescapeddoublequotecharacterinacommandlinescript.Thefollowingexample,courtesyof
JeroenBrink,showshowtoprintalllinesinafilesurroundedbydoublequotes:
gawk"{print\"\042\"$0\"\042\"}"file

Next:VerySimple,Previous:Runninggawk,Up:GettingStarted[Contents][Index]

1.2DatafilesfortheExamples
ManyoftheexamplesinthisWebpagetaketheirinputfromtwosampledatafiles.Thefirst,maillist,representsalist
ofpeoplesnamestogetherwiththeiremailaddressesandinformationaboutthosepeople.Theseconddatafile,called
inventoryshipped,containsinformationaboutmonthlyshipments.Inbothfiles,eachlineisconsideredtobeone
record.
https://www.gnu.org/software/gawk/manual/gawk.html

32/479

29/7/2016

TheGNUAwkUsersGuide

Inmaillist,eachrecordcontainsthenameofaperson,his/herphonenumber,his/heremailaddress,andacodefor
his/herrelationshipwiththeauthorofthelist.Thecolumnsarealignedusingspaces.AnAinthelastcolumnmeans
thatthepersonisanacquaintance.AnFinthelastcolumnmeansthatthepersonisafriend.AnRmeansthatthe
personisarelative:
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]

Thedatafileinventoryshippedrepresentsinformationaboutshipmentsduringtheyear.Eachrecordcontainsthe
month,thenumberofgreencratesshipped,thenumberofredboxesshipped,thenumberoforangebagsshipped,and
thenumberofbluepackagesshipped,respectively.Thereare16entries,coveringthe12monthsoflastyearandthefirst
fourmonthsofthecurrentyear.Anemptylineseparatesthedataforthetwoyears:
Jan132515115
Feb153224226
Mar152434228
Apr315263420
May163429208
Jun314275492
Jul243467436
Aug153447316
Sep135537277
Oct295468525
Nov208782577
Dec173561401
Jan213664620
Feb265880652
Mar247570495
Apr217074514

Thesamplefilesareincludedinthegawkdistribution,inthedirectoryawklib/eg/data.
Next:TwoRules,Previous:SampleDataFiles,Up:GettingStarted[Contents][Index]

1.3SomeSimpleExamples
Thefollowingcommandrunsasimpleawkprogramthatsearchestheinputfilemaillistforthecharacterstringli(a
groupingofcharactersisusuallycalledastringthetermstringisbasedonsimilarusageinEnglish,suchasastringof
pearlsorastringofcarsinatrain):
awk'/li/{print$0}'maillist

Whenlinescontainingliarefound,theyareprintedbecauseprint$0meansprintthecurrentline.(Justprintby
itselfmeansthesamething,sowecouldhavewrittenthatinstead.)
Youwillnoticethatslashes(/)surroundthestringliinthe awkprogram.Theslashesindicatethatliisthepattern
tosearchfor.Thistypeofpatterniscalledaregularexpression,whichiscoveredinmoredetaillater(seeRegexp).The
patternisallowedtomatchpartsofwords.Therearesinglequotesaroundtheawkprogramsothattheshellwont
interpretanyofitasspecialshellcharacters.
Hereiswhatthisprogramprints:
https://www.gnu.org/software/gawk/manual/gawk.html

33/479

29/7/2016

TheGNUAwkUsersGuide

$awk'/li/{print$0}'maillist
|[email protected]
|[email protected]
|[email protected]
|[email protected]

Inanawkrule,eitherthepatternortheactioncanbeomitted,butnotboth.Ifthepatternisomitted,thentheactionis
performedforeveryinputline.Iftheactionisomitted,thedefaultactionistoprintalllinesthatmatchthepattern.
Thus,wecouldleaveouttheaction(theprintstatementandthebraces)inthepreviousexampleandtheresultwould
bethesame:awkprintsalllinesmatchingthepatternli.Bycomparison,omittingtheprintstatementbutretainingthe
bracesmakesanemptyactionthatdoesnothing(i.e.,nolinesareprinted).
Manypracticalawkprogramsarejustalineortwolong.Followingisacollectionofuseful,shortprogramstogetyou
started.Someoftheseprogramscontainconstructsthathaventbeencoveredyet.(Thedescriptionoftheprogramwill
giveyouagoodideaofwhatisgoingon,butyoullneedtoreadtherestoftheWebpagetobecomeanawkexpert!)
Mostoftheexamplesuseadatafilenameddata.Thisisjustaplaceholderifyouusetheseprogramsyourself,
substituteyourownfilenamesfordata.Forfuturereference,notethatthereisoftenmorethanonewaytodothingsin
awk.Atsomepoint,youmaywanttolookbackattheseexamplesandseeifyoucancomeupwithdifferentwaystodo
thesamethingsshownhere:
Printeverylinethatislongerthan80characters:
awk'length($0)>80'data

Thesolerulehasarelationalexpressionasitspatternandhasnoactionsoitusesthedefaultaction,printingthe
record.
Printthelengthofthelongestinputline:
awk'{if(length($0)>max)max=length($0)}
END{printmax}'data

ThecodeassociatedwithENDexecutesafterallinputhasbeenreaditstheothersideofthecointoBEGIN.
Printthelengthofthelongestlineindata:
expanddata|awk'{if(x<length($0))x=length($0)}
END{print"maximumlinelengthis"x}'

Thisexamplediffersslightlyfromthepreviousone:theinputisprocessedbytheexpandutilitytochangeTABs
intospaces,sothewidthscomparedareactuallytherightmargincolumns,asopposedtothenumberofinput
charactersoneachline.
Printeverylinethathasatleastonefield:
awk'NF>0'data

Thisisaneasywaytodeleteblanklinesfromafile(orrather,tocreateanewfilesimilartotheoldfilebutfrom
whichtheblanklineshavebeenremoved).
Printsevenrandomnumbersfrom0to100,inclusive:
awk'BEGIN{for(i=1;i<=7;i++)
printint(101*rand())}'

Printthetotalnumberofbytesusedbyfiles:
lslfiles|awk'{x+=$5}
END{print"totalbytes:"x}'

Printthetotalnumberofkilobytesusedbyfiles:
lslfiles|awk'{x+=$5}
https://www.gnu.org/software/gawk/manual/gawk.html

34/479

29/7/2016

TheGNUAwkUsersGuide

END{print"totalKbytes:",x/1024}'

Printasortedlistoftheloginnamesofallusers:
awkF:'{print$1}'/etc/passwd|sort

Countthelinesinafile:
awk'END{printNR}'data

Printtheevennumberedlinesinthedatafile:
awk'NR%2==0'data

IfyouusedtheexpressionNR%2==1instead,theprogramwouldprinttheoddnumberedlines.
Next:MoreComplex,Previous:VerySimple,Up:GettingStarted[Contents][Index]

1.4AnExamplewithTwoRules
Theawkutilityreadstheinputfilesonelineatatime.Foreachline,awktriesthepatternsofeachrule.Ifseveralpatterns
match,thenseveralactionsexecuteintheorderinwhichtheyappearintheawkprogram.Ifnopatternsmatch,thenno
actionsrun.
Afterprocessingalltherulesthatmatchtheline(andperhapstherearenone),awkreadsthenextline.(However,see
NextStatement,andalsoseeNextfileStatement.)Thiscontinuesuntiltheprogramreachestheendofthefile.For
example,thefollowingawkprogramcontainstworules:
/12/{print$0}
/21/{print$0}

Thefirstrulehasthestring12asthepatternandprint$0astheaction.Thesecondrulehasthestring21asthe
patternandalsohasprint$0astheaction.Eachrulesactionisenclosedinitsownpairofbraces.
Thisprogramprintseverylinethatcontainsthestring12orthestring21.Ifalinecontainsbothstrings,itisprinted
twice,oncebyeachrule.
Thisiswhathappensifwerunthisprogramonourtwosampledatafiles,maillistandinventoryshipped:
$awk'/12/{print$0}
>/21/{print$0}'maillistinventoryshipped
|[email protected]
|[email protected]
|[email protected]
|[email protected]
|[email protected]
|Jan213664620
|Apr217074514

NotehowthelinebeginningwithJeanPaulin maillistwasprintedtwice,onceforeachrule.
Next:Statements/Lines,Previous:TwoRules,Up:GettingStarted[Contents][Index]

1.5AMoreComplexExample
Nowthatwevemasteredsomesimpletasks,letslookatwhattypicalawkprogramsdo.Thisexampleshowshowawk
canbeusedtosummarize,select,andrearrangetheoutputofanotherutility.Itusesfeaturesthathaventbeencovered
yet,sodontworryifyoudontunderstandallthedetails:
lsl|awk'$6=="Nov"{sum+=$5}
END{printsum}'
https://www.gnu.org/software/gawk/manual/gawk.html

35/479

29/7/2016

TheGNUAwkUsersGuide

Thiscommandprintsthetotalnumberofbytesinallthefilesinthecurrentdirectorythatwerelastmodifiedin
November(ofanyyear).Thelslpartofthisexampleisasystemcommandthatgivesyoualistingofthefilesina
directory,includingeachfilessizeandthedatethefilewaslastmodified.Itsoutputlookslikethis:
rwrr1arnolduser1933Nov713:05Makefile
rwrr1arnolduser10809Nov713:03awk.h
rwrr1arnolduser983Apr1312:14awk.tab.h
rwrr1arnolduser31869Jun1512:20awkgram.y
rwrr1arnolduser22414Nov713:03awk1.c
rwrr1arnolduser37455Nov713:03awk2.c
rwrr1arnolduser27511Dec913:07awk3.c
rwrr1arnolduser7989Nov713:03awk4.c

Thefirstfieldcontainsreadwritepermissions,thesecondfieldcontainsthenumberoflinkstothefile,andthethird
fieldidentifiesthefilesowner.Thefourthfieldidentifiesthefilesgroup.Thefifthfieldcontainsthefilessizeinbytes.
Thesixth,seventh,andeighthfieldscontainthemonth,day,andtime,respectively,thatthefilewaslastmodified.
Finally,theninthfieldcontainsthefilename.
The$6=="Nov"inourawkprogramisanexpressionthattestswhetherthesixthfieldoftheoutputfromlsl
matchesthestringNov.EachtimealinehasthestringNovforitssixthfield,awkperformstheactionsum+=$5.
Thisaddsthefifthfield(thefilessize)tothevariablesum.Asaresult,whenawkhasfinishedreadingalltheinputlines,
sumisthetotalofthesizesofthefileswhoselinesmatchedthepattern.(Thisworksbecauseawkvariablesare
automaticallyinitializedtozero.)
Afterthelastlineofoutputfromlshasbeenprocessed,theENDruleexecutesandprintsthevalueofsum.Inthis
example,thevalueofsumis80600.
Thesemoreadvancedawktechniquesarecoveredinlatersections(seeActionOverview).Beforeyoucanmoveonto
moreadvancedawkprogramming,youhavetoknowhowawkinterpretsyourinputanddisplaysyouroutput.By
manipulatingfieldsandusingprintstatements,youcanproducesomeveryusefulandimpressivelookingreports.
Next:OtherFeatures,Previous:MoreComplex,Up:GettingStarted[Contents][Index]

1.6awkStatementsVersusLines
Mostoften,eachlineinanawkprogramisaseparatestatementorseparaterule,likethis:
awk'/12/{print$0}
/21/{print$0}'maillistinventoryshipped

However,gawkignoresnewlinesafteranyofthefollowingsymbolsandkeywords:
,{?:||&&doelse

Anewlineatanyotherpointisconsideredtheendofthestatement.9
Ifyouwouldliketosplitasinglestatementintotwolinesatapointwhereanewlinewouldterminateit,youcan
continueitbyendingthefirstlinewithabackslashcharacter(\).Thebackslashmustbethefinalcharacterontheline
inordertoberecognizedasacontinuationcharacter.Abackslashisallowedanywhereinthestatement,eveninthe
middleofastringorregularexpression.Forexample:
awk'/Thisregularexpressionistoolong,socontinueit\
onthenextline/{print$1}'

Wehavegenerallynotusedbackslashcontinuationinoursampleprograms.gawkplacesnolimitonthelengthofaline,
sobackslashcontinuationisneverstrictlynecessaryitjustmakesprogramsmorereadable.Forthissamereason,as
wellasforclarity,wehavekeptmoststatementsshortintheprogramspresentedthroughouttheWebpage.Backslash
continuationismostusefulwhenyourawkprogramisinaseparatesourcefileinsteadofenteredfromthecommand
line.Youshouldalsonotethatmanyawkimplementationsaremoreparticularaboutwhereyoumayusebackslash
continuation.Forexample,theymaynotallowyoutosplitastringconstantusingbackslashcontinuation.Thus,for
https://www.gnu.org/software/gawk/manual/gawk.html

36/479

29/7/2016

TheGNUAwkUsersGuide

maximumportabilityofyourawkprograms,itisbestnottosplityourlinesinthemiddleofaregularexpressionora
string.
CAUTION:BackslashcontinuationdoesnotworkasdescribedwiththeCshell.Itworksforawkprograms
infilesandforoneshotprograms,providedyouareusingaPOSIXcompliantshell,suchastheUnix
BourneshellorBash.ButtheCshellbehavesdifferently!Thereyoumustusetwobackslashesinarow,
followedbyanewline.NotealsothatwhenusingtheCshell,everynewlineinyourawkprogrammustbe
escapedwithabackslash.Toillustrate:
%awk'BEGIN{\
?print\\
?"hello,world"\
?}'
|hello,world

Here,the%and?aretheCshellsprimaryandsecondaryprompts,analogoustothestandardshells$
and>.
ComparethepreviousexampletohowitisdonewithaPOSIXcompliantshell:
$awk'BEGIN{
>print\
>"hello,world"
>}'
|hello,world
awkisalineorientedlanguage.Eachrulesactionhastobeginonthesamelineasthepattern.Tohavethepatternand

actiononseparatelines,youmustusebackslashcontinuationthereisnootheroption.
Anotherthingtokeepinmindisthatbackslashcontinuationandcommentsdonotmix.Assoonasawkseesthe#that
startsacomment,itignoreseverythingontherestoftheline.Forexample:
$gawk'BEGIN{print"dontpanic"#afriendly\
>BEGINrule
>}'
errorgawk:cmd.line:2:BEGINrule
errorgawk:cmd.line:2:^syntaxerror

Inthiscase,itlookslikethebackslashwouldcontinuethecommentontothenextline.However,thebackslashnewline
combinationisneverevennoticedbecauseitishiddeninsidethecomment.Thus,theBEGINisnotedasasyntaxerror.
Whenawkstatementswithinoneruleareshort,youmightwanttoputmorethanoneofthemonaline.Thisis
accomplishedbyseparatingthestatementswithasemicolon(;).Thisalsoappliestotherulesthemselves.Thus,the
programshownatthestartofthissectioncouldalsobewrittenthisway:
/12/{print$0};/21/{print$0}

NOTE:Therequirementthatstatesthatrulesonthesamelinemustbeseparatedwithasemicolonwasnot
intheoriginalawklanguageitwasaddedforconsistencywiththetreatmentofstatementswithinanaction.
Next:When,Previous:Statements/Lines,Up:GettingStarted[Contents][Index]

1.7OtherFeaturesofawk
Theawklanguageprovidesanumberofpredefined,orbuiltin,variablesthatyourprogramscanusetogetinformation
fromawk.Thereareothervariablesyourprogramcansetaswelltocontrolhowawkprocessesyourdata.
Inaddition,awkprovidesanumberofbuiltinfunctionsfordoingcommoncomputationalandstringrelatedoperations.
gawkprovidesbuiltinfunctionsforworkingwithtimestamps,performingbitmanipulation,forruntimestring
translation(internationalization),determiningthetypeofavariable,andarraysorting.
https://www.gnu.org/software/gawk/manual/gawk.html

37/479

29/7/2016

TheGNUAwkUsersGuide

Aswedevelopourpresentationoftheawklanguage,wewillintroducemostofthevariablesandmanyofthefunctions.
TheyaredescribedsystematicallyinBuiltinVariables,andinBuiltin.
Next:IntroSummary,Previous:OtherFeatures,Up:GettingStarted[Contents][Index]

1.8WhentoUseawk
Nowthatyouveseensomeofwhatawkcando,youmightwonderhowawkcouldbeusefulforyou.Byusingutility
programs,advancedpatterns,fieldseparators,arithmeticstatements,andotherselectioncriteria,youcanproducemuch
morecomplexoutput.Theawklanguageisveryusefulforproducingreportsfromlargeamountsofrawdata,suchas
summarizinginformationfromtheoutputofotherutilityprogramslikels.(SeeMoreComplex.)
Programswrittenwithawkareusuallymuchsmallerthantheywouldbeinotherlanguages.Thismakesawkprograms
easytocomposeanduse.Often,awkprogramscanbequicklycomposedatyourkeyboard,usedonce,andthrownaway.
Becauseawkprogramsareinterpreted,youcanavoidthe(usuallylengthy)compilationpartofthetypicaleditcompile
testdebugcycleofsoftwaredevelopment.
Complexprogramshavebeenwritteninawk,includingacompleteretargetableassemblerforeightbitmicroprocessors
(seeGlossary,formoreinformation),andamicrocodeassemblerforaspecialpurposePrologcomputer.Theoriginal
awkscapabilitieswerestrainedbytasksofsuchcomplexity,butmodernversionsaremorecapable.
Ifyoufindyourselfwritingawkscriptsofmorethan,say,afewhundredlines,youmightconsiderusingadifferent
programminglanguage.Theshellisgoodatstringandpatternmatchinginaddition,itallowspowerfuluseofthe
systemutilities.Pythonoffersanicebalancebetweenhighleveleaseofprogrammingandaccesstosystemfacilities.10
Previous:When,Up:GettingStarted[Contents][Index]

1.9Summary
Programsinawkconsistofpatternactionpairs.
Anactionwithoutapatternalwaysruns.Thedefaultactionforapatternwithoutoneis{print$0}.
Useeitherawk'program'filesor awkfprogramfilefilestorun awk.
Youmayusethespecial#!headerlinetocreateawkprogramsthataredirectlyexecutable.
Commentsinawkprogramsstartwith#andcontinuetotheendofthesameline.
Beawareofquotingissueswhenwritingawkprogramsaspartofalargershellscript(orMSWindowsbatchfile).
Youmayusebackslashcontinuationtocontinueasourceline.Linesareautomaticallycontinuedafteracomma,
openbrace,questionmark,colon,||,&&,do,andelse.
Next:Regexp,Previous:GettingStarted,Up:Top[Contents][Index]

2Runningawkandgawk
Thischaptercovershowtorunawk,bothPOSIXstandardandgawkspecificcommandlineoptions,andwhatawkand
gawkdowithnonoptionarguments.Itthenproceedstocoverhowgawksearchesforsourcefiles,readingstandardinput
alongwithotherfiles,gawksenvironmentvariables,gawksexitstatus,usingincludefiles,andobsoleteand
undocumentedoptionsand/orfeatures.
ManyoftheoptionsandfeaturesdescribedherearediscussedinmoredetaillaterintheWebpagefeelfreetoskipover
thingsinthischapterthatdontinterestyourightnow.
CommandLine:
Options:
OtherArguments:

Howtorunawk.
Commandlineoptionsandtheirmeanings.
Inputfilenamesandvariableassignments.

https://www.gnu.org/software/gawk/manual/gawk.html

38/479

29/7/2016

TheGNUAwkUsersGuide

NamingStandardInput: Howtospecifystandardinputwithotherfiles.
EnvironmentVariables: Theenvironmentvariablesgawkuses.
ExitStatus:
gawksexitstatus.
IncludeFiles:
Includingotherfilesintoyourprogram.
LoadingSharedLibraries: Loadingsharedlibrariesintoyourprogram.
Obsolete:
ObsoleteOptionsand/orfeatures.
Undocumented:
UndocumentedOptionsandFeatures.
InvokingSummary:
Invocationsummary.
Next:Options,Up:InvokingGawk[Contents][Index]

2.1Invokingawk
Therearetwowaystorunawkwithanexplicitprogramorwithoneormoreprogramfiles.Herearetemplatesforboth
ofthemitemsenclosedin[]inthesetemplatesareoptional:
awk[options]fprogfile[]file
awk[options][]'program'file

InadditiontotraditionaloneletterPOSIXstyleoptions,gawkalsosupportsGNUlongoptions.
Itispossibletoinvokeawkwithanemptyprogram:
awk''datafile1datafile2

Doingsomakeslittlesense,thoughawkexitssilentlywhengivenanemptyprogram.(d.c.)Iflinthasbeenspecified
onthecommandline,gawkissuesawarningthattheprogramisempty.
Next:OtherArguments,Previous:CommandLine,Up:InvokingGawk[Contents][Index]

2.2CommandLineOptions
Optionsbeginwithadashandconsistofasinglecharacter.GNUstylelongoptionsconsistoftwodashesanda
keyword.Thekeywordcanbeabbreviated,aslongastheabbreviationallowstheoptiontobeuniquelyidentified.Ifthe
optiontakesanargument,eitherthekeywordisimmediatelyfollowedbyanequalssign(=)andtheargumentsvalue,
orthekeywordandtheargumentsvalueareseparatedbywhitespace.Ifaparticularoptionwithavalueisgivenmore
thanonce,itisthelastvaluethatcounts.
EachlongoptionforgawkhasacorrespondingPOSIXstyleshortoption.Thelongandshortoptionsareinterchangeable
inallcontexts.ThefollowinglistdescribesoptionsmandatedbythePOSIXstandard:
Ffs
fieldseparatorfs

SettheFSvariabletofs(seeFieldSeparators).
fsourcefile
filesourcefile

Readtheawkprogramsourcefromsourcefileinsteadofinthefirstnonoptionargument.Thisoptionmaybe
givenmultipletimestheawkprogramconsistsoftheconcatenationofthecontentsofeachspecifiedsourcefile.
vvar=val
assignvar=val

Setthevariablevartothevaluevalbeforeexecutionoftheprogrambegins.Suchvariablevaluesareavailable
insidetheBEGINrule(seeOtherArguments).
https://www.gnu.org/software/gawk/manual/gawk.html

39/479

29/7/2016

TheGNUAwkUsersGuide

Thevoptioncanonlysetonevariable,butitcanbeusedmorethanonce,settinganothervariableeachtime,like
this:awkvfoo=1vbar=2.
CAUTION:Usingvtosetthevaluesofthebuiltinvariablesmayleadtosurprisingresults.awk
willresetthevaluesofthosevariablesasitneedsto,possiblyignoringanyinitialvalueyoumayhave
given.
Wgawkopt

Provideanimplementationspecificoption.ThisisthePOSIXconventionforprovidingimplementationspecific
options.TheseoptionsalsohavecorrespondingGNUstylelongoptions.Notethatthelongoptionsmaybe
abbreviated,aslongastheabbreviationsremainunique.Thefulllistofgawkspecificoptionsisprovidednext.

Signaltheendofthecommandlineoptions.Thefollowingargumentsarenottreatedasoptionsevenifthey
beginwith.ThisinterpretationoffollowsthePOSIXargumentparsingconventions.
Thisisusefulifyouhavefilenamesthatstartwith,orinshellscripts,ifyouhavefilenamesthatwillbe
specifiedbytheuserthatcouldstartwith.Itisalsousefulforpassingoptionsontotheawkprogramsee
GetoptFunction.
Thefollowinglistdescribesgawkspecificoptions:
b
charactersasbytes

Causegawktotreatallinputdataassinglebytecharacters.Inaddition,alloutputwrittenwithprintorprintfis
treatedassinglebytecharacters.
Normally,gawkfollowsthePOSIXstandardandattemptstoprocessitsinputdataaccordingtothecurrentlocale
(seeLocales).Thiscanofteninvolveconvertingmultibytecharactersintowidecharacters(internally),andcan
leadtoproblemsorconfusioniftheinputdatadoesnotcontainvalidmultibytecharacters.Thisoptionisaneasy
waytotellgawk,Handsoffmydata!
c
traditional

Specifycompatibilitymode,inwhichtheGNUextensionstotheawklanguagearedisabled,sothatgawkbehaves
justlikeBWKawk.SeePOSIX/GNU,whichsummarizestheextensions.AlsoseeCompatibilityMode.
C
copyright

PrinttheshortversionoftheGeneralPublicLicenseandthenexit.
d[file]
dumpvariables[=file]

Printasortedlistofglobalvariables,theirtypes,andfinalvaluestofile.Ifnofileisprovided,printthislisttoa
filenamedawkvars.outinthecurrentdirectory.Nospaceisallowedbetweenthedandfile,iffileissupplied.
Havingalistofallglobalvariablesisagoodwaytolookfortypographicalerrorsinyourprograms.Youwould
alsousethisoptionifyouhavealargeprogramwithalotoffunctions,andyouwanttobesurethatyour
functionsdontinadvertentlyuseglobalvariablesthatyoumeanttobelocal.(Thisisaparticularlyeasymistake
tomakewithsimplevariablenameslikei,j,etc.)
D[file]
debug[=file]

Enabledebuggingofawkprograms(seeDebugging).Bydefault,thedebuggerreadscommandsinteractivelyfrom
https://www.gnu.org/software/gawk/manual/gawk.html

40/479

29/7/2016

TheGNUAwkUsersGuide

thekeyboard(standardinput).Theoptionalfileargumentallowsyoutospecifyafilewithalistofcommandsfor
thedebuggertoexecutenoninteractively.NospaceisallowedbetweentheDandfile,iffileissupplied.
eprogramtext
sourceprogramtext

Provideprogramsourcecodeintheprogramtext.Thisoptionallowsyoutomixsourcecodeinfileswithsource
codethatyouenteronthecommandline.Thisisparticularlyusefulwhenyouhavelibraryfunctionsthatyou
wanttousefromyourcommandlineprograms(seeAWKPATHVariable).
Efile
execfile

Similartof,readawkprogramtextfromfile.Therearetwodifferencesfromf:
Thisoptionterminatesoptionprocessinganythingelseonthecommandlineispassedondirectlytothe
awkprogram.
Commandlinevariableassignmentsoftheformvar=valuearedisallowed.
ThisoptionisparticularlynecessaryforWorldWideWebCGIapplicationsthatpassargumentsthroughtheURL
usingthisoptionpreventsamalicious(orother)userfrompassinginoptions,assignments,orawksourcecode
(viae)totheCGIapplication.11Thisoptionshouldbeusedwith#!scripts(seeExecutableScripts),likeso:
#!/usr/local/bin/gawkE
awkprogramhere
g
genpot

AnalyzethesourceprogramandgenerateaGNUgettextportableobjecttemplatefileonstandardoutputforall
stringconstantsthathavebeenmarkedfortranslation.SeeInternationalization,forinformationaboutthisoption.
h
help

Printausagemessagesummarizingtheshortandlongstyleoptionsthatgawkacceptsandthenexit.
isourcefile
includesourcefile

Readanawksourcelibraryfromsourcefile.Thisoptioniscompletelyequivalenttousingthe@includedirective
insideyourprogram.Itisverysimilartothefoption,buttherearetwoimportantdifferences.First,wheniis
used,theprogramsourceisnotloadedifithasbeenpreviouslyloaded,whereaswithf,gawkalwaysloadsthe
file.Second,becausethisoptionisintendedtobeusedwithcodelibraries,gawkdoesnotrecognizesuchfilesas
constitutingmainprograminput.Thus,afterprocessinganiargument,gawkstillexpectstofindthemainsource
codeviathefoptionoronthecommandline.
lext
loadext

Loadadynamicextensionnamedext.Extensionsarestoredassystemsharedlibraries.Thisoptionsearchesfor
thelibraryusingtheAWKLIBPATHenvironmentvariable.Thecorrectlibrarysuffixforyourplatformwillbe
suppliedbydefault,soitneednotbespecifiedintheextensionname.Theextensioninitializationroutineshould
benameddl_load().Analternativeistousethe@loadkeywordinsidetheprogramtoloadasharedlibrary.This
advancedfeatureisdescribedindetailinDynamicExtensions.
L[value]
lint[=value]

https://www.gnu.org/software/gawk/manual/gawk.html

41/479

29/7/2016

TheGNUAwkUsersGuide

Warnaboutconstructsthataredubiousornonportabletootherawkimplementations.Nospaceisallowedbetween
theLandvalue,ifvalueissupplied.Somewarningsareissuedwhengawkfirstreadsyourprogram.Othersare
issuedatruntime,asyourprogramexecutes.Withanoptionalargumentoffatal,lintwarningsbecomefatal
errors.Thismaybedrastic,butitsusewillcertainlyencouragethedevelopmentofcleanerawkprograms.Withan
optionalargumentofinvalid,onlywarningsaboutthingsthatareactuallyinvalidareissued.(Thisisnotfully
implementedyet.)
Somewarningsareonlyprintedonce,evenifthedubiousconstructstheywarnaboutoccurmultipletimesinyour
awkprogram.Thus,wheneliminatingproblemspointedoutbylint,youshouldtakecaretosearchforall
occurrencesofeachinappropriateconstruct.Asawkprogramsareusuallyshort,doingsoisnotburdensome.
M
bignum

Forcearbitraryprecisionarithmeticonnumbers.ThisoptionhasnoeffectifgawkisnotcompiledtousetheGNU
MPFRandMPlibraries(seeArbitraryPrecisionArithmetic).
n
nondecimaldata

Enableautomaticinterpretationofoctalandhexadecimalvaluesininputdata(seeNondecimalData).
CAUTION:Thisoptioncanseverelybreakoldprograms.Usewithcare.Alsonotethatthisoption
maydisappearinafutureversionofgawk.
N
uselcnumeric

Forcetheuseofthelocalesdecimalpointcharacterwhenparsingnumericinputdata(seeLocales).
o[file]
prettyprint[=file]

Enableprettyprintingofawkprograms.Bydefault,theoutputprogramiscreatedinafilenamedawkprof.out(see
Profiling).Theoptionalfileargumentallowsyoutospecifyadifferentfilenamefortheoutput.Nospaceis
allowedbetweentheoandfile,iffileissupplied.
NOTE:Duetothewaygawkhasevolved,withthisoptionyourprogramstillexecutes.Thiswill
changeinthenextmajorrelease,suchthatgawkwillonlyprettyprinttheprogramandnotrunit.
O
optimize

Enablesomeoptimizationsontheinternalrepresentationoftheprogram.Atthemoment,thisincludesjustsimple
constantfolding.
p[file]
profile[=file]

Enableprofilingofawkprograms(seeProfiling).Bydefault,profilesarecreatedinafilenamedawkprof.out.The
optionalfileargumentallowsyoutospecifyadifferentfilenamefortheprofilefile.Nospaceisallowedbetween
thepandfile,iffileissupplied.
Theprofilecontainsexecutioncountsforeachstatementintheprogramintheleftmargin,andfunctioncall
countsforeachfunction.
P
posix

OperateinstrictPOSIXmode.Thisdisablesallgawkextensions(justliketraditional)anddisablesall
extensionsnotallowedbyPOSIX.SeeCommonExtensions,forasummaryoftheextensionsingawkthatare
https://www.gnu.org/software/gawk/manual/gawk.html

42/479

29/7/2016

TheGNUAwkUsersGuide

disabledbythisoption.Also,thefollowingadditionalrestrictionsapply:
NewlinesdonotactaswhitespacetoseparatefieldswhenFSisequaltoasinglespace(seeFields).
Newlinesarenotallowedafter?or:(seeConditionalExp).
SpecifyingFtonthecommandlinedoesnotsetthevalueofFStobeasingleTABcharacter(seeField
Separators).
Thelocalesdecimalpointcharacterisusedforparsinginputdata(seeLocales).
Ifyousupplybothtraditionalandposixonthecommandline,posixtakesprecedence.gawkissuesa
warningifbothoptionsaresupplied.
r
reinterval

Allowintervalexpressions(seeRegexpOperators)inregexps.Thisisnowgawksdefaultbehavior.Nevertheless,
thisoptionremains(bothforbackwardcompatibilityandforuseincombinationwithtraditional).
S
sandbox

Disablethesystem()function,inputredirectionswithgetline,outputredirectionswithprintandprintf,and
dynamicextensions.Thisisparticularlyusefulwhenyouwanttorunawkscriptsfromquestionablesourcesand
needtomakesurethescriptscantaccessyoursystem(otherthanthespecifiedinputdatafile).
t
lintold

WarnaboutconstructsthatarenotavailableintheoriginalversionofawkfromVersion7Unix(seeV7/SVR3.1).
V
version

Printversioninformationforthisparticularcopyofgawk.Thisallowsyoutodetermineifyourcopyofgawkisup
todatewithrespecttowhatevertheFreeSoftwareFoundationiscurrentlydistributing.Itisalsousefulforbug
reports(seeBugs).
Aslongasprogramtexthasbeensupplied,anyotheroptionsareflaggedasinvalidwithawarningmessagebutare
otherwiseignored.
Incompatibilitymode,asaspecialcase,ifthevalueoffssuppliedtotheFoptionist,thenFSissettotheTAB
character("\t").Thisistrueonlyfortraditionalandnotforposix(seeFieldSeparators).
Thefoptionmaybeusedmorethanonceonthecommandline.Ifitis,awkreadsitsprogramsourcefromallofthe
namedfiles,asiftheyhadbeenconcatenatedtogetherintoonebigfile.Thisisusefulforcreatinglibrariesofawk
functions.Thesefunctionscanbewrittenonceandthenretrievedfromastandardplace,insteadofhavingtobe
includedineachindividualprogram.Theioptionissimilarinthisregard.(AsmentionedinDefinitionSyntax,
functionnamesmustbeunique.)
Withstandardawk,libraryfunctionscanstillbeused,eveniftheprogramisenteredatthekeyboard,byspecifyingf
/dev/tty.Aftertypingyourprogram,typeCtrld(theendoffilecharacter)toterminateit.(Youmayalsousefto
readprogramsourcefromthestandardinput,butthenyouwillnotbeabletoalsousethestandardinputasasourceof
data.)
Becauseitisclumsyusingthestandardawkmechanismstomixsourcefileandcommandlineawkprograms,gawk
providestheeoption.Thisdoesnotrequireyoutopreemptthestandardinputforyoursourcecodeitallowsyouto
easilymixcommandlineandlibrarysourcecode(seeAWKPATHVariable).Aswithf,theeandioptionsmayalso
beusedmultipletimesonthecommandline.
Ifnoforeoptionisspecified,thengawkusesthefirstnonoptioncommandlineargumentasthetextoftheprogram
sourcecode.
https://www.gnu.org/software/gawk/manual/gawk.html

43/479

29/7/2016

TheGNUAwkUsersGuide

IftheenvironmentvariablePOSIXLY_CORRECTexists,thengawkbehavesinstrictPOSIXmode,exactlyasifyouhad
suppliedposix.ManyGNUprogramslookforthisenvironmentvariabletosuppressextensionsthatconflictwith
POSIX,butgawkbehavesdifferently:itsuppressesallextensions,eventhosethatdonotconflictwithPOSIX,and
behavesinstrictPOSIXmode.IflintissuppliedonthecommandlineandgawkturnsonPOSIXmodebecauseof
POSIXLY_CORRECT,thenitissuesawarningmessageindicatingthatPOSIXmodeisineffect.Youwouldtypicallysetthis
variableinyourshellsstartupfile.ForaBournecompatibleshell(suchasBash),youwouldaddtheselinestothe
.profilefileinyourhomedirectory:
POSIXLY_CORRECT=true
exportPOSIXLY_CORRECT

ForaCshellcompatibleshell,12youwouldaddthislinetothe.loginfileinyourhomedirectory:
setenvPOSIXLY_CORRECTtrue

HavingPOSIXLY_CORRECTsetisnotrecommendedfordailyuse,butitisgoodfortestingtheportabilityofyourprograms
tootherenvironments.
Next:NamingStandardInput,Previous:Options,Up:InvokingGawk[Contents][Index]

2.3OtherCommandLineArguments
Anyadditionalargumentsonthecommandlinearenormallytreatedasinputfilestobeprocessedintheorderspecified.
However,anargumentthathastheformvar=value,assignsthevaluevaluetothevariablevaritdoesnotspecifyafile
atall.(SeeAssignmentOptions.)Inthefollowingexample,count=1isavariableassignment,notafilename:
awkfprogram.awkfile1count=1file2

AllthecommandlineargumentsaremadeavailabletoyourawkprogramintheARGVarray(seeBuiltinVariables).
Commandlineoptionsandtheprogramtext(ifpresent)areomittedfromARGV.Allotherarguments,includingvariable
assignments,areincluded.AseachelementofARGVisprocessed,gawksetsARGINDtotheindexinARGVofthecurrent
element.
ChangingARGCandARGVinyourawkprogramletsyoucontrolhowawkprocessestheinputfilesthisisdescribedinmore
detailinARGCandARGV.
Thedistinctionbetweenfilenameargumentsandvariableassignmentargumentsismadewhenawkisabouttoopenthe
nextinputfile.Atthatpointinexecution,itchecksthefilenametoseewhetheritisreallyavariableassignmentifso,
awksetsthevariableinsteadofreadingafile.
Therefore,thevariablesactuallyreceivethegivenvaluesafterallpreviouslyspecifiedfileshavebeenread.In
particular,thevaluesofvariablesassignedinthisfashionarenotavailableinsideaBEGINrule(seeBEGIN/END),
becausesuchrulesarerunbeforeawkbeginsscanningtheargumentlist.
Thevariablevaluesgivenonthecommandlineareprocessedforescapesequences(seeEscapeSequences).(d.c.)
Insomeveryearlyimplementationsofawk,whenavariableassignmentoccurredbeforeanyfilenames,theassignment
wouldhappenbeforetheBEGINrulewasexecuted.awksbehaviorwasthusinconsistentsomecommandline
assignmentswereavailableinsidetheBEGINrule,whileotherswerenot.Unfortunately,someapplicationscameto
dependuponthisfeature.Whenawkwaschangedtobemoreconsistent,thevoptionwasaddedtoaccommodate
applicationsthatdependedupontheoldbehavior.
ThevariableassignmentfeatureismostusefulforassigningtovariablessuchasRS,OFS,andORS,whichcontrolinput
andoutputformats,beforescanningthedatafiles.Itisalsousefulforcontrollingstateifmultiplepassesareneeded
overadatafile.Forexample:
awk'pass==1{pass1stuff}
pass==2{pass2stuff}'pass=1mydatapass=2mydata
https://www.gnu.org/software/gawk/manual/gawk.html

44/479

29/7/2016

TheGNUAwkUsersGuide

Giventhevariableassignmentfeature,theFoptionforsettingthevalueofFSisnotstrictlynecessary.Itremainsfor
historicalcompatibility.
Next:EnvironmentVariables,Previous:OtherArguments,Up:InvokingGawk[Contents][Index]

2.4NamingStandardInput
Often,youmaywishtoreadstandardinputtogetherwithotherfiles.Forexample,youmaywishtoreadonefile,read
standardinputcomingfromapipe,andthenreadanotherfile.
Thewaytonamethestandardinput,withallversionsofawk,istouseasingle,standaloneminussignordash,.For
example:
some_command|awkfmyprog.awkfile1file2

Here,awkfirstreadsfile1,thenitreadstheoutputofsome_command,andfinallyitreadsfile2.
Youmayalsouse""tonamestandardinputwhenreadingfileswithgetline(seeGetline/File).
Inaddition,gawkallowsyoutospecifythespecialfilename/dev/stdin,bothonthecommandlineandwithgetline.
Someotherversionsofawkalsosupportthis,butitisnotstandard.(Someoperatingsystemsprovidea/dev/stdinfilein
thefilesystemhowever,gawkalwaysprocessesthisfilenameitself.)
Next:ExitStatus,Previous:NamingStandardInput,Up:InvokingGawk[Contents][Index]

2.5TheEnvironmentVariablesgawkUses
Anumberofenvironmentvariablesinfluencehowgawkbehaves.
AWKPATHVariable:
Searchingdirectoriesforawkprograms.
AWKLIBPATHVariable:
Searchingdirectoriesforawksharedlibraries.
OtherEnvironmentVariables: Theenvironmentvariables.
Next:AWKLIBPATHVariable,Up:EnvironmentVariables[Contents][Index]
2.5.1TheAWKPATHEnvironmentVariable
Inmostawkimplementations,youmustsupplyaprecisepathnameforeachprogramfile,unlessthefileisinthecurrent
directory.Butwithgawk,ifthefilenamesuppliedtotheforioptionsdoesnotcontainadirectoryseparator/,then
gawksearchesalistofdirectories(calledthesearchpath)onebyone,lookingforafilewiththespecifiedname.
Thesearchpathisastringconsistingofdirectorynamesseparatedbycolons.13gawkgetsitssearchpathfromthe
AWKPATHenvironmentvariable.Ifthatvariabledoesnotexist,orifithasanemptyvalue,gawkusesadefaultpath
(describedshortly).
Thesearchpathfeatureisparticularlyhelpfulforbuildinglibrariesofusefulawkfunctions.Thelibraryfilescanbe
placedinastandarddirectoryinthedefaultpathandthenspecifiedonthecommandlinewithashortfilename.
Otherwise,youwouldhavetotypethefullfilenameforeachfile.
Byusingtheiorfoptions,yourcommandlineawkprogramscanusefacilitiesinawklibraryfiles(seeLibrary
Functions).Pathsearchingisnotdoneifgawkisincompatibilitymode.Thisistrueforbothtraditionalandposix.
SeeOptions.
Ifthesourcecodefileisnotfoundaftertheinitialsearch,thepathissearchedagainafteraddingthesuffix.awktothe
filename.
https://www.gnu.org/software/gawk/manual/gawk.html

45/479

29/7/2016

TheGNUAwkUsersGuide

gawkspathsearchmechanismissimilartotheshells.(SeeTheBourneAgainSHellmanual.)Ittreatsanullentryinthe

pathasindicatingthecurrentdirectory.(Anullentryisindicatedbystartingorendingthepathwithacolonorby
placingtwocolonsnexttoeachother[::].)
NOTE:Toincludethecurrentdirectoryinthepath,eitherplace.asanentryinthepathorwriteanull
entryinthepath.
Differentpastversionsofgawkwouldalsolookexplicitlyinthecurrentdirectory,eitherbeforeorafterthe
pathsearch.Asofversion4.1.2,thisnolongerhappensifyouwishtolookinthecurrentdirectory,you
mustinclude.eitherasaseparateentryorasanullentryinthesearchpath.
ThedefaultvalueforAWKPATHis.:/usr/local/share/awk.14Since.isincludedatthebeginning,gawksearchesfirstin
thecurrentdirectoryandthenin/usr/local/share/awk.Inpractice,thismeansthatyouwillrarelyneedtochangethe
valueofAWKPATH.
gawkplacesthevalueofthesearchpaththatitusedintoENVIRON["AWKPATH"].Thisprovidesaccesstotheactualsearch
pathvaluefromwithinanawkprogram.

AlthoughyoucanchangeENVIRON["AWKPATH"]withinyourawkprogram,thishasnoeffectontherunningprograms
behavior.Thismakessense:theAWKPATHenvironmentvariableisusedtofindtheprogramsourcefiles.Onceyour
programisrunning,allthefileshavebeenfound,andgawknolongerneedstouseAWKPATH.
Next:OtherEnvironmentVariables,Previous:AWKPATHVariable,Up:EnvironmentVariables[Contents][Index]
2.5.2TheAWKLIBPATHEnvironmentVariable
TheAWKLIBPATHenvironmentvariableissimilartotheAWKPATHvariable,butitisusedtosearchforloadableextensions
(storedassystemsharedlibraries)specifiedwiththeloptionratherthanforsourcefiles.Iftheextensionisnotfound,
thepathissearchedagainafteraddingtheappropriatesharedlibrarysuffixfortheplatform.Forexample,on
GNU/Linuxsystems,thesuffix.soisused.Thesearchpathspecifiedisalsousedforextensionsloadedviathe@load
keyword(seeLoadingSharedLibraries).
IfAWKLIBPATHdoesnotexistintheenvironment,orifithasanemptyvalue,gawkusesadefaultpaththisistypically
/usr/local/lib/gawk,althoughitcanvarydependinguponhowgawkwasbuilt.
gawkplacesthevalueofthesearchpaththatitusedintoENVIRON["AWKLIBPATH"].Thisprovidesaccesstotheactual
searchpathvaluefromwithinanawkprogram.

Previous:AWKLIBPATHVariable,Up:EnvironmentVariables[Contents][Index]
2.5.3OtherEnvironmentVariables
Anumberofotherenvironmentvariablesaffectgawksbehavior,buttheyaremorespecialized.Thoseinthefollowing
listaremeanttobeusedbyregularusers:
GAWK_MSEC_SLEEP

Specifiestheintervalbetweenconnectionretries,inmilliseconds.Onsystemsthatdonotsupporttheusleep()
systemcall,thevalueisroundeduptoanintegralnumberofseconds.
GAWK_READ_TIMEOUT

Specifiesthetime,inmilliseconds,forgawktowaitforinputbeforereturningwithanerror.SeeReadTimeout.
GAWK_SOCK_RETRIES

ControlsthenumberoftimesgawkattemptstoretryatwowayTCP/IP(socket)connectionbeforegivingup.See
https://www.gnu.org/software/gawk/manual/gawk.html

46/479

29/7/2016

TheGNUAwkUsersGuide

TCP/IPNetworking.
POSIXLY_CORRECT

CausesgawktoswitchtoPOSIXcompatibilitymode,disablingalltraditionalandGNUextensions.SeeOptions.
Theenvironmentvariablesinthefollowinglistaremeantforusebythegawkdevelopersfortestingandtuning.Theyare
subjecttochange.Thevariablesare:
AWKBUFSIZE

ThisvariableonlyaffectsgawkonPOSIXcompliantsystems.Withavalueofexact,gawkusesthesizeofeach
inputfileasthesizeofthememorybuffertoallocateforI/O.Otherwise,thevalueshouldbeanumber,andgawk
usesthatnumberasthesizeofthebuffertoallocate.(Whenthisvariableisnotset,gawkusesthesmallerofthe
filessizeandthedefaultblocksize,whichisusuallythefilesystemsI/Oblocksize.)
AWK_HASH

Ifthisvariableexistswithavalueofgst,gawkswitchestousingthehashfunctionfromGNUSmalltalkfor
managingarrays.Thisfunctionmaybemarginallyfasterthanthestandardfunction.
AWKREADFUNC

Ifthisvariableexists,gawkswitchestoreadingsourcefilesonelineatatime,insteadofreadinginblocks.This
existsfordebuggingproblemsonfilesystemsonnonPOSIXoperatingsystemswhereI/Oisperformedin
records,notinblocks.
GAWK_MSG_SRC

Ifthisvariableexists,gawkincludesthefilenameandlinenumberwithinthegawksourcecodefromwhich
warningand/orfatalmessagesaregenerated.Itspurposeistohelpisolatethesourceofamessage,asthereare
multipleplacesthatproducethesamewarningorerrormessage.
GAWK_NO_DFA

Ifthisvariableexists,gawkdoesnotusetheDFAregexpmatcherfordoesitmatchkindsoftests.Thiscancause
gawktobeslower.Itspurposeistohelpisolatedifferencesbetweenthetworegexpmatchersthatgawkuses
internally.(Therearentsupposedtobedifferences,butoccasionallytheoryandpracticedontcoordinatewith
eachother.)
GAWK_NO_PP_RUN

Whengawkisinvokedwiththeprettyprintoption,itwillnotruntheprogramifthisenvironmentvariable
exists.
CAUTION:Thisvariablewillnotsurviveintothenextmajorrelease.
GAWK_STACKSIZE

Thisspecifiestheamountbywhichgawkshouldgrowitsinternalevaluationstack,whenneeded.
INT_CHAIN_MAX

Thisspecifiesintendedmaximumnumberofitemsgawkwillmaintainonahashchainformanagingarrays
indexedbyintegers.
STR_CHAIN_MAX

Thisspecifiesintendedmaximumnumberofitemsgawkwillmaintainonahashchainformanagingarrays
indexedbystrings.
TIDYMEM
https://www.gnu.org/software/gawk/manual/gawk.html

47/479

29/7/2016

TheGNUAwkUsersGuide

Ifthisvariableexists,gawkusesthemtrace()librarycallsfromtheGNUClibrarytohelptrackdownpossible
memoryleaks.
Next:IncludeFiles,Previous:EnvironmentVariables,Up:InvokingGawk[Contents][Index]

2.6gawksExitStatus
Iftheexitstatementisusedwithavalue(seeExitStatement),thengawkexitswiththenumericvaluegiventoit.
Otherwise,iftherewerenoproblemsduringexecution,gawkexitswiththevalueoftheCconstantEXIT_SUCCESS.Thisis
usuallyzero.
Ifanerroroccurs,gawkexitswiththevalueoftheCconstantEXIT_FAILURE.Thisisusuallyone.
Ifgawkexitsbecauseofafatalerror,theexitstatusistwo.OnnonPOSIXsystems,thisvaluemaybemappedto
EXIT_FAILURE.
Next:LoadingSharedLibraries,Previous:ExitStatus,Up:InvokingGawk[Contents][Index]

2.7IncludingOtherFilesintoYourProgram
Thissectiondescribesafeaturethatisspecifictogawk.
The@includekeywordcanbeusedtoreadexternalawksourcefiles.Thisgivesyoutheabilitytosplitlargeawksource
filesintosmaller,moremanageablepieces,andalsoletsyoureusecommonawkcodefromvariousawkscripts.Inother
words,youcangrouptogetherawkfunctionsusedtocarryoutspecifictasksintoexternalfiles.Thesefilescanbeused
justlikefunctionlibraries,usingthe@includekeywordinconjunctionwiththeAWKPATHenvironmentvariable.Notethat
sourcefilesmayalsobeincludedusingtheioption.
Letsseeanexample.Wellstartwithtwo(trivial)awkscripts,namelytest1andtest2.Hereisthetest1script:
BEGIN{
print"Thisisscripttest1."
}

andhereistest2:
@include"test1"
BEGIN{
print"Thisisscripttest2."
}

Runninggawkwithtest2producesthefollowingresult:
$gawkftest2
|Thisisscripttest1.
|Thisisscripttest2.
gawkrunsthetest2script,[email protected],toincludeexternalawksourcefiles,
youjustuse@includefollowedbythenameofthefiletobeincluded,enclosedindoublequotes.

NOTE:Keepinmindthatthisisalanguageconstructandthefilenamecannotbeastringvariable,but
ratherjustaliteralstringconstantindoublequotes.
Thefilestobeincludedmaybenestede.g.,givenathirdscript,namelytest3:
@include"test2"
BEGIN{
print"Thisisscripttest3."
https://www.gnu.org/software/gawk/manual/gawk.html

48/479

29/7/2016

TheGNUAwkUsersGuide

Runninggawkwiththetest3scriptproducesthefollowingresults:
$gawkftest3
|Thisisscripttest1.
|Thisisscripttest2.
|Thisisscripttest3.

Thefilenamecan,ofcourse,beapathname.Forexample:
@include"../io_funcs"

and:
@include"/usr/awklib/network"

arebothvalid.TheAWKPATHenvironmentvariablecanbeofgreatvaluewhenusing@include.Thesamerulesfortheuse
oftheAWKPATHvariableincommandlinefilesearches(seeAWKPATHVariable)applyto@includealso.
Thisisveryhelpfulinconstructinggawkfunctionlibraries.Ifyouhavealargescriptwithuseful,generalpurposeawk
functions,youcanbreakitdownintolibraryfilesandputthosefilesinaspecialdirectory.Youcanthenincludethose
libraries,eitherbyusingthefullpathnamesofthefiles,orbysettingtheAWKPATHenvironmentvariableaccordingly
andthenusing@includewithjustthefilepartofthefullpathname.Ofcourse,youcankeeplibraryfilesinmorethan
onedirectorythemorecomplextheworkingenvironmentis,themoredirectoriesyoumayneedtoorganizethefilesto
beincluded.
Giventheabilitytospecifymultiplefoptions,[email protected],the
@includekeywordcanhelpyouinconstructingselfcontainedgawkprograms,thusreducingtheneedforwriting
complexandtediouscommandlines.Inparticular,@includeisveryusefulforwritingCGIscriptstoberunfromweb
pages.
AsmentionedinAWKPATHVariable,thecurrentdirectoryisalwayssearchedfirstforsourcefiles,beforesearchingin
AWKPATHthisalsoappliestofilesnamedwith@include.
Next:Obsolete,Previous:IncludeFiles,Up:InvokingGawk[Contents][Index]

2.8LoadingDynamicExtensionsintoYourProgram
Thissectiondescribesafeaturethatisspecifictogawk.
The@loadkeywordcanbeusedtoreadexternalawkextensions(storedassystemsharedlibraries).Thisallowsyouto
linkincompiledcodethatmayoffersuperiorperformanceand/orgiveyouaccesstoextendedcapabilitiesnotsupported
bytheawklanguage.TheAWKLIBPATHvariableisusedtosearchfortheextension.Using@loadiscompletelyequivalent
tousingthelcommandlineoption.
IftheextensionisnotinitiallyfoundinAWKLIBPATH,anothersearchisconductedafterappendingtheplatformsdefault
sharedlibrarysuffixtothefilename.Forexample,onGNU/Linuxsystems,thesuffix.soisused:
$gawk'@load"ordchr";BEGIN{printchr(65)}'
|A

Thisisequivalenttothefollowingexample:
$gawklordchr'BEGIN{printchr(65)}'
|A

Forcommandlineusage,theloptionismoreconvenient,but@loadisusefulforembeddinginsideanawksourcefile
thatrequiresaccesstoanextension.
https://www.gnu.org/software/gawk/manual/gawk.html

49/479

29/7/2016

TheGNUAwkUsersGuide

DynamicExtensions,describeshowtowriteextensions(inCorC++)thatcanbeloadedwitheither@loadorthel
option.Italsodescribestheordchrextension.
Next:Undocumented,Previous:LoadingSharedLibraries,Up:InvokingGawk[Contents][Index]

2.9ObsoleteOptionsand/orFeatures
Thissectiondescribesfeaturesand/orcommandlineoptionsfrompreviousreleasesofgawkthateitherarenotavailable
inthecurrentversionorarestillsupportedbutdeprecated(meaningthattheywillnotbeinthenextrelease).
Theprocessrelatedspecialfiles/dev/pid,/dev/ppid,/dev/pgrpid,and/dev/userweredeprecatedingawk3.1,butstill
worked.Asofversion4.0,theyarenolongerinterpretedspeciallybygawk.(UsePROCINFOinsteadseeAutoset.)
Next:InvokingSummary,Previous:Obsolete,Up:InvokingGawk[Contents][Index]

2.10UndocumentedOptionsandFeatures
UsetheSource,Luke!
ObiWan
Thissectionintentionallyleftblank.
Previous:Undocumented,Up:InvokingGawk[Contents][Index]

2.11Summary
Useeitherawk'program'filesor awkfprogramfilefilestorun awk.
Thethreestandardoptionsforallversionsofawkaref,F,andv.gawksuppliestheseandmanyothers,aswell
ascorrespondingGNUstylelongoptions.
Nonoptioncommandlineargumentsareusuallytreatedasfilenames,unlesstheyhavetheformvar=value,in
whichcasetheyaretakenasvariableassignmentstobeperformedatthatpointinprocessingtheinput.
Allnonoptioncommandlinearguments,excludingtheprogramtext,areplacedintheARGVarray.AdjustingARGC
andARGVaffectshowawkprocessesinput.
Youcanuseasingleminussign()torefertostandardinputonthecommandline.gawkalsoletsyouusethe
specialfilename/dev/stdin.
gawkpaysattentiontoanumberofenvironmentvariables.AWKPATH,AWKLIBPATH,andPOSIXLY_CORRECTarethemost
importantones.
gawksexitstatusconveysinformationtotheprogramthatinvokedit.Usetheexitstatementfromwithinanawk
programtosettheexitstatus.
gawkallowsyoutoincludeotherawksourcefilesintoyourprogramusingthe@includestatementand/orthei
andfcommandlineoptions.
gawkallowsyoutoloadadditionalfunctionswritteninCorC++usingthe@loadstatementand/ortheloption.
(Thisadvancedfeatureisdescribedlater,inDynamicExtensions.)
Next:ReadingFiles,Previous:InvokingGawk,Up:Top[Contents][Index]

3RegularExpressions
Aregularexpression,orregexp,isawayofdescribingasetofstrings.Becauseregularexpressionsaresucha
fundamentalpartofawkprogramming,theirformatandusedeserveaseparatechapter.
Aregularexpressionenclosedinslashes(/)isanawkpatternthatmatcheseveryinputrecordwhosetextbelongsto
https://www.gnu.org/software/gawk/manual/gawk.html

50/479

29/7/2016

TheGNUAwkUsersGuide

thatset.Thesimplestregularexpressionisasequenceofletters,numbers,orboth.Sucharegexpmatchesanystring
thatcontainsthatsequence.Thus,theregexpfoomatchesanystringcontainingfoo.Thus,thepattern/foo/matches
anyinputrecordcontainingthethreeadjacentcharactersfooanywhereintherecord.Otherkindsofregexpsletyou
specifymorecomplicatedclassesofstrings.
Initially,theexamplesinthischapteraresimple.Asweexplainmoreabouthowregularexpressionswork,wepresent
morecomplicatedinstances.
RegexpUsage:
HowtoUseRegularExpressions.
EscapeSequences:
Howtowritenonprintingcharacters.
RegexpOperators:
RegularExpressionOperators.
BracketExpressions:
Whatcangobetween[...].
LeftmostLongest:
Howmuchtextmatches.
ComputedRegexps:
UsingDynamicRegexps.
GNURegexpOperators: OperatorsspecifictoGNUsoftware.
Casesensitivity:
Howtodocaseinsensitivematching.
RegexpSummary:
Regularexpressionssummary.
Next:EscapeSequences,Up:Regexp[Contents][Index]

3.1HowtoUseRegularExpressions
Aregularexpressioncanbeusedasapatternbyenclosingitinslashes.Thentheregularexpressionistestedagainstthe
entiretextofeachrecord.(Normally,itonlyneedstomatchsomepartofthetextinordertosucceed.)Forexample,the
followingprintsthesecondfieldofeachrecordwherethestringliappearsanywhereintherecord:
$awk'/li/{print$2}'maillist
|5555553
|5550542
|5556699
|5553430

Regularexpressionscanalsobeusedinmatchingexpressions.Theseexpressionsallowyoutospecifythestringto
matchagainstitneednotbetheentirecurrentinputrecord.Thetwooperators~and !~performregularexpression
comparisons.Expressionsusingtheseoperatorscanbeusedaspatterns,orinif,while,for,anddostatements.(See
Statements.)Forexample,thefollowingistrueiftheexpressionexp(takenasastring)matchesregexp:
exp~/regexp/

Thisexamplematches,orselects,allinputrecordswiththeuppercaseletterJsomewhereinthefirstfield:
$awk'$1~/J/'inventoryshipped
|Jan132515115
|Jun314275492
|Jul243467436
|Jan213664620

Sodoesthis:
awk'{if($1~/J/)print}'inventoryshipped

Thisnextexampleistrueiftheexpressionexp(takenasacharacterstring)doesnotmatchregexp:
exp!~/regexp/

Thefollowingexamplematches,orselects,allinputrecordswhosefirstfielddoesnotcontaintheuppercaseletterJ:
$awk'$1!~/J/'inventoryshipped
|Feb153224226
|Mar152434228
https://www.gnu.org/software/gawk/manual/gawk.html

51/479

29/7/2016

TheGNUAwkUsersGuide

|Apr315263420
|May163429208

Whenaregexpisenclosedinslashes,suchas/foo/,wecallitaregexpconstant,muchlike5.27isanumericconstant
and"foo"isastringconstant.
Next:RegexpOperators,Previous:RegexpUsage,Up:Regexp[Contents][Index]

3.2EscapeSequences
Somecharacterscannotbeincludedliterallyinstringconstants("foo")orregexpconstants(/foo/).Instead,theyshould
berepresentedwithescapesequences,whicharecharactersequencesbeginningwithabackslash(\).Oneuseofan
escapesequenceistoincludeadoublequotecharacterinastringconstant.Becauseaplaindoublequoteendsthestring,
youmustuse\"torepresentanactualdoublequotecharacterasapartofthestring.Forexample:
$awk'BEGIN{print"Hesaid\"hi!\"toher."}'
|Hesaid"hi!"toher.

Thebackslashcharacteritselfisanothercharacterthatcannotbeincludednormallyyoumustwrite\\toputone
backslashinthestringorregexp.Thus,thestringwhosecontentsarethetwocharacters"and\mustbewritten
"\"\\".
OtherescapesequencesrepresentunprintablecharacterssuchasTABornewline.Thereisnothingtostopyoufrom
enteringmostunprintablecharactersdirectlyinastringconstantorregexpconstant,buttheymaylookugly.
Thefollowinglistpresentsalltheescapesequencesusedinawkandwhattheyrepresent.Unlessnotedotherwise,all
theseescapesequencesapplytobothstringconstantsandregexpconstants:
\\

Aliteralbackslash,\.
\a

Thealertcharacter,Ctrlg,ASCIIcode7(BEL).(Thisoftenmakessomesortofaudiblenoise.)
\b

Backspace,Ctrlh,ASCIIcode8(BS).
\f

Formfeed,Ctrll,ASCIIcode12(FF).
\n

Newline,Ctrlj,ASCIIcode10(LF).
\r

Carriagereturn,Ctrlm,ASCIIcode13(CR).
\t

HorizontalTAB,Ctrli,ASCIIcode9(HT).
\v

VerticalTAB,Ctrlk,ASCIIcode11(VT).
https://www.gnu.org/software/gawk/manual/gawk.html

52/479

29/7/2016

TheGNUAwkUsersGuide

\nnn

Theoctalvaluennn,wherennnstandsfor1to3digitsbetween0and7.Forexample,thecodefortheASCII
ESC(escape)characteris\033.
\xhh

Thehexadecimalvaluehh,wherehhstandsforasequenceofhexadecimaldigits(09,andeitherAFor
af).LikethesameconstructinISOC,theescapesequencecontinuesuntilthefirstnonhexadecimaldigitis
seen.(c.e.)However,usingmorethantwohexadecimaldigitsproducesundefinedresults.(The\xescape
sequenceisnotallowedinPOSIXawk.)
CAUTION:Thenextmajorreleaseofgawkwillchange,suchthatamaximumoftwohexadecimal
digitsfollowingthe\xwillbeused.
\/

Aliteralslash(necessaryforregexpconstantsonly).Thissequenceisusedwhenyouwanttowritearegexp
constantthatcontainsaslash(suchas/.*:\/home\/[[:alnum:]]+:.*/the[[:alnum:]]notationisdiscussedin
BracketExpressions).Becausetheregexpisdelimitedbyslashes,youneedtoescapeanyslashthatispartofthe
pattern,inordertotellawktokeepprocessingtherestoftheregexp.
\"

Aliteraldoublequote(necessaryforstringconstantsonly).Thissequenceisusedwhenyouwanttowriteastring
constantthatcontainsadoublequote(suchas"Hesaid\"hi!\"toher.").Becausethestringisdelimitedby
doublequotes,youneedtoescapeanyquotethatispartofthestring,inordertotellawktokeepprocessingthe
restofthestring.
Ingawk,anumberofadditionaltwocharactersequencesthatbeginwithabackslashhavespecialmeaninginregexps.
SeeGNURegexpOperators.
Inaregexp,abackslashbeforeanycharacterthatisnotinthepreviouslistandnotlistedinGNURegexpOperators,
meansthatthenextcharactershouldbetakenliterally,evenifitwouldnormallybearegexpoperator.Forexample,
/a\+b/matchesthethreecharactersa+b.
Forcompleteportability,donotuseabackslashbeforeanycharacternotshowninthepreviouslistorthatisnotan
operator.
BackslashBeforeRegularCharacters
Ifyouplaceabackslashinastringconstantbeforesomethingthatisnotoneofthecharacterspreviouslylisted,
POSIXawkpurposelyleaveswhathappensasundefined.Therearetwochoices:
Stripthebackslashout
ThisiswhatBWKawkandgawkbothdo.Forexample,"a\qc"isthesameas"aqc".(Becausethisissuchaneasy
bugbothtointroduceandtomiss,gawkwarnsyouaboutit.)ConsiderFS="[\t]+\|[\t]+"tousevertical
barssurroundedbywhitespaceasthefieldseparator.Thereshouldbetwobackslashesinthestring:FS="
[\t]+\\|[\t]+".)
Leavethebackslashalone
Someotherawkimplementationsdothis.Insuchimplementations,typing"a\qc"isthesameastyping"a\\qc".
Tosummarize:
Theescapesequencesintheprecedinglistarealwaysprocessedfirst,forbothstringconstantsandregexp
constants.Thishappensveryearly,assoonasawkreadsyourprogram.
https://www.gnu.org/software/gawk/manual/gawk.html

53/479

29/7/2016

TheGNUAwkUsersGuide

gawkprocessesbothregexpconstantsanddynamicregexps(seeComputedRegexps),forthespecialoperators

listedinGNURegexpOperators.
Abackslashbeforeanyothercharactermeanstotreatthatcharacterliterally.
EscapeSequencesforMetacharacters
Supposeyouuseanoctalorhexadecimalescapetorepresentaregexpmetacharacter.(SeeRegexpOperators.)Does
awktreatthecharacterasaliteralcharacterorasaregexpoperator?
Historically,suchcharactersweretakenliterally.(d.c.)However,thePOSIXstandardindicatesthattheyshouldbe
treatedasrealmetacharacters,whichiswhatgawkdoes.Incompatibilitymode(seeOptions),gawktreatsthecharacters
representedbyoctalandhexadecimalescapesequencesliterallywhenusedinregexpconstants.Thus,/a\52b/is
equivalentto/a\*b/.

Next:BracketExpressions,Previous:EscapeSequences,Up:Regexp[Contents][Index]

3.3RegularExpressionOperators
Youcancombineregularexpressionswithspecialcharacters,calledregularexpressionoperatorsormetacharacters,to
increasethepowerandversatilityofregularexpressions.
TheescapesequencesdescribedearlierinEscapeSequences,arevalidinsidearegexp.Theyareintroducedbya\and
arerecognizedandconvertedintocorrespondingrealcharactersastheveryfirststepinprocessingregexps.
Hereisalistofmetacharacters.Allcharactersthatarenotescapesequencesandthatarenotlistedherestandfor
themselves:
\

Thissuppressesthespecialmeaningofacharacterwhenmatching.Forexample,\$matchesthecharacter $.
^

Thismatchesthebeginningofastring.^@chaptermatches @chapteratthebeginningofastring,forexample,
andcanbeusedtoidentifychapterbeginningsinTexinfosourcefiles.The^isknownasananchor,becauseit
anchorsthepatterntomatchonlyatthebeginningofthestring.
Itisimportanttorealizethat^doesnotmatchthebeginningofaline(thepointrightaftera\nnewline
character)embeddedinastring.Theconditionisnottrueinthefollowingexample:
if("line1\nLINE2"~/^L/)
$

Thisissimilarto^,butitmatchesonlyattheendofastring.Forexample,p$matchesarecordthatendswith
ap.The$isananchoranddoesnotmatchtheendofaline(thepointrightbeforea\nnewlinecharacter)
embeddedinastring.Theconditioninthefollowingexampleisnottrue:
if("line1\nLINE2"~/1$/)
.(period)

Thismatchesanysinglecharacter,includingthenewlinecharacter.Forexample,.Pmatchesanysingle
characterfollowedbyaPinastring.Usingconcatenation,wecanmakearegularexpressionsuchasU.A,
whichmatchesanythreecharactersequencethatbeginswithUandendswithA.
InstrictPOSIXmode(seeOptions),.doesnotmatchtheNULcharacter,whichisacharacterwithallbitsequal
tozero.Otherwise,NULisjustanothercharacter.OtherversionsofawkmaynotbeabletomatchtheNUL
https://www.gnu.org/software/gawk/manual/gawk.html

54/479

29/7/2016

TheGNUAwkUsersGuide

character.
[]

Thisiscalledabracketexpression.15Itmatchesanyoneofthecharactersthatareenclosedinthesquarebrackets.
Forexample,[MVX]matchesanyoneofthecharactersM,V,orXinastring.Afulldiscussionofwhatcanbe
insidethesquarebracketsofabracketexpressionisgiveninBracketExpressions.
[^]

Thisisacomplementedbracketexpression.Thefirstcharacterafterthe[mustbea^.Itmatchesany
charactersexceptthoseinthesquarebrackets.Forexample,[^awk]matchesanycharacterthatisnotan a,w,
ork.
|

Thisisthealternationoperatoranditisusedtospecifyalternatives.The|hasthelowestprecedenceofallthe
regularexpressionoperators.Forexample,^P|[aeiouy]matchesanystringthatmatcheseither^Por
[aeiouy].ThismeansitmatchesanystringthatstartswithPorcontains(anywherewithinit)alowercase
Englishvowel.
Thealternationappliestothelargestpossibleregexpsoneitherside.
()

Parenthesesareusedforgroupinginregularexpressions,asinarithmetic.Theycanbeusedtoconcatenate
regularexpressionscontainingthealternationoperator,|.Forexample,@(samp|code)\{[^}]+\}matchesboth
@code{foo}and@samp{bar}.(TheseareTexinfoformattingcontrolsequences.The+isexplainedfurtheron
inthislist.)
*

Thissymbolmeansthattheprecedingregularexpressionshouldberepeatedasmanytimesasnecessarytofinda
match.Forexample,ph*appliesthe *symboltotheprecedinghandlooksformatchesofonepfollowed
byanynumberofhs.Thisalsomatchesjustpifnohsarepresent.
Therearetwosubtlepointstounderstandabouthow*works.First,the*appliesonlytothesinglepreceding
regularexpressioncomponent(e.g.,inph*,itappliesjusttotheh).Tocause*toapplytoalarger
subexpression,useparentheses:(ph)*matchesph,phph,phphph,andsoon.
Second,*findsasmanyrepetitionsaspossible.Ifthetexttobematchedisphhhhhhhhhhhhhhooey,ph*
matchesallofthehs.
+

Thissymbolissimilarto*,exceptthattheprecedingexpressionmustbematchedatleastonce.Thismeansthat
wh+ywouldmatch whyandwhhy,butnotwy,whereaswh*ywouldmatchallthree.
?

Thissymbolissimilarto*,exceptthattheprecedingexpressioncanbematchedeitheronceornotatall.For
example,fe?dmatches fedandfd,butnothingelse.
{n}
{n,}
{n,m}

Oneortwonumbersinsidebracesdenoteanintervalexpression.Ifthereisonenumberinthebraces,the
precedingregexpisrepeatedntimes.Iftherearetwonumbersseparatedbyacomma,theprecedingregexpis
repeatedntomtimes.Ifthereisonenumberfollowedbyacomma,thentheprecedingregexpisrepeatedatleast
https://www.gnu.org/software/gawk/manual/gawk.html

55/479

29/7/2016

TheGNUAwkUsersGuide

ntimes:
wh{3}y

Matcheswhhhy,butnotwhyorwhhhhy.
wh{3,5}y

Matcheswhhhy,whhhhy,orwhhhhhyonly.
wh{2,}y

Matcheswhhy,whhhy,andsoon.
Intervalexpressionswerenottraditionallyavailableinawk.TheywereaddedaspartofthePOSIXstandardto
makeawkandegrepconsistentwitheachother.
Initially,becauseoldprogramsmayuse{and}inregexpconstants,gawkdidnotmatchintervalexpressionsin
regexps.
However,beginningwithversion4.0,gawkdoesmatchintervalexpressionsbydefault.Thisisbecause
compatibilitywithPOSIXhasbecomemoreimportanttomostgawkusersthancompatibilitywitholdprograms.
Forprogramsthatuse{and}inregexpconstants,itisgoodpracticetoalwaysescapethemwithabackslash.
Thentheregexpconstantsarevalidandworkthewayyouwantthemto,usinganyversionofawk.16
Finally,when{and}appearinregexpconstantsinawaythatcannotbeinterpretedasanintervalexpression
(suchas/q{a}/),thentheystandforthemselves.
Inregularexpressions,the*,+,and?operators,aswellasthebraces{and},havethehighestprecedence,
followedbyconcatenation,andfinallyby|.Asinarithmetic,parenthesescanchangehowoperatorsaregrouped.
InPOSIXawkandgawk,the*,+,and?operatorsstandforthemselveswhenthereisnothingintheregexpthat
precedesthem.Forexample,/+/matchesaliteralplussign.However,manyotherversionsofawktreatsuchausageas
asyntaxerror.
Ifgawkisincompatibilitymode(seeOptions),intervalexpressionsarenotavailableinregularexpressions.
Next:LeftmostLongest,Previous:RegexpOperators,Up:Regexp[Contents][Index]

3.4UsingBracketExpressions
Asmentionedearlier,abracketexpressionmatchesanycharacteramongthoselistedbetweentheopeningandclosing
squarebrackets.
Withinabracketexpression,arangeexpressionconsistsoftwocharactersseparatedbyahyphen.Itmatchesanysingle
characterthatsortsbetweenthetwocharacters,baseduponthesystemsnativecharacterset.Forexample,[09]is
equivalentto[0123456789].(SeeRangesandLocales,foranexplanationofhowthePOSIXstandardandgawkhave
changedovertime.Thisismainlyofhistoricalinterest.)
Toincludeoneofthecharacters\,],,or^inabracketexpression,puta\infrontofit.Forexample:
[d\]]

matcheseitherdor].Additionally,ifyouplace]rightaftertheopening[,theclosingbracketistreatedasoneof
thecharacterstobematched.
Thetreatmentof\inbracketexpressionsiscompatiblewithotherawkimplementationsandisalsomandatedby
POSIX.TheregularexpressionsinawkareasupersetofthePOSIXspecificationforExtendedRegularExpressions
https://www.gnu.org/software/gawk/manual/gawk.html

56/479

29/7/2016

TheGNUAwkUsersGuide

(EREs).POSIXEREsarebasedontheregularexpressionsacceptedbythetraditionalegreputility.
CharacterclassesareafeatureintroducedinthePOSIXstandard.Acharacterclassisaspecialnotationfordescribing
listsofcharactersthathaveaspecificattribute,buttheactualcharacterscanvaryfromcountrytocountryand/orfrom
charactersettocharacterset.Forexample,thenotionofwhatisanalphabeticcharacterdiffersbetweentheUnited
StatesandFrance.
Acharacterclassisonlyvalidinaregexpinsidethebracketsofabracketexpression.Characterclassesconsistof[:,a
keyworddenotingtheclass,and:].Table3.1liststhecharacterclassesdefinedbythePOSIXstandard.
Class
[:alnum:]
[:alpha:]
[:blank:]
[:cntrl:]
[:digit:]
[:graph:]
[:lower:]
[:print:]
[:punct:]
[:space:]
[:upper:]
[:xdigit:]

Meaning
Alphanumericcharacters
Alphabeticcharacters
SpaceandTABcharacters
Controlcharacters
Numericcharacters
Charactersthatarebothprintableandvisible(aspaceisprintablebutnotvisible,whereasanais
both)
Lowercasealphabeticcharacters
Printablecharacters(charactersthatarenotcontrolcharacters)
Punctuationcharacters(charactersthatarenotletters,digits,controlcharacters,orspacecharacters)
Spacecharacters(suchasspace,TAB,andformfeed,tonameafew)
Uppercasealphabeticcharacters
Charactersthatarehexadecimaldigits

Table3.1:POSIXcharacterclasses
Forexample,beforethePOSIXstandard,youhadtowrite/[AZaz09]/tomatchalphanumericcharacters.Ifyour
charactersethadotheralphabeticcharactersinit,thiswouldnotmatchthem.WiththePOSIXcharacterclasses,you
canwrite/[[:alnum:]]/tomatchthealphabeticandnumericcharactersinyourcharacterset.
Someutilitiesthatmatchregularexpressionsprovideanonstandard[:ascii:]characterclassawkdoesnot.However,
youcansimulatesuchaconstructusing[\x00\x7F].Thismatchesallvaluesnumericallybetweenzeroand127,
whichisthedefinedrangeoftheASCIIcharacterset.Useacomplementedcharacterlist([^\x00\x7F])tomatchany
singlebytecharactersthatarenotintheASCIIrange.
Twoadditionalspecialsequencescanappearinbracketexpressions.TheseapplytononASCIIcharactersets,which
canhavesinglesymbols(calledcollatingelements)thatarerepresentedwithmorethanonecharacter.Theycanalso
haveseveralcharactersthatareequivalentforcollating,orsorting,purposes.(Forexample,inFrench,aplaineanda
graveaccentedareequivalent.)Thesesequencesare:
Collatingsymbols
Multicharactercollatingelementsenclosedbetween[.and.].Forexample,ifchisacollatingelement,then
[[.ch.]]isaregexpthatmatchesthiscollatingelement,whereas[ch]isaregexpthatmatcheseithercor
h.
Equivalenceclasses
Localespecificnamesforalistofcharactersthatareequal.Thenameisenclosedbetween[=and=].For
example,thenameemightbeusedtorepresentallofe,,,and.Inthiscase,[[=e=]]isaregexp
thatmatchesanyofe,,,or.
ThesefeaturesareveryvaluableinnonEnglishspeakinglocales.
CAUTION:Thelibraryfunctionsthatgawkusesforregularexpressionmatchingcurrentlyrecognizeonly
https://www.gnu.org/software/gawk/manual/gawk.html

57/479

29/7/2016

TheGNUAwkUsersGuide

POSIXcharacterclassestheydonotrecognizecollatingsymbolsorequivalenceclasses.
Next:ComputedRegexps,Previous:BracketExpressions,Up:Regexp[Contents][Index]

3.5HowMuchTextMatches?
Considerthefollowing:
echoaaaabcd|awk'{sub(/a+/,"<A>");print}'

Thisexampleusesthesub()functiontomakeachangetotheinputrecord.(sub()replacesthefirstinstanceofanytext
matchedbythefirstargumentwiththestringprovidedasthesecondargumentseeStringFunctions.)Here,theregexp
/a+/indicatesoneormoreacharacters,andthereplacementtextis<A>.
Theinputcontainsfouracharacters.awk(andPOSIX)regularexpressionsalwaysmatchtheleftmost,longest
sequenceofinputcharactersthatcanmatch.Thus,allfouracharactersarereplacedwith<A>inthisexample:
$echoaaaabcd|awk'{sub(/a+/,"<A>");print}'
|<A>bcd

Forsimplematch/nomatchtests,thisisnotsoimportant.Butwhendoingtextmatchingandsubstitutionswiththe
match(),sub(),gsub(),andgensub()functions,itisveryimportant.Understandingthisprincipleisalsoimportantfor
regexpbasedrecordandfieldsplitting(seeRecords,andalsoseeFieldSeparators).
Next:GNURegexpOperators,Previous:LeftmostLongest,Up:Regexp[Contents][Index]

3.6UsingDynamicRegexps
Therighthandsideofa~or!~operatorneednotbearegexpconstant(i.e.,astringofcharactersbetweenslashes).It
maybeanyexpression.Theexpressionisevaluatedandconvertedtoastringifnecessarythecontentsofthestringare
thenusedastheregexp.Aregexpcomputedinthiswayiscalledadynamicregexporacomputedregexp:
BEGIN{digits_regexp="[[:digit:]]+"}
$0~digits_regexp{print}

Thissetsdigits_regexptoaregexpthatdescribesoneormoredigits,andtestswhethertheinputrecordmatchesthis
regexp.
NOTE:Whenusingthe~and!~operators,beawarethatthereisadifferencebetweenaregexp
constantenclosedinslashesandastringconstantenclosedindoublequotes.Ifyouaregoingtouseastring
constant,youhavetounderstandthatthestringis,inessence,scannedtwice:thefirsttimewhenawkreads
yourprogram,andthesecondtimewhenitgoestomatchthestringonthelefthandsideoftheoperatorwith
thepatternontheright.Thisistrueofanystringvaluedexpression(suchasdigits_regexp,showninthe
previousexample),notjuststringconstants.
Whatdifferencedoesitmakeifthestringisscannedtwice?Theanswerhastodowithescapesequences,and
particularlywithbackslashes.Togetabackslashintoaregularexpressioninsideastring,youhavetotypetwo
backslashes.
Forexample,/\*/isaregexpconstantforaliteral*.Onlyonebackslashisneeded.Todothesamethingwithastring,
youhavetotype"\\*".Thefirstbackslashescapesthesecondonesothatthestringactuallycontainsthetwocharacters
\and*.
Giventhatyoucanusebothregexpandstringconstantstodescriberegularexpressions,whichshouldyouuse?The
answerisregexpconstants,forseveralreasons:
Stringconstantsaremorecomplicatedtowriteandmoredifficulttoread.Usingregexpconstantsmakesyour
https://www.gnu.org/software/gawk/manual/gawk.html

58/479

29/7/2016

TheGNUAwkUsersGuide

programslesserrorprone.Notunderstandingthedifferencebetweenthetwokindsofconstantsisacommon
sourceoferrors.
Itismoreefficienttouseregexpconstants.awkcannotethatyouhavesuppliedaregexpandstoreitinternallyin
aformthatmakespatternmatchingmoreefficient.Whenusingastringconstant,awkmustfirstconvertthestring
intothisinternalformandthenperformthepatternmatching.
Usingregexpconstantsisbetterformitshowsclearlythatyouintendaregexpmatch.
Using\ninBracketExpressionsofDynamicRegexps
Someolderversionsofawkdonotallowthenewlinecharactertobeusedinsideabracketexpressionforadynamic
regexp:
$awk'$0~"[\t\n]"'
errorawk:newlineincharacterclass[
error]...
errorsourcelinenumber1
errorcontextis
error$0~"[>>>\t\n]"<<<

Butanewlineinaregexpconstantworkswithnoproblem:
$awk'$0~/[\t\n]/'
hereisasampleline
|hereisasampleline
Ctrld
gawkdoesnothavethisproblem,anditisntlikelytooccurofteninpractice,butitsworthnotingforfuturereference.

Next:Casesensitivity,Previous:ComputedRegexps,Up:Regexp[Contents][Index]

3.7gawkSpecificRegexpOperators
GNUsoftwarethatdealswithregularexpressionsprovidesanumberofadditionalregexpoperators.Theseoperatorsare
describedinthissectionandarespecifictogawktheyarenotavailableinotherawkimplementations.Mostofthe
additionaloperatorsdealwithwordmatching.Forourpurposes,awordisasequenceofoneormoreletters,digits,or
underscores(_):
\s

Matchesanywhitespacecharacter.Thinkofitasshorthandfor[[:space:]].
\S

Matchesanycharacterthatisnotwhitespace.Thinkofitasshorthandfor[^[:space:]].
\w

Matchesanywordconstituentcharacterthatis,itmatchesanyletter,digit,orunderscore.Thinkofitas
shorthandfor[[:alnum:]_].
\W

Matchesanycharacterthatisnotwordconstituent.Thinkofitasshorthandfor[^[:alnum:]_].
\<

Matchestheemptystringatthebeginningofaword.Forexample,/\<away/matchesawaybutnotstowaway.
\>
https://www.gnu.org/software/gawk/manual/gawk.html

59/479

29/7/2016

TheGNUAwkUsersGuide

Matchestheemptystringattheendofaword.Forexample,/stow\>/matchesstowbutnotstowaway.
\y

Matchestheemptystringateitherthebeginningortheendofaword(i.e.,thewordboundary).Forexample,
\yballs?\ymatcheseither ballorballs,asaseparateword.
\B

Matchestheemptystringthatoccursbetweentwowordconstituentcharacters.Forexample,/\Brat\B/matches
crate,butitdoesnotmatchdirtyrat.\Bisessentiallytheoppositeof\y.
Therearetwootheroperatorsthatworkonbuffers.InEmacs,abufferis,naturally,anEmacsbuffer.OtherGNU
programs,includinggawk,considertheentirestringtomatchasthebuffer.Theoperatorsare:
\`

Matchestheemptystringatthebeginningofabuffer(string)
\'

Matchestheemptystringattheendofabuffer(string)
Because^and$alwaysworkintermsofthebeginningandendofstrings,theseoperatorsdontaddanynew
capabilitiesforawk.TheyareprovidedforcompatibilitywithotherGNUsoftware.
InotherGNUsoftware,thewordboundaryoperatoris\b.However,thatconflictswiththeawklanguagesdefinition
of\basbackspace,so gawkusesadifferentletter.Analternativemethodwouldhavebeentorequiretwobackslashes
intheGNUoperators,butthiswasdeemedtooconfusing.Thecurrentmethodofusing\yfortheGNU\bappearsto
bethelesseroftwoevils.
Thevariouscommandlineoptions(seeOptions)controlhowgawkinterpretscharactersinregexps:
Nooptions
Inthedefaultcase,gawkprovidesallthefacilitiesofPOSIXregexpsandthepreviouslydescribedGNUregexp
operators.GNUregexpoperatorsdescribedinRegexpOperators.
posix

MatchonlyPOSIXregexpstheGNUoperatorsarenotspecial(e.g.,\wmatchesaliteralw).Interval
expressionsareallowed.
traditional

MatchtraditionalUnixawkregexps.TheGNUoperatorsarenotspecial,andintervalexpressionsarenot
available.BecauseBWKawksupportsthem,thePOSIXcharacterclasses([[:alnum:]],etc.)areavailable.
Charactersdescribedbyoctalandhexadecimalescapesequencesaretreatedliterally,eveniftheyrepresent
regexpmetacharacters.
reinterval

Allowintervalexpressionsinregexps,iftraditionalhasbeenprovided.Otherwise,intervalexpressionsare
availablebydefault.
Next:RegexpSummary,Previous:GNURegexpOperators,Up:Regexp[Contents][Index]

3.8CaseSensitivityinMatching
Caseisnormallysignificantinregularexpressions,bothwhenmatchingordinarycharacters(i.e.,notmetacharacters)
https://www.gnu.org/software/gawk/manual/gawk.html

60/479

29/7/2016

TheGNUAwkUsersGuide

andinsidebracketexpressions.Thus,awinaregularexpressionmatchesonlyalowercasewandnotanuppercase
W.
Thesimplestwaytodoacaseindependentmatchistouseabracketexpressionforexample,[Ww].However,this
canbecumbersomeifyouneedtouseitoften,anditcanmaketheregularexpressionshardertoread.Therearetwo
alternativesthatyoumightprefer.
Onewaytoperformacaseinsensitivematchataparticularpointintheprogramistoconvertthedatatoasinglecase,
usingthetolower()ortoupper()builtinstringfunctions(whichwehaventdiscussedyetseeStringFunctions).For
example:
tolower($1)~/foo/{}

convertsthefirstfieldtolowercasebeforematchingagainstit.ThisworksinanyPOSIXcompliantawk.
Anothermethod,specifictogawk,istosetthevariableIGNORECASEtoanonzerovalue(seeBuiltinVariables).When
IGNORECASEisnotzero,allregexpandstringoperationsignorecase.
ChangingthevalueofIGNORECASEdynamicallycontrolsthecasesensitivityoftheprogramasitruns.Caseissignificant
bydefaultbecauseIGNORECASE(likemostvariables)isinitializedtozero:
x="aB"
if(x~/ab/)#thistestwillfail
IGNORECASE=1
if(x~/ab/)#nowitwillsucceed

Ingeneral,youcannotuseIGNORECASEtomakecertainrulescaseinsensitiveandotherrulescasesensitive,asthereisno
straightforwardwaytosetIGNORECASEjustforthepatternofaparticularrule.17Todothis,useeitherbracket
expressionsortolower().However,onethingyoucandowithIGNORECASEonlyisdynamicallyturncasesensitivityon
oroffforalltherulesatonce.
IGNORECASEcanbesetonthecommandlineorinaBEGINrule(seeOtherArgumentsalsoseeUsingBEGIN/END).
SettingIGNORECASEfromthecommandlineisawaytomakeaprogramcaseinsensitivewithouthavingtoeditit.

Inmultibytelocales,theequivalencesbetweenupperandlowercasecharactersaretestedbasedonthewidecharacter
valuesofthelocalescharacterset.Otherwise,thecharactersaretestedbasedontheISO88591(ISOLatin1)
characterset.Thischaractersetisasupersetofthetraditional128ASCIIcharacters,whichalsoprovidesanumberof
characterssuitableforusewithEuropeanlanguages.18
ThevalueofIGNORECASEhasnoeffectifgawkisincompatibilitymode(seeOptions).Caseisalwayssignificantin
compatibilitymode.
Previous:Casesensitivity,Up:Regexp[Contents][Index]

3.9Summary
Regularexpressionsdescribesetsofstringstobematched.Inawk,regularexpressionconstantsarewritten
enclosedbetweenslashes://.
Regexpconstantsmaybeusedstandaloneinpatternsandinconditionalexpressions,oraspartofmatching
expressionsusingthe~and!~operators.
Escapesequencesletyourepresentnonprintablecharactersandalsoletyourepresentregexpmetacharactersas
literalcharacterstobematched.
Regexpoperatorsprovidegrouping,alternation,andrepetition.
Bracketexpressionsgiveyouashorthandforspecifyingsetsofcharactersthatcanmatchataparticularpointina
regexp.Withinbracketexpressions,POSIXcharacterclassesletyouspecifycertaingroupsofcharactersina
localeindependentfashion.
Regularexpressionsmatchtheleftmostlongesttextinthestringbeingmatched.Thismattersforcaseswhereyou
https://www.gnu.org/software/gawk/manual/gawk.html

61/479

29/7/2016

TheGNUAwkUsersGuide

needtoknowtheextentofthematch,suchasfortextsubstitutionandwhentherecordseparatorisaregexp.
Matchingexpressionsmayusedynamicregexps(i.e.,stringvaluestreatedasregularexpressions).
gawksIGNORECASEvariableletsyoucontrolthecasesensitivityofregexpmatching.Inotherawkversions,use
tolower()ortoupper().
Next:Printing,Previous:Regexp,Up:Top[Contents][Index]

4ReadingInputFiles
Inthetypicalawkprogram,awkreadsallinputeitherfromthestandardinput(bydefault,thisisthekeyboard,butoftenit
isapipefromanothercommand)orfromfileswhosenamesyouspecifyontheawkcommandline.Ifyouspecifyinput
files,awkreadstheminorder,processingallthedatafromonebeforegoingontothenext.Thenameofthecurrent
inputfilecanbefoundinthepredefinedvariableFILENAME(seeBuiltinVariables).
Theinputisreadinunitscalledrecords,andisprocessedbytherulesofyourprogramonerecordatatime.Bydefault,
eachrecordisoneline.Eachrecordisautomaticallysplitintochunkscalledfields.Thismakesitmoreconvenientfor
programstoworkonthepartsofarecord.
Onrareoccasions,youmayneedtousethegetlinecommand.Thegetlinecommandisvaluablebothbecauseitcando
explicitinputfromanynumberoffiles,andbecausethefilesusedwithitdonothavetobenamedontheawkcommand
line(seeGetline).
Records:
Controllinghowdataissplitintorecords.
Fields:
Anintroductiontofields.
NonconstantFields:
NonconstantFieldNumbers.
ChangingFields:
ChangingtheContentsofaField.
FieldSeparators:
Thefieldseparatorandhowtochangeit.
ConstantSize:
Readingconstantwidthdata.
SplittingByContent:
DefiningFieldsByContent
MultipleLine:
Readingmultilinerecords.
Getline:
Readingfilesunderexplicitprogramcontrolusingthegetlinefunction.
ReadTimeout:
Readinginputwithatimeout.
Commandlinedirectories: Whathappensifyouputadirectoryonthecommandline.
InputSummary:
Inputsummary.
InputExercises:
Exercises.
Next:Fields,Up:ReadingFiles[Contents][Index]

4.1HowInputIsSplitintoRecords
awkdividestheinputforyourprogramintorecordsandfields.Itkeepstrackofthenumberofrecordsthathavebeen
readsofarfromthecurrentinputfile.ThisvalueisstoredinapredefinedvariablecalledFNR,whichisresettozero
everytimeanewfileisstarted.Anotherpredefinedvariable,NR,recordsthetotalnumberofinputrecordsreadsofar

fromalldatafiles.Itstartsatzero,butisneverautomaticallyresettozero.
awksplitrecords: Howstandardawksplitsrecords.
gawksplitrecords: Howgawksplitsrecords.
Next:gawksplitrecords,Up:Records[Contents][Index]
4.1.1RecordSplittingwithStandardawk
Recordsareseparatedbyacharactercalledtherecordseparator.Bydefault,therecordseparatoristhenewline
https://www.gnu.org/software/gawk/manual/gawk.html

62/479

29/7/2016

TheGNUAwkUsersGuide

character.Thisiswhyrecordsare,bydefault,singlelines.Touseadifferentcharacterfortherecordseparator,simply
assignthatcharactertothepredefinedvariableRS.
Likeanyothervariable,thevalueofRScanbechangedintheawkprogramwiththeassignmentoperator,=(see
AssignmentOps).Thenewrecordseparatorcharactershouldbeenclosedinquotationmarks,whichindicateastring
constant.Often,therighttimetodothisisatthebeginningofexecution,beforeanyinputisprocessed,sothatthevery
firstrecordisreadwiththeproperseparator.Todothis,usethespecialBEGINpattern(seeBEGIN/END).Forexample:
awk'BEGIN{RS="u"}
{print$0}'maillist

changesthevalueofRStou,beforereadinganyinput.Thenewvalueisastringwhosefirstcharacteristheletteru
asaresult,recordsareseparatedbytheletteru.Thentheinputfileisread,andthesecondruleintheawkprogram(the
actionwithnopattern)printseachrecord.Becauseeachprintstatementaddsanewlineattheendofitsoutput,thisawk
programcopiestheinputwitheachuchangedtoanewline.Herearetheresultsofrunningtheprogramonmaillist:
$awk'BEGIN{RS="u"}
>{print$0}'maillist
|Amelia5555553amelia.zodiac
|sq
|[email protected]
|Anthony5553412anthony.assert
|[email protected]
|Becky5557685becky.algebrar
|[email protected]
|[email protected]
|Broderick5550542broderick.aliq
|[email protected]
|Camilla5552912camilla.inf
|sar
|[email protected]
|Fabi
|s5551234fabi
|s.
|ndevicesim
|s@
|cb.ed
|F
|J
|lie5556699j
|lie.perscr
|[email protected]
|Martin5556480martin.codicib
|[email protected]
|Sam
|el5553430sam
|el.lanceolis@sh
|.ed
|A
|JeanPa
|l5552127jeanpa
|l.campanor
|m@ny
|.ed
|R
|

NotethattheentryforthenameBillisnotsplit.Intheoriginaldatafile(seeSampleDataFiles),thelinelookslike
this:
[email protected]

Itcontainsnou,sothereisnoreasontosplittherecord,unliketheothers,whicheachhaveoneormoreoccurrences
oftheu.Infact,thisrecordistreatedaspartofthepreviousrecordthenewlineseparatingthemintheoutputisthe
originalnewlineinthedatafile,nottheoneaddedbyawkwhenitprintedtherecord!
https://www.gnu.org/software/gawk/manual/gawk.html

63/479

29/7/2016

TheGNUAwkUsersGuide

Anotherwaytochangetherecordseparatorisonthecommandline,usingthevariableassignmentfeature(seeOther
Arguments):
awk'{print$0}'RS="u"maillist

ThissetsRStoubeforeprocessingmaillist.
Usinganalphabeticcharactersuchasufortherecordseparatorishighlylikelytoproducestrangeresults.Usingan
unusualcharactersuchas/ismorelikelytoproducecorrectbehaviorinthemajorityofcases,butthereareno
guarantees.Themoralis:KnowYourData.
Whenusingregularcharactersastherecordseparator,thereisoneunusualcasethatoccurswhengawkisbeingfully
POSIXcompliant(seeOptions).Then,thefollowing(extreme)pipelineprintsasurprising1:
$echo|gawkposix'BEGIN{RS="a"};{printNF}'
|1

Thereisonefield,consistingofanewline.ThevalueofthebuiltinvariableNFisthenumberoffieldsinthecurrent
record.(Inthenormalcase,gawktreatsthenewlineaswhitespace,printing0astheresult.Mostotherversionsofawk
alsoactthisway.)
Reachingtheendofaninputfileterminatesthecurrentinputrecord,evenifthelastcharacterinthefileisnotthe
characterinRS.(d.c.)
Theemptystring""(astringwithoutanycharacters)hasaspecialmeaningasthevalueofRS.Itmeansthatrecordsare
separatedbyoneormoreblanklinesandnothingelse.SeeMultipleLine,formoredetails.
IfyouchangethevalueofRSinthemiddleofanawkrun,thenewvalueisusedtodelimitsubsequentrecords,butthe
recordcurrentlybeingprocessed,aswellasrecordsalreadyprocessed,arenotaffected.
Aftertheendoftherecordhasbeendetermined,gawksetsthevariableRTtothetextintheinputthatmatchedRS.
Previous:awksplitrecords,Up:Records[Contents][Index]
4.1.2RecordSplittingwithgawk
Whenusinggawk,thevalueofRSisnotlimitedtoaonecharacterstring.Itcanbeanyregularexpression(seeRegexp).
(c.e.)Ingeneral,eachrecordendsatthenextstringthatmatchestheregularexpressionthenextrecordstartsattheend
ofthematchingstring.Thisgeneralruleisactuallyatworkintheusualcase,whereRScontainsjustanewline:arecord
endsatthebeginningofthenextmatchingstring(thenextnewlineintheinput),andthefollowingrecordstartsjust
aftertheendofthisstring(atthefirstcharacterofthefollowingline).Thenewline,becauseitmatchesRS,isnotpartof
eitherrecord.
WhenRSisasinglecharacter,RTcontainsthesamesinglecharacter.However,whenRSisaregularexpression,RT
containstheactualinputtextthatmatchedtheregularexpression.
IftheinputfileendswithoutanytextmatchingRS,gawksetsRTtothenullstring.
Thefollowingexampleillustratesbothofthesefeatures.ItsetsRSequaltoaregularexpressionthatmatcheseithera
newlineoraseriesofoneormoreuppercaseletterswithoptionalleadingand/ortrailingwhitespace:
$echorecord1AAAArecord2BBBBrecord3|
>gawk'BEGIN{RS="\n|(*[[:upper:]]+*)"}
>{print"Record=",$0,"andRT=["RT"]"}'
|Record=record1andRT=[AAAA]
|Record=record2andRT=[BBBB]
|Record=record3andRT=[
|]

ThesquarebracketsdelineatethecontentsofRT,lettingyouseetheleadingandtrailingwhitespace.Thefinalvalueof
https://www.gnu.org/software/gawk/manual/gawk.html

64/479

29/7/2016

TheGNUAwkUsersGuide

RTisanewline.SeeSimpleSed,foramoreusefulexampleofRSasaregexpandRT.

IfyousetRStoaregularexpressionthatallowsoptionaltrailingtext,suchasRS="abc(XYZ)?",itispossible,dueto
implementationconstraints,thatgawkmaymatchtheleadingpartoftheregularexpression,butnotthetrailingpart,
particularlyiftheinputtextthatcouldmatchthetrailingpartisfairlylong.gawkattemptstoavoidthisproblem,but
currently,theresnoguaranteethatthiswillneverhappen.
NOTE:Rememberthatinawk,the^and$anchormetacharactersmatchthebeginningandendofa
string,andnotthebeginningandendofaline.Asaresult,somethinglikeRS="^[[:upper:]]"canonly
matchatthebeginningofafile.Thisisbecausegawkviewstheinputfileasonelongstringthathappensto
containnewlinecharacters.ItisthusbesttoavoidanchormetacharactersinthevalueofRS.
TheuseofRSasaregularexpressionandtheRTvariablearegawkextensionstheyarenotavailableincompatibility
mode(seeOptions).Incompatibilitymode,onlythefirstcharacterofthevalueofRSdeterminestheendoftherecord.
RS="\0"IsNotPortable

Therearetimeswhenyoumightwanttotreatanentiredatafileasasinglerecord.Theonlywaytomakethishappen
istogiveRSavaluethatyouknowdoesntoccurintheinputfile.Thisishardtodoinageneralway,suchthata
programalwaysworksforarbitraryinputfiles.
Youmightthinkthatfortextfiles,theNULcharacter,whichconsistsofacharacterwithallbitsequaltozero,isagood
valuetouseforRSinthiscase:
BEGIN{RS="\0"}#wholefilebecomesonerecord?
gawkinfactacceptsthis,andusestheNULcharacterfortherecordseparator.Thisworksforcertainspecialfiles,such
as/proc/environonGNU/Linuxsystems,wheretheNULcharacterisinfacttherecordseparator.However,thisusage
isnotportabletomostotherawkimplementations.

Almostallotherawkimplementations19storestringsinternallyasCstylestrings.CstringsusetheNULcharacteras
thestringterminator.Ineffect,thismeansthatRS="\0"isthesameasRS="".(d.c.)
IthappensthatrecentversionsofmawkcanusetheNULcharacterasarecordseparator.However,thisisaspecialcase:
mawkdoesnotallowembeddedNULcharactersinstrings.(Thismaychangeinafutureversionofmawk.)
SeeReadfileFunction,foraninterestingwaytoreadwholefiles.Ifyouareusinggawk,seeExtensionSample
Readfile,foranotheroption.

Next:NonconstantFields,Previous:Records,Up:ReadingFiles[Contents][Index]

4.2ExaminingFields
Whenawkreadsaninputrecord,therecordisautomaticallyparsedorseparatedbytheawkutilityintochunkscalled
fields.Bydefault,fieldsareseparatedbywhitespace,likewordsinaline.Whitespaceinawkmeansanystringofoneor
morespaces,TABs,ornewlines20othercharactersthatareconsideredwhitespacebyotherlanguages(suchas
formfeed,verticaltab,etc.)arenotconsideredwhitespacebyawk.
Thepurposeoffieldsistomakeitmoreconvenientforyoutorefertothesepiecesoftherecord.Youdonthavetouse
themyoucanoperateonthewholerecordifyouwantbutfieldsarewhatmakesimpleawkprogramssopowerful.
Youuseadollarsign($)torefertoafieldinanawkprogram,followedbythenumberofthefieldyouwant.Thus,$1
referstothefirstfield,$2tothesecond,andsoon.(UnlikeintheUnixshells,thefieldnumbersarenotlimitedtosingle
digits.$127isthe127thfieldintherecord.)Forexample,supposethefollowingisalineofinput:
Thisseemslikeaprettyniceexample.
https://www.gnu.org/software/gawk/manual/gawk.html

65/479

29/7/2016

TheGNUAwkUsersGuide

Herethefirstfield,or$1,isThis,thesecondfield,or$2,isseems,andsoon.Notethatthelastfield,$7,is
example..Becausethereisnospacebetweentheeandthe.,theperiodisconsideredpartoftheseventhfield.
NFisapredefinedvariablewhosevalueisthenumberoffieldsinthecurrentrecord.awkautomaticallyupdatesthevalue
ofNFeachtimeitreadsarecord.Nomatterhowmanyfieldsthereare,thelastfieldinarecordcanberepresentedby
$NF.So,$NFisthesameas$7,whichisexample..Ifyoutrytoreferenceafieldbeyondthelastone(suchas$8when

therecordhasonlysevenfields),yougettheemptystring.(Ifusedinanumericoperation,yougetzero.)
Theuseof$0,whichlookslikeareferencetothezerothfield,isaspecialcase:itrepresentsthewholeinputrecord.
Useitwhenyouarenotinterestedinspecificfields.Herearesomemoreexamples:
$awk'$1~/li/{print$0}'maillist
|[email protected]
|[email protected]

Thisexampleprintseachrecordinthefilemaillistwhosefirstfieldcontainsthestringli.
Bycontrast,thefollowingexamplelooksforliintheentirerecordandprintsthefirstandlastfieldsforeach
matchinginputrecord:
$awk'/li/{print$1,$NF}'maillist
|AmeliaF
|BroderickR
|JulieF
|SamuelA

Next:ChangingFields,Previous:Fields,Up:ReadingFiles[Contents][Index]

4.3NonconstantFieldNumbers
Afieldnumberneednotbeaconstant.Anyexpressionintheawklanguagecanbeusedaftera$torefertoafield.The
valueoftheexpressionspecifiesthefieldnumber.Ifthevalueisastring,ratherthananumber,itisconvertedtoa
number.Considerthisexample:
awk'{print$NR}'

RecallthatNRisthenumberofrecordsreadsofar:oneinthefirstrecord,twointhesecond,andsoon.Sothisexample
printsthefirstfieldofthefirstrecord,thesecondfieldofthesecondrecord,andsoon.Forthetwentiethrecord,field
number20isprintedmostlikely,therecordhasfewerthan20fields,sothisprintsablankline.Hereisanother
exampleofusingexpressionsasfieldnumbers:
awk'{print$(2*2)}'maillist
awkevaluatestheexpression(2*2)andusesitsvalueasthenumberofthefieldtoprint.The*represents
multiplication,sotheexpression2*2evaluatestofour.Theparenthesesareusedsothatthemultiplicationisdone
beforethe$operationtheyarenecessarywheneverthereisabinaryoperator21inthefieldnumberexpression.This

example,then,printsthetypeofrelationship(thefourthfield)foreverylineofthefilemaillist.(Alloftheawk
operatorsarelisted,inorderofdecreasingprecedence,inPrecedence.)
Ifthefieldnumberyoucomputeiszero,yougettheentirerecord.Thus,$(22)hasthesamevalueas$0.Negative
fieldnumbersarenotallowedtryingtoreferenceoneusuallyterminatestheprogram.(ThePOSIXstandarddoesnot
definewhathappenswhenyoureferenceanegativefieldnumber.gawknoticesthisandterminatesyourprogram.Other
awkimplementationsmaybehavedifferently.)
AsmentionedinFields,awkstoresthecurrentrecordsnumberoffieldsinthebuiltinvariableNF(alsoseeBuiltin
Variables).Thus,theexpression$NFisnotaspecialfeatureitisthedirectconsequenceofevaluatingNFandusingits
valueasafieldnumber.

https://www.gnu.org/software/gawk/manual/gawk.html

66/479

29/7/2016

TheGNUAwkUsersGuide

Next:FieldSeparators,Previous:NonconstantFields,Up:ReadingFiles[Contents][Index]

4.4ChangingtheContentsofaField
Thecontentsofafield,asseenbyawk,canbechangedwithinanawkprogramthischangeswhatawkperceivesasthe
currentinputrecord.(Theactualinputisuntouchedawknevermodifiestheinputfile.)Considerthefollowingexample
anditsoutput:
$awk'{nboxes=$3;$3=$310
>printnboxes,$3}'inventoryshipped
|2515
|3222
|2414

Theprogramfirstsavestheoriginalvalueoffieldthreeinthevariablenboxes.Thesignrepresentssubtraction,so
thisprogramreassignsfieldthree,$3,astheoriginalvalueoffieldthreeminusten:$310.(SeeArithmeticOps.)
Thenitprintstheoriginalandnewvaluesforfieldthree.(Someoneinthewarehousemadeaconsistentmistakewhile
inventoryingtheredboxes.)
Forthistowork,thetextin$3mustmakesenseasanumberthestringofcharactersmustbeconvertedtoanumberfor
thecomputertodoarithmeticonit.Thenumberresultingfromthesubtractionisconvertedbacktoastringofcharacters
thatthenbecomesfieldthree.SeeConversion.
Whenthevalueofafieldischanged(asperceivedbyawk),thetextoftheinputrecordisrecalculatedtocontainthenew
fieldwheretheoldonewas.Inotherwords,$0changestoreflectthealteredfield.Thus,thisprogramprintsacopyof
theinputfile,with10subtractedfromthesecondfieldofeachline:
$awk'{$2=$210;print$0}'inventoryshipped
|Jan32515115
|Feb53224226
|Mar52434228

Itisalsopossibletoassigncontentstofieldsthatareoutofrange.Forexample:
$awk'{$6=($5+$4+$3+$2)
>print$6}'inventoryshipped
|168
|297
|301

Wevejustcreated$6,whosevalueisthesumoffields$2,$3,$4,and$5.The+signrepresentsaddition.Forthefile
inventoryshipped,$6representsthetotalnumberofparcelsshippedforaparticularmonth.
Creatinganewfieldchangesawksinternalcopyofthecurrentinputrecord,whichisthevalueof$0.Thus,ifyoudo
print$0afteraddingafield,therecordprintedincludesthenewfield,withtheappropriatenumberoffieldseparators
betweenitandthepreviouslyexistingfields.
ThisrecomputationaffectsandisaffectedbyNF(thenumberoffieldsseeFields).Forexample,thevalueofNFissetto
thenumberofthehighestfieldyoucreate.Theexactformatof$0isalsoaffectedbyafeaturethathasnotbeen
discussedyet:theoutputfieldseparator,OFS,usedtoseparatethefields(seeOutputSeparators).
Note,however,thatmerelyreferencinganoutofrangefielddoesnotchangethevalueofeither$0orNF.Referencing
anoutofrangefieldonlyproducesanemptystring.Forexample:
if($(NF+1)!="")
print"can'thappen"
else
print"everythingisnormal"

https://www.gnu.org/software/gawk/manual/gawk.html

67/479

29/7/2016

TheGNUAwkUsersGuide

shouldprinteverythingisnormal,becauseNF+1iscertaintobeoutofrange.(SeeIfStatement,formoreinformation
aboutawksifelsestatements.SeeTypingandComparison,formoreinformationaboutthe!=operator.)
Itisimportanttonotethatmakinganassignmenttoanexistingfieldchangesthevalueof$0butdoesnotchangethe
valueofNF,evenwhenyouassigntheemptystringtoafield.Forexample:
$echoabcd|awk'{OFS=":";$2=""
>print$0;printNF}'
|a::c:d
|4

Thefieldisstillthereitjusthasanemptyvalue,delimitedbythetwocolonsbetweenaandc.Thisexampleshows
whathappensifyoucreateanewfield:
$echoabcd|awk'{OFS=":";$2="";$6="new"
>print$0;printNF}'
|a::c:d::new
|6

Theinterveningfield,$5,iscreatedwithanemptyvalue(indicatedbythesecondpairofadjacentcolons),andNFis
updatedwiththevaluesix.
DecrementingNFthrowsawaythevaluesofthefieldsafterthenewvalueofNFandrecomputes$0.(d.c.)Hereisan
example:
$echoabcdef|awk'{print"NF=",NF;
>NF=3;print$0}'
|NF=6
|abc

CAUTION:Someversionsofawkdontrebuild$0whenNFisdecremented.
Finally,therearetimeswhenitisconvenienttoforceawktorebuildtheentirerecord,usingthecurrentvaluesofthe
fieldsandOFS.Todothis,usetheseeminglyinnocuousassignment:
$1=$1#forcerecordtobereconstituted
print$0#orwhateverelsewith$0

Thisforcesawktorebuildtherecord.Itdoeshelptoaddacomment,asweveshownhere.
Thereisaflipsidetotherelationshipbetween$0andthefields.Anyassignmentto$0causestherecordtobereparsed
intofieldsusingthecurrentvalueofFS.Thisalsoappliestoanybuiltinfunctionthatupdates$0,suchassub()and
gsub()(seeStringFunctions).
Understanding$0
Itisimportanttorememberthat$0isthefullrecord,exactlyasitwasreadfromtheinput.Thisincludesanyleadingor
trailingwhitespace,andtheexactwhitespace(orothercharacters)thatseparatesthefields.
ItisacommonerrortotrytochangethefieldseparatorsinarecordsimplybysettingFSandOFS,andthenexpectinga
plainprintorprint$0toprintthemodifiedrecord.
Butthisdoesnotwork,becausenothingwasdonetochangetherecorditself.Instead,youmustforcetherecordtobe
rebuilt,typicallywithastatementsuchas$1=$1,asdescribedearlier.

Next:ConstantSize,Previous:ChangingFields,Up:ReadingFiles[Contents][Index]

4.5SpecifyingHowFieldsAreSeparated
https://www.gnu.org/software/gawk/manual/gawk.html

68/479

29/7/2016

TheGNUAwkUsersGuide

DefaultFieldSplitting:
Howfieldsarenormallyseparated.
RegexpFieldSplitting:
Usingregexpsasthefieldseparator.
SingleCharacterFields:
Makingeachcharacteraseparatefield.
CommandLineFieldSeparator: SettingFSfromthecommandline.
FullLineFields:
Makingthefulllinebeasinglefield.
FieldSplittingSummary:
Somefinalpointsandasummarytable.
Thefieldseparator,whichiseitherasinglecharacteroraregularexpression,controlsthewayawksplitsaninputrecord
intofields.awkscanstheinputrecordforcharactersequencesthatmatchtheseparatorthefieldsthemselvesarethetext
betweenthematches.
Intheexamplesthatfollow,weusethebulletsymbol()torepresentspacesintheoutput.Ifthefieldseparatorisoo,
thenthefollowingline:
moogoogaipan

issplitintothreefields:m,g,andgaipan.Notetheleadingspacesinthevaluesofthesecondandthirdfields.
ThefieldseparatorisrepresentedbythepredefinedvariableFS.Shellprogrammerstakenote:awkdoesnotusethename
IFSthatisusedbythePOSIXcompliantshells(suchastheUnixBourneshell,sh,orBash).
ThevalueofFScanbechangedintheawkprogramwiththeassignmentoperator,=(seeAssignmentOps).Often,the
righttimetodothisisatthebeginningofexecutionbeforeanyinputhasbeenprocessed,sothattheveryfirstrecordis
readwiththeproperseparator.Todothis,usethespecialBEGINpattern(seeBEGIN/END).Forexample,herewesetthe
valueofFStothestring",":
awk'BEGIN{FS=","};{print$2}'

Giventheinputline:
JohnQ.Smith,29OakSt.,Walamazoo,MI42139

thisawkprogramextractsandprintsthestring29OakSt..
Sometimestheinputdatacontainsseparatorcharactersthatdontseparatefieldsthewayyouthoughttheywould.For
instance,thepersonsnameintheexamplewejustusedmighthaveatitleorsuffixattached,suchas:
JohnQ.Smith,LXIX,29OakSt.,Walamazoo,MI42139

ThesameprogramwouldextractLXIXinsteadof29OakSt..Ifyouwereexpectingtheprogramtoprintthe
address,youwouldbesurprised.Themoralistochooseyourdatalayoutandseparatorcharacterscarefullytoprevent
suchproblems.(Ifthedataisnotinaformthatiseasytoprocess,perhapsyoucanmassageitfirstwithaseparateawk
program.)
Next:RegexpFieldSplitting,Up:FieldSeparators[Contents][Index]
4.5.1WhitespaceNormallySeparatesFields
Fieldsarenormallyseparatedbywhitespacesequences(spaces,TABs,andnewlines),notbysinglespaces.Twospaces
inarowdonotdelimitanemptyfield.ThedefaultvalueofthefieldseparatorFSisastringcontainingasinglespace,
"".Ifawkinterpretedthisvalueintheusualway,eachspacecharacterwouldseparatefields,sotwospacesinarow
wouldmakeanemptyfieldbetweenthem.ThereasonthisdoesnothappenisthatasinglespaceasthevalueofFSisa
specialcaseitistakentospecifythedefaultmannerofdelimitingfields.
IfFSisanyothersinglecharacter,suchas",",theneachoccurrenceofthatcharacterseparatestwofields.Two
consecutiveoccurrencesdelimitanemptyfield.Ifthecharacteroccursatthebeginningortheendoftheline,thattoo
delimitsanemptyfield.Thespacecharacteristheonlysinglecharacterthatdoesnotfollowtheserules.
https://www.gnu.org/software/gawk/manual/gawk.html

69/479

29/7/2016

TheGNUAwkUsersGuide

Next:SingleCharacterFields,Previous:DefaultFieldSplitting,Up:FieldSeparators[Contents][Index]
4.5.2UsingRegularExpressionstoSeparateFields
TheprevioussubsectiondiscussedtheuseofsinglecharactersorsimplestringsasthevalueofFS.Moregenerally,the
valueofFSmaybeastringcontaininganyregularexpression.Inthiscase,eachmatchintherecordfortheregular
expressionseparatesfields.Forexample,theassignment:
FS=",\t"

makeseveryareaofaninputlinethatconsistsofacommafollowedbyaspaceandaTABintoafieldseparator.
Foralesstrivialexampleofaregularexpression,tryusingsinglespacestoseparatefieldsthewaysinglecommasare
used.FScanbesetto"[]"(leftbracket,space,rightbracket).Thisregularexpressionmatchesasinglespaceand
nothingelse(seeRegexp).
ThereisanimportantdifferencebetweenthetwocasesofFS=""(asinglespace)and FS="[\t\n]+"(aregular
expressionmatchingoneormorespaces,TABs,ornewlines).ForbothvaluesofFS,fieldsareseparatedbyruns
(multipleadjacentoccurrences)ofspaces,TABs,and/ornewlines.However,whenthevalueofFSis"",awkfirststrips
leadingandtrailingwhitespacefromtherecordandthendecideswherethefieldsare.Forexample,thefollowing
pipelineprintsb:
$echo'abcd'|awk'{print$2}'
|b

However,thispipelineprintsa(notetheextraspacesaroundeachletter):
$echo'abcd'|awk'BEGIN{FS="[\t\n]+"}
>{print$2}'
|a

Inthiscase,thefirstfieldisnull,orempty.
Thestrippingofleadingandtrailingwhitespacealsocomesintoplaywhenever$0isrecomputed.Forinstance,study
thispipeline:
$echo'abcd'|awk'{print;$2=$2;print}'
|abcd
|abcd

Thefirstprintstatementprintstherecordasitwasread,withleadingwhitespaceintact.Theassignmentto$2rebuilds
$0byconcatenating$1through$NFtogether,separatedbythevalueofOFS(whichisaspacebydefault).Becausethe
leadingwhitespacewasignoredwhenfinding$1,itisnotpartofthenew$0.Finally,thelastprintstatementprintsthe
new$0.
Thereisanadditionalsubtletytobeawareofwhenusingregularexpressionsforfieldsplitting.Itisnotwellspecified
inthePOSIXstandard,oranywhereelse,what^meanswhensplittingfields.Doesthe^matchonlyatthebeginning
oftheentirerecord?Oriseachfieldseparatoranewstring?Itturnsoutthatdifferentawkversionsanswerthisquestion
differently,andyoushouldnotrelyonanyspecificbehaviorinyourprograms.(d.c.)
Asapointofinformation,BWKawkallows^tomatchonlyatthebeginningoftherecord. gawkalsoworksthisway.
Forexample:
$echo'xxAAxxBxxC'|
>gawkF'(^x+)|(+)''{for(i=1;i<=NF;i++)
>printf">%s<\n",$i}'
|><
|>AA<
|>xxBxx<
|>C<

https://www.gnu.org/software/gawk/manual/gawk.html

70/479

29/7/2016

TheGNUAwkUsersGuide

Next:CommandLineFieldSeparator,Previous:RegexpFieldSplitting,Up:FieldSeparators[Contents][Index]
4.5.3MakingEachCharacteraSeparateField
Therearetimeswhenyoumaywanttoexamineeachcharacterofarecordseparately.Thiscanbedoneingawkby
simplyassigningthenullstring("")toFS.(c.e.)Inthiscase,eachindividualcharacterintherecordbecomesaseparate
field.Forexample:
$echoab|gawk'BEGIN{FS=""}
>{
>for(i=1;i<=NF;i=i+1)
>print"Field",i,"is",$i
>}'
|Field1isa
|Field2is
|Field3isb

Traditionally,thebehaviorofFSequalto""wasnotdefined.Inthiscase,mostversionsofUnixawksimplytreatthe
entirerecordasonlyhavingonefield.(d.c.)Incompatibilitymode(seeOptions),ifFSisthenullstring,thengawkalso
behavesthisway.
Next:FullLineFields,Previous:SingleCharacterFields,Up:FieldSeparators[Contents][Index]
4.5.4SettingFSfromtheCommandLine
FScanbesetonthecommandline.UsetheFoptiontodoso.Forexample:
awkF,'program'inputfiles

setsFStothe,character.NoticethattheoptionusesanuppercaseFinsteadofalowercasef.Thelatteroption(f)
specifiesafilecontaininganawkprogram.
ThevalueusedfortheargumenttoFisprocessedinexactlythesamewayasassignmentstothepredefinedvariableFS.
Anyspecialcharactersinthefieldseparatormustbeescapedappropriately.Forexample,tousea\asthefield
separatoronthecommandline,youwouldhavetotype:
#sameasFS="\\"
awkF\\\\''files

Because\isusedforquotingintheshell,awkseesF\\.Thenawkprocessesthe\\forescapecharacters(see
EscapeSequences),finallyyieldingasingle\touseforthefieldseparator.
Asaspecialcase,incompatibilitymode(seeOptions),iftheargumenttoFist,thenFSissettotheTABcharacter.If
youtypeF\tattheshell,withoutanyquotes,the\getsdeleted,soawkfiguresthatyoureallywantyourfieldstobe
separatedwithTABsandnotts.UsevFS="t"orF"[t]"onthecommandlineifyoureallydowanttoseparate
yourfieldswithts.UseF'\t'whennotincompatibilitymodetospecifythatTABsseparatefields.
Asanexample,letsuseanawkprogramfilecallededu.awkthatcontainsthepattern/edu/andtheactionprint$1:
/edu/{print$1}

LetsalsosetFStobethecharacterandruntheprogramonthefilemaillist.Thefollowingcommandprintsalist
ofthenamesofthepeoplethatworkatorattendauniversity,andthefirstthreedigitsoftheirphonenumbers:
$awkFfedu.awkmaillist
|Fabius555
|Samuel555
|Jean

Notethethirdlineofoutput.Thethirdlineintheoriginalfilelookedlikethis:
https://www.gnu.org/software/gawk/manual/gawk.html

71/479

29/7/2016

TheGNUAwkUsersGuide

[email protected]

Theaspartofthepersonsnamewasusedasthefieldseparator,insteadoftheinthephonenumberthatwas
originallyintended.Thisdemonstrateswhyyouhavetobecarefulinchoosingyourfieldandrecordseparators.
PerhapsthemostcommonuseofasinglecharacterasthefieldseparatoroccurswhenprocessingtheUnixsystem
passwordfile.OnmanyUnixsystems,eachuserhasaseparateentryinthesystempasswordfile,withonelineperuser.
Theinformationintheselinesisseparatedbycolons.Thefirstfieldistheusersloginnameandthesecondistheusers
encryptedorshadowpassword.(Ashadowpasswordisindicatedbythepresenceofasinglexinthesecondfield.)A
passwordfileentrymightlooklikethis:
arnold:x:2076:10:ArnoldRobbins:/home/arnold:/bin/bash

Thefollowingprogramsearchesthesystempasswordfileandprintstheentriesforuserswhosefullnameisnot
indicated:
awkF:'$5==""'/etc/passwd

Next:FieldSplittingSummary,Previous:CommandLineFieldSeparator,Up:FieldSeparators[Contents][Index]
4.5.5MakingtheFullLineBeaSingleField
Occasionally,itsusefultotreatthewholeinputlineasasinglefield.Thiscanbedoneeasilyandportablysimplyby
settingFSto"\n"(anewline):22
awkF'\n''program'files

Whenyoudothis,$1isthesameas$0.
ChangingFSDoesNotAffecttheFields
AccordingtothePOSIXstandard,awkissupposedtobehaveasifeachrecordissplitintofieldsatthetimeitisread.
Inparticular,thismeansthatifyouchangethevalueofFSafterarecordisread,thevaluesofthefields(i.e.,howthey
weresplit)shouldreflecttheoldvalueofFS,notthenewone.
However,manyolderimplementationsofawkdonotworkthisway.Instead,theydefersplittingthefieldsuntilafield
isactuallyreferenced.ThefieldsaresplitusingthecurrentvalueofFS!(d.c.)Thisbehaviorcanbedifficultto
diagnose.Thefollowingexampleillustratesthedifferencebetweenthetwomethods:
sed1q/etc/passwd|awk'{FS=":";print$1}'

whichusuallyprints:
root

onanincorrectimplementationofawk,whilegawkprintsthefullfirstlineofthefile,somethinglike:
root:x:0:0:Root:/:

(Thesed23commandprintsjustthefirstlineof/etc/passwd.)

Previous:FullLineFields,Up:FieldSeparators[Contents][Index]
4.5.6FieldSplittingSummary
ItisimportanttorememberthatwhenyouassignastringconstantasthevalueofFS,itundergoesnormalawkstring
processing.Forexample,withUnixawkandgawk,theassignmentFS="\.."assignsthecharacterstring".."toFS
https://www.gnu.org/software/gawk/manual/gawk.html

72/479

29/7/2016

TheGNUAwkUsersGuide

(thebackslashisstripped).Thiscreatesaregexpmeaningfieldsareseparatedbyoccurrencesofanytwocharacters.If
insteadyouwantfieldstobeseparatedbyaliteralperiodfollowedbyanysinglecharacter,useFS="\\..".
Thefollowinglistsummarizeshowfieldsaresplit,basedonthevalueofFS(==meansisequalto):
FS==""

Fieldsareseparatedbyrunsofwhitespace.Leadingandtrailingwhitespaceareignored.Thisisthedefault.
FS==anyothersinglecharacter

Fieldsareseparatedbyeachoccurrenceofthecharacter.Multiplesuccessiveoccurrencesdelimitemptyfields,as
doleadingandtrailingoccurrences.Thecharactercanevenbearegexpmetacharacteritdoesnotneedtobe
escaped.
FS==regexp

Fieldsareseparatedbyoccurrencesofcharactersthatmatchregexp.Leadingandtrailingmatchesofregexp
delimitemptyfields.
FS==""

Eachindividualcharacterintherecordbecomesaseparatefield.(Thisisacommonextensionitisnotspecified
bythePOSIXstandard.)
FSandIGNORECASE

TheIGNORECASEvariable(seeUsermodified)affectsfieldsplittingonlywhenthevalueofFSisaregexp.Ithasno
effectwhenFSisasinglecharacter,evenifthatcharacterisaletter.Thus,inthefollowingcode:
FS="c"
IGNORECASE=1
$0="aCa"
print$1

TheoutputisaCa.Ifyoureallywanttosplitfieldsonanalphabeticcharacterwhileignoringcase,usearegexpthat
willdoitforyou(e.g.,FS="[c]").Inthiscase,IGNORECASEwilltakeeffect.

Next:SplittingByContent,Previous:FieldSeparators,Up:ReadingFiles[Contents][Index]

4.6ReadingFixedWidthData
Thissectiondiscussesanadvancedfeatureofgawk.Ifyouareanoviceawkuser,youmightwanttoskipitonthefirst
reading.
gawkprovidesafacilityfordealingwithfixedwidthfieldswithnodistinctivefieldseparator.Forexample,dataofthis

naturearisesintheinputforoldFortranprogramswherenumbersareruntogether,orintheoutputofprogramsthatdid
notanticipatetheuseoftheiroutputasinputforotherprograms.
Anexampleofthelatterisatablewhereallthecolumnsarelinedupbytheuseofavariablenumberofspacesand
emptyfieldsarejustspaces.Clearly,awksnormalfieldsplittingbasedonFSdoesnotworkwellinthiscase.Althougha
portableawkprogramcanuseaseriesofsubstr()callson$0(seeStringFunctions),thisisawkwardandinefficientfor
alargenumberoffields.
Thesplittingofaninputrecordintofixedwidthfieldsisspecifiedbyassigningastringcontainingspaceseparated
numberstothebuiltinvariableFIELDWIDTHS.Eachnumberspecifiesthewidthofthefield,includingcolumnsbetween
fields.Ifyouwanttoignorethecolumnsbetweenfields,youcanspecifythewidthasaseparatefieldthatis
subsequentlyignored.Itisafatalerrortosupplyafieldwidththathasanegativevalue.Thefollowingdataistheoutput
https://www.gnu.org/software/gawk/manual/gawk.html

73/479

29/7/2016

TheGNUAwkUsersGuide

oftheUnixwutility.ItisusefultoillustratetheuseofFIELDWIDTHS:
10:06pmup21days,14:04,23users
UserttyloginidleJCPUPCPUwhat
hzuottyV08:58pm95vip24.tex
hzangttyV36:37pm50csh
eklyettyV59:53pm71emthes.tex
dporteinttyV68:17pm1:47csh
gierdttyD310:00pm1elm
davettyD49:47pm44w
brentttyp026Jun914:4626:464:41bash
davettyq426Jun9115days4646wnewmail

Thefollowingprogramtakesthisinput,convertstheidletimetonumberofseconds,andprintsoutthefirsttwofields
andthecalculatedidletime:
BEGIN{FIELDWIDTHS="961067735"}
NR>2{
idle=$4
sub(/^+/,"",idle)#stripleadingspaces
if(idle=="")
idle=0
if(idle~/:/){
split(idle,t,":")
idle=t[1]*60+t[2]
}
if(idle~/days/)
idle*=24*60*60
print$1,$2,idle
}

NOTE:Theprecedingprogramusesanumberofawkfeaturesthathaventbeenintroducedyet.
Runningtheprogramonthedataproducesthefollowingresults:
hzuottyV00
hzangttyV350
eklyettyV50
dporteinttyV6107
gierdttyD31
davettyD40
brentttyp0286
davettyq41296000

Another(possiblymorepractical)exampleoffixedwidthinputdataistheinputfromadeckofballotingcards.Insome
partsoftheUnitedStates,votersmarktheirchoicesbypunchingholesincomputercards.Thesecardsarethen
processedtocountthevotesforanyparticularcandidateoronanyparticularissue.Becauseavotermaychoosenotto
voteonsomeissue,anycolumnonthecardmaybeempty.Anawkprogramforprocessingsuchdatacouldusethe
FIELDWIDTHSfeaturetosimplifyreadingthedata.(Ofcourse,gettinggawktorunonasystemwithcardreadersisanother
story!)
AssigningavaluetoFScausesgawktouseFSforfieldsplittingagain.UseFS=FStomakethishappen,without
havingtoknowthecurrentvalueofFS.Inordertotellwhichkindoffieldsplittingisineffect,usePROCINFO["FS"](see
Autoset).Thevalueis"FS"ifregularfieldsplittingisbeingused,or"FIELDWIDTHS"iffixedwidthfieldsplittingis
beingused:
if(PROCINFO["FS"]=="FS")
regularfieldsplitting
elseif(PROCINFO["FS"]=="FIELDWIDTHS")
fixedwidthfieldsplitting
else
contentbasedfieldsplitting(seenextsection)

ThisinformationisusefulwhenwritingafunctionthatneedstotemporarilychangeFSorFIELDWIDTHS,readsome
https://www.gnu.org/software/gawk/manual/gawk.html

74/479

29/7/2016

TheGNUAwkUsersGuide

records,andthenrestoretheoriginalsettings(seePasswdFunctions,foranexampleofsuchafunction).
Next:MultipleLine,Previous:ConstantSize,Up:ReadingFiles[Contents][Index]

4.7DefiningFieldsbyContent
Thissectiondiscussesanadvancedfeatureofgawk.Ifyouareanoviceawkuser,youmightwanttoskipitonthefirst
reading.
Normally,whenusingFS,gawkdefinesthefieldsasthepartsoftherecordthatoccurinbetweeneachfieldseparator.In
otherwords,FSdefineswhatafieldisnot,insteadofwhatafieldis.However,therearetimeswhenyoureallywantto
definethefieldsbywhattheyare,andnotbywhattheyarenot.
Themostnotorioussuchcaseissocalledcommaseparatedvalues(CSV)data.Manyspreadsheetprograms,for
example,canexporttheirdataintotextfiles,whereeachrecordisterminatedwithanewline,andfieldsareseparatedby
commas.Ifcommasonlyseparatedthedata,therewouldntbeanissue.Theproblemcomeswhenoneofthefields
containsanembeddedcomma.Insuchcases,mostprogramsembedthefieldindoublequotes.24So,wemighthave
datalikethis:
Robbins,Arnold,"1234APrettyStreet,NE",MyTown,MyState,123456789,USA

TheFPATvariableoffersasolutionforcaseslikethis.ThevalueofFPATshouldbeastringthatprovidesaregular
expression.Thisregularexpressiondescribesthecontentsofeachfield.
InthecaseofCSVdataaspresentedhere,eachfieldiseitheranythingthatisnotacomma,oradoublequote,
anythingthatisnotadoublequote,andaclosingdoublequote.Ifwrittenasaregularexpressionconstant(see
Regexp),wewouldhave/([^,]+)|("[^"]+")/.Writingthisasastringrequiresustoescapethedoublequotes,leading
to:
FPAT="([^,]+)|(\"[^\"]+\")"

Puttingthistouse,hereisasimpleprogramtoparsethedata:
BEGIN{
FPAT="([^,]+)|(\"[^\"]+\")"
}
{
print"NF=",NF
for(i=1;i<=NF;i++){
printf("$%d=<%s>\n",i,$i)
}
}

Whenrun,wegetthefollowing:
$gawkfsimplecsv.awkaddresses.csv
NF=7
$1=<Robbins>
$2=<Arnold>
$3=<"1234APrettyStreet,NE">
$4=<MyTown>
$5=<MyState>
$6=<123456789>
$7=<USA>

Notetheembeddedcommainthevalueof$3.
AstraightforwardimprovementwhenprocessingCSVdataofthissortwouldbetoremovethequoteswhentheyoccur,
withsomethinglikethis:
https://www.gnu.org/software/gawk/manual/gawk.html

75/479

29/7/2016

TheGNUAwkUsersGuide

if(substr($i,1,1)=="\""){
len=length($i)
$i=substr($i,2,len2)#Gettextwithinthetwoquotes
}

AswithFS,theIGNORECASEvariable(seeUsermodified)affectsfieldsplittingwithFPAT.
AssigningavaluetoFPAToverridesfieldsplittingwithFSandwithFIELDWIDTHS.SimilartoFIELDWIDTHS,thevalueof
PROCINFO["FS"]willbe"FPAT"ifcontentbasedfieldsplittingisbeingused.
NOTE:SomeprogramsexportCSVdatathatcontainsembeddednewlinesbetweenthedoublequotes.
gawkprovidesnowaytodealwiththis.EventhoughaformalspecificationforCSVdataexists,thereisnt
muchmoretobedonetheFPATmechanismprovidesanelegantsolutionforthemajorityofcases,andthe
gawkdevelopersaresatisfiedwiththat.
Aswritten,theregexpusedforFPATrequiresthateachfieldcontainatleastonecharacter.Astraightforward
modification(changingthefirst+to*)allowsfieldstobeempty:
FPAT="([^,]*)|(\"[^\"]+\")"

Finally,thepatsplit()functionmakesthesamefunctionalityavailableforsplittingregularstrings(seeString
Functions).
Torecap,gawkprovidesthreeindependentmethodstosplitinputrecordsintofields.Themechanismusedisbasedon
whichofthethreevariablesFS,FIELDWIDTHS,orFPATwaslastassignedto.
Next:Getline,Previous:SplittingByContent,Up:ReadingFiles[Contents][Index]

4.8MultipleLineRecords
Insomedatabases,asinglelinecannotconvenientlyholdalltheinformationinoneentry.Insuchcases,youcanuse
multilinerecords.Thefirststepindoingthisistochooseyourdataformat.
Onetechniqueistouseanunusualcharacterorstringtoseparaterecords.Forexample,youcouldusetheformfeed
character(written\finawk,asinC)toseparatethem,makingeachrecordapageofthefile.Todothis,justsetthe
variableRSto"\f"(astringcontainingtheformfeedcharacter).Anyothercharactercouldequallywellbeused,aslong
asitwontbepartofthedatainarecord.
Anothertechniqueistohaveblanklinesseparaterecords.Byaspecialdispensation,anemptystringasthevalueofRS
indicatesthatrecordsareseparatedbyoneormoreblanklines.WhenRSissettotheemptystring,eachrecordalways
endsatthefirstblanklineencountered.Thenextrecorddoesntstartuntilthefirstnonblanklinethatfollows.Nomatter
howmanyblanklinesappearinarow,theyallactasonerecordseparator.(Blanklinesmustbecompletelyemptylines
thatcontainonlywhitespacedonotcount.)
YoucanachievethesameeffectasRS=""byassigningthestring"\n\n+"toRS.Thisregexpmatchesthenewlineat
theendoftherecordandoneormoreblanklinesaftertherecord.Inaddition,aregularexpressionalwaysmatchesthe
longestpossiblesequencewhenthereisachoice(seeLeftmostLongest).So,thenextrecorddoesntstartuntilthefirst
nonblanklinethatfollowsnomatterhowmanyblanklinesappearinarow,theyareconsideredonerecordseparator.
However,thereisanimportantdifferencebetweenRS=""andRS="\n\n+".Inthefirstcase,leadingnewlinesin
theinputdatafileareignored,andifafileendswithoutextrablanklinesafterthelastrecord,thefinalnewlineis
removedfromtherecord.Inthesecondcase,thisspecialprocessingisnotdone.(d.c.)
Nowthattheinputisseparatedintorecords,thesecondstepistoseparatethefieldsintherecords.Onewaytodothisis
todivideeachofthelinesintofieldsinthenormalmanner.Thishappensbydefaultastheresultofaspecialfeature.
WhenRSissettotheemptystringandFSissettoasinglecharacter,thenewlinecharacteralwaysactsasafield
separator.ThisisinadditiontowhateverfieldseparationsresultfromFS.25
https://www.gnu.org/software/gawk/manual/gawk.html

76/479

29/7/2016

TheGNUAwkUsersGuide

Theoriginalmotivationforthisspecialexceptionwasprobablytoprovideusefulbehaviorinthedefaultcase(i.e.,FSis
equalto"").Thisfeaturecanbeaproblemifyoureallydontwantthenewlinecharactertoseparatefields,because
thereisnowaytopreventit.However,youcanworkaroundthisbyusingthesplit()functiontobreakuptherecord
manually(seeStringFunctions).Ifyouhaveasinglecharacterfieldseparator,youcanworkaroundthespecialfeature
inadifferentway,bymakingFSintoaregexpforthatsinglecharacter.Forexample,ifthefieldseparatorisapercent
character,insteadofFS="%",useFS="[%]".
Anotherwaytoseparatefieldsistoputeachfieldonaseparateline:todothis,justsetthevariableFStothestring"\n".
(Thissinglecharacterseparatormatchesasinglenewline.)Apracticalexampleofadatafileorganizedthiswaymight
beamailinglist,whereblanklinesseparatetheentries.Consideramailinglistinafilenamedaddresses,whichlooks
likethis:
JaneDoe
123MainStreet
Anywhere,SE123456789
JohnSmith
456TreelinedAvenue
Smallville,MW987654321

Asimpleprogramtoprocessthisfileisasfollows:
#addrs.awksimplemailinglistprogram
#Recordsareseparatedbyblanklines.
#Eachlineisonefield.
BEGIN{RS="";FS="\n"}
{
print"Nameis:",$1
print"Addressis:",$2
print"CityandStateare:",$3
print""
}

Runningtheprogramproducesthefollowingoutput:
$awkfaddrs.awkaddresses
|Nameis:JaneDoe
|Addressis:123MainStreet
|CityandStateare:Anywhere,SE123456789
|
|Nameis:JohnSmith
|Addressis:456TreelinedAvenue
|CityandStateare:Smallville,MW987654321
|

SeeLabelsProgram,foramorerealisticprogramdealingwithaddresslists.Thefollowinglistsummarizeshowrecords
aresplit,basedonthevalueofRS:
RS=="\n"

Recordsareseparatedbythenewlinecharacter(\n).Ineffect,everylineinthedatafileisaseparaterecord,
includingblanklines.Thisisthedefault.
RS==anysinglecharacter

Recordsareseparatedbyeachoccurrenceofthecharacter.Multiplesuccessiveoccurrencesdelimitempty
records.
RS==""

https://www.gnu.org/software/gawk/manual/gawk.html

77/479

29/7/2016

TheGNUAwkUsersGuide

Recordsareseparatedbyrunsofblanklines.WhenFSisasinglecharacter,thenthenewlinecharacteralways
servesasafieldseparator,inadditiontowhatevervalueFSmayhave.Leadingandtrailingnewlinesinafileare
ignored.
RS==regexp

Recordsareseparatedbyoccurrencesofcharactersthatmatchregexp.Leadingandtrailingmatchesofregexp
delimitemptyrecords.(ThisisagawkextensionitisnotspecifiedbythePOSIXstandard.)
Ifnotincompatibilitymode(seeOptions),gawksetsRTtotheinputtextthatmatchedthevaluespecifiedbyRS.Butif
theinputfileendedwithoutanytextthatmatchesRS,thengawksetsRTtothenullstring.
Next:ReadTimeout,Previous:MultipleLine,Up:ReadingFiles[Contents][Index]

4.9ExplicitInputwithgetline
Sofarwehavebeengettingourinputdatafromawksmaininputstreameitherthestandardinput(usuallyyour
keyboard,sometimestheoutputfromanotherprogram)orthefilesspecifiedonthecommandline.Theawklanguage
hasaspecialbuiltincommandcalledgetlinethatcanbeusedtoreadinputunderyourexplicitcontrol.
Thegetlinecommandisusedinseveraldifferentwaysandshouldnotbeusedbybeginners.Theexamplesthatfollow
theexplanationofthegetlinecommandincludematerialthathasnotbeencoveredyet.Therefore,comebackandstudy
thegetlinecommandafteryouhavereviewedtherestofthisWebpageandhaveagoodknowledgeofhowawkworks.
Thegetlinecommandreturns1ifitfindsarecordand0ifitencounterstheendofthefile.Ifthereissomeerrorin
gettingarecord,suchasafilethatcannotbeopened,thengetlinereturns1.Inthiscase,gawksetsthevariableERRNOto
astringdescribingtheerrorthatoccurred.
Inthefollowingexamples,commandstandsforastringvaluethatrepresentsashellcommand.
NOTE:Whensandboxisspecified(seeOptions),readinglinesfromfiles,pipes,andcoprocessesis
disabled.
PlainGetline:
Usinggetlinewithnoarguments.
Getline/Variable:
Usinggetlineintoavariable.
Getline/File:
Usinggetlinefromafile.
Getline/Variable/File:
Usinggetlineintoavariablefromafile.
Getline/Pipe:
Usinggetlinefromapipe.
Getline/Variable/Pipe:
Usinggetlineintoavariablefromapipe.
Getline/Coprocess:
Usinggetlinefromacoprocess.
Getline/Variable/Coprocess: Usinggetlineintoavariablefromacoprocess.
GetlineNotes:
Importantthingstoknowaboutgetline.
GetlineSummary:
SummaryofgetlineVariants.
Next:Getline/Variable,Up:Getline[Contents][Index]
4.9.1UsinggetlinewithNoArguments
Thegetlinecommandcanbeusedwithoutargumentstoreadinputfromthecurrentinputfile.Allitdoesinthiscaseis
readthenextinputrecordandsplititupintofields.Thisisusefulifyouvefinishedprocessingthecurrentrecord,but
wanttodosomespecialprocessingonthenextrecordrightnow.Forexample:
#Removetextbetween/*and*/,inclusive
{
if((i=index($0,"/*"))!=0){
out=substr($0,1,i1)#leadingpartofthestring
https://www.gnu.org/software/gawk/manual/gawk.html

78/479

29/7/2016

TheGNUAwkUsersGuide

rest=substr($0,i+2)#...*/...
j=index(rest,"*/")#is*/intrailingpart?
if(j>0){
rest=substr(rest,j+2)#removecomment
}else{
while(j==0){
#getmoretext
if(getline<=0){
print("unexpectedEOForerror:",ERRNO)>"/dev/stderr"
exit
}
#buildupthelineusingstringconcatenation
rest=rest$0
j=index(rest,"*/")#is*/intrailingpart?
if(j!=0){
rest=substr(rest,j+2)
break
}
}
}
#builduptheoutputlineusingstringconcatenation
$0=outrest
}
print$0
}

ThisawkprogramdeletesCstylecomments(/**/)fromtheinput.Itusesanumberoffeatureswehaventcovered
yet,includingstringconcatenation(seeConcatenation)andtheindex()andsubstr()builtinfunctions(seeString
Functions).Byreplacingtheprint$0withotherstatements,youcouldperformmorecomplicatedprocessingonthe
decommentedinput,suchassearchingformatchesofaregularexpression.(Thisprogramhasasubtleproblemitdoes
notworkifonecommentendsandanotherbeginsonthesameline.)
ThisformofthegetlinecommandsetsNF,NR,FNR,RT,andthevalueof$0.
NOTE:Thenewvalueof$0isusedtotestthepatternsofanysubsequentrules.Theoriginalvalueof$0
thattriggeredtherulethatexecutedgetlineislost.Bycontrast,thenextstatementreadsanewrecordbut
immediatelybeginsprocessingitnormally,startingwiththefirstruleintheprogram.SeeNextStatement.
Next:Getline/File,Previous:PlainGetline,Up:Getline[Contents][Index]
4.9.2UsinggetlineintoaVariable
Youcanusegetlinevartoreadthenextrecordfromawksinputintothevariablevar.Nootherprocessingisdone.
Forexample,supposethenextlineisacommentoraspecialstring,andyouwanttoreaditwithouttriggeringanyrules.
Thisformofgetlineallowsyoutoreadthatlineandstoreitinavariablesothatthemainreadalineandcheckeach
ruleloopofawkneverseesit.Thefollowingexampleswapseverytwolinesofinput:
{
if((getlinetmp)>0){
printtmp
print$0
}else
print$0
}

Ittakesthefollowinglist:
wan
tew
free
phore

andproducestheseresults:
https://www.gnu.org/software/gawk/manual/gawk.html

79/479

29/7/2016

TheGNUAwkUsersGuide

tew
wan
phore
free

ThegetlinecommandusedinthiswaysetsonlythevariablesNR,FNR,andRT(and,ofcourse,var).Therecordisnot
splitintofields,sothevaluesofthefields(including$0)andthevalueofNFdonotchange.
Next:Getline/Variable/File,Previous:Getline/Variable,Up:Getline[Contents][Index]
4.9.3UsinggetlinefromaFile
Usegetline<filetoreadthenextrecordfromfile.Here,fileisastringvaluedexpressionthatspecifiesthefile
name.<fileiscalledaredirectionbecauseitdirectsinputtocomefromadifferentplace.Forexample,thefollowing
programreadsitsinputrecordfromthefilesecondary.inputwhenitencountersafirstfieldwithavalueequalto10in
thecurrentinputfile:
{
if($1==10){
getline<"secondary.input"
print
}else
print
}

Becausethemaininputstreamisnotused,thevaluesofNRandFNRarenotchanged.However,therecorditreadsissplit
intofieldsinthenormalmanner,sothevaluesof$0andtheotherfieldsarechanged,resultinginanewvalueofNF.RT
isalsoset.
AccordingtoPOSIX,getline<expressionisambiguousifexpressioncontainsunparenthesizedoperatorsotherthan
$forexample,getline<dir"/"fileisambiguousbecausetheconcatenationoperator(notdiscussedyetsee
Concatenation)isnotparenthesized.Youshouldwriteitasgetline<(dir"/"file)ifyouwantyourprogramtobe
portabletoallawkimplementations.
Next:Getline/Pipe,Previous:Getline/File,Up:Getline[Contents][Index]
4.9.4UsinggetlineintoaVariablefromaFile
Usegetlinevar<filetoreadinputfromthefilefile,andputitinthevariablevar.Asearlier,fileisastringvalued
expressionthatspecifiesthefilefromwhichtoread.
Inthisversionofgetline,noneofthepredefinedvariablesarechangedandtherecordisnotsplitintofields.Theonly
variablechangedisvar.26Forexample,thefollowingprogramcopiesalltheinputfilestotheoutput,exceptforrecords
thatsay@includefilename.Sucharecordisreplacedbythecontentsofthefilefilename:
{
if(NF==2&&$1=="@include"){
while((getlineline<$2)>0)
printline
close($2)
}else
print
}

Noteherehowthenameoftheextrainputfileisnotbuiltintotheprogramitistakendirectlyfromthedata,
specificallyfromthesecondfieldonthe@includeline.
Theclose()functioniscalledtoensurethatiftwoidentical@includelinesappearintheinput,theentirespecifiedfile
isincludedtwice.SeeCloseFilesAndPipes.
https://www.gnu.org/software/gawk/manual/gawk.html

80/479

29/7/2016

TheGNUAwkUsersGuide

Onedeficiencyofthisprogramisthatitdoesnotprocessnested@includestatements(i.e.,@includestatementsin
includedfiles)thewayatruemacropreprocessorwould.SeeIgawkProgram,foraprogramthatdoeshandlenested
@includestatements.
Next:Getline/Variable/Pipe,Previous:Getline/Variable/File,Up:Getline[Contents][Index]
4.9.5UsinggetlinefromaPipe
Omnisciencehasmuchtorecommendit.Failingthat,attentiontodetailswouldbeuseful.
BrianKernighan
Theoutputofacommandcanalsobepipedintogetline,usingcommand|getline.Inthiscase,thestringcommandis
runasashellcommandanditsoutputispipedintoawktobeusedasinput.Thisformofgetlinereadsonerecordata
timefromthepipe.Forexample,thefollowingprogramcopiesitsinputtoitsoutput,exceptforlinesthatbeginwith
@execute,whicharereplacedbytheoutputproducedbyrunningtherestofthelineasashellcommand:
{
if($1=="@execute"){
tmp=substr($0,10)#Remove"@execute"
while((tmp|getline)>0)
print
close(tmp)
}else
print
}

Theclose()functioniscalledtoensurethatiftwoidentical@executelinesappearintheinput,thecommandisrunfor
eachone.SeeCloseFilesAndPipes.Giventheinput:
foo
bar
baz
@executewho
bletch

theprogrammightproduce:
foo
bar
baz
arnoldttyv0Jul1314:22
miriamttyp0Jul1314:23(murphy:0)
billttyp1Jul1314:23(murphy:0)
bletch

Noticethatthisprogramranthecommandwhoandprintedtheresult.(Ifyoutrythisprogramyourself,youwillof
coursegetdifferentresults,dependinguponwhoisloggedinonyoursystem.)
Thisvariationofgetlinesplitstherecordintofields,setsthevalueofNF,andrecomputesthevalueof$0.Thevaluesof
NRandFNRarenotchanged.RTisset.
AccordingtoPOSIX,expression|getlineisambiguousifexpressioncontainsunparenthesizedoperatorsotherthan
$forexample,"echo""date"|getlineisambiguousbecausetheconcatenationoperatorisnotparenthesized.
Youshouldwriteitas("echo""date")|getlineifyouwantyourprogramtobeportabletoallawk
implementations.
NOTE:Unfortunately,gawkhasnotbeenconsistentinitstreatmentofaconstructlike"echo""date"|
getline.Mostversions,includingthecurrentversion,treatitatas("echo""date")|getline.(Thisis
alsohowBWKawkbehaves.)Someversionsinsteadtreatitas"echo"("date"|getline).(Thisishow
mawkbehaves.)Inshort,alwaysuseexplicitparentheses,andthenyouwonthavetoworry.
https://www.gnu.org/software/gawk/manual/gawk.html

81/479

29/7/2016

TheGNUAwkUsersGuide

Next:Getline/Coprocess,Previous:Getline/Pipe,Up:Getline[Contents][Index]
4.9.6UsinggetlineintoaVariablefromaPipe
Whenyouusecommand|getlinevar,theoutputofcommandissentthroughapipetogetlineandintothevariable
var.Forexample,thefollowingprogramreadsthecurrentdateandtimeintothevariablecurrent_time,usingthedate
utility,andthenprintsit:
BEGIN{
"date"|getlinecurrent_time
close("date")
print"Reportprintedon"current_time
}

Inthisversionofgetline,noneofthepredefinedvariablesarechangedandtherecordisnotsplitintofields.However,
RTisset.
Next:Getline/Variable/Coprocess,Previous:Getline/Variable/Pipe,Up:Getline[Contents][Index]
4.9.7UsinggetlinefromaCoprocess
Readinginputintogetlinefromapipeisaonewayoperation.Thecommandthatisstartedwithcommand|getline
onlysendsdatatoyourawkprogram.
Onoccasion,youmightwanttosenddatatoanotherprogramforprocessingandthenreadtheresultsback.gawkallows
youtostartacoprocess,withwhichtwowaycommunicationsarepossible.Thisisdonewiththe|&operator.
Typically,youwritedatatothecoprocessfirstandthenreadtheresultsback,asshowninthefollowing:
print"somequery"|&"db_server"
"db_server"|&getline

whichsendsaquerytodb_serverandthenreadstheresults.
ThevaluesofNRandFNRarenotchanged,becausethemaininputstreamisnotused.However,therecordissplitinto
fieldsinthenormalmanner,thuschangingthevaluesof$0,oftheotherfields,andofNFandRT.
Coprocessesareanadvancedfeature.Theyarediscussedhereonlybecausethisisthesectionongetline.SeeTwoway
I/O,wherecoprocessesarediscussedinmoredetail.
Next:GetlineNotes,Previous:Getline/Coprocess,Up:Getline[Contents][Index]
4.9.8UsinggetlineintoaVariablefromaCoprocess
Whenyouusecommand|&getlinevar,theoutputfromthecoprocesscommandissentthroughatwowaypipeto
getlineandintothevariablevar.
Inthisversionofgetline,noneofthepredefinedvariablesarechangedandtherecordisnotsplitintofields.Theonly
variablechangedisvar.However,RTisset.
Next:GetlineSummary,Previous:Getline/Variable/Coprocess,Up:Getline[Contents][Index]
4.9.9PointstoRememberAboutgetline
Herearesomemiscellaneouspointsaboutgetlinethatyoushouldbearinmind:
https://www.gnu.org/software/gawk/manual/gawk.html

82/479

29/7/2016

TheGNUAwkUsersGuide

Whengetlinechangesthevalueof$0andNF,awkdoesnotautomaticallyjumptothestartoftheprogramand
starttestingthenewrecordagainsteverypattern.However,thenewrecordistestedagainstanysubsequentrules.
Someveryoldawkimplementationslimitthenumberofpipelinesthatanawkprogrammayhaveopentojustone.
Ingawk,thereisnosuchlimit.Youcanopenasmanypipelines(andcoprocesses)astheunderlyingoperating
systempermits.
AninterestingsideeffectoccursifyouusegetlinewithoutaredirectioninsideaBEGINrule.Becausean
unredirectedgetlinereadsfromthecommandlinedatafiles,thefirstgetlinecommandcausesawktosetthe
valueofFILENAME.Normally,FILENAMEdoesnothaveavalueinsideBEGINrules,becauseyouhavenotyetstarted
toprocessthecommandlinedatafiles.(d.c.)(SeeBEGIN/ENDalsoseeAutoset.)
UsingFILENAMEwithgetline(getline<FILENAME)islikelytobeasourceofconfusion.awkopensaseparate
inputstreamfromthecurrentinputfile.However,bynotusingavariable,$0andNFarestillupdated.Ifyoure
doingthis,itsprobablybyaccident,andyoushouldreconsiderwhatitisyouretryingtoaccomplish.
GetlineSummary,presentsatablesummarizingthegetlinevariantsandwhichvariablestheycanaffect.Itis
worthnotingthatthosevariantsthatdonotuseredirectioncancauseFILENAMEtobeupdatediftheycauseawkto
startreadinganewinputfile.
Ifthevariablebeingassignedisanexpressionwithsideeffects,differentversionsofawkbehavedifferentlyupon
encounteringendoffile.Someversionsdontevaluatetheexpressionmanyversions(includinggawk)do.Hereis
anexample,courtesyofDuncanMoore:
BEGIN{
system("echo1>f")
while((getlinea[++c]<"f")>0){}
printc
}

Here,thesideeffectisthe++c.Iscincrementedifendoffileisencounteredbeforetheelementinaisassigned?
gawktreatsgetlinelikeafunctioncall,andevaluatestheexpressiona[++c]beforeattemptingtoreadfromf.
However,someversionsofawkonlyevaluatetheexpressiononcetheyknowthatthereisastringvaluetobe

assigned.
Previous:GetlineNotes,Up:Getline[Contents][Index]
4.9.10SummaryofgetlineVariants
Table4.1summarizestheeightvariantsofgetline,listingwhichpredefinedvariablesaresetbyeachone,andwhether
thevariantisstandardoragawkextension.Note:foreachvariant,gawksetstheRTpredefinedvariable.
Variant

Effect
Sets$0,NF,FNR,NR,andRT
getlinevar
Setsvar,FNR,NR,andRT
getline<file
Sets$0,NF,andRT
getlinevar<file
SetsvarandRT
command|getline
Sets$0,NF,andRT
command|getlinevar SetsvarandRT
command|&getline
Sets$0,NF,andRT
command|&getlinevar SetsvarandRT
getline

awk/gawk
awk
awk
awk
awk
awk
awk
gawk
gawk

Table4.1:getlinevariantsandwhattheyset
Next:Commandlinedirectories,Previous:Getline,Up:ReadingFiles[Contents][Index]

4.10ReadingInputwithaTimeout
https://www.gnu.org/software/gawk/manual/gawk.html

83/479

29/7/2016

TheGNUAwkUsersGuide

Thissectiondescribesafeaturethatisspecifictogawk.
Youmayspecifyatimeoutinmillisecondsforreadinginputfromthekeyboard,apipe,ortwowaycommunication,
includingTCP/IPsockets.Thiscanbedoneonaperinput,percommand,orperconnectionbasis,bysettingaspecial
elementinthePROCINFOarray(seeAutoset):
PROCINFO["input_name","READ_TIMEOUT"]=timeoutinmilliseconds

Whenset,thiscausesgawktotimeoutandreturnfailureifnodataisavailabletoreadwithinthespecifiedtimeout
period.Forexample,aTCPclientcandecidetogiveuponreceivinganyresponsefromtheserverafteracertain
amountoftime:
Service="/inet/tcp/0/localhost/daytime"
PROCINFO[Service,"READ_TIMEOUT"]=100
if((Service|&getline)>0)
print$0
elseif(ERRNO!="")
printERRNO

Hereishowtoreadinteractivelyfromtheuser27withoutwaitingformorethanfiveseconds:
PROCINFO["/dev/stdin","READ_TIMEOUT"]=5000
while((getline<"/dev/stdin")>0)
print$0
gawkterminatesthereadoperationifinputdoesnotarriveafterwaitingforthetimeoutperiod,returnsfailure,andsets
ERRNOtoanappropriatestringvalue.Anegativeorzerovalueforthetimeoutisthesameasspecifyingnotimeoutatall.

Atimeoutcanalsobesetforreadingfromthekeyboardintheimplicitloopthatreadsinputrecordsandmatchesthem
againstpatterns,likeso:
$gawk'BEGIN{PROCINFO["","READ_TIMEOUT"]=5000}
>{print"Youentered:"$0}'
gawk
|Youentered:gawk

Inthiscase,failuretorespondwithinfivesecondsresultsinthefollowingerrormessage:
errorgawk:cmd.line:2:(FILENAME=FNR=1)fatal:errorreadinginputfile`':Connectiontimedout

Thetimeoutcanbesetorchangedatanytime,andwilltakeeffectonthenextattempttoreadfromtheinputdevice.In
thefollowingexample,westartwithatimeoutvalueofonesecond,andprogressivelyreduceitbyonetenthofa
seconduntilwewaitindefinitelyfortheinputtoarrive:
PROCINFO[Service,"READ_TIMEOUT"]=1000
while((Service|&getline)>0){
print$0
PROCINFO[Service,"READ_TIMEOUT"]=100
}

NOTE:Youshouldnotassumethatthereadoperationwillblockexactlyafterthetenthrecordhasbeen
printed.Itispossiblethatgawkwillreadandbuffermorethanonerecordsworthofdatathefirsttime.
Becauseofthis,changingthevalueoftimeoutlikeintheprecedingexampleisnotveryuseful.
IfthePROCINFOelementisnotpresentandtheGAWK_READ_TIMEOUTenvironmentvariableexists,gawkusesitsvalueto
initializethetimeoutvalue.Theexclusiveuseoftheenvironmentvariabletospecifytimeouthasthedisadvantageof
notbeingabletocontrolitonapercommandorperconnectionbasis.
gawkconsidersatimeouteventtobeanerroreventhoughtheattempttoreadfromtheunderlyingdevicemaysucceedin

alaterattempt.Thisisalimitation,anditalsomeansthatyoucannotusethistomultiplexinputfromtwoormore
sources.
https://www.gnu.org/software/gawk/manual/gawk.html

84/479

29/7/2016

TheGNUAwkUsersGuide

Assigningatimeoutvaluepreventsreadoperationsfromblockingindefinitely.Butbearinmindthatthereareother
waysgawkcanstallwaitingforaninputdevicetobeready.Anetworkclientcansometimestakealongtimetoestablish
aconnectionbeforeitcanstartreadinganydata,ortheattempttoopenaFIFOspecialfileforreadingcanblock
indefinitelyuntilsomeotherprocessopensitforwriting.
Next:InputSummary,Previous:ReadTimeout,Up:ReadingFiles[Contents][Index]

4.11DirectoriesontheCommandLine
AccordingtothePOSIXstandard,filesnamedontheawkcommandlinemustbetextfilesitisafatalerroriftheyare
not.Mostversionsofawktreatadirectoryonthecommandlineasafatalerror.
Bydefault,gawkproducesawarningforadirectoryonthecommandline,butotherwiseignoresit.Thismakesiteasier
touseshellwildcardswithyourawkprogram:
$gawkfwhizprog.awk*Directoriescouldkillthisprogram

Ifeitheroftheposixortraditionaloptionsisgiven,thengawkrevertstotreatingadirectoryonthecommandline
asafatalerror.
SeeExtensionSampleReaddir,forawaytotreatdirectoriesasusabledatafromanawkprogram.
Next:InputExercises,Previous:Commandlinedirectories,Up:ReadingFiles[Contents][Index]

4.12Summary
InputissplitintorecordsbasedonthevalueofRS.Thepossibilitiesareasfollows:
ValueofRS
Recordsarespliton
awk/gawk
awk
Anysinglecharacter Thatcharacter
Theemptystring("") Runsoftwoormorenewlines awk
Aregexp
Textthatmatchestheregexp gawk
FNRindicateshowmanyrecordshavebeenreadfromthecurrentinputfileNRindicateshowmanyrecordshave
beenreadintotal.
gawksetsRTtothetextmatchedbyRS.
Aftersplittingtheinputintorecords,awkfurthersplitstherecordsintoindividualfields,named$1,$2,andsoon.
$0isthewholerecord,andNFindicateshowmanyfieldsthereare.Thedefaultwaytosplitfieldsisbetween
whitespacecharacters.
Fieldsmaybereferencedusingavariable,asin$NF.Fieldsmayalsobeassignedvalues,whichcausesthevalue
of$0toberecomputedwhenitislaterreferenced.AssigningtoafieldwithanumbergreaterthanNFcreatesthe
fieldandrebuildstherecord,usingOFStoseparatethefields.IncrementingNFdoesthesamething.Decrementing
NFthrowsawayfieldsandrebuildstherecord.
Fieldsplittingismorecomplicatedthanrecordsplitting:
Fieldseparatorvalue
Fieldsaresplit
awk/gawk
FS==""
awk
Onrunsofwhitespace
FS==anysinglecharacter
awk
Onthatcharacter
FS==regexp
awk
Ontextmatchingtheregexp
Suchthateachindividualcharacterisa
FS==""
gawk
separatefield
FIELDWIDTHS==listofcolumns
gawk
Basedoncharacterposition
Onthetextsurroundingtextmatchingthe
FPAT==regexp
gawk
regexp
UsingFS="\n"causestheentirerecordtobeasinglefield(assumingthatnewlinesseparaterecords).
FSmaybesetfromthecommandlineusingtheFoption.Thiscanalsobedoneusingcommandlinevariable
https://www.gnu.org/software/gawk/manual/gawk.html

85/479

29/7/2016

TheGNUAwkUsersGuide

assignment.
UsePROCINFO["FS"]toseehowfieldsarebeingsplit.
Usegetlineinitsvariousformstoreadadditionalrecordsfromthedefaultinputstream,fromafile,orfroma
pipeorcoprocess.
UsePROCINFO[file,"READ_TIMEOUT"]tocausereadstotimeoutforfile.
DirectoriesonthecommandlinearefatalforstandardawkgawkignoresthemifnotinPOSIXmode.
Previous:InputSummary,Up:ReadingFiles[Contents][Index]

4.13Exercises
1.UsingtheFIELDWIDTHSvariable(seeConstantSize),writeaprogramtoreadelectiondata,whereeachrecord
representsonevotersvotes.Comeupwithawaytodefinewhichcolumnsareassociatedwitheachballotitem,
andprintthetotalvotes,includingabstentions,foreachitem.
2.PlainGetline,presentedaprogramtoremoveCstylecomments(/**/)fromtheinput.Thatprogramdoes
notworkifonecommentendsononelineandanotheronestartslateronthesameline.Thatcanbefixedby
makingonesimplechange.Whatisit?
Next:Expressions,Previous:ReadingFiles,Up:Top[Contents][Index]

5PrintingOutput
Oneofthemostcommonprogrammingactionsistoprint,oroutput,someoralloftheinput.Usetheprintstatement
forsimpleoutput,andtheprintfstatementforfancierformatting.Theprintstatementisnotlimitedwhencomputing
whichvaluestoprint.However,withtwoexceptions,youcannotspecifyhowtoprintthemhowmanycolumns,
whethertouseexponentialnotationornot,andsoon.(Fortheexceptions,seeOutputSeparators,andOFMT.)For
printingwithspecifications,youneedtheprintfstatement(seePrintf).
Besidesbasicandformattedprinting,thischapteralsocoversI/Oredirectionstofilesandpipes,introducesthespecial
filenamesthatgawkprocessesinternally,anddiscussestheclose()builtinfunction.
Print:
Theprintstatement.
PrintExamples:
Simpleexamplesofprintstatements.
OutputSeparators:
Theoutputseparatorsandhowtochangethem.
OFMT:
ControllingNumericOutputWithprint.
Printf:
Theprintfstatement.
Redirection:
Howtoredirectoutputtomultiplefilesandpipes.
SpecialFD:
SpecialfilesforI/O.
SpecialFiles:
Filenameinterpretationingawk.gawkallowsaccesstoinheritedfiledescriptors.
CloseFilesAndPipes: ClosingInputandOutputFilesandPipes.
OutputSummary:
Outputsummary.
OutputExercises:
Exercises.
Next:PrintExamples,Up:Printing[Contents][Index]

5.1TheprintStatement
Usetheprintstatementtoproduceoutputwithsimple,standardizedformatting.Youspecifyonlythestringsor
numberstoprint,inalistseparatedbycommas.Theyareoutput,separatedbysinglespaces,followedbyanewline.The
statementlookslikethis:
printitem1,item2,

https://www.gnu.org/software/gawk/manual/gawk.html

86/479

29/7/2016

TheGNUAwkUsersGuide

Theentirelistofitemsmaybeoptionallyenclosedinparentheses.Theparenthesesarenecessaryifanyoftheitem
expressionsusesthe>relationaloperatorotherwiseitcouldbeconfusedwithanoutputredirection(seeRedirection).
Theitemstoprintcanbeconstantstringsornumbers,fieldsofthecurrentrecord(suchas$1),variables,oranyawk
expression.Numericvaluesareconvertedtostringsandthenprinted.
Thesimplestatementprintwithnoitemsisequivalenttoprint$0:itprintstheentirecurrentrecord.Toprinta
blankline,useprint"".Toprintafixedpieceoftext,useastringconstant,suchas"Don'tPanic",asoneitem.Ifyou
forgettousethedoublequotecharacters,yourtextistakenasanawkexpression,andyouwillprobablygetanerror.
Keepinmindthataspaceisprintedbetweenanytwoitems.
Notethattheprintstatementisastatementandnotanexpressionyoucantuseitinthepatternpartofapattern
actionstatement,forexample.
Next:OutputSeparators,Previous:Print,Up:Printing[Contents][Index]

5.2printStatementExamples
Eachprintstatementmakesatleastonelineofoutput.However,itisntlimitedtoonlyoneline.Ifanitemvalueisa
stringcontaininganewline,thenewlineisoutputalongwiththerestofthestring.Asingle printstatementcanmake
anynumberoflinesthisway.
Thefollowingisanexampleofprintingastringthatcontainsembeddednewlines(the\nisanescapesequence,used
torepresentthenewlinecharacterseeEscapeSequences):
$awk'BEGIN{print"lineone\nlinetwo\nlinethree"}'
|lineone
|linetwo
|linethree

Thenextexample,whichisrunontheinventoryshippedfile,printsthefirsttwofieldsofeachinputrecord,witha
spacebetweenthem:
$awk'{print$1,$2}'inventoryshipped
|Jan13
|Feb15
|Mar15

Acommonmistakeinusingtheprintstatementistoomitthecommabetweentwoitems.Thisoftenhastheeffectof
makingtheitemsruntogetherintheoutput,withnospace.Thereasonforthisisthatjuxtaposingtwostringexpressions
inawkmeanstoconcatenatethem.Hereisthesameprogram,withoutthecomma:
$awk'{print$1$2}'inventoryshipped
|Jan13
|Feb15
|Mar15

Tosomeoneunfamiliarwiththeinventoryshippedfile,neitherexamplesoutputmakesmuchsense.Aheadinglineat
thebeginningwouldmakeitclearer.Letsaddsomeheadingstoourtableofmonths($1)andgreencratesshipped($2).
WedothisusingaBEGINrule(seeBEGIN/END)sothattheheadingsareonlyprintedonce:
awk'BEGIN{print"MonthCrates"
print""}
{print$1,$2}'inventoryshipped

Whenrun,theprogramprintsthefollowing:
MonthCrates

https://www.gnu.org/software/gawk/manual/gawk.html

87/479

29/7/2016

TheGNUAwkUsersGuide

Jan13
Feb15
Mar15

Theonlyproblem,however,isthattheheadingsandthetabledatadontlineup!Wecanfixthisbyprintingsome
spacesbetweenthetwofields:
awk'BEGIN{print"MonthCrates"
print""}
{print$1,"",$2}'inventoryshipped

Liningupcolumnsthiswaycangetprettycomplicatedwhentherearemanycolumnstofix.Countingspacesfortwoor
threecolumnsissimple,butanymorethanthiscantakeupalotoftime.Thisiswhytheprintfstatementwascreated
(seePrintf)oneofitsspecialtiesisliningupcolumnsofdata.
NOTE:Youcancontinueeitheraprintorprintfstatementsimplybyputtinganewlineafteranycomma
(seeStatements/Lines).
Next:OFMT,Previous:PrintExamples,Up:Printing[Contents][Index]

5.3OutputSeparators
Asmentionedpreviously,aprintstatementcontainsalistofitemsseparatedbycommas.Intheoutput,theitemsare
normallyseparatedbysinglespaces.However,thisdoesntneedtobethecaseasinglespaceissimplythedefault.Any
stringofcharactersmaybeusedastheoutputfieldseparatorbysettingthepredefinedvariableOFS.Theinitialvalueof
thisvariableisthestring""(i.e.,asinglespace).
Theoutputfromanentireprintstatementiscalledanoutputrecord.Eachprintstatementoutputsoneoutputrecord,
andthenoutputsastringcalledtheoutputrecordseparator(orORS).TheinitialvalueofORSisthestring"\n"(i.e.,a
newlinecharacter).Thus,eachprintstatementnormallymakesaseparateline.
Inordertochangehowoutputfieldsandrecordsareseparated,assignnewvaluestothevariablesOFSandORS.The
usualplacetodothisisintheBEGINrule(seeBEGIN/END),sothatithappensbeforeanyinputisprocessed.Itcanalso
bedonewithassignmentsonthecommandline,beforethenamesoftheinputfiles,orusingthevcommandline
option(seeOptions).Thefollowingexampleprintsthefirstandsecondfieldsofeachinputrecord,separatedbya
semicolon,withablanklineaddedaftereachnewline:
$awk'BEGIN{OFS=";";ORS="\n\n"}
>{print$1,$2}'maillist
|Amelia;5555553
|
|Anthony;5553412
|
|Becky;5557685
|
|Bill;5551675
|
|Broderick;5550542
|
|Camilla;5552912
|
|Fabius;5551234
|
|Julie;5556699
|
|Martin;5556480
|
|Samuel;5553430
|
|JeanPaul;5552127
|
https://www.gnu.org/software/gawk/manual/gawk.html

88/479

29/7/2016

TheGNUAwkUsersGuide

IfthevalueofORSdoesnotcontainanewline,theprogramsoutputrunstogetheronasingleline.
Next:Printf,Previous:OutputSeparators,Up:Printing[Contents][Index]

5.4ControllingNumericOutputwithprint
Whenprintingnumericvalueswiththeprintstatement,awkinternallyconvertseachnumbertoastringofcharacters
andprintsthatstring.awkusesthesprintf()functiontodothisconversion(seeStringFunctions).Fornow,itsuffices
tosaythatthesprintf()functionacceptsaformatspecificationthattellsithowtoformatnumbers(orstrings),andthat
thereareanumberofdifferentwaysinwhichnumberscanbeformatted.Thedifferentformatspecificationsare
discussedmorefullyinControlLetters.
ThepredefinedvariableOFMTcontainstheformatspecificationthatprintuseswithsprintf()whenitwantstoconverta
numbertoastringforprinting.ThedefaultvalueofOFMTis"%.6g".Thewayprintprintsnumberscanbechangedby
supplyingadifferentformatspecificationforthevalueofOFMT,asshowninthefollowingexample:
$awk'BEGIN{
>OFMT="%.0f"#printnumbersasintegers(rounds)
>print17.23,17.54}'
|1718

AccordingtothePOSIXstandard,awksbehaviorisundefinedifOFMTcontainsanythingbutafloatingpointconversion
specification.(d.c.)
Next:Redirection,Previous:OFMT,Up:Printing[Contents][Index]

5.5UsingprintfStatementsforFancierPrinting
Formoreprecisecontrolovertheoutputformatthanwhatisprovidedbyprint,useprintf.Withprintfyoucan
specifythewidthtouseforeachitem,aswellasvariousformattingchoicesfornumbers(suchaswhatoutputbaseto
use,whethertoprintanexponent,whethertoprintasign,andhowmanydigitstoprintafterthedecimalpoint).
BasicPrintf:
Syntaxoftheprintfstatement.
ControlLetters: Formatcontrolletters.
FormatModifiers: Formatspecificationmodifiers.
PrintfExamples: Severalexamples.
Next:ControlLetters,Up:Printf[Contents][Index]
5.5.1IntroductiontotheprintfStatement
Asimpleprintfstatementlookslikethis:
printfformat,item1,item2,

Asforprint,theentirelistofargumentsmayoptionallybeenclosedinparentheses.Heretoo,theparenthesesare
necessaryifanyoftheitemexpressionsusesthe>relationaloperatorotherwise,itcanbeconfusedwithanoutput
redirection(seeRedirection).
Thedifferencebetweenprintfandprintistheformatargument.Thisisanexpressionwhosevalueistakenasastring
itspecifieshowtooutputeachoftheotherarguments.Itiscalledtheformatstring.
TheformatstringisverysimilartothatintheISOClibraryfunctionprintf().Mostofformatistexttooutput
verbatim.Scatteredamongthistextareformatspecifiersoneperitem.Eachformatspecifiersaystooutputthenext
itemintheargumentlistatthatplaceintheformat.
https://www.gnu.org/software/gawk/manual/gawk.html

89/479

29/7/2016

TheGNUAwkUsersGuide

Theprintfstatementdoesnotautomaticallyappendanewlinetoitsoutput.Itoutputsonlywhattheformatstring
specifies.Soifanewlineisneeded,youmustincludeoneintheformatstring.TheoutputseparatorvariablesOFSand
ORShavenoeffectonprintfstatements.Forexample:
$awk'BEGIN{
>ORS="\nOUCH!\n";OFS="+"
>msg="Don\47tPanic!"
>printf"%s\n",msg
>}'
|Don'tPanic!

Here,neitherthe+northeOUCH!appearsintheoutputmessage.
Next:FormatModifiers,Previous:BasicPrintf,Up:Printf[Contents][Index]
5.5.2FormatControlLetters
Aformatspecifierstartswiththecharacter%andendswithaformatcontrolletterittellstheprintfstatementhow
tooutputoneitem.Theformatcontrolletterspecifieswhatkindofvaluetoprint.Therestoftheformatspecifieris
madeupofoptionalmodifiersthatcontrolhowtoprintthevalue,suchasthefieldwidth.Hereisalistoftheformat
controlletters:
%c

Printanumberasacharacterthus,printf"%c",65outputstheletterA.Theoutputforastringvalueisthe
firstcharacterofthestring.
NOTE:ThePOSIXstandardsaysthefirstcharacterofastringisprinted.Inlocaleswithmultibyte
characters,gawkattemptstoconverttheleadingbytesofthestringintoavalidwidecharacterand
thentoprintthemultibyteencodingofthatcharacter.Similarly,whenprintinganumericvalue,gawk
allowsthevaluetobewithinthenumericrangeofvaluesthatcanbeheldinawidecharacter.Ifthe
conversiontomultibyteencodingfails,gawkusestheloweightbitsofthevalueasthecharacterto
print.
Otherawkversionsgenerallyrestrictthemselvestoprintingthefirstbyteofastringortonumeric
valueswithintherangeofasinglebyte(0255).
%d,%i

Printadecimalinteger.Thetwocontrollettersareequivalent.(The%ispecificationisforcompatibilitywith
ISOC.)
%e,%E

Printanumberinscientific(exponential)notation.Forexample:
printf"%4.3e\n",1950

prints1.950e+03,withatotaloffoursignificantfigures,threeofwhichfollowthedecimalpoint.(The4.3
representstwomodifiers,discussedinthenextsubsection.)%EusesEinsteadofeintheoutput.
%f

Printanumberinfloatingpointnotation.Forexample:
printf"%4.3f",1950

prints1950.000,withatotaloffoursignificantfigures,threeofwhichfollowthedecimalpoint.(The4.3
representstwomodifiers,discussedinthenextsubsection.)
https://www.gnu.org/software/gawk/manual/gawk.html

90/479

29/7/2016

TheGNUAwkUsersGuide

OnsystemssupportingIEEE754floatingpointformat,valuesrepresentingnegativeinfinityareformattedas
inforinfinity,andpositiveinfinityasinforinfinity.Thespecialnotanumbervalueformatsas
nanornan(seeMathDefinitions).
%F

Like%f,buttheinfinityandnotanumbervaluesarespelledusinguppercaseletters.
The%FformatisaPOSIXextensiontoISOCnotallsystemssupportit.Onthosethatdont,gawkuses%f
instead.
%g,%G

Printanumberineitherscientificnotationorinfloatingpointnotation,whicheverusesfewercharactersifthe
resultisprintedinscientificnotation,%GusesEinsteadofe.
%o

Printanunsignedoctalinteger(seeNondecimalnumbers).
%s

Printastring.
%u

Printanunsigneddecimalinteger.(Thisformatisofmarginaluse,becauseallnumbersinawkarefloatingpointit
isprovidedprimarilyforcompatibilitywithC.)
%x,%X

Printanunsignedhexadecimalinteger%XusesthelettersAthroughFinsteadof athroughf(see
Nondecimalnumbers).
%%

Printasingle%.Thisdoesnotconsumeanargumentanditignoresanymodifiers.
NOTE:WhenusingtheintegerformatcontrollettersforvaluesthatareoutsidetherangeofthewidestC
integertype,gawkswitchestothe%gformatspecifier.Iflintisprovidedonthecommandline(see
Options),gawkwarnsaboutthis.Otherversionsofawkmayprintinvalidvaluesordosomethingelse
entirely.(d.c.)
Next:PrintfExamples,Previous:ControlLetters,Up:Printf[Contents][Index]
5.5.3ModifiersforprintfFormats
Aformatspecificationcanalsoincludemodifiersthatcancontrolhowmuchoftheitemsvalueisprinted,aswellas
howmuchspaceitgets.Themodifierscomebetweenthe%andtheformatcontrolletter.Weusethebulletsymbol
inthefollowingexamplestorepresentspacesintheoutput.Herearethepossiblemodifiers,intheorderinwhichthey
mayappear:
N$

Anintegerconstantfollowedbya$isapositionalspecifier.Normally,formatspecificationsareappliedto
argumentsintheordergivenintheformatstring.Withapositionalspecifier,theformatspecificationisappliedto
aspecificargument,insteadofwhatwouldbethenextargumentinthelist.Positionalspecifiersbegincounting
withone.Thus:
https://www.gnu.org/software/gawk/manual/gawk.html

91/479

29/7/2016

TheGNUAwkUsersGuide

printf"%s%s\n","don't","panic"
printf"%2$s%1$s\n","panic","don't"

printsthefamousfriendlymessagetwice.
Atfirstglance,thisfeaturedoesntseemtobeofmuchuse.Itisinfactagawkextension,intendedforusein
translatingmessagesatruntime.SeePrintfOrdering,whichdescribeshowandwhytousepositionalspecifiers.
Fornow,weignorethem.
(Minus)

Theminussign,usedbeforethewidthmodifier(seelateroninthislist),saystoleftjustifytheargumentwithin
itsspecifiedwidth.Normally,theargumentisprintedrightjustifiedinthespecifiedwidth.Thus:
printf"%4s","foo"

printsfoo.
space
Fornumericconversions,prefixpositivevalueswithaspaceandnegativevalueswithaminussign.
+

Theplussign,usedbeforethewidthmodifier(seelateroninthislist),saystoalwayssupplyasignfornumeric
conversions,evenifthedatatoformatispositive.The+overridesthespacemodifier.
#

Useanalternativeformforcertaincontrolletters.For%o,supplyaleadingzero.For%xand%X,supplya
leading0xor0Xforanonzeroresult.For%e,%E,%f,and%F,theresultalwayscontainsadecimalpoint.
For%gand%G,trailingzerosarenotremovedfromtheresult.
0

Aleading0(zero)actsasaflagindicatingthatoutputshouldbepaddedwithzerosinsteadofspaces.This
appliesonlytothenumericoutputformats.Thisflagonlyhasaneffectwhenthefieldwidthiswiderthanthe
valuetoprint.
'

AsinglequoteorapostrophecharacterisaPOSIXextensiontoISOC.Itindicatesthattheintegerpartofa
floatingpointvalue,ortheentirepartofanintegerdecimalvalue,shouldhaveathousandsseparatorcharacterin
it.Thisonlyworksinlocalesthatsupportsuchcharacters.Forexample:
$catthousands.awkShowsourceprogram
|BEGIN{printf"%'d\n",1234567}
$LC_ALL=Cgawkfthousands.awk
|1234567Resultsin"C"locale
$LC_ALL=en_US.UTF8gawkfthousands.awk
|1,234,567ResultsinUSEnglishUTFlocale

Formoreinformationaboutlocalesandinternationalizationissues,seeLocales.
NOTE:The'flagisanicefeature,butitsusecomplicatesthings:itbecomesdifficulttouseitin
commandlineprograms.Forinformationonappropriatequotingtricks,seeQuoting.
width
Thisisanumberspecifyingthedesiredminimumwidthofafield.Insertinganynumberbetweenthe%signand
theformatcontrolcharacterforcesthefieldtoexpandtothiswidth.Thedefaultwaytodothisistopadwith
spacesontheleft.Forexample:
https://www.gnu.org/software/gawk/manual/gawk.html

92/479

29/7/2016

TheGNUAwkUsersGuide

printf"%4s","foo"

printsfoo.
Thevalueofwidthisaminimumwidth,notamaximum.Iftheitemvaluerequiresmorethanwidthcharacters,it
canbeaswideasnecessary.Thus,thefollowing:
printf"%4s","foobar"

printsfoobar.
Precedingthewidthwithaminussigncausestheoutputtobepaddedwithspacesontheright,insteadofonthe
left.
.prec

Aperiodfollowedbyanintegerconstantspecifiestheprecisiontousewhenprinting.Themeaningofthe
precisionvariesbycontrolletter:
%d,%i,%o,%u,%x,%X

Minimumnumberofdigitstoprint.
%e,%E,%f,%F

Numberofdigitstotherightofthedecimalpoint.
%g,%G

Maximumnumberofsignificantdigits.
%s

Maximumnumberofcharactersfromthestringthatshouldprint.
Thus,thefollowing:
printf"%.4s","foobar"

printsfoob.
TheClibraryprintfsdynamicwidthandpreccapability(e.g.,"%*.*s")issupported.Insteadofsupplyingexplicit
widthand/orprecvaluesintheformatstring,theyarepassedintheargumentlist.Forexample:
w=5
p=3
s="abcdefg"
printf"%*.*s\n",w,p,s

isexactlyequivalentto:
s="abcdefg"
printf"%5.3s\n",s

Bothprogramsoutputabc.Earlierversionsofawkdidnotsupportthiscapability.Ifyoumustusesuchaversion,you
maysimulatethisfeaturebyusingconcatenationtobuilduptheformatstring,likeso:
w=5
p=3
s="abcdefg"
printf"%"w"."p"s\n",s

Thisisnotparticularlyeasytoread,butitdoeswork.
https://www.gnu.org/software/gawk/manual/gawk.html

93/479

29/7/2016

TheGNUAwkUsersGuide

Cprogrammersmaybeusedtosupplyingadditionalmodifiers(h,j,l,L,t,andz)inprintfformatstrings.
Thesearenotvalidinawk.Mostawkimplementationssilentlyignorethem.Iflintisprovidedonthecommandline
(seeOptions),gawkwarnsabouttheiruse.Ifposixissupplied,theiruseisafatalerror.
Previous:FormatModifiers,Up:Printf[Contents][Index]
5.5.4ExamplesUsingprintf
Thefollowingsimpleexampleshowshowtouseprintftomakeanalignedtable:
awk'{printf"%10s%s\n",$1,$2}'maillist

Thiscommandprintsthenamesofthepeople($1)inthefilemaillistasastringof10charactersthatareleftjustified.
Italsoprintsthephonenumbers($2)nextontheline.Thisproducesanalignedtwocolumntableofnamesandphone
numbers,asshownhere:
$awk'{printf"%10s%s\n",$1,$2}'maillist
|Amelia5555553
|Anthony5553412
|Becky5557685
|Bill5551675
|Broderick5550542
|Camilla5552912
|Fabius5551234
|Julie5556699
|Martin5556480
|Samuel5553430
|JeanPaul5552127

Inthiscase,thephonenumbershadtobeprintedasstringsbecausethenumbersareseparatedbydashes.Printingthe
phonenumbersasnumberswouldhaveproducedjustthefirstthreedigits:555.Thiswouldhavebeenpretty
confusing.
Itwasntnecessarytospecifyawidthforthephonenumbersbecausetheyarelastontheirlines.Theydontneedto
havespacesafterthem.
Thetablecouldbemadetolookevennicerbyaddingheadingstothetopsofthecolumns.ThisisdoneusingaBEGIN
rule(seeBEGIN/END)sothattheheadersareonlyprintedonce,atthebeginningoftheawkprogram:
awk'BEGIN{print"NameNumber"
print""}
{printf"%10s%s\n",$1,$2}'maillist

Theprecedingexamplemixesprintandprintfstatementsinthesameprogram.Usingjustprintfstatementscan
producethesameresults:
awk'BEGIN{printf"%10s%s\n","Name","Number"
printf"%10s%s\n","",""}
{printf"%10s%s\n",$1,$2}'maillist

Printingeachcolumnheadingwiththesameformatspecificationusedforthecolumnelementsensuresthatthe
headingsarealignedjustlikethecolumns.
Thefactthatthesameformatspecificationisusedthreetimescanbeemphasizedbystoringitinavariable,likethis:
awk'BEGIN{format="%10s%s\n"
printfformat,"Name","Number"
printfformat,"",""}
{printfformat,$1,$2}'maillist

Next:SpecialFD,Previous:Printf,Up:Printing[Contents][Index]
https://www.gnu.org/software/gawk/manual/gawk.html

94/479

29/7/2016

TheGNUAwkUsersGuide

5.6RedirectingOutputofprintandprintf
Sofar,theoutputfromprintandprintfhasgonetothestandardoutput,usuallythescreen.Bothprintandprintfcan
alsosendtheiroutputtootherplaces.Thisiscalledredirection.
NOTE:Whensandboxisspecified(seeOptions),redirectingoutputtofiles,pipes,andcoprocessesis
disabled.
Aredirectionappearsaftertheprintorprintfstatement.Redirectionsinawkarewrittenjustlikeredirectionsinshell
commands,exceptthattheyarewritteninsidetheawkprogram.
Therearefourformsofoutputredirection:outputtoafile,outputappendedtoafile,outputthroughapipetoanother
command,andoutputtoacoprocess.Weshowthemallfortheprintstatement,buttheyworkidenticallyforprintf:
printitems>outputfile

Thisredirectionprintstheitemsintotheoutputfilenamedoutputfile.Thefilenameoutputfilecanbeany
expression.Itsvalueischangedtoastringandthenusedasafilename(seeExpressions).
Whenthistypeofredirectionisused,theoutputfileiserasedbeforethefirstoutputiswrittentoit.Subsequent
writestothesameoutputfiledonoteraseoutputfile,butappendtoit.(Thisisdifferentfromhowyouuse
redirectionsinshellscripts.)Ifoutputfiledoesnotexist,itiscreated.Forexample,hereishowanawkprogram
canwritealistofpeoplesnamestoonefilenamednamelist,andalistofphonenumberstoanotherfilenamed
phonelist:
$awk'{print$2>"phonelist"
>print$1>"namelist"}'maillist
$catphonelist
|5555553
|5553412

$catnamelist
|Amelia
|Anthony

Eachoutputfilecontainsonenameornumberperline.
printitems>>outputfile

Thisredirectionprintstheitemsintothepreexistingoutputfilenamedoutputfile.Thedifferencebetweenthis
andthesingle>redirectionisthattheoldcontents(ifany)ofoutputfilearenoterased.Instead,theawkoutput
isappendedtothefile.Ifoutputfiledoesnotexist,thenitiscreated.
printitems|command

Itispossibletosendoutputtoanotherprogramthroughapipeinsteadofintoafile.Thisredirectionopensapipe
tocommand,andwritesthevaluesofitemsthroughthispipetoanotherprocesscreatedtoexecutecommand.
Theredirectionargumentcommandisactuallyanawkexpression.Itsvalueisconvertedtoastringwhosecontents
givetheshellcommandtoberun.Forexample,thefollowingproducestwofiles,oneunsortedlistofpeoples
names,andonelistsortedinreversealphabeticalorder:
awk'{print$1>"names.unsorted"
command="sortr>names.sorted"
print$1|command}'maillist

Theunsortedlistiswrittenwithanordinaryredirection,whilethesortedlistiswrittenbypipingthroughthesort
utility.
Thenextexampleusesredirectiontomailamessagetothemailinglistbugsystem.Thismightbeusefulwhen
troubleisencounteredinanawkscriptrunperiodicallyforsystemmaintenance:
https://www.gnu.org/software/gawk/manual/gawk.html

95/479

29/7/2016

TheGNUAwkUsersGuide

report="mailbugsystem"
print("Awkscriptfailed:",$0)|report
print("atrecordnumber",FNR,"of",FILENAME)|report
close(report)

Theclose()functioniscalledherebecauseitsagoodideatoclosethepipeassoonasalltheintendedoutputhas
beensenttoit.SeeCloseFilesAndPipes,formoreinformation.
Thisexamplealsoillustratestheuseofavariabletorepresentafileorcommanditisnotnecessarytoalways
useastringconstant.Usingavariableisgenerallyagoodidea,because(ifyoumeantorefertothatsamefileor
command)awkrequiresthatthestringvaluebewrittenidenticallyeverytime.
printitems|&command

Thisredirectionprintstheitemstotheinputofcommand.Thedifferencebetweenthisandthesingle|
redirectionisthattheoutputfromcommandcanbereadwithgetline.Thus,commandisacoprocess,which
workstogetherwithbutissubsidiarytotheawkprogram.
Thisfeatureisagawkextension,andisnotavailableinPOSIXawk.SeeGetline/Coprocess,forabriefdiscussion.
SeeTwowayI/O,foramorecompletediscussion.
Redirectingoutputusing>,>>,|,or|&asksthesystemtoopenafile,pipe,orcoprocessonlyiftheparticularfile
orcommandyouspecifyhasnotalreadybeenwrittentobyyourprogramorifithasbeenclosedsinceitwaslast
writtento.
Itisacommonerrortouse>redirectionforthefirstprinttoafile,andthentouse>>forsubsequentoutput:
#clearthefile
print"Don'tpanic">"guide.txt"

#append
print"Avoidimprobabilitygenerators">>"guide.txt"

Thisisindeedhowredirectionsmustbeusedfromtheshell.Butinawk,itisntnecessary.Inthiskindofcase,aprogram
shoulduse>foralltheprintstatements,becausetheoutputfileisonlyopenedonce.(Ithappensthatifyoumix>
and>>outputisproducedintheexpectedorder.However,mixingtheoperatorsforthesamefileisdefinitelypoor
style,andisconfusingtoreadersofyourprogram.)
Asmentionedearlier(seeGetlineNotes),manyManyolderawkimplementationslimitthenumberofpipelinesthatan
awkprogrammayhaveopentojustone!Ingawk,thereisnosuchlimit.gawkallowsaprogramtoopenasmanypipelines
astheunderlyingoperatingsystempermits.
Pipingintosh
Aparticularlypowerfulwaytouseredirectionistobuildcommandlinesandpipethemintotheshell,sh.For
example,supposeyouhavealistoffilesbroughtoverfromasystemwhereallthefilenamesarestoredinuppercase,
andyouwishtorenamethemtohavenamesinalllowercase.Thefollowingprogramisbothsimpleandefficient:
{printf("mv%s%s\n",$0,tolower($0))|"sh"}
END{close("sh")}

Thetolower()functionreturnsitsargumentstringwithalluppercasecharactersconvertedtolowercase(seeString
Functions).Theprogrambuildsupalistofcommandlines,usingthemvutilitytorenamethefiles.Itthensendsthelist
totheshellforexecution.
SeeShellQuoting,forafunctionthatcanhelpingeneratingcommandlinestobefedtotheshell.

Next:SpecialFiles,Previous:Redirection,Up:Printing[Contents][Index]
https://www.gnu.org/software/gawk/manual/gawk.html

96/479

29/7/2016

TheGNUAwkUsersGuide

5.7SpecialFilesforStandardPreopenedDataStreams
Runningprogramsconventionallyhavethreeinputandoutputstreamsalreadyavailabletothemforreadingandwriting.
Theseareknownasthestandardinput,standardoutput,andstandarderroroutput.Theseopenstreams(andanyother
openfilesorpipes)areoftenreferredtobythetechnicaltermfiledescriptors.
Thesestreamsare,bydefault,connectedtoyourkeyboardandscreen,buttheyareoftenredirectedwiththeshell,via
the<,<<,>,>>,>&,and|operators.Standarderroristypicallyusedforwritingerrormessagesthereason
therearetwoseparatestreams,standardoutputandstandarderror,issothattheycanberedirectedseparately.
Intraditionalimplementationsofawk,theonlywaytowriteanerrormessagetostandarderrorinanawkprogramisas
follows:
print"Seriouserrordetected!"|"cat1>&2"

Thisworksbyopeningapipelinetoashellcommandthatcanaccessthestandarderrorstreamthatitinheritsfromthe
awkprocess.Thisisfarfromelegant,anditalsorequiresaseparateprocess.Sopeoplewritingawkprogramsoftendont
dothis.Instead,theysendtheerrormessagestothescreen,likethis:
print"Seriouserrordetected!">"/dev/tty"

(/dev/ttyisaspecialfilesuppliedbytheoperatingsystemthatisconnectedtoyourkeyboardandscreen.Itrepresents
theterminal,28whichonmodernsystemsisakeyboardandscreen,notaserialconsole.)Thisgenerallyhasthesame
effect,butnotalways:althoughthestandarderrorstreamisusuallythescreen,itcanberedirectedwhenthathappens,
writingtothescreenisnotcorrect.Infact,ifawkisrunfromabackgroundjob,itmaynothaveaterminalatall.Then
opening/dev/ttyfails.
gawk,BWKawk,andmawkprovidespecialfilenamesforaccessingthethreestandardstreams.Ifthefilenamematches
oneofthesespecialnameswhengawk(oroneoftheothers)redirectsinputoroutput,thenitdirectlyusesthedescriptor
thatthefilenamestandsfor.Thesespecialfilenamesworkforalloperatingsystemsthatgawkhasbeenportedto,not

justthosethatarePOSIXcompliant:
/dev/stdin

Thestandardinput(filedescriptor0).
/dev/stdout

Thestandardoutput(filedescriptor1).
/dev/stderr

Thestandarderroroutput(filedescriptor2).
Withthesefacilities,theproperwaytowriteanerrormessagethenbecomes:
print"Seriouserrordetected!">"/dev/stderr"

Notetheuseofquotesaroundthefilename.Likewithanyotherredirection,thevaluemustbeastring.Itisacommon
errortoomitthequotes,whichleadstoconfusingresults.
gawkdoesnottreatthesefilenamesasspecialwheninPOSIXcompatibilitymode.However,becauseBWKawk
supportsthem,gawkdoessupportthemevenwheninvokedwiththetraditionaloption(seeOptions).

Next:CloseFilesAndPipes,Previous:SpecialFD,Up:Printing[Contents][Index]

5.8SpecialFilenamesingawk
Besidesaccesstostandardinput,standardoutput,andstandarderror,gawkprovidesaccesstoanyopenfiledescriptor.
https://www.gnu.org/software/gawk/manual/gawk.html

97/479

29/7/2016

TheGNUAwkUsersGuide

Additionally,therearespecialfilenamesreservedforTCP/IPnetworking.
OtherInheritedFiles: Accessingotheropenfileswithgawk.
SpecialNetwork:
Specialfilesfornetworkcommunications.
SpecialCaveats:
Thingstowatchoutfor.
Next:SpecialNetwork,Up:SpecialFiles[Contents][Index]
5.8.1AccessingOtherOpenFileswithgawk
Besidesthe/dev/stdin,/dev/stdout,and/dev/stderrspecialfilenamesmentionedearlier,gawkprovidessyntaxfor
accessinganyotherinheritedopenfile:
/dev/fd/N

ThefileassociatedwithfiledescriptorN.Suchafilemustbeopenedbytheprograminitiatingtheawkexecution
(typicallytheshell).Unlessspecialpainsaretakenintheshellfromwhichgawkisinvoked,onlydescriptors0,1,
and2areavailable.
Thefilenames/dev/stdin,/dev/stdout,and/dev/stderrareessentiallyaliasesfor/dev/fd/0,/dev/fd/1,and
/dev/fd/2,respectively.However,thosenamesaremoreselfexplanatory.
Notethatusingclose()onafilenameoftheform"/dev/fd/N",forfiledescriptornumbersabovetwo,doesactually
closethegivenfiledescriptor.
Next:SpecialCaveats,Previous:OtherInheritedFiles,Up:SpecialFiles[Contents][Index]
5.8.2SpecialFilesforNetworkCommunications
gawkprogramscanopenatwowayTCP/IPconnection,actingaseitheraclientoraserver.Thisisdoneusingaspecial

filenameoftheform:
/nettype/protocol/localport/remotehost/remoteport

Thenettypeisoneofinet,inet4,orinet6.Theprotocolisoneoftcporudp,andtheotherfieldsrepresentthe
otheressentialpiecesofinformationformakinganetworkingconnection.Thesefilenamesareusedwiththe|&
operatorforcommunicatingwithacoprocess(seeTwowayI/O).Thisisanadvancedfeature,mentionedhereonlyfor
completeness.FulldiscussionisdelayeduntilTCP/IPNetworking.
Previous:SpecialNetwork,Up:SpecialFiles[Contents][Index]
5.8.3SpecialFilenameCaveats
Herearesomethingstobearinmindwhenusingthespecialfilenamesthatgawkprovides:
RecognitionofthefilenamesforthethreestandardpreopenedfilesisdisabledonlyinPOSIXmode.
Recognitionoftheotherspecialfilenamesisdisabledifgawkisincompatibilitymode(eithertraditionalor
posixseeOptions).
gawkalwaysinterpretsthesespecialfilenames.Forexample,using/dev/fd/4foroutputactuallywritesonfile
descriptor4,andnotonanewfiledescriptorthatisdup()edfromfiledescriptor4.Mostofthetimethisdoesnot
matterhowever,itisimportanttonotcloseanyofthefilesrelatedtofiledescriptors0,1,and2.Doingsoresults
inunpredictablebehavior.
Next:OutputSummary,Previous:SpecialFiles,Up:Printing[Contents][Index]
https://www.gnu.org/software/gawk/manual/gawk.html

98/479

29/7/2016

TheGNUAwkUsersGuide

5.9ClosingInputandOutputRedirections
Ifthesamefilenameorthesameshellcommandisusedwithgetlinemorethanonceduringtheexecutionofanawk
program(seeGetline),thefileisopened(orthecommandisexecuted)thefirsttimeonly.Atthattime,thefirstrecord
ofinputisreadfromthatfileorcommand.Thenexttimethesamefileorcommandisusedwithgetline,anotherrecord
isreadfromit,andsoon.
Similarly,whenafileorpipeisopenedforoutput,awkremembersthefilenameorcommandassociatedwithit,and
subsequentwritestothesamefileorcommandareappendedtothepreviouswrites.Thefileorpipestaysopenuntilawk
exits.
Thisimpliesthatspecialstepsarenecessaryinordertoreadthesamefileagainfromthebeginning,ortorerunashell
command(ratherthanreadingmoreoutputfromthesamecommand).Theclose()functionmakesthesethings
possible:
close(filename)

or:
close(command)

Theargumentfilenameorcommandcanbeanyexpression.Itsvaluemustexactlymatchthestringthatwasusedto
openthefileorstartthecommand(spacesandotherirrelevantcharactersincluded).Forexample,ifyouopenapipe
withthis:
"sortrnames"|getlinefoo

thenyoumustcloseitwiththis:
close("sortrnames")

Oncethisfunctioncallisexecuted,thenextgetlinefromthatfileorcommand,orthenextprintorprintftothatfile
orcommand,reopensthefileorrerunsthecommand.Becausetheexpressionthatyouusetocloseafileorpipeline
mustexactlymatchtheexpressionusedtoopenthefileorrunthecommand,itisgoodpracticetouseavariabletostore
thefilenameorcommand.Thepreviousexamplebecomesthefollowing:
sortcom="sortrnames"
sortcom|getlinefoo

close(sortcom)

Thishelpsavoidhardtofindtypographicalerrorsinyourawkprograms.Herearesomeofthereasonsforclosingan
outputfile:
Towriteafileandreaditbacklateroninthesameawkprogram.Closethefileafterwritingit,thenbeginreading
itwithgetline.
Towritenumerousfiles,successively,inthesameawkprogram.Ifthefilesarentclosed,eventuallyawkmay
exceedasystemlimitonthenumberofopenfilesinoneprocess.Itisbesttocloseeachonewhentheprogram
hasfinishedwritingit.
Tomakeacommandfinish.Whenoutputisredirectedthroughapipe,thecommandreadingthepipenormally
continuestotrytoreadinputaslongasthepipeisopen.Oftenthismeansthecommandcannotreallydoitswork
untilthepipeisclosed.Forexample,ifoutputisredirectedtothemailprogram,themessageisnotactuallysent
untilthepipeisclosed.
Torunthesameprogramasecondtime,withthesamearguments.Thisisnotthesamethingasgivingmoreinput
tothefirstrun!
Forexample,supposeaprogrampipesoutputtothemailprogram.Ifitoutputsseverallinesredirectedtothispipe
withoutclosingit,theymakeasinglemessageofseverallines.Bycontrast,iftheprogramclosesthepipeafter
eachlineofoutput,theneachlinemakesaseparatemessage.
https://www.gnu.org/software/gawk/manual/gawk.html

99/479

29/7/2016

TheGNUAwkUsersGuide

Ifyouusemorefilesthanthesystemallowsyoutohaveopen,gawkattemptstomultiplextheavailableopenfilesamong
yourdatafiles.gawksabilitytodothisdependsuponthefacilitiesofyouroperatingsystem,soitmaynotalwayswork.
Itisthereforebothgoodpracticeandgoodportabilityadvicetoalwaysuseclose()onyourfileswhenyouaredone
withthem.Infact,ifyouareusingalotofpipes,itisessentialthatyouclosecommandswhendone.Forexample,
considersomethinglikethis:
{

command=("grep"$1"/some/file|my_progq"$3)
while((command|getline)>0){
processoutputofcommand
}
#needclose(command)here
}

Thisexamplecreatesanewpipelinebasedondataineachrecord.Withoutthecalltoclose()indicatedinthecomment,
awkcreateschildprocessestorunthecommands,untiliteventuallyrunsoutoffiledescriptorsformorepipelines.
Eventhougheachcommandhasfinished(asindicatedbytheendoffilereturnstatusfromgetline),thechildprocessis
notterminated29moreimportantly,thefiledescriptorforthepipeisnotclosedandreleaseduntilclose()iscalledor
awkexits.
close()silentlydoesnothingifgivenanargumentthatdoesnotrepresentafile,pipe,orcoprocessthatwasopened
witharedirection.Insuchacase,itreturnsanegativevalue,indicatinganerror.Inaddition,gawksetsERRNOtoastring

indicatingtheerror.
Notealsothatclose(FILENAME)hasnomagiceffectsontheimplicitloopthatreadsthroughthefilesnamedonthe
commandline.Itis,morelikely,acloseofafilethatwasneveropenedwitharedirection,soawksilentlydoesnothing,
exceptreturnanegativevalue.
Whenusingthe|&operatortocommunicatewithacoprocess,itisoccasionallyusefultobeabletocloseoneendof
thetwowaypipewithoutclosingtheother.Thisisdonebysupplyingasecondargumenttoclose().Asinanyother
calltoclose(),thefirstargumentisthenameofthecommandorspecialfileusedtostartthecoprocess.Thesecond
argumentshouldbeastring,witheitherofthevalues"to"or"from".Casedoesnotmatter.Asthisisanadvanced
feature,discussionisdelayeduntilTwowayI/O,whichdescribesitinmoredetailandgivesanexample.
Usingclose()sReturnValue
InmanyolderversionsofUnixawk,theclose()functionisactuallyastatement.(d.c.)Itisasyntaxerrortotryand
usethereturnvaluefromclose():
command=""
command|getlineinfo
retval=close(command)#syntaxerrorinmanyUnixawks
gawktreatsclose()asafunction.Thereturnvalueis1iftheargumentnamessomethingthatwasneveropenedwitha
redirection,orifthereisasystemproblemclosingthefileorprocess.Inthesecases,gawksetsthepredefinedvariable
ERRNOtoastringdescribingtheproblem.

Ingawk,whenclosingapipeorcoprocess(inputoroutput),thereturnvalueistheexitstatusofthecommand.30
Otherwise,itisthereturnvaluefromthesystemsclose()orfclose()Cfunctionswhenclosinginputoroutputfiles,
respectively.Thisvalueiszeroiftheclosesucceeds,or1ifitfails.
ThePOSIXstandardisveryvagueitsaysthatclose()returnszeroonsuccessandanonzerovalueotherwise.In
general,differentimplementationsvaryinwhattheyreportwhenclosingpipesthus,thereturnvaluecannotbeused
portably.(d.c.)InPOSIXmode(seeOptions),gawkjustreturnszerowhenclosingapipe.

Next:OutputExercises,Previous:CloseFilesAndPipes,Up:Printing[Contents][Index]
https://www.gnu.org/software/gawk/manual/gawk.html

100/479

29/7/2016

TheGNUAwkUsersGuide

5.10Summary
Theprintstatementprintscommaseparatedexpressions.EachexpressionisseparatedbythevalueofOFSand
terminatedbythevalueofORS.OFMTprovidestheconversionformatfornumericvaluesfortheprintstatement.
Theprintfstatementprovidesfinergrainedcontroloveroutput,withformatcontrollettersfordifferentdata
typesandvariousflagsthatmodifythebehavioroftheformatcontrolletters.
Outputfrombothprintandprintfmayberedirectedtofiles,pipes,andcoprocesses.
gawkprovidesspecialfilenamesforaccesstostandardinput,output,anderror,andfornetworkcommunications.
Useclose()tocloseopenfile,pipe,andcoprocessredirections.Forcoprocesses,itispossibletocloseonlyone
directionofthecommunications.
Previous:OutputSummary,Up:Printing[Contents][Index]

5.11Exercises
1.Rewritetheprogram:
awk'BEGIN{print"MonthCrates"
print""}
{print$1,"",$2}'inventoryshipped

fromOutputSeparators,byusinganewvalueofOFS.
2.Usetheprintfstatementtolineuptheheadingsandtabledatafortheinventoryshippedexamplethatwas
coveredinPrint.
3.Whathappensifyouforgetthedoublequoteswhenredirectingoutput,asfollows:
BEGIN{print"Seriouserrordetected!">/dev/stderr}

Next:PatternsandActions,Previous:Printing,Up:Top[Contents][Index]

6Expressions
Expressionsarethebasicbuildingblocksofawkpatternsandactions.Anexpressionevaluatestoavaluethatyoucan
print,test,orpasstoafunction.Additionally,anexpressioncanassignanewvaluetoavariableorafieldbyusingan
assignmentoperator.
Anexpressioncanserveasapatternoractionstatementonitsown.Mostotherkindsofstatementscontainoneormore
expressionsthatspecifythedataonwhichtooperate.Asinotherlanguages,expressionsinawkcanincludevariables,
arrayreferences,constants,andfunctioncalls,aswellascombinationsofthesewithvariousoperators.
Values:
Constants,Variables,andRegularExpressions.
AllOperators:
gawksoperators.
TruthValuesandConditions: Testingfortrueandfalse.
FunctionCalls:
Afunctioncallisanexpression.
Precedence:
Howvariousoperatorsnest.
Locales:
Howthelocaleaffectsthings.
ExpressionsSummary:
Expressionssummary.
Next:AllOperators,Up:Expressions[Contents][Index]

6.1Constants,Variables,andConversions
Expressionsarebuiltupfromvaluesandtheoperationsperformeduponthem.Thissectiondescribestheelementary
objectsthatprovidethevaluesusedinexpressions.
https://www.gnu.org/software/gawk/manual/gawk.html

101/479

29/7/2016

TheGNUAwkUsersGuide

Constants:
String,numericandregexpconstants.
UsingConstantRegexps: Whenandhowtousearegexpconstant.
Variables:
Variablesgivenamestovaluesforlateruse.
Conversion:
Theconversionofstringstonumbersandviceversa.
Next:UsingConstantRegexps,Up:Values[Contents][Index]
6.1.1ConstantExpressions
Thesimplesttypeofexpressionistheconstant,whichalwayshasthesamevalue.Therearethreetypesofconstants:
numeric,string,andregularexpression.
Eachisusedintheappropriatecontextwhenyouneedadatavaluethatisntgoingtochange.Numericconstantscan
havedifferentforms,butareinternallystoredinanidenticalmanner.
ScalarConstants:
Numericandstringconstants.
Nondecimalnumbers: Whatareoctalandhexnumbers.
RegexpConstants:
RegularExpressionconstants.
Next:Nondecimalnumbers,Up:Constants[Contents][Index]
6.1.1.1NumericandStringConstants
Anumericconstantstandsforanumber.Thisnumbercanbeaninteger,adecimalfraction,oranumberinscientific
(exponential)notation.31Herearesomeexamplesofnumericconstantsthatallhavethesamevalue:
105
1.05e+2
1050e1

Astringconstantconsistsofasequenceofcharactersenclosedindoublequotationmarks.Forexample:
"parrot"

representsthestringwhosecontentsareparrot.Stringsingawkcanbeofanylength,andtheycancontainanyofthe
possibleeightbitASCIIcharacters,includingASCIINUL(charactercodezero).Otherawkimplementationsmayhave
difficultywithsomecharactercodes.
Next:RegexpConstants,Previous:ScalarConstants,Up:Constants[Contents][Index]
6.1.1.2OctalandHexadecimalNumbers
Inawk,allnumbersareindecimal(i.e.,base10).Manyotherprogramminglanguagesallowyoutospecifynumbersin
otherbases,oftenoctal(base8)andhexadecimal(base16).Inoctal,thenumbersgo0,1,2,3,4,5,6,7,10,11,12,and
soon.Justas11indecimalis1times10plus1,so11inoctalis1times8plus1.Thisequals9indecimal.In
hexadecimal,thereare16digits.Becausetheeverydaydecimalnumbersystemonlyhastendigits(09),theletters
athroughfareusedtorepresenttherest.(CaseinthelettersisusuallyirrelevanthexadecimalaandAhavethe
samevalue.)Thus,11inhexadecimalis1times16plus1,whichequals17indecimal.
Justbylookingatplain11,youcanttellwhatbaseitsin.So,inC,C++,andotherlanguagesderivedfromC,thereis
aspecialnotationtosignifythebase.Octalnumbersstartwithaleading0,andhexadecimalnumbersstartwitha
leading0xor0X:
11

https://www.gnu.org/software/gawk/manual/gawk.html

102/479

29/7/2016

TheGNUAwkUsersGuide

Decimalvalue11
011

Octal11,decimalvalue9
0x11

Hexadecimal11,decimalvalue17
Thisexampleshowsthedifference:
$gawk'BEGIN{printf"%d,%d,%d\n",011,11,0x11}'
|9,11,17

Beingabletouseoctalandhexadecimalconstantsinyourprogramsismostusefulwhenworkingwithdatathatcannot
berepresentedconvenientlyascharactersorasregularnumbers,suchasbinarydataofvarioussorts.
gawkallowstheuseofoctalandhexadecimalconstantsinyourprogramtext.However,suchnumbersintheinputdata
arenottreateddifferentlydoingsobydefaultwouldbreakoldprograms.(Ifyoureallyneedtodothis,usethenon
decimaldatacommandlineoptionseeNondecimalData.)Ifyouhaveoctalorhexadecimaldata,youcanusethe
strtonum()function(seeStringFunctions)toconvertthedataintoanumber.Mostofthetime,youwillwanttouse

octalorhexadecimalconstantswhenworkingwiththebuiltinbitmanipulationfunctionsseeBitwiseFunctions,for
moreinformation.
UnlikeinsomeearlyCimplementations,8and9arenotvalidinoctalconstants.Forexample,gawktreats018as
decimal18:
$gawk'BEGIN{print"021is",021;print018}'
|021is17
|18

Octalandhexadecimalsourcecodeconstantsareagawkextension.Ifgawkisincompatibilitymode(seeOptions),they
arenotavailable.
AConstantsBaseDoesNotAffectItsValue
Onceanumericconstanthasbeenconvertedinternallyintoanumber,gawknolongerrememberswhattheoriginal
formoftheconstantwastheinternalvalueisalwaysused.Thishasparticularconsequencesforconversionof
numberstostrings:
$gawk'BEGIN{printf"0x11is<%s>\n",0x11}'
|0x11is<17>

Previous:Nondecimalnumbers,Up:Constants[Contents][Index]
6.1.1.3RegularExpressionConstants
Aregexpconstantisaregularexpressiondescriptionenclosedinslashes,suchas/^beginningandend$/.Mostregexps
usedinawkprogramsareconstant,butthe~and!~matchingoperatorscanalsomatchcomputedordynamicregexps
(whicharetypicallyjustordinarystringsorvariablesthatcontainaregexp,butcouldbemorecomplexexpressions).
Next:Variables,Previous:Constants,Up:Values[Contents][Index]
6.1.2UsingRegularExpressionConstants
Whenusedontherighthandsideofthe~or!~operators,aregexpconstantmerelystandsfortheregexpthatistobe
https://www.gnu.org/software/gawk/manual/gawk.html

103/479

29/7/2016

TheGNUAwkUsersGuide

matched.However,regexpconstants(suchas/foo/)maybeusedlikesimpleexpressions.Whenaregexpconstant
appearsbyitself,ithasthesamemeaningasifitappearedinapattern(i.e.,($0~/foo/)).(d.c.)SeeExpression
Patterns.Thismeansthatthefollowingtwocodesegments:
if($0~/barfly/||$0~/camelot/)
print"found"

and:
if(/barfly/||/camelot/)
print"found"

areexactlyequivalent.OneratherbizarreconsequenceofthisruleisthatthefollowingBooleanexpressionisvalid,but
doesnotdowhatitsauthorprobablyintended:
#Notethat/foo/isontheleftofthe~
if(/foo/~$1)print"foundfoo"

Thiscodeisobviouslytesting$1foramatchagainsttheregexp/foo/.Butinfact,theexpression/foo/~$1really
means($0~/foo/)~$1.Inotherwords,firstmatchtheinputrecordagainsttheregexp/foo/.Theresultiseither
zeroorone,dependinguponthesuccessorfailureofthematch.Thatresultisthenmatchedagainstthefirstfieldinthe
record.Becauseitisunlikelythatyouwouldeverreallywanttomakethiskindoftest,gawkissuesawarningwhenit
seesthisconstructinaprogram.Anotherconsequenceofthisruleisthattheassignmentstatement:
matches=/foo/

assignseitherzerooronetothevariablematches,dependinguponthecontentsofthecurrentinputrecord.
Constantregularexpressionsarealsousedasthefirstargumentforthegensub(),sub(),andgsub()functions,asthe
secondargumentofthematch()function,andasthethirdargumentofthesplit()andpatsplit()functions(seeString
Functions).Modernimplementationsofawk,includinggawk,allowthethirdargumentofsplit()tobearegexp
constant,butsomeolderimplementationsdonot.(d.c.)Becausesomebuiltinfunctionsacceptregexpconstantsas
arguments,confusioncanarisewhenattemptingtouseregexpconstantsasargumentstouserdefinedfunctions(see
Userdefined).Forexample:
functionmysub(pat,repl,str,global)
{
if(global)
gsub(pat,repl,str)
else
sub(pat,repl,str)
returnstr
}
{

text="hi!hiyourself!"
mysub(/hi/,"howdy",text,1)

Inthisexample,theprogrammerwantstopassaregexpconstanttotheuserdefinedfunctionmysub(),whichinturn
passesitontoeithersub()orgsub().However,whatreallyhappensisthatthepatparameterisassignedavalueof
eitheroneorzero,dependinguponwhetherornot$0matches/hi/.gawkissuesawarningwhenitseesaregexp
constantusedasaparametertoauserdefinedfunction,becausepassingatruthvalueinthiswayisprobablynotwhat
wasintended.
Next:Conversion,Previous:UsingConstantRegexps,Up:Values[Contents][Index]
6.1.3Variables
https://www.gnu.org/software/gawk/manual/gawk.html

104/479

29/7/2016

TheGNUAwkUsersGuide

Variablesarewaysofstoringvaluesatonepointinyourprogramforuselaterinanotherpartofyourprogram.They
canbemanipulatedentirelywithintheprogramtext,andtheycanalsobeassignedvaluesontheawkcommandline.
UsingVariables: Usingvariablesinyourprograms.
Assignment
Settingvariablesonthecommandlineandasummaryofcommandlinesyntax.Thisisan

Options:
advancedmethodofinput.
Next:AssignmentOptions,Up:Variables[Contents][Index]
6.1.3.1UsingVariablesinaProgram
Variablesletyougivenamestovaluesandrefertothemlater.Variableshavealreadybeenusedinmanyofthe
examples.Thenameofavariablemustbeasequenceofletters,digits,orunderscores,anditmaynotbeginwithadigit.
Here,aletterisanyoneofthe52upperandlowercaseEnglishletters.Othercharactersthatmaybedefinedaslettersin
nonEnglishlocalesarenotvalidinvariablenames.CaseissignificantinvariablenamesaandAaredistinctvariables.
Avariablenameisavalidexpressionbyitselfitrepresentsthevariablescurrentvalue.Variablesaregivennewvalues
withassignmentoperators,incrementoperators,anddecrementoperators(seeAssignmentOps).Inaddition,thesub()
andgsub()functionscanchangeavariablesvalue,andthematch(),split(),andpatsplit()functionscanchangethe
contentsoftheirarrayparameters(seeStringFunctions).
Afewvariableshavespecialbuiltinmeanings,suchasFS(thefieldseparator)andNF(thenumberoffieldsinthe
currentinputrecord).SeeBuiltinVariables,foralistofthepredefinedvariables.Thesepredefinedvariablescanbe
usedandassignedjustlikeallothervariables,buttheirvaluesarealsousedorchangedautomaticallybyawk.All
predefinedvariablesnamesareentirelyuppercase.
Variablesinawkcanbeassignedeithernumericorstringvalues.Thekindofvalueavariableholdscanchangeoverthe
lifeofaprogram.Bydefault,variablesareinitializedtotheemptystring,whichiszeroifconvertedtoanumber.There
isnoneedtoexplicitlyinitializeavariableinawk,whichiswhatyouwoulddoinCandinmostothertraditional
languages.
Previous:UsingVariables,Up:Variables[Contents][Index]
6.1.3.2AssigningVariablesontheCommandLine
Anyawkvariablecanbesetbyincludingavariableassignmentamongtheargumentsonthecommandlinewhenawkis
invoked(seeOtherArguments).Suchanassignmenthasthefollowingform:
variable=text

Withit,avariableisseteitheratthebeginningoftheawkrunorinbetweeninputfiles.Whentheassignmentispreceded
withthevoption,asinthefollowing:
vvariable=text

thevariableissetattheverybeginning,evenbeforetheBEGINrulesexecute.Thevoptionanditsassignmentmust
precedeallthefilenamearguments,aswellastheprogramtext.(SeeOptions,formoreinformationaboutthev
option.)Otherwise,thevariableassignmentisperformedatatimedeterminedbyitspositionamongtheinputfile
argumentsaftertheprocessingoftheprecedinginputfileargument.Forexample:
awk'{print$n}'n=4inventoryshippedn=2maillist

printsthevalueoffieldnumbernforallinputrecords.Beforethefirstfileisread,thecommandlinesetsthevariablen
equaltofour.Thiscausesthefourthfieldtobeprintedinlinesfrominventoryshipped.Afterthefirstfilehasfinished,
butbeforethesecondfileisstarted,nissettotwo,sothatthesecondfieldisprintedinlinesfrommaillist:
$awk'{print$n}'n=4inventoryshippedn=2maillist
https://www.gnu.org/software/gawk/manual/gawk.html

105/479

29/7/2016

TheGNUAwkUsersGuide

|15
|24

|5555553
|5553412

CommandlineargumentsaremadeavailableforexplicitexaminationbytheawkprogramintheARGVarray(seeARGC
andARGV).awkprocessesthevaluesofcommandlineassignmentsforescapesequences(seeEscapeSequences).
(d.c.)
Previous:Variables,Up:Values[Contents][Index]
6.1.4ConversionofStringsandNumbers
Numbertostringandstringtonumberconversionaregenerallystraightforward.Therecanbesubtletiestobeawareof
thissectiondiscussesthisimportantfacetofawk.
StringsAndNumbers:
HowawkConvertsBetweenStringsAndNumbers.
Localeinfluencesconversions: Howthelocalemayaffectconversions.
Next:Localeinfluencesconversions,Up:Conversion[Contents][Index]
6.1.4.1HowawkConvertsBetweenStringsandNumbers
Stringsareconvertedtonumbersandnumbersareconvertedtostrings,ifthecontextoftheawkprogramdemandsit.
Forexample,ifthevalueofeitherfooorbarintheexpressionfoo+barhappenstobeastring,itisconvertedtoa
numberbeforetheadditionisperformed.Ifnumericvaluesappearinstringconcatenation,theyareconvertedtostrings.
Considerthefollowing:
two=2;three=3
print(twothree)+4

Thisprintsthe(numeric)value27.Thenumericvaluesofthevariablestwoandthreeareconvertedtostringsand
concatenatedtogether.Theresultingstringisconvertedbacktothenumber23,towhich4isthenadded.
If,forsomereason,youneedtoforceanumbertobeconvertedtoastring,concatenatethatnumberwiththeempty
string,"".Toforceastringtobeconvertedtoanumber,addzerotothatstring.Astringisconvertedtoanumberby
interpretinganynumericprefixofthestringasnumerals:"2.5"convertsto2.5,"1e3"convertsto1,000,and"25fix"
hasanumericvalueof25.Stringsthatcantbeinterpretedasvalidnumbersconverttozero.
TheexactmannerinwhichnumbersareconvertedintostringsiscontrolledbytheawkpredefinedvariableCONVFMT(see
BuiltinVariables).Numbersareconvertedusingthesprintf()functionwithCONVFMTastheformatspecifier(seeString
Functions).
CONVFMTsdefaultvalueis"%.6g",whichcreatesavaluewithatmostsixsignificantdigits.Forsomeapplications,you

mightwanttochangeittospecifymoreprecision.Onmostmodernmachines,17digitsisusuallyenoughtocapturea
floatingpointnumbersvalueexactly.32
StrangeresultscanoccurifyousetCONVFMTtoastringthatdoesnttellsprintf()howtoformatfloatingpointnumbers
inausefulway.Forexample,ifyouforgetthe%intheformat,awkconvertsallnumberstothesameconstantstring.
Asaspecialcase,ifanumberisaninteger,thentheresultofconvertingittoastringisalwaysaninteger,nomatter
whatthevalueofCONVFMTmaybe.Giventhefollowingcodefragment:
CONVFMT="%2.2f"
a=12
b=a""
https://www.gnu.org/software/gawk/manual/gawk.html

106/479

29/7/2016

TheGNUAwkUsersGuide

bhasthevalue"12",not"12.00".(d.c.)

PrePOSIXawkUsedOFMTforStringConversion
PriortothePOSIXstandard,awkusedthevalueofOFMTforconvertingnumberstostrings.OFMTspecifiestheoutput
formattousewhenprintingnumberswithprint.CONVFMTwasintroducedinordertoseparatethesemanticsof
conversionfromthesemanticsofprinting.BothCONVFMTandOFMThavethesamedefaultvalue:"%.6g".Inthevast
majorityofcases,oldawkprogramsdonotchangetheirbehavior.SeePrint,formoreinformationontheprint
statement.

Previous:StringsAndNumbers,Up:Conversion[Contents][Index]
6.1.4.2LocalesCanInfluenceConversion
Whereyouarecanmatterwhenitcomestoconvertingbetweennumbersandstrings.Thelocalcharactersetand
languagethelocalecanaffectnumericformats.Inparticular,forawkprograms,itaffectsthedecimalpointcharacter
andthethousandsseparatorcharacter.The"C"locale,andmostEnglishlanguagelocales,usetheperiodcharacter(.)
asthedecimalpointanddonthaveathousandsseparator.However,many(ifnotmost)EuropeanandnonEnglish
localesusethecomma(,)asthedecimalpointcharacter.Europeanlocalesoftenuseeitheraspaceoraperiodasthe
thousandsseparator,iftheyhaveone.
ThePOSIXstandardsaysthatawkalwaysusestheperiodasthedecimalpointwhenreadingtheawkprogramsource
code,andforcommandlinevariableassignments(seeOtherArguments).However,wheninterpretinginputdata,for
printandprintfoutput,andfornumbertostringconversion,thelocaldecimalpointcharacterisused.(d.c.)Inall
cases,numbersinsourcecodeandininputdatacannothaveathousandsseparator.Herearesomeexamplesindicating
thedifferenceinbehavior,onaGNU/Linuxsystem:
$exportPOSIXLY_CORRECT=1ForcePOSIXbehavior
$gawk'BEGIN{printf"%g\n",3.1415927}'
|3.14159
$LC_ALL=en_DK.utf8gawk'BEGIN{printf"%g\n",3.1415927}'
|3,14159
$echo4,321|gawk'{print$1+1}'
|5
$echo4,321|LC_ALL=en_DK.utf8gawk'{print$1+1}'
|5,321

Theen_DK.utf8localeisforEnglishinDenmark,wherethecommaactsasthedecimalpointseparator.Inthenormal
"C"locale,gawktreats4,321as4,whileintheDanishlocale,itstreatedasthefullnumberincludingthefractional
part,4.321.
Someearlierversionsofgawkfullycompliedwiththisaspectofthestandard.However,manyusersinnonEnglish
localescomplainedaboutthisbehavior,becausetheirdatausedaperiodasthedecimalpoint,sothedefaultbehavior
wasrestoredtouseaperiodasthedecimalpointcharacter.Youcanusetheuselcnumericoption(seeOptions)to
forcegawktousethelocalesdecimalpointcharacter.(gawkalsousesthelocalesdecimalpointcharacterwhenin
POSIXmode,eitherviaposixorthePOSIXLY_CORRECTenvironmentvariable,asshownpreviously.)
Table6.1describesthecasesinwhichthelocalesdecimalpointcharacterisusedandwhenaperiodisused.Someof
thesefeatureshavenotbeendescribedyet.
Feature
%'g
%g

Input
strtonum()

Default
Uselocale
Useperiod
Useperiod
Useperiod

posixoruselcnumeric

Uselocale
Uselocale
Uselocale
Uselocale

https://www.gnu.org/software/gawk/manual/gawk.html

107/479

29/7/2016

TheGNUAwkUsersGuide

Table6.1:Localedecimalpointversusaperiod
Finally,moderndayformalstandardsandtheIEEEstandardfloatingpointrepresentationcanhaveanunusualbut
importanteffectonthewaygawkconvertssomespecialstringvaluestonumbers.ThedetailsarepresentedinPOSIX
FloatingPointProblems.
Next:TruthValuesandConditions,Previous:Values,Up:Expressions[Contents][Index]

6.2Operators:DoingSomethingwithValues
Thissectionintroducestheoperatorsthatmakeuseofthevaluesprovidedbyconstantsandvariables.
ArithmeticOps: Arithmeticoperations(+,,etc.)
Concatenation: Concatenatingstrings.
AssignmentOps: Changingthevalueofavariableorafield.
IncrementOps: Incrementingthenumericvalueofavariable.
Next:Concatenation,Up:AllOperators[Contents][Index]
6.2.1ArithmeticOperators
Theawklanguageusesthecommonarithmeticoperatorswhenevaluatingexpressions.Allofthesearithmeticoperators
follownormalprecedencerulesandworkasyouwouldexpectthemto.
Thefollowingexampleusesafilenamedgrades,whichcontainsalistofstudentnamesaswellasthreetestscoresper
student(itsasmallclass):
Pat1009758
Sandy847293
Chris729289

Thisprogramtakesthefilegradesandprintstheaverageofthescores:
$awk'{sum=$2+$3+$4;avg=sum/3
>print$1,avg}'grades
|Pat85
|Sandy83
|Chris84.3333

Thefollowinglistprovidesthearithmeticoperatorsinawk,inorderfromthehighestprecedencetothelowest:
x^y
x**y

Exponentiationxraisedtotheypower.2^3hasthevalueeightthecharactersequence**isequivalentto
^.(c.e.)
x

Negation.
+x

Unaryplustheexpressionisconvertedtoanumber.
x*y

Multiplication.
https://www.gnu.org/software/gawk/manual/gawk.html

108/479

29/7/2016

TheGNUAwkUsersGuide

x/y

Divisionbecauseallnumbersinawkarefloatingpointnumbers,theresultisnotroundedtoaninteger3/4
hasthevalue0.75.(Itisacommonmistake,especiallyforCprogrammers,toforgetthatallnumbersinawkare
floatingpoint,andthatdivisionofintegerlookingconstantsproducesarealnumber,notaninteger.)
x%y

Remainderfurtherdiscussionisprovidedinthetext,justafterthislist.
x+y

Addition.
xy

Subtraction.
Unaryplusandminushavethesameprecedence,themultiplicationoperatorsallhavethesameprecedence,and
additionandsubtractionhavethesameprecedence.
Whencomputingtheremainderofx%y,thequotientisroundedtowardzerotoanintegerandmultipliedbyy.This
resultissubtractedfromxthisoperationissometimesknownastruncmod.Thefollowingrelationalwaysholds:
b*int(a/b)+(a%b)==a

Onepossiblyundesirableeffectofthisdefinitionofremainderisthatx%yisnegativeifxisnegative.Thus:
17%8=1

Inotherawkimplementations,thesignednessoftheremaindermaybemachinedependent.
NOTE:ThePOSIXstandardonlyspecifiestheuseof^forexponentiation.Formaximumportability,do
notusethe**operator.
Next:AssignmentOps,Previous:ArithmeticOps,Up:AllOperators[Contents][Index]
6.2.2StringConcatenation
Itseemedlikeagoodideaatthetime.
BrianKernighan
Thereisonlyonestringoperation:concatenation.Itdoesnothaveaspecificoperatortorepresentit.Instead,
concatenationisperformedbywritingexpressionsnexttooneanother,withnooperator.Forexample:
$awk'{print"Fieldnumberone:"$1}'maillist
|Fieldnumberone:Amelia
|Fieldnumberone:Anthony

Withoutthespaceinthestringconstantafterthe:,thelinerunstogether.Forexample:
$awk'{print"Fieldnumberone:"$1}'maillist
|Fieldnumberone:Amelia
|Fieldnumberone:Anthony

Becausestringconcatenationdoesnothaveanexplicitoperator,itisoftennecessarytoensurethatithappensatthe
righttimebyusingparenthesestoenclosetheitemstoconcatenate.Forexample,youmightexpectthatthefollowing
codefragmentconcatenatesfileandname:
https://www.gnu.org/software/gawk/manual/gawk.html

109/479

29/7/2016

TheGNUAwkUsersGuide

file="file"
name="name"
print"somethingmeaningful">filename

ThisproducesasyntaxerrorwithsomeversionsofUnixawk.33Itisnecessarytousethefollowing:
print"somethingmeaningful">(filename)

Parenthesesshouldbeusedaroundconcatenationinallbutthemostcommoncontexts,suchasontherighthandsideof
=.Becarefulaboutthekindsofexpressionsusedinstringconcatenation.Inparticular,theorderofevaluationof
expressionsusedforconcatenationisundefinedintheawklanguage.Considerthisexample:
BEGIN{
a="don't"
print(a""(a="panic"))
}

Itisnotdefinedwhetherthesecondassignmenttoahappensbeforeorafterthevalueofaisretrievedforproducingthe
concatenatedvalue.Theresultcouldbeeitherdon'tpanic,orpanicpanic.
Theprecedenceofconcatenation,whenmixedwithotheroperators,isoftencounterintuitive.Considerthisexample:
$awk'BEGIN{print12""24}'
|1224

Thisobviouslyisconcatenating12,aspace,and24.Butwheredidthespacedisappearto?Theanswerliesinthe
combinationofoperatorprecedencesandawksautomaticconversionrules.Togetthedesiredresult,writetheprogram
thisway:
$awk'BEGIN{print12""(24)}'
|1224

Thisforcesawktotreattheonthe24asunary.Otherwise,itsparsedasfollows:
12(""24)
12(024)
12(24)
1224
Asmentionedearlier,whenmixingconcatenationwithotheroperators,parenthesize.Otherwise,youreneverquitesure
whatyoullget.
Next:IncrementOps,Previous:Concatenation,Up:AllOperators[Contents][Index]
6.2.3AssignmentExpressions
Anassignmentisanexpressionthatstoresa(usuallydifferent)valueintoavariable.Forexample,letsassignthevalue
onetothevariablez:
z=1

Afterthisexpressionisexecuted,thevariablezhasthevalueone.Whateveroldvaluezhadbeforetheassignmentis
forgotten.
Assignmentscanalsostorestringvalues.Forexample,thefollowingstoresthevalue"thisfoodisgood"inthe
variablemessage:
thing="food"
predicate="good"
message="this"thing"is"predicate
https://www.gnu.org/software/gawk/manual/gawk.html

110/479

29/7/2016

TheGNUAwkUsersGuide

Thisalsoillustratesstringconcatenation.The=signiscalledanassignmentoperator.Itisthesimplestassignment
operatorbecausethevalueoftherighthandoperandisstoredunchanged.Mostoperators(addition,concatenation,and
soon)havenoeffectexcepttocomputeavalue.Ifthevalueisntused,theresnoreasontousetheoperator.An
assignmentoperatorisdifferentitdoesproduceavalue,butevenifyouignoreit,theassignmentstillmakesitselffelt
throughthealterationofthevariable.Wecallthisasideeffect.
Thelefthandoperandofanassignmentneednotbeavariable(seeVariables)itcanalsobeafield(seeChanging
Fields)oranarrayelement(seeArrays).Theseareallcalledlvalues,whichmeanstheycanappearonthelefthandside
ofanassignmentoperator.Therighthandoperandmaybeanyexpressionitproducesthenewvaluethattheassignment
storesinthespecifiedvariable,field,orarrayelement.(Suchvaluesarecalledrvalues.)
Itisimportanttonotethatvariablesdonothavepermanenttypes.Avariablestypeissimplythetypeofwhatevervalue
waslastassignedtoit.Inthefollowingprogramfragment,thevariablefoohasanumericvalueatfirst,andastring
valuelateron:
foo=1
printfoo
foo="bar"
printfoo

Whenthesecondassignmentgivesfooastringvalue,thefactthatitpreviouslyhadanumericvalueisforgotten.
Stringvaluesthatdonotbeginwithadigithaveanumericvalueofzero.Afterexecutingthefollowingcode,thevalue
offooisfive:
foo="astring"
foo=foo+5

NOTE:Usingavariableasanumberandthenlaterasastringcanbeconfusingandispoorprogramming
style.Theprevioustwoexamplesillustratehowawkworks,nothowyoushouldwriteyourprograms!
Anassignmentisanexpression,soithasavaluethesamevaluethatisassigned.Thus,z=1isanexpressionwith
thevalueone.Oneconsequenceofthisisthatyoucanwritemultipleassignmentstogether,suchas:
x=y=z=5

Thisexamplestoresthevaluefiveinallthreevariables(x,y,andz).Itdoessobecausethevalueofz=5,whichis
five,isstoredintoyandthenthevalueofy=z=5,whichisfive,isstoredintox.
Assignmentsmaybeusedanywhereanexpressioniscalledfor.Forexample,itisvalidtowritex!=(y=1)tosety
toone,andthentestwhetherxequalsone.Butthisstyletendstomakeprogramshardtoreadsuchnestingof
assignmentsshouldbeavoided,exceptperhapsinaoneshotprogram.
Asidefrom=,thereareseveralotherassignmentoperatorsthatdoarithmeticwiththeoldvalueofthevariable.For
example,theoperator+=computesanewvaluebyaddingtherighthandvaluetotheoldvalueofthevariable.Thus,
thefollowingassignmentaddsfivetothevalueoffoo:
foo+=5

Thisisequivalenttothefollowing:
foo=foo+5

Usewhichevermakesthemeaningofyourprogramclearer.
Therearesituationswhereusing+=(oranyassignmentoperator)isnotthesameassimplyrepeatingthelefthand
operandintherighthandexpression.Forexample:
#ThankstoPatRankinforthisexample
BEGIN{
foo[rand()]+=5
for(xinfoo)
https://www.gnu.org/software/gawk/manual/gawk.html

111/479

29/7/2016

TheGNUAwkUsersGuide

printx,foo[x]
bar[rand()]=bar[rand()]+5
for(xinbar)
printx,bar[x]
}

Theindicesofbararepracticallyguaranteedtobedifferent,becauserand()returnsdifferentvalueseachtimeitis
called.(Arraysandtherand()functionhaventbeencoveredyet.SeeArrays,andseeNumericFunctions,formore
information.)Thisexampleillustratesanimportantfactaboutassignmentoperators:thelefthandexpressionisonly
evaluatedonce.
Itisuptotheimplementationastowhichexpressionisevaluatedfirst,thelefthandortherighthand.Considerthis
example:
i=1
a[i+=2]=i+1

Thevalueofa[3]couldbeeithertwoorfour.
Table6.2liststhearithmeticassignmentoperators.Ineachcase,therighthandoperandisanexpressionwhosevalueis
convertedtoanumber.
Operator
lvalue+=increment
lvalue=decrement
lvalue*=coefficient
lvalue/=divisor
lvalue%=modulus
lvalue^=power
lvalue**=power

Effect
Addincrementtothevalueoflvalue.
Subtractdecrementfromthevalueoflvalue.
Multiplythevalueoflvaluebycoefficient.
Dividethevalueoflvaluebydivisor.
Setlvaluetoitsremainderbymodulus.
Raiselvaluetothepowerpower.
Raiselvaluetothepowerpower.(c.e.)

Table6.2:Arithmeticassignmentoperators
NOTE:Onlythe^=operatorisspecifiedbyPOSIX.Formaximumportability,donotusethe**=
operator.
SyntacticAmbiguitiesBetween/=andRegularExpressions
Thereisasyntacticambiguitybetweenthe/=assignmentoperatorandregexpconstantswhosefirstcharacterisan=.
(d.c.)Thisismostnotableinsomecommercialawkversions.Forexample:
$awk/==//dev/null
errorawk:syntaxerroratsourceline1
errorcontextis
error>>>/=<<<
errorawk:bailingoutatsourceline1

Aworkaroundis:
awk'/[=]=/'/dev/null
gawkdoesnothavethisproblemBWKawkandmawkalsodonot.

Previous:AssignmentOps,Up:AllOperators[Contents][Index]
6.2.4IncrementandDecrementOperators
https://www.gnu.org/software/gawk/manual/gawk.html

112/479

29/7/2016

TheGNUAwkUsersGuide

Incrementanddecrementoperatorsincreaseordecreasethevalueofavariablebyone.Anassignmentoperatorcando
thesamething,sotheincrementoperatorsaddnopowertotheawklanguagehowever,theyareconvenient
abbreviationsforverycommonoperations.
Theoperatorusedforaddingoneiswritten++.Itcanbeusedtoincrementavariableeitherbeforeoraftertakingits
value.Topreincrementavariablev,write++v.Thisaddsonetothevalueofvthatnewvalueisalsothevalueofthe
expression.(Theassignmentexpressionv+=1iscompletelyequivalent.)Writingthe++afterthevariablespecifies
postincrement.Thisincrementsthevariablevaluejustthesamethedifferenceisthatthevalueoftheincrement
expressionitselfisthevariablesoldvalue.Thus,iffoohasthevaluefour,thentheexpressionfoo++hasthevalue
four,butitchangesthevalueoffootofive.Inotherwords,theoperatorreturnstheoldvalueofthevariable,butwith
thesideeffectofincrementingit.
Thepostincrementfoo++isnearlythesameaswriting(foo+=1)1.Itisnotperfectlyequivalentbecauseall
numbersinawkarefloatingpointinfloatingpoint,foo+11doesnotnecessarilyequalfoo.Butthedifferenceis
minuteaslongasyousticktonumbersthatarefairlysmall(lessthan10e12).
Fieldsandarrayelementsareincrementedjustlikevariables.(Use$(i++)whenyouwanttodoafieldreferenceanda
variableincrementatthesametime.Theparenthesesarenecessarybecauseoftheprecedenceofthefieldreference
operator$.)
Thedecrementoperatorworksjustlike++,exceptthatitsubtractsoneinsteadofaddingit.Aswith++,itcanbe
usedbeforethelvaluetopredecrementorafterittopostdecrement.Followingisasummaryofincrementand
decrementexpressions:
++lvalue

Incrementlvalue,returningthenewvalueasthevalueoftheexpression.
lvalue++

Incrementlvalue,returningtheoldvalueoflvalueasthevalueoftheexpression.
lvalue

Decrementlvalue,returningthenewvalueasthevalueoftheexpression.(Thisexpressionislike++lvalue,but
insteadofadding,itsubtracts.)
lvalue

Decrementlvalue,returningtheoldvalueoflvalueasthevalueoftheexpression.(Thisexpressionislike
lvalue++,butinsteadofadding,itsubtracts.)
OperatorEvaluationOrder
Doctor,ithurtswhenIdothis!
Thendontdothat!
GrouchoMarx
Whathappensforsomethinglikethefollowing?
b=6
printb+=b++

Orsomethingevenstranger?
b=6
b+=++b+b++
printb

Inotherwords,whendothevarioussideeffectsprescribedbythepostfixoperators(b++)takeeffect?Whenside
https://www.gnu.org/software/gawk/manual/gawk.html

113/479

29/7/2016

TheGNUAwkUsersGuide

effectshappenisimplementationdefined.Inotherwords,itisuptotheparticularversionofawk.Theresultforthe
firstexamplemaybe12or13,andforthesecond,itmaybe22or23.
Inshort,doingthingslikethisisnotrecommendedanddefinitelynotanythingthatyoucanrelyuponforportability.
Youshouldavoidsuchthingsinyourownprograms.

Next:FunctionCalls,Previous:AllOperators,Up:Expressions[Contents][Index]

6.3TruthValuesandConditions
Incertaincontexts,expressionvaluesalsoserveastruthvaluesi.e.,theydeterminewhatshouldhappennextasthe
programruns.Thissectiondescribeshowawkdefinestrueandfalseandhowvaluesarecompared.
TruthValues:
Typingand
Comparison:
BooleanOps:
ConditionalExp:

Whatistrueandwhatisfalse.
Howvariablesacquiretypesandhowthisaffectscomparisonofnumbersandstringswith<,

etc.
Combiningcomparisonexpressionsusingbooleanoperators||(or),&&(and)and!

(not).
Conditionalexpressionsselectbetweentwosubexpressionsundercontrolofathird

subexpression.

Next:TypingandComparison,Up:TruthValuesandConditions[Contents][Index]
6.3.1TrueandFalseinawk
Manyprogramminglanguageshaveaspecialrepresentationfortheconceptsoftrueandfalse.Suchlanguages
usuallyusethespecialconstantstrueandfalse,orperhapstheiruppercaseequivalents.However,awkisdifferent.It
borrowsaverysimpleconceptoftrueandfalsefromC.Inawk,anynonzeronumericvalueoranynonemptystring
valueistrue.Anyothervalue(zeroorthenullstring,"")isfalse.ThefollowingprogramprintsAstrangetruth
valuethreetimes:
BEGIN{
if(3.1415927)
print"Astrangetruthvalue"
if("FourScoreAndSevenYearsAgo")
print"Astrangetruthvalue"
if(j=57)
print"Astrangetruthvalue"
}

Thereisasurprisingconsequenceofthenonzeroornonnullrule:thestringconstant"0"isactuallytrue,becauseitis
nonnull.(d.c.)
Next:BooleanOps,Previous:TruthValues,Up:TruthValuesandConditions[Contents][Index]
6.3.2VariableTypingandComparisonExpressions
TheGuideisdefinitive.Realityisfrequentlyinaccurate.
DouglasAdams,TheHitchhikersGuidetotheGalaxy
Unlikeinotherprogramminglanguages,inawkvariablesdonothaveafixedtype.Instead,theycanbeeitheranumber
orastring,dependinguponthevaluethatisassignedtothem.Welooknowathowvariablesaretyped,andhowawk
comparesvariables.
https://www.gnu.org/software/gawk/manual/gawk.html

114/479

29/7/2016

TheGNUAwkUsersGuide

VariableTyping:
Stringtypeversusnumerictype.
ComparisonOperators:
Thecomparisonoperators.
POSIXStringComparison: StringcomparisonwithPOSIXrules.
Next:ComparisonOperators,Up:TypingandComparison[Contents][Index]
6.3.2.1StringTypeversusNumericType
ThePOSIXstandardintroducedtheconceptofanumericstring,whichissimplyastringthatlookslikeanumberfor
example,"+2".Thisconceptisusedfordeterminingthetypeofavariable.Thetypeofthevariableisimportant
becausethetypesoftwovariablesdeterminehowtheyarecompared.Variabletypingfollowstheserules:
Anumericconstantortheresultofanumericoperationhasthenumericattribute.
Astringconstantortheresultofastringoperationhasthestringattribute.
Fields,getlineinput,FILENAME,ARGVelements,ENVIRONelements,andtheelementsofanarraycreatedby
match(),split(),andpatsplit()thatarenumericstringshavethestrnumattribute.Otherwise,theyhavethe
stringattribute.Uninitializedvariablesalsohavethestrnumattribute.
Attributespropagateacrossassignmentsbutarenotchangedbyanyuse.
Thelastruleisparticularlyimportant.Inthefollowingprogram,ahasnumerictype,eventhoughitislaterusedina
stringoperation:
BEGIN{
a=12.345
b=a"isacutenumber"
printb
}

Whentwooperandsarecompared,eitherstringcomparisonornumericcomparisonmaybeused.Thisdependsuponthe
attributesoftheoperands,accordingtothefollowingsymmetricmatrix:
+
|STRINGNUMERICSTRNUM
+
|
STRING|stringstringstring
|
NUMERIC|stringnumericnumeric
|
STRNUM|stringnumericnumeric
+
Thebasicideaisthatuserinputthatlooksnumericandonlyuserinputshouldbetreatedasnumeric,eventhoughit
isactuallymadeofcharactersandisthereforealsoastring.Thus,forexample,thestringconstant"+3.14",whenit
appearsinprogramsourcecode,isastringeventhoughitlooksnumericandisnevertreatedasanumberfor
comparisonpurposes.
Inshort,whenoneoperandisapurestring,suchasastringconstant,thenastringcomparisonisperformed.
Otherwise,anumericcomparisonisperformed.
Thispointbearsadditionalemphasis:Alluserinputismadeofcharacters,andsoisfirstandforemostofstringtype
inputstringsthatlooknumericareadditionallygiventhestrnumattribute.Thus,thesixcharacterinputstring+3.14
receivesthestrnumattribute.Incontrast,theeightcharacters"+3.14"appearinginprogramtextcompriseastring
constant.Thefollowingexamplesprint1whenthecomparisonbetweenthetwodifferentconstantsistrue,and0
otherwise:
$echo'+3.14'|awk'{print($0=="+3.14")}'True
|1
$echo'+3.14'|awk'{print($0=="+3.14")}'False
https://www.gnu.org/software/gawk/manual/gawk.html

115/479

29/7/2016

TheGNUAwkUsersGuide

|0
$echo'+3.14'|awk'{print($0=="3.14")}'False
|0
$echo'+3.14'|awk'{print($0==3.14)}'True
|1
$echo'+3.14'|awk'{print($1=="+3.14")}'False
|0
$echo'+3.14'|awk'{print($1=="+3.14")}'True
|1
$echo'+3.14'|awk'{print($1=="3.14")}'False
|0
$echo'+3.14'|awk'{print($1==3.14)}'True
|1

Next:POSIXStringComparison,Previous:VariableTyping,Up:TypingandComparison[Contents][Index]
6.3.2.2ComparisonOperators
Comparisonexpressionscomparestringsornumbersforrelationshipssuchasequality.Theyarewrittenusingrelational
operators,whichareasupersetofthoseinC.Table6.3describesthem.
Expression
x<y
x<=y
x>y
x>=y
x==y
x!=y
x~y
x!~y
subscriptinarray

Result
Trueifxislessthany
Trueifxislessthanorequaltoy
Trueifxisgreaterthany
Trueifxisgreaterthanorequaltoy
Trueifxisequaltoy
Trueifxisnotequaltoy
Trueifthestringxmatchestheregexpdenotedbyy
Trueifthestringxdoesnotmatchtheregexpdenotedbyy
Trueifthearrayarrayhasanelementwiththesubscript
subscript

Table6.3:Relationaloperators
Comparisonexpressionshavethevalueoneiftrueandzeroiffalse.Whencomparingoperandsofmixedtypes,numeric
operandsareconvertedtostringsusingthevalueofCONVFMT(seeConversion).
Stringsarecomparedbycomparingthefirstcharacterofeach,thenthesecondcharacterofeach,andsoon.Thus,"10"
islessthan"9".Iftherearetwostringswhereoneisaprefixoftheother,theshorterstringislessthanthelongerone.
Thus,"abc"islessthan"abcd".
Itisveryeasytoaccidentallymistypethe==operatorandleaveoffoneofthe=characters.Theresultisstillvalid
awkcode,buttheprogramdoesnotdowhatisintended:
if(a=b)#oops!shouldbea==b

else

Unlessbhappenstobezeroorthenullstring,theifpartofthetestalwayssucceeds.Becausetheoperatorsareso
similar,thiskindoferrorisverydifficulttospotwhenscanningthesourcecode.
Thefollowinglistofexpressionsillustratesthekindsofcomparisonsawkperforms,aswellaswhattheresultofeach
comparisonis:
1.5<=2.0

https://www.gnu.org/software/gawk/manual/gawk.html

116/479

29/7/2016

TheGNUAwkUsersGuide

Numericcomparison(true)
"abc">="xyz"

Stringcomparison(false)
1.5!="+2"

Stringcomparison(true)
"1e2"<"3"

Stringcomparison(true)
a=2;b="2"
a==b

Stringcomparison(true)
a=2;b="+2"
a==b

Stringcomparison(false)
Inthisexample:
$echo1e23|awk'{print($1<$2)?"true":"false"}'
|false

theresultisfalsebecauseboth$1and$2areuserinput.Theyarenumericstringsthereforebothhavethestrnum
attribute,dictatinganumericcomparison.Thepurposeofthecomparisonrulesandtheuseofnumericstringsisto
attempttoproducethebehaviorthatisleastsurprising,whilestilldoingtherightthing.
Stringcomparisonsandregularexpressioncomparisonsareverydifferent.Forexample:
x=="foo"

hasthevalueone,oristrueifthevariablexispreciselyfoo.Bycontrast:
x~/foo/

hasthevalueoneifxcontainsfoo,suchas"Oh,whatafoolamI!".
Therighthandoperandofthe~and!~operatorsmaybeeitheraregexpconstant(//)oranordinaryexpression.In
thelattercase,thevalueoftheexpressionasastringisusedasadynamicregexp(seeRegexpUsagealsosee
ComputedRegexps).
Aconstantregularexpressioninslashesbyitselfisalsoanexpression./regexp/isanabbreviationforthefollowing
comparisonexpression:
$0~/regexp/

Onespecialplacewhere/foo/isnotanabbreviationfor$0~/foo/iswhenitistherighthandoperandof~or!~.
SeeUsingConstantRegexps,wherethisisdiscussedinmoredetail.
Previous:ComparisonOperators,Up:TypingandComparison[Contents][Index]
6.3.2.3StringComparisonwithPOSIXRules
ThePOSIXstandardsaysthatstringcomparisonisperformedbasedonthelocalescollatingorder.Thisistheorderin
whichcharacterssort,asdefinedbythelocale(formorediscussion,seeLocales).Thisorderisusuallyverydifferent
https://www.gnu.org/software/gawk/manual/gawk.html

117/479

29/7/2016

TheGNUAwkUsersGuide

fromtheresultsobtainedwhendoingstraightcharacterbycharactercomparison.34
Becausethisbehaviordiffersconsiderablyfromexistingpractice,gawkonlyimplementsitwheninPOSIXmode(see
Options).Hereisanexampletoillustratethedifference,inanen_US.UTF8locale:
$gawk'BEGIN{printf("ABC<abc=%s\n",
>("ABC"<"abc"?"TRUE":"FALSE"))}'
|ABC<abc=TRUE
$gawkposix'BEGIN{printf("ABC<abc=%s\n",
>("ABC"<"abc"?"TRUE":"FALSE"))}'
|ABC<abc=FALSE

Next:ConditionalExp,Previous:TypingandComparison,Up:TruthValuesandConditions[Contents][Index]
6.3.3BooleanExpressions
ABooleanexpressionisacombinationofcomparisonexpressionsormatchingexpressions,usingtheBooleanoperators
or(||),and(&&),andnot(!),alongwithparenthesestocontrolnesting.ThetruthvalueoftheBoolean
expressioniscomputedbycombiningthetruthvaluesofthecomponentexpressions.Booleanexpressionsarealso
referredtoaslogicalexpressions.Thetermsareequivalent.
Booleanexpressionscanbeusedwherevercomparisonandmatchingexpressionscanbeused.Theycanbeusedinif,
while,do,andforstatements(seeStatements).Theyhavenumericvalues(oneiftrue,zeroiffalse)thatcomeintoplay
iftheresultoftheBooleanexpressionisstoredinavariableorusedinarithmetic.
Inaddition,everyBooleanexpressionisalsoavalidpattern,soyoucanuseoneasapatterntocontroltheexecutionof
rules.TheBooleanoperatorsare:
boolean1&&boolean2

Trueifbothboolean1andboolean2aretrue.Forexample,thefollowingstatementprintsthecurrentinputrecord
ifitcontainsbotheduandli:
if($0~/edu/&&$0~/li/)print

Thesubexpressionboolean2isevaluatedonlyifboolean1istrue.Thiscanmakeadifferencewhenboolean2
containsexpressionsthathavesideeffects.Inthecaseof$0~/foo/&&($2==bar++),thevariablebarisnot
incrementedifthereisnosubstringfoointherecord.
boolean1||boolean2

Trueifatleastoneofboolean1orboolean2istrue.Forexample,thefollowingstatementprintsallrecordsinthe
inputthatcontaineithereduorli:
if($0~/edu/||$0~/li/)print

Thesubexpressionboolean2isevaluatedonlyifboolean1isfalse.Thiscanmakeadifferencewhenboolean2
containsexpressionsthathavesideeffects.(Thus,thistestneverreallydistinguishesrecordsthatcontainboth
eduandliassoonaseduismatched,thefulltestsucceeds.)
!boolean

Trueifbooleanisfalse.Forexample,thefollowingprogramprintsnohome!intheunusualeventthattheHOME
environmentvariableisnotdefined:
BEGIN{if(!("HOME"inENVIRON))
print"nohome!"}

(TheinoperatorisdescribedinReferencetoElements.)
https://www.gnu.org/software/gawk/manual/gawk.html

118/479

29/7/2016

TheGNUAwkUsersGuide

The&&and||operatorsarecalledshortcircuitoperatorsbecauseofthewaytheywork.Evaluationofthefull
expressionisshortcircuitediftheresultcanbedeterminedpartwaythroughitsevaluation.
Statementsthatendwith&&or||canbecontinuedsimplybyputtinganewlineafterthem.Butyoucannotputa
newlineinfrontofeitheroftheseoperatorswithoutusingbackslashcontinuation(seeStatements/Lines).
Theactualvalueofanexpressionusingthe!operatoriseitheroneorzero,dependinguponthetruthvalueofthe
expressionitisappliedto.The!operatorisoftenusefulforchangingthesenseofaflagvariablefromfalsetotrueand
backagain.Forexample,thefollowingprogramisonewaytoprintlinesinbetweenspecialbracketinglines:
$1=="START"{interested=!interested;next}
interested{print}
$1=="END"{interested=!interested;next}

Thevariableinterested,aswithallawkvariables,startsoutinitializedtozero,whichisalsofalse.Whenalineisseen
whosefirstfieldisSTART,thevalueofinterestedistoggledtotrue,using!.Thenextruleprintslinesaslongas
interestedistrue.WhenalineisseenwhosefirstfieldisEND,interestedistoggledbacktofalse.35
Mostcommonly,the!operatorisusedintheconditionsofifandwhilestatements,whereitoftenmakesmoresense
tophrasethelogicinthenegative:
if(!somecondition||someothercondition){
dowhateverprocessing
}

NOTE:ThenextstatementisdiscussedinNextStatement.nexttellsawktoskiptherestoftherules,get
thenextrecord,andstartprocessingtherulesoveragainatthetop.Thereasonitsthereistoavoidprinting
thebracketingSTARTandENDlines.
Previous:BooleanOps,Up:TruthValuesandConditions[Contents][Index]
6.3.4ConditionalExpressions
Aconditionalexpressionisaspecialkindofexpressionthathasthreeoperands.Itallowsyoutouseoneexpressions
valuetoselectoneoftwootherexpressions.TheconditionalexpressioninawkisthesameasintheClanguage,as
shownhere:
selector?iftrueexp:iffalseexp

Therearethreesubexpressions.Thefirst,selector,isalwayscomputedfirst.Ifitistrue(notzeroornotnull),thenif
trueexpiscomputednext,anditsvaluebecomesthevalueofthewholeexpression.Otherwise,iffalseexpiscomputed
next,anditsvaluebecomesthevalueofthewholeexpression.Forexample,thefollowingexpressionproducesthe
absolutevalueofx:
x>=0?x:x

Eachtimetheconditionalexpressioniscomputed,onlyoneofiftrueexpandiffalseexpisusedtheotherisignored.
Thisisimportantwhentheexpressionshavesideeffects.Forexample,thisconditionalexpressionexamineselementi
ofeitherarrayaorarrayb,andincrementsi:
x==y?a[i++]:b[i++]

Thisisguaranteedtoincrementiexactlyonce,becauseeachtimeonlyoneofthetwoincrementexpressionsis
executedandtheotherisnot.SeeArrays,formoreinformationaboutarrays.
Asaminorgawkextension,astatementthatuses?:canbecontinuedsimplybyputtinganewlineaftereither
character.However,puttinganewlineinfrontofeithercharacterdoesnotworkwithoutusingbackslashcontinuation
(seeStatements/Lines).Ifposixisspecified(seeOptions),thisextensionisdisabled.
https://www.gnu.org/software/gawk/manual/gawk.html

119/479

29/7/2016

TheGNUAwkUsersGuide

Next:Precedence,Previous:TruthValuesandConditions,Up:Expressions[Contents][Index]

6.4FunctionCalls
Afunctionisanameforaparticularcalculation.Thisenablesyoutoaskforitbynameatanypointintheprogram.For
example,thefunctionsqrt()computesthesquarerootofanumber.
Afixedsetoffunctionsarebuiltin,whichmeanstheyareavailableineveryawkprogram.Thesqrt()functionisoneof
these.SeeBuiltin,foralistofbuiltinfunctionsandtheirdescriptions.Inaddition,youcandefinefunctionsforusein
yourprogram.SeeUserdefined,forinstructionsonhowtodothis.Finally,gawkletsyouwritefunctionsinCorC++
thatmaybecalledfromyourprogram(seeDynamicExtensions).
Thewaytouseafunctioniswithafunctioncallexpression,whichconsistsofthefunctionnamefollowedimmediately
byalistofargumentsinparentheses.Theargumentsareexpressionsthatprovidetherawmaterialsforthefunctions
calculations.Whenthereismorethanoneargument,theyareseparatedbycommas.Iftherearenoarguments,justwrite
()afterthefunctionname.Thefollowingexamplesshowfunctioncallswithandwithoutarguments:
sqrt(x^2+y^2)oneargument
atan2(y,x)twoarguments
rand()noarguments

CAUTION:Donotputanyspacebetweenthefunctionnameandtheopeningparenthesis!Auserdefined
functionnamelooksjustlikethenameofavariableaspacewouldmaketheexpressionlooklike
concatenationofavariablewithanexpressioninsideparentheses.Withbuiltinfunctions,spacebeforethe
parenthesisisharmless,butitisbestnottogetintothehabitofusingspacetoavoidmistakeswithuser
definedfunctions.
Eachfunctionexpectsaparticularnumberofarguments.Forexample,thesqrt()functionmustbecalledwithasingle
argument,thenumberofwhichtotakethesquareroot:
sqrt(argument)

Someofthebuiltinfunctionshaveoneormoreoptionalarguments.Ifthoseargumentsarenotsupplied,thefunctions
useareasonabledefaultvalue.SeeBuiltin,forfulldetails.Ifargumentsareomittedincallstouserdefinedfunctions,
thenthoseargumentsaretreatedaslocalvariables.Suchlocalvariablesactliketheemptystringifreferencedwherea
stringvalueisrequired,andlikezeroifreferencedwhereanumericvalueisrequired(seeUserdefined).
Asanadvancedfeature,gawkprovidesindirectfunctioncalls,whichisawaytochoosethefunctiontocallatruntime,
insteadofwhenyouwritethesourcecodetoyourprogram.WedeferdiscussionofthisfeatureuntillaterseeIndirect
Calls.
Likeeveryotherexpression,thefunctioncallhasavalue,oftencalledthereturnvalue,whichiscomputedbythe
functionbasedontheargumentsyougiveit.Inthisexample,thereturnvalueofsqrt(argument)isthesquarerootof
argument.Thefollowingprogramreadsnumbers,onenumberperline,andprintsthesquarerootofeachone:
$awk'{print"Thesquarerootof",$1,"is",sqrt($1)}'
1
|Thesquarerootof1is1
3
|Thesquarerootof3is1.73205
5
|Thesquarerootof5is2.23607
Ctrld

Afunctioncanalsohavesideeffects,suchasassigningvaluestocertainvariablesordoingI/O.Thisprogramshows
howthematch()function(seeStringFunctions)changesthevariablesRSTARTandRLENGTH:
{
if(match($1,$2))
printRSTART,RLENGTH
else
https://www.gnu.org/software/gawk/manual/gawk.html

120/479

29/7/2016

TheGNUAwkUsersGuide

print"nomatch"
}

Hereisasamplerun:
$awkfmatchit.awk
aaccddc+
|32
foobar
|nomatch
abcdefge
|51

Next:Locales,Previous:FunctionCalls,Up:Expressions[Contents][Index]

6.5OperatorPrecedence(HowOperatorsNest)
Operatorprecedencedetermineshowoperatorsaregroupedwhendifferentoperatorsappearclosebyinoneexpression.
Forexample,*hashigherprecedencethan+thus,a+b*cmeanstomultiply bandc,andthenaddatothe
product(i.e.,a+(b*c)).
Thenormalprecedenceoftheoperatorscanbeoverruledbyusingparentheses.Thinkoftheprecedencerulesassaying
wheretheparenthesesareassumedtobe.Infact,itiswisetoalwaysuseparentheseswheneverthereisanunusual
combinationofoperators,becauseotherpeoplewhoreadtheprogrammaynotrememberwhattheprecedenceisinthis
case.Evenexperiencedprogrammersoccasionallyforgettheexactrules,whichleadstomistakes.Explicitparentheses
helppreventanysuchmistakes.
Whenoperatorsofequalprecedenceareusedtogether,theleftmostoperatorgroupsfirst,exceptfortheassignment,
conditional,andexponentiationoperators,whichgroupintheoppositeorder.Thus,ab+cgroupsas(ab)+c
anda=b=cgroupsasa=(b=c).
Normallytheprecedenceofprefixunaryoperatorsdoesnotmatter,becausethereisonlyonewaytointerpretthem:
innermostfirst.Thus,$++imeans$(++i)and++$xmeans++($x).However,whenanotheroperatorfollowsthe
operand,thentheprecedenceoftheunaryoperatorscanmatter.$x^2means($x)^2,butx^2means(x^2),
becausehaslowerprecedencethan^,whereas$hashigherprecedence.Also,operatorscannotbecombinedina
waythatviolatestheprecedencerulesforexample,$$0++isnotavalidexpressionbecausethefirst$hashigher
precedencethanthe++toavoidtheproblemtheexpressioncanberewrittenas$($0++).
Thislistpresentsawksoperators,inorderofhighesttolowestprecedence:
()

Grouping.
$

Fieldreference.
++

Increment,decrement.
^**

Exponentiation.Theseoperatorsgrouprighttoleft.
+!

Unaryplus,minus,logicalnot.
*/%
https://www.gnu.org/software/gawk/manual/gawk.html

121/479

29/7/2016

TheGNUAwkUsersGuide

Multiplication,division,remainder.
+

Addition,subtraction.
Stringconcatenation
Thereisnospecialsymbolforconcatenation.Theoperandsaresimplywrittensidebyside(seeConcatenation).
<<===!=>>=>>||&

Relationalandredirection.Therelationaloperatorsandtheredirectionshavethesameprecedencelevel.
Characterssuchas>servebothasrelationalsandasredirectionsthecontextdistinguishesbetweenthetwo
meanings.
NotethattheI/Oredirectionoperatorsinprintandprintfstatementsbelongtothestatementlevel,notto
expressions.Theredirectiondoesnotproduceanexpressionthatcouldbetheoperandofanotheroperator.Asa
result,itdoesnotmakesensetousearedirectionoperatornearanotheroperatoroflowerprecedencewithout
parentheses.Suchcombinations(e.g.,printfoo>a?b:c)resultinsyntaxerrors.Thecorrectwaytowrite
thisstatementisprintfoo>(a?b:c).
~!~

Matching,nonmatching.
in

Arraymembership.
&&

Logicaland.
||

Logicalor.
?:

Conditional.Thisoperatorgroupsrighttoleft.
=+==*=/=%=^=**=

Assignment.Theseoperatorsgrouprighttoleft.
NOTE:The|&,**,and**=operatorsarenotspecifiedbyPOSIX.Formaximumportability,donot
usethem.
Next:ExpressionsSummary,Previous:Precedence,Up:Expressions[Contents][Index]

6.6WhereYouAreMakesaDifference
Modernsystemssupportthenotionoflocales:awaytotellthesystemaboutthelocalcharactersetandlanguage.The
ISOCstandarddefinesadefault"C"locale,whichisanenvironmentthatistypicalofwhatmanyCprogrammersare
usedto.
Onceuponatime,thelocalesettingusedtoaffectregexpmatching,butthisisnolongertrue(seeRangesandLocales).
Localescanaffectrecordsplitting.ForthenormalcaseofRS="\n",thelocaleislargelyirrelevant.Forothersingle
https://www.gnu.org/software/gawk/manual/gawk.html

122/479

29/7/2016

TheGNUAwkUsersGuide

characterrecordseparators,settingLC_ALL=Cintheenvironmentwillgiveyoumuchbetterperformancewhenreading
records.Otherwise,gawkhastomakeseveralfunctioncalls,perinputcharacter,tofindtherecordterminator.
Localescanaffecthowdatesandtimesareformatted(seeTimeFunctions).Forexample,acommonwaytoabbreviate
thedateSeptember4,2015,intheUnitedStatesis9/4/15.InmanycountriesinEurope,however,itisabbreviated
4.9.15.Thus,the%xspecificationina "US"localemightproduce9/4/15,whileina"EUROPE"locale,itmight
produce4.9.15.
AccordingtoPOSIX,stringcomparisonisalsoaffectedbylocales(similartoregularexpressions).Thedetailsare
presentedinPOSIXStringComparison.
Finally,thelocaleaffectsthevalueofthedecimalpointcharacterusedwhengawkparsesinputdata.Thisisdiscussedin
detailinConversion.
Previous:Locales,Up:Expressions[Contents][Index]

6.7Summary
Expressionsarethebasicelementsofcomputationinprograms.Theyarebuiltfromconstants,variables,function
calls,andcombinationsofthevariouskindsofvalueswithoperators.
awksuppliesthreekindsofconstants:numeric,string,andregexp.gawkletsyouspecifynumericconstantsinoctal
andhexadecimal(bases8and16)aswellasdecimal(base10).Incertaincontexts,astandaloneregexpconstant
suchas/foo/hasthesamemeaningas$0~/foo/.
Variablesholdvaluesbetweenusesincomputations.Anumberofbuiltinvariablesprovideinformationtoyour
awkprogram,andanumberofothersletyoucontrolhowawkbehaves.
Numbersareautomaticallyconvertedtostrings,andstringstonumbers,asneededbyawk.Numericvaluesare
convertedasiftheywereformattedwithsprintf()usingtheformatinCONVFMT.Localescaninfluencethe
conversions.
awkprovidestheusualarithmeticoperators(addition,subtraction,multiplication,division,modulus),andunary
plusandminus.Italsoprovidescomparisonoperators,Booleanoperators,anarraymembershiptestingoperator,
andregexpmatchingoperators.Stringconcatenationisaccomplishedbyplacingtwoexpressionsnexttoeach
otherthereisnoexplicitoperator.Thethreeoperand?:operatorprovidesanifelsetestwithinexpressions.
Assignmentoperatorsprovideconvenientshorthandsforcommonarithmeticoperations.
Inawk,avalueisconsideredtobetrueifitisnonzeroornonnull.Otherwise,thevalueisfalse.
Avariablestypeissetuponeachassignmentandmaychangeoveritslifetime.Thetypedetermineshowit
behavesincomparisons(stringornumeric).
Functioncallsreturnavaluethatmaybeusedaspartofalargerexpression.Expressionsusedtopassparameter
valuesarefullyevaluatedbeforethefunctioniscalled.awkprovidesbuiltinanduserdefinedfunctionsthisis
describedinFunctions.
Operatorprecedencespecifiestheorderinwhichoperationsareperformed,unlessexplicitlyoverriddenby
parentheses.awksoperatorprecedenceiscompatiblewiththatofC.
Localescanaffecttheformatofdataasoutputbyanawkprogram,andoccasionallytheformatfordatareadas
input.
Next:Arrays,Previous:Expressions,Up:Top[Contents][Index]

7Patterns,Actions,andVariables
Asyouhavealreadyseen,eachawkstatementconsistsofapatternwithanassociatedaction.Thischapterdescribeshow
youbuildpatternsandactions,whatkindsofthingsyoucandowithinactions,andawkspredefinedvariables.
Thepatternactionrulesandthestatementsavailableforusewithinactionsformthecoreofawkprogramming.Ina
sense,everythingcovereduptoherehasbeenthefoundationthatprogramsarebuiltontopof.Nowitstimetostart
buildingsomethinguseful.
https://www.gnu.org/software/gawk/manual/gawk.html

123/479

29/7/2016

TheGNUAwkUsersGuide

PatternOverview:
Whatgoesintoapattern.
UsingShellVariables: Howtouseshellvariableswithawk.
ActionOverview:
Whatgoesintoanaction.
Statements:
Describesthevariouscontrolstatementsindetail.
BuiltinVariables:
Summarizesthepredefinedvariables.
PatternActionSummary: PatternsandActionssummary.
Next:UsingShellVariables,Up:PatternsandActions[Contents][Index]

7.1PatternElements
RegexpPatterns:
Usingregexpsaspatterns.
ExpressionPatterns:
Anyexpressioncanbeusedasapattern.
Ranges:
Pairsofpatternsspecifyrecordranges.
BEGIN/END:
Specifyinginitializationandcleanuprules.
BEGINFILE/ENDFILE: Twospecialpatternsforadvancedcontrol.
Empty:
Theemptypattern,whichmatcheseveryrecord.
Patternsinawkcontroltheexecutionofrulesaruleisexecutedwhenitspatternmatchesthecurrentinputrecord.The
followingisasummaryofthetypesofawkpatterns:
/regularexpression/

Aregularexpression.Itmatcheswhenthetextoftheinputrecordfitstheregularexpression.(SeeRegexp.)
expression

Asingleexpression.Itmatcheswhenitsvalueisnonzero(ifanumber)ornonnull(ifastring).(SeeExpression
Patterns.)
begpat,endpat

Apairofpatternsseparatedbyacomma,specifyingarangeofrecords.Therangeincludesboththeinitialrecord
thatmatchesbegpatandthefinalrecordthatmatchesendpat.(SeeRanges.)
BEGIN
END

Specialpatternsforyoutosupplystartuporcleanupactionsforyourawkprogram.(SeeBEGIN/END.)
BEGINFILE
ENDFILE

Specialpatternsforyoutosupplystartuporcleanupactionstobedoneonaperfilebasis.(See
BEGINFILE/ENDFILE.)
empty

Theemptypatternmatcheseveryinputrecord.(SeeEmpty.)
Next:ExpressionPatterns,Up:PatternOverview[Contents][Index]
7.1.1RegularExpressionsasPatterns
Regularexpressionsareoneofthefirstkindsofpatternspresentedinthisbook.Thiskindofpatternissimplyaregexp
constantinthepatternpartofarule.Itsmeaningis$0~/pattern/.Thepatternmatcheswhentheinputrecord
matchestheregexp.Forexample:
https://www.gnu.org/software/gawk/manual/gawk.html

124/479

29/7/2016

TheGNUAwkUsersGuide

/foo|bar|baz/{buzzwords++}
END{printbuzzwords,"buzzwordsseen"}

Next:Ranges,Previous:RegexpPatterns,Up:PatternOverview[Contents][Index]
7.1.2ExpressionsasPatterns
Anyawkexpressionisvalidasanawkpattern.Thepatternmatchesiftheexpressionsvalueisnonzero(ifanumber)or
nonnull(ifastring).Theexpressionisreevaluatedeachtimetheruleistestedagainstanewinputrecord.Ifthe
expressionusesfieldssuchas$1,thevaluedependsdirectlyonthenewinputrecordstextotherwise,itdependson
onlywhathashappenedsofarintheexecutionoftheawkprogram.
Comparisonexpressions,usingthecomparisonoperatorsdescribedinTypingandComparison,areaverycommonkind
ofpattern.Regexpmatchingandnonmatchingarealsoverycommonexpressions.Theleftoperandofthe~and!~
operatorsisastring.Therightoperandiseitheraconstantregularexpressionenclosedinslashes(/regexp/),orany
expressionwhosestringvalueisusedasadynamicregularexpression(seeComputedRegexps).Thefollowingexample
printsthesecondfieldofeachinputrecordwhosefirstfieldispreciselyli:
$awk'$1=="li"{print$2}'maillist

(Thereisnooutput,becausethereisnopersonwiththeexactnameli.)Contrastthiswiththefollowingregular
expressionmatch,whichacceptsanyrecordwithafirstfieldthatcontainsli:
$awk'$1~/li/{print$2}'maillist
|5555553
|5556699

Aregexpconstantasapatternisalsoaspecialcaseofanexpressionpattern.Theexpression/li/hasthevalueoneif
liappearsinthecurrentinputrecord.Thus,asapattern, /li/matchesanyrecordcontainingli.
Booleanexpressionsarealsocommonlyusedaspatterns.Whetherthepatternmatchesaninputrecorddependson
whetheritssubexpressionsmatch.Forexample,thefollowingcommandprintsalltherecordsinmaillistthatcontain
botheduandli:
$awk'/edu/&&/li/'maillist
|[email protected]

Thefollowingcommandprintsallrecordsinmaillistthatcontaineithereduorli(orboth,ofcourse):
$awk'/edu/||/li/'maillist
|[email protected]
|[email protected]
|[email protected]
|[email protected]
|[email protected]
|[email protected]

Thefollowingcommandprintsallrecordsinmaillistthatdonotcontainthestringli:
$awk'!/li/'maillist
|[email protected]
|[email protected]
|[email protected]
|[email protected]
|[email protected]
|[email protected]
|[email protected]

ThesubexpressionsofaBooleanoperatorinapatterncanbeconstantregularexpressions,comparisons,oranyother
awkexpressions.Rangepatternsarenotexpressions,sotheycannotappearinsideBooleanpatterns.Likewise,the
specialpatternsBEGIN,END,BEGINFILE,andENDFILE,whichnevermatchanyinputrecord,arenotexpressionsandcannot
appearinsideBooleanpatterns.
https://www.gnu.org/software/gawk/manual/gawk.html

125/479

29/7/2016

TheGNUAwkUsersGuide

TheprecedenceofthedifferentoperatorsthatcanappearinpatternsisdescribedinPrecedence.
Next:BEGIN/END,Previous:ExpressionPatterns,Up:PatternOverview[Contents][Index]
7.1.3SpecifyingRecordRangeswithPatterns
Arangepatternismadeoftwopatternsseparatedbyacomma,intheformbegpat,endpat.Itisusedtomatchranges
ofconsecutiveinputrecords.Thefirstpattern,begpat,controlswheretherangebegins,whileendpatcontrolswherethe
patternends.Forexample,thefollowing:
awk'$1=="on",$1=="off"'myfile

printseveryrecordinmyfilebetweenon/offpairs,inclusive.
Arangepatternstartsoutbymatchingbegpatagainsteveryinputrecord.Whenarecordmatchesbegpat,therange
patternisturnedon,andtherangepatternmatchesthisrecordaswell.Aslongastherangepatternstaysturnedon,it
automaticallymatcheseveryinputrecordread.Therangepatternalsomatchesendpatagainsteveryinputrecordwhen
thissucceeds,therangepatternisturnedoffagainforthefollowingrecord.Thentherangepatterngoesbackto
checkingbegpatagainsteachrecord.
Therecordthatturnsontherangepatternandtheonethatturnsitoffbothmatchtherangepattern.Ifyoudontwantto
operateontheserecords,youcanwriteifstatementsintherulesactiontodistinguishthemfromtherecordsyouare
interestedin.
Itispossibleforapatterntobeturnedonandoffbythesamerecord.Iftherecordsatisfiesbothconditions,thenthe
actionisexecutedforjustthatrecord.Forexample,supposethereistextbetweentwoidenticalmarkers(e.g.,the%
symbol),eachonitsownline,thatshouldbeignored.Afirstattemptwouldbetocombinearangepatternthatdescribes
thedelimitedtextwiththenextstatement(notdiscussedyet,seeNextStatement).Thiscausesawktoskipanyfurther
processingofthecurrentrecordandstartoveragainwiththenextinputrecord.Suchaprogramlookslikethis:
/^%$/,/^%$/{next}
{print}

Thisprogramfailsbecausetherangepatternisbothturnedonandturnedoffbythefirstline,whichjusthasa%onit.
Toaccomplishthistask,writetheprograminthefollowingmanner,usingaflag:
/^%$/{skip=!skip;next}
skip==1{next}#skiplineswith`skip'set

Inarangepattern,thecomma(,)hasthelowestprecedenceofalltheoperators(i.e.,itisevaluatedlast).Thus,the
followingprogramattemptstocombinearangepatternwithanother,simplertest:
echoYes|awk'/1/,/2/||/Yes/'

Theintentofthisprogramis(/1/,/2/)||/Yes/.However,awkinterpretsthisas/1/,(/2/||/Yes/).Thiscannot
bechangedorworkedaroundrangepatternsdonotcombinewithotherpatterns:
$echoYes|gawk'(/1/,/2/)||/Yes/'
errorgawk:cmd.line:1:(/1/,/2/)||/Yes/
errorgawk:cmd.line:1:^syntaxerror

Asaminorpointofinterest,althoughitispoorstyle,POSIXallowsyoutoputanewlineafterthecommainarange
pattern.(d.c.)
Next:BEGINFILE/ENDFILE,Previous:Ranges,Up:PatternOverview[Contents][Index]
7.1.4TheBEGINandENDSpecialPatterns
https://www.gnu.org/software/gawk/manual/gawk.html

126/479

29/7/2016

TheGNUAwkUsersGuide

Allthepatternsdescribedsofarareformatchinginputrecords.TheBEGINandENDspecialpatternsaredifferent.They
supplystartupandcleanupactionsforawkprograms.BEGINandENDrulesmusthaveactionsthereisnodefaultaction
fortheserulesbecausethereisnocurrentrecordwhentheyrun.BEGINandENDrulesareoftenreferredtoasBEGINand
ENDblocksbylongtimeawkprogrammers.
UsingBEGIN/END: HowandwhytouseBEGIN/ENDrules.
I/OAndBEGIN/END: I/OissuesinBEGIN/ENDrules.
Next:I/OAndBEGIN/END,Up:BEGIN/END[Contents][Index]
7.1.4.1StartupandCleanupActions
ABEGINruleisexecutedonceonly,beforethefirstinputrecordisread.Likewise,anENDruleisexecutedonceonly,
afteralltheinputisread.Forexample:
$awk'
>BEGIN{print"Analysisof\"li\""}
>/li/{++n}
>END{print"\"li\"appearsin",n,"records."}'maillist
|Analysisof"li"
|"li"appearsin4records.

Thisprogramfindsthenumberofrecordsintheinputfilemaillistthatcontainthestringli.TheBEGINruleprintsa
titleforthereport.ThereisnoneedtousetheBEGINruletoinitializethecounterntozero,asawkdoesthisautomatically
(seeVariables).Thesecondruleincrementsthevariableneverytimearecordcontainingthepatternliisread.The
ENDruleprintsthevalueofnattheendoftherun.
ThespecialpatternsBEGINandENDcannotbeusedinrangesorwithBooleanoperators(indeed,theycannotbeusedwith
anyoperators).AnawkprogrammayhavemultipleBEGINand/orENDrules.Theyareexecutedintheorderinwhichthey
appear:alltheBEGINrulesatstartupandalltheENDrulesattermination.BEGINandENDrulesmaybeintermixedwith
otherrules.Thisfeaturewasaddedinthe1987versionofawkandisincludedinthePOSIXstandard.Theoriginal
(1978)versionofawkrequiredtheBEGINruletobeplacedatthebeginningoftheprogram,theENDruletobeplacedat
theend,andonlyallowedoneofeach.Thisisnolongerrequired,butitisagoodideatofollowthistemplateintermsof
programorganizationandreadability.
MultipleBEGINandENDrulesareusefulforwritinglibraryfunctions,becauseeachlibraryfilecanhaveitsownBEGIN
and/orENDruletodoitsowninitializationand/orcleanup.Theorderinwhichlibraryfunctionsarenamedonthe
commandlinecontrolstheorderinwhichtheirBEGINandENDrulesareexecuted.Therefore,youhavetobecareful
whenwritingsuchrulesinlibraryfilessothattheorderinwhichtheyareexecuteddoesntmatter.SeeOptions,for
moreinformationonusinglibraryfunctions.SeeLibraryFunctions,foranumberofusefullibraryfunctions.
IfanawkprogramhasonlyBEGINrulesandnootherrules,thentheprogramexitsaftertheBEGINrulesarerun.36
However,ifanENDruleexists,thentheinputisread,eveniftherearenootherrulesintheprogram.Thisisnecessaryin
casetheENDrulecheckstheFNRandNRvariables.
Previous:UsingBEGIN/END,Up:BEGIN/END[Contents][Index]
7.1.4.2Input/OutputfromBEGINandENDRules
Thereareseveral(sometimessubtle)pointstobeawareofwhendoingI/OfromaBEGINorENDrule.Thefirsthastodo
withthevalueof$0inaBEGINrule.BecauseBEGINrulesareexecutedbeforeanyinputisread,theresimplyisnoinput
record,andthereforenofields,whenexecutingBEGINrules.Referencesto$0andthefieldsyieldanullstringorzero,
dependinguponthecontext.Onewaytogive$0arealvalueistoexecuteagetlinecommandwithoutavariable(see
Getline).Anotherwayissimplytoassignavalueto$0.
Thesecondpointissimilartothefirst,butfromtheotherdirection.Traditionally,duelargelytoimplementationissues,
$0andNFwereundefinedinsideanENDrule.ThePOSIXstandardspecifiesthatNFisavailableinanENDrule.Itcontains
https://www.gnu.org/software/gawk/manual/gawk.html

127/479

29/7/2016

TheGNUAwkUsersGuide

thenumberoffieldsfromthelastinputrecord.Mostprobablyduetoanoversight,thestandarddoesnotsaythat$0is
alsopreserved,althoughlogicallyonewouldthinkthatitshouldbe.Infact,allofBWKawk,mawk,andgawkpreservethe
valueof$0foruseinENDrules.Beaware,however,thatsomeotherimplementationsandmanyolderversionsofUnix
awkdonot.
Thethirdpointfollowsfromthefirsttwo.Themeaningofprintinsidea BEGINorENDruleisthesameasalways:
print$0.If$0isthenullstring,thenthisprintsanemptyrecord.Manylongtimeawkprogrammersuseanunadorned
printinBEGINandENDrules,tomeanprint"",relyingon$0beingnull.Althoughonemightgenerallygetaway
withthisinBEGINrules,itisaverybadideainENDrules,atleastingawk.Itisalsopoorstyle,becauseifanemptylineis
neededintheoutput,theprogramshouldprintoneexplicitly.
Finally,thenextandnextfilestatementsarenotallowedinaBEGINrule,becausetheimplicitreadarecordandmatch
againsttherulesloophasnotstartedyet.Similarly,thosestatementsarenotvalidinanENDrule,becausealltheinput
hasbeenread.(SeeNextStatement,andseeNextfileStatement,.)
Next:Empty,Previous:BEGIN/END,Up:PatternOverview[Contents][Index]
7.1.5TheBEGINFILEandENDFILESpecialPatterns
Thissectiondescribesagawkspecificfeature.
Twospecialkindsofrule,BEGINFILEandENDFILE,giveyouhooksintogawkscommandlinefileprocessingloop.As
withtheBEGINandENDrules(seeBEGIN/END),allBEGINFILErulesinaprogramaremerged,intheordertheyareread
bygawk,andallENDFILErulesaremergedaswell.
ThebodyoftheBEGINFILErulesisexecutedjustbeforegawkreadsthefirstrecordfromafile.FILENAMEissettothe
nameofthecurrentfile,andFNRissettozero.
TheBEGINFILEruleprovidesyoutheopportunitytoaccomplishtwotasksthatwouldotherwisebedifficultorimpossible
toperform:
Youcantestifthefileisreadable.Normally,itisafatalerrorifafilenamedonthecommandlinecannotbe
openedforreading.However,youcanbypassthefatalerrorandmoveontothenextfileonthecommandline.
YoudothisbycheckingiftheERRNOvariableisnottheemptystringifso,thengawkwasnotabletoopenthefile.
Inthiscase,yourprogramcanexecutethenextfilestatement(seeNextfileStatement).Thiscausesgawktoskip
thefileentirely.Otherwise,gawkexitswiththeusualfatalerror.
Ifyouhavewrittenextensionsthatmodifytherecordhandling(byinsertinganinputparserseeInputParsers),
youcaninvokethematthispoint,beforegawkhasstartedprocessingthefile.(Thisisaveryadvancedfeature,
currentlyusedonlybythegawkextlibproject.)
TheENDFILEruleiscalledwhengawkhasfinishedprocessingthelastrecordinaninputfile.Forthelastinputfile,itwill
becalledbeforeanyENDrules.TheENDFILEruleisexecutedevenforemptyinputfiles.
Normally,whenanerroroccurswhenreadinginputinthenormalinputprocessingloop,theerrorisfatal.However,if
anENDFILEruleispresent,theerrorbecomesnonfatal,andinsteadERRNOisset.Thismakesitpossibletocatchand
processI/Oerrorsattheleveloftheawkprogram.
Thenextstatement(seeNextStatement)isnotallowedinsideeitheraBEGINFILEoranENDFILErule.Thenextfile
statementisallowedonlyinsideaBEGINFILErule,notinsideanENDFILErule.
Thegetlinestatement(seeGetline)isrestrictedinsidebothBEGINFILEandENDFILE:onlyredirectedformsofgetline
areallowed.
BEGINFILEandENDFILEaregawkextensions.Inmostotherawkimplementations,orifgawkisincompatibilitymode(see

Options),theyarenotspecial.
https://www.gnu.org/software/gawk/manual/gawk.html

128/479

29/7/2016

TheGNUAwkUsersGuide

Previous:BEGINFILE/ENDFILE,Up:PatternOverview[Contents][Index]
7.1.6TheEmptyPattern
Anempty(i.e.,nonexistent)patternisconsideredtomatcheveryinputrecord.Forexample,theprogram:
awk'{print$1}'maillist

printsthefirstfieldofeveryrecord.
Next:ActionOverview,Previous:PatternOverview,Up:PatternsandActions[Contents][Index]

7.2UsingShellVariablesinPrograms
awkprogramsareoftenusedascomponentsinlargerprogramswritteninshell.Forexample,itisverycommontousea
shellvariabletoholdapatternthattheawkprogramsearchesfor.Therearetwowaystogetthevalueoftheshell
variableintothebodyoftheawkprogram.

Acommonmethodistouseshellquotingtosubstitutethevariablesvalueintotheprograminsidethescript.For
example,considerthefollowingprogram:
printf"Entersearchpattern:"
readpattern
awk"/$pattern/"'{nmatches++}
END{printnmatches,"found"}'/path/to/data

Theawkprogramconsistsoftwopiecesofquotedtextthatareconcatenatedtogethertoformtheprogram.Thefirstpart
isdoublequoted,whichallowssubstitutionofthepatternshellvariableinsidethequotes.Thesecondpartissingle
quoted.
Variablesubstitutionviaquotingworks,butcanpotentiallybemessy.Itrequiresagoodunderstandingoftheshells
quotingrules(seeQuoting),anditsoftendifficulttocorrectlymatchupthequoteswhenreadingtheprogram.
Abettermethodistouseawksvariableassignmentfeature(seeAssignmentOptions)toassigntheshellvariablesvalue
toanawkvariable.Thenusedynamicregexpstomatchthepattern(seeComputedRegexps).Thefollowingshowshow
toredothepreviousexampleusingthistechnique:
printf"Entersearchpattern:"
readpattern
awkvpat="$pattern"'$0~pat{nmatches++}
END{printnmatches,"found"}'/path/to/data

Now,theawkprogramisjustonesinglequotedstring.Theassignmentvpat="$pattern"stillrequiresdoublequotes,
incasethereiswhitespaceinthevalueof$pattern.Theawkvariablepatcouldbenamedpatterntoo,butthatwouldbe
moreconfusing.Usingavariablealsoprovidesmoreflexibility,asthevariablecanbeusedanywhereinsidethe
programforprinting,asanarraysubscript,orforanyotherusewithoutrequiringthequotingtricksateverypointin
theprogram.
Next:Statements,Previous:UsingShellVariables,Up:PatternsandActions[Contents][Index]

7.3Actions
Anawkprogramorscriptconsistsofaseriesofrulesandfunctiondefinitionsinterspersed.(Functionsaredescribed
later.SeeUserdefined.)Arulecontainsapatternandanaction,eitherofwhich(butnotboth)maybeomitted.The
purposeoftheactionistotellawkwhattodoonceamatchforthepatternisfound.Thus,inoutline,anawkprogram
generallylookslikethis:
https://www.gnu.org/software/gawk/manual/gawk.html

129/479

29/7/2016

TheGNUAwkUsersGuide

[pattern]{action}
pattern[{action}]

functionname(args){}

Anactionconsistsofoneormoreawkstatements,enclosedinbraces({}).Eachstatementspecifiesonethingtodo.
Thestatementsareseparatedbynewlinesorsemicolons.Thebracesaroundanactionmustbeusedeveniftheaction
containsonlyonestatement,orifitcontainsnostatementsatall.However,ifyouomittheactionentirely,omitthe
bracesaswell.Anomittedactionisequivalentto{print$0}:
/foo/{}matchfoo,donothingemptyaction
/foo/matchfoo,printtherecordomittedaction

Thefollowingtypesofstatementsaresupportedinawk:
Expressions
Callfunctionsorassignvaluestovariables(seeExpressions).Executingthiskindofstatementsimplycomputes
thevalueoftheexpression.Thisisusefulwhentheexpressionhassideeffects(seeAssignmentOps).
Controlstatements
Specifythecontrolflowofawkprograms.TheawklanguagegivesyouClikeconstructs(if,for,while,anddo)as
wellasafewspecialones(seeStatements).
Compoundstatements
Encloseoneormorestatementsinbraces.Acompoundstatementisusedinordertoputseveralstatements
togetherinthebodyofanif,while,do,orforstatement.
Inputstatements
Usethegetlinecommand(seeGetline).Alsosuppliedinawkarethenextstatement(seeNextStatement)andthe
nextfilestatement(seeNextfileStatement).
Outputstatements
Suchasprintandprintf.SeePrinting.
Deletionstatements
Fordeletingarrayelements.SeeDelete.
Next:BuiltinVariables,Previous:ActionOverview,Up:PatternsandActions[Contents][Index]

7.4ControlStatementsinActions
Controlstatements,suchasif,while,andsoon,controltheflowofexecutioninawkprograms.Mostofawkscontrol
statementsarepatternedaftersimilarstatementsinC.
Allthecontrolstatementsstartwithspecialkeywords,suchasifandwhile,todistinguishthemfromsimple
expressions.Manycontrolstatementscontainotherstatements.Forexample,theifstatementcontainsanother
statementthatmayormaynotbeexecuted.Thecontainedstatementiscalledthebody.Toincludemorethanone
statementinthebody,groupthemintoasinglecompoundstatementwithbraces,separatingthemwithnewlinesor
semicolons.
IfStatement:

Conditionallyexecutesomeawkstatements.

https://www.gnu.org/software/gawk/manual/gawk.html

130/479

29/7/2016

TheGNUAwkUsersGuide

WhileStatement: Loopuntilsomeconditionissatisfied.
DoStatement:
Dospecifiedactionwhileloopinguntilsomeconditionissatisfied.
ForStatement:
Anotherloopingstatement,thatprovidesinitializationandincrementclauses.
SwitchStatement: Switch/caseevaluationforconditionalexecutionofstatementsbasedonavalue.
BreakStatement: Immediatelyexittheinnermostenclosingloop.
ContinueStatement: Skiptotheendoftheinnermostenclosingloop.
NextStatement:
Stopprocessingthecurrentinputrecord.
NextfileStatement: Stopprocessingthecurrentfile.
ExitStatement:
Stopexecutionofawk.
Next:WhileStatement,Up:Statements[Contents][Index]
7.4.1TheifelseStatement
Theifelsestatementisawksdecisionmakingstatement.Itlookslikethis:
if(condition)thenbody[elseelsebody]

Theconditionisanexpressionthatcontrolswhattherestofthestatementdoes.Iftheconditionistrue,thenbodyis
executedotherwise,elsebodyisexecuted.Theelsepartofthestatementisoptional.Theconditionisconsideredfalse
ifitsvalueiszeroorthenullstringotherwise,theconditionistrue.Refertothefollowing:
if(x%2==0)
print"xiseven"
else
print"xisodd"

Inthisexample,iftheexpressionx%2==0istrue(i.e.,ifthevalueofxisevenlydivisiblebytwo),thenthefirst
printstatementisexecutedotherwise,thesecondprintstatementisexecuted.Iftheelsekeywordappearsonthesame
lineasthenbodyandthenbodyisnotacompoundstatement(i.e.,notsurroundedbybraces),thenasemicolonmust
separatethenbodyfromtheelse.Toillustratethis,thepreviousexamplecanberewrittenas:
if(x%2==0)print"xiseven";else
print"xisodd"

Ifthe;isleftout,awkcantinterpretthestatementanditproducesasyntaxerror.Dontactuallywriteprogramsthis
way,becauseahumanreadermightfailtoseetheelseifitisnotthefirstthingonitsline.
Next:DoStatement,Previous:IfStatement,Up:Statements[Contents][Index]
7.4.2ThewhileStatement
Inprogramming,aloopisapartofaprogramthatcanbeexecutedtwoormoretimesinsuccession.Thewhile
statementisthesimplestloopingstatementinawk.Itrepeatedlyexecutesastatementaslongasaconditionistrue.For
example:
while(condition)
body

bodyisastatementcalledthebodyoftheloop,andconditionisanexpressionthatcontrolshowlongtheloopkeeps
running.Thefirstthingthewhilestatementdoesistestthecondition.Iftheconditionistrue,itexecutesthestatement
body.Afterbodyhasbeenexecuted,conditionistestedagain,andifitisstilltrue,bodyexecutesagain.Thisprocess
repeatsuntiltheconditionisnolongertrue.Iftheconditionisinitiallyfalse,thebodyoftheloopneverexecutesand
awkcontinueswiththestatementfollowingtheloop.Thisexampleprintsthefirstthreefieldsofeachrecord,oneper
line:
awk'
https://www.gnu.org/software/gawk/manual/gawk.html

131/479

29/7/2016

TheGNUAwkUsersGuide

{
i=1
while(i<=3){
print$i
i++
}
}'inventoryshipped

Thebodyofthisloopisacompoundstatementenclosedinbraces,containingtwostatements.Theloopworksinthe
followingmanner:first,thevalueofiissettoone.Then,thewhilestatementtestswhetheriislessthanorequalto
three.Thisistruewheniequalsone,sotheithfieldisprinted.Thenthei++incrementsthevalueofiandtheloop
repeats.Theloopterminateswhenireachesfour.
Anewlineisnotrequiredbetweentheconditionandthebodyhowever,usingonemakestheprogramclearerunlessthe
bodyisacompoundstatementorelseisverysimple.Thenewlineaftertheopenbracethatbeginsthecompound
statementisnotrequiredeither,buttheprogramishardertoreadwithoutit.
Next:ForStatement,Previous:WhileStatement,Up:Statements[Contents][Index]
7.4.3ThedowhileStatement
Thedoloopisavariationofthewhileloopingstatement.Thedoloopexecutesthebodyonceandthenrepeatsthebody
aslongastheconditionistrue.Itlookslikethis:
do
body
while(condition)

Eveniftheconditionisfalseatthestart,thebodyexecutesatleastonce(andonlyonce,unlessexecutingbodymakes
conditiontrue).Contrastthiswiththecorrespondingwhilestatement:
while(condition)
body

Thisstatementdoesnotexecutethebodyevenonceiftheconditionisfalsetobeginwith.Thefollowingisanexample
ofadostatement:
{
i=1
do{
print$0
i++
}while(i<=10)
}

Thisprogramprintseachinputrecord10times.However,itisntaveryrealisticexample,becauseinthiscasean
ordinarywhilewoulddojustaswell.Thissituationreflectsactualexperienceonlyoccasionallyistherearealusefora
dostatement.
Next:SwitchStatement,Previous:DoStatement,Up:Statements[Contents][Index]
7.4.4TheforStatement
Theforstatementmakesitmoreconvenienttocountiterationsofaloop.Thegeneralformoftheforstatementlooks
likethis:
for(initialization;condition;increment)
body

Theinitialization,condition,andincrementpartsarearbitraryawkexpressions,andbodystandsforanyawkstatement.
https://www.gnu.org/software/gawk/manual/gawk.html

132/479

29/7/2016

TheGNUAwkUsersGuide

Theforstatementstartsbyexecutinginitialization.Then,aslongastheconditionistrue,itrepeatedlyexecutesbody
andthenincrement.Typically,initializationsetsavariabletoeitherzeroorone,incrementaddsonetoit,andcondition
comparesitagainstthedesirednumberofiterations.Forexample:
awk'
{
for(i=1;i<=3;i++)
print$i
}'inventoryshipped

Thisprintsthefirstthreefieldsofeachinputrecord,withonefieldperline.
Itisntpossibletosetmorethanonevariableintheinitializationpartwithoutusingamultipleassignmentstatement
suchasx=y=0.Thismakessenseonlyifalltheinitialvaluesareequal.(Butitispossibletoinitializeadditional
variablesbywritingtheirassignmentsasseparatestatementsprecedingtheforloop.)
Thesameistrueoftheincrementpart.Incrementingadditionalvariablesrequiresseparatestatementsattheendofthe
loop.TheCcompoundexpression,usingCscommaoperator,isusefulinthiscontext,butitisnotsupportedinawk.
Mostoften,incrementisanincrementexpression,asinthepreviousexample.Butthisisnotrequireditcanbeany
expressionwhatsoever.Forexample,thefollowingstatementprintsallthepowersoftwobetween1and100:
for(i=1;i<=100;i*=2)
printi

Ifthereisnothingtobedone,anyofthethreeexpressionsintheparenthesesfollowingtheforkeywordmaybe
omitted.Thus,for(;x>0;)isequivalenttowhile(x>0).Iftheconditionisomitted,itistreatedastrue,
effectivelyyieldinganinfiniteloop(i.e.,aloopthatneverterminates).
Inmostcases,aforloopisanabbreviationforawhileloop,asshownhere:
initialization
while(condition){
body
increment
}

Theonlyexceptioniswhenthecontinuestatement(seeContinueStatement)isusedinsidetheloop.Changingafor
statementtoawhilestatementinthiswaycanchangetheeffectofthecontinuestatementinsidetheloop.
Theawklanguagehasaforstatementinadditiontoawhilestatementbecauseaforloopisoftenbothlessworktotype
andmorenaturaltothinkof.Countingthenumberofiterationsisverycommoninloops.Itcanbeeasiertothinkofthis
countingaspartofloopingratherthanassomethingtodoinsidetheloop.
Thereisanalternativeversionoftheforloop,foriteratingoveralltheindicesofanarray:
for(iinarray)
dosomethingwitharray[i]

SeeScanninganArray,formoreinformationonthisversionoftheforloop.
Next:BreakStatement,Previous:ForStatement,Up:Statements[Contents][Index]
7.4.5TheswitchStatement
Thissectiondescribesagawkspecificfeature.Ifgawkisincompatibilitymode(seeOptions),itisnotavailable.
Theswitchstatementallowstheevaluationofanexpressionandtheexecutionofstatementsbasedonacasematch.
Casestatementsarecheckedforamatchintheordertheyaredefined.Ifnosuitablecaseisfound,thedefaultsectionis
executed,ifsupplied.
https://www.gnu.org/software/gawk/manual/gawk.html

133/479

29/7/2016

TheGNUAwkUsersGuide

Eachcasecontainsasingleconstant,beitnumeric,string,orregexp.Theswitchexpressionisevaluated,andtheneach
casesconstantiscomparedagainsttheresultinturn.Thetypeofconstantdeterminesthecomparison:numericorstring
dotheusualcomparisons.Aregexpconstantdoesaregularexpressionmatchagainstthestringvalueoftheoriginal
expression.Thegeneralformoftheswitchstatementlookslikethis:
switch(expression){
casevalueorregularexpression:
casebody
default:
defaultbody
}

ControlflowintheswitchstatementworksasitdoesinC.Onceamatchtoagivencaseismade,thecasestatement
bodiesexecuteuntilabreak,continue,next,nextfile,orexitisencountered,ortheendoftheswitchstatementitself.
Forexample:
while((c=getopt(ARGC,ARGV,"aksx"))!=1){
switch(c){
case"a":
#reportsizeofallfiles
all_files=TRUE;
break
case"k":
BLOCK_SIZE=1024#1Kblocksize
break
case"s":
#dosumsonly
sum_only=TRUE
break
case"x":
#don'tcrossfilesystems
fts_flags=or(fts_flags,FTS_XDEV)
break
case"?":
default:
usage()
break
}
}

Notethatifnoneofthestatementsspecifiedherehaltexecutionofamatchedcasestatement,executionfallsthroughto
thenextcaseuntilexecutionhalts.Inthisexample,thecasefor"?"fallsthroughtothedefaultcase,whichistocalla
functionnamedusage().(Thegetopt()functionbeingcalledhereisdescribedinGetoptFunction.)
Next:ContinueStatement,Previous:SwitchStatement,Up:Statements[Contents][Index]
7.4.6ThebreakStatement
Thebreakstatementjumpsoutoftheinnermostfor,while,ordoloopthatenclosesit.Thefollowingexamplefindsthe
smallestdivisorofanyinteger,andalsoidentifiesprimenumbers:
#findsmallestdivisorofnum
{
num=$1
for(divisor=2;divisor*divisor<=num;divisor++){
if(num%divisor==0)
break
}
if(num%divisor==0)
printf"Smallestdivisorof%dis%d\n",num,divisor
else
printf"%disprime\n",num
}
https://www.gnu.org/software/gawk/manual/gawk.html

134/479

29/7/2016

TheGNUAwkUsersGuide

Whentheremainderiszerointhefirstifstatement,awkimmediatelybreaksoutofthecontainingforloop.Thismeans
thatawkproceedsimmediatelytothestatementfollowingtheloopandcontinuesprocessing.(Thisisverydifferentfrom
theexitstatement,whichstopstheentireawkprogram.SeeExitStatement.)
Thefollowingprogramillustrateshowtheconditionofafororwhilestatementcouldbereplacedwithabreakinside
anif:
#findsmallestdivisorofnum
{
num=$1
for(divisor=2;;divisor++){
if(num%divisor==0){
printf"Smallestdivisorof%dis%d\n",num,divisor
break
}
if(divisor*divisor>num){
printf"%disprime\n",num
break
}
}
}

Thebreakstatementisalsousedtobreakoutoftheswitchstatement.ThisisdiscussedinSwitchStatement.
Thebreakstatementhasnomeaningwhenusedoutsidethebodyofalooporswitch.However,althoughitwasnever
documented,historicalimplementationsofawktreatedthebreakstatementoutsideofaloopasifitwereanext
statement(seeNextStatement).(d.c.)RecentversionsofBWKawknolongerallowthisusage,nordoesgawk.
Next:NextStatement,Previous:BreakStatement,Up:Statements[Contents][Index]
7.4.7ThecontinueStatement
Similartobreak,thecontinuestatementisusedonlyinsidefor,while,anddoloops.Itskipsovertherestoftheloop
body,causingthenextcyclearoundthelooptobeginimmediately.Contrastthiswithbreak,whichjumpsoutofthe
loopaltogether.
Thecontinuestatementinaforloopdirectsawktoskiptherestofthebodyoftheloopandresumeexecutionwiththe
incrementexpressionoftheforstatement.Thefollowingprogramillustratesthisfact:
BEGIN{
for(x=0;x<=20;x++){
if(x==5)
continue
printf"%d",x
}
print""
}

Thisprogramprintsallthenumbersfrom0to20exceptfor5,forwhichtheprintfisskipped.Becausetheincrement
x++isnotskipped, xdoesnotremainstuckat5.Contrasttheforloopfromthepreviousexamplewiththefollowing
whileloop:
BEGIN{
x=0
while(x<=20){
if(x==5)
continue
printf"%d",x
x++
}
print""
}
https://www.gnu.org/software/gawk/manual/gawk.html

135/479

29/7/2016

TheGNUAwkUsersGuide

Thisprogramloopsforeveroncexreaches5,becausetheincrement(x++)isneverreached.
Thecontinuestatementhasnospecialmeaningwithrespecttotheswitchstatement,nordoesithaveanymeaning
whenusedoutsidethebodyofaloop.Historicalversionsofawktreatedacontinuestatementoutsidealoopthesame
waytheytreatedabreakstatementoutsidealoop:asifitwereanextstatement(seeNextStatement).(d.c.)Recent
versionsofBWKawknolongerworkthisway,nordoesgawk.
Next:NextfileStatement,Previous:ContinueStatement,Up:Statements[Contents][Index]
7.4.8ThenextStatement
Thenextstatementforcesawktoimmediatelystopprocessingthecurrentrecordandgoontothenextrecord.This
meansthatnofurtherrulesareexecutedforthecurrentrecord,andtherestofthecurrentrulesactionisntexecuted.
Contrastthiswiththeeffectofthegetlinefunction(seeGetline).Thatalsocausesawktoreadthenextrecord
immediately,butitdoesnotaltertheflowofcontrolinanyway(i.e.,therestofthecurrentactionexecuteswithanew
inputrecord).
Atthehighestlevel,awkprogramexecutionisaloopthatreadsaninputrecordandthentestseachrulespatternagainst
it.Ifyouthinkofthisloopasaforstatementwhosebodycontainstherules,thenthenextstatementisanalogoustoa
continuestatement.Itskipstotheendofthebodyofthisimplicitloopandexecutestheincrement(whichreadsanother
record).
Forexample,supposeanawkprogramworksonlyonrecordswithfourfields,anditshouldntfailwhengivenbadinput.
Toavoidcomplicatingtherestoftheprogram,writeaweedoutrulenearthebeginning,inthefollowingmanner:
NF!=4{
printf("%s:%d:skipped:NF!=4\n",FILENAME,FNR)>"/dev/stderr"
next
}

Becauseofthenextstatement,theprogramssubsequentruleswontseethebadrecord.Theerrormessageisredirected
tothestandarderroroutputstream,aserrormessagesshouldbe.Formoredetail,seeSpecialFiles.
Ifthenextstatementcausestheendoftheinputtobereached,thenthecodeinanyENDrulesisexecuted.See
BEGIN/END.
ThenextstatementisnotallowedinsideBEGINFILEandENDFILErules.SeeBEGINFILE/ENDFILE.
AccordingtothePOSIXstandard,thebehaviorisundefinedifthenextstatementisusedinaBEGINorENDrule.gawk
treatsitasasyntaxerror.AlthoughPOSIXdoesnotdisallowit,mostotherawkimplementationsdontallowthenext
statementinsidefunctionbodies(seeUserdefined).Justaswithanyothernextstatement,anextstatementinsidea
functionbodyreadsthenextrecordandstartsprocessingitwiththefirstruleintheprogram.
Next:ExitStatement,Previous:NextStatement,Up:Statements[Contents][Index]
7.4.9ThenextfileStatement
Thenextfilestatementissimilartothenextstatement.However,insteadofabandoningprocessingofthecurrent
record,thenextfilestatementinstructsawktostopprocessingthecurrentdatafile.
Uponexecutionofthenextfilestatement,FILENAMEisupdatedtothenameofthenextdatafilelistedonthecommand
line,FNRisresettoone,andprocessingstartsoverwiththefirstruleintheprogram.Ifthenextfilestatementcausesthe
endoftheinputtobereached,thenthecodeinanyENDrulesisexecuted.Anexceptiontothisiswhennextfileis
invokedduringexecutionofanystatementinanENDruleinthiscase,itcausestheprogramtostopimmediately.See
BEGIN/END.
https://www.gnu.org/software/gawk/manual/gawk.html

136/479

29/7/2016

TheGNUAwkUsersGuide

Thenextfilestatementisusefulwhentherearemanydatafilestoprocessbutitisntnecessarytoprocesseveryrecord
ineveryfile.Withoutnextfile,inordertomoveontothenextdatafile,aprogramwouldhavetocontinuescanningthe
unwantedrecords.Thenextfilestatementaccomplishesthismuchmoreefficiently.
Ingawk,executionofnextfilecausesadditionalthingstohappen:anyENDFILErulesareexecutedifgawkisnotcurrently
inanENDorBEGINFILErule,ARGINDisincremented,andanyBEGINFILErulesareexecuted.(ARGINDhasntbeenintroduced
yet.SeeBuiltinVariables.)
Withgawk,nextfileisusefulinsideaBEGINFILEruletoskipoverafilethatwouldotherwisecausegawktoexitwitha
fatalerror.Inthiscase,ENDFILErulesarenotexecuted.SeeBEGINFILE/ENDFILE.
Althoughitmightseemthatclose(FILENAME)wouldaccomplishthesameasnextfile,thisisnttrue.close()is
reservedforclosingfiles,pipes,andcoprocessesthatareopenedwithredirections.Itisnotrelatedtothemain
processingthatawkdoeswiththefileslistedinARGV.
NOTE:Formanyyears,nextfilewasacommonextension.InSeptember2012,itwasacceptedfor
inclusionintothePOSIXstandard.SeetheAustinGroupwebsite.
ThecurrentversionofBWKawkandmawkalsosupportnextfile.However,theydontallowthenextfilestatement
insidefunctionbodies(seeUserdefined).gawkdoesanextfileinsideafunctionbodyreadsthenextrecordandstarts
processingitwiththefirstruleintheprogram,justasanyothernextfilestatement.
Previous:NextfileStatement,Up:Statements[Contents][Index]
7.4.10TheexitStatement
Theexitstatementcausesawktoimmediatelystopexecutingthecurrentruleandtostopprocessinginputany
remaininginputisignored.Theexitstatementiswrittenasfollows:
exit[returncode]

WhenanexitstatementisexecutedfromaBEGINrule,theprogramstopsprocessingeverythingimmediately.Noinput
recordsareread.However,ifanENDruleispresent,aspartofexecutingtheexitstatement,theENDruleisexecuted(see
BEGIN/END).IfexitisusedinthebodyofanENDrule,itcausestheprogramtostopimmediately.
AnexitstatementthatisnotpartofaBEGINorENDrulestopstheexecutionofanyfurtherautomaticrulesforthecurrent
record,skipsreadinganyremaininginputrecords,andexecutestheENDruleifthereisone.gawkalsoskipsanyENDFILE
rulestheydonotexecute.
Insuchacase,ifyoudontwanttheENDruletodoitsjob,setavariabletoanonzerovaluebeforetheexitstatement
andcheckthatvariableintheENDrule.SeeAssertFunction,foranexamplethatdoesthis.
Ifanargumentissuppliedtoexit,itsvalueisusedastheexitstatuscodefortheawkprocess.Ifnoargumentis
supplied,exitcausesawktoreturnasuccessstatus.Inthecasewhereanargumentissuppliedtoafirstexit
statement,andthenexitiscalledasecondtimefromanENDrulewithnoargument,awkusesthepreviouslysupplied
exitvalue.(d.c.)SeeExitStatus,formoreinformation.
Forexample,supposeanerrorconditionoccursthatisdifficultorimpossibletohandle.Conventionally,programs
reportthisbyexitingwithanonzerostatus.Anawkprogramcandothisusinganexitstatementwithanonzero
argument,asshowninthefollowingexample:
BEGIN{
if(("date"|getlinedate_now)<=0){
print"Can'tgetsystemdate">"/dev/stderr"
exit1
}
print"currentdateis",date_now
close("date")
}
https://www.gnu.org/software/gawk/manual/gawk.html

137/479

29/7/2016

TheGNUAwkUsersGuide

NOTE:Forfullportability,exitvaluesshouldbebetweenzeroand126,inclusive.Negativevalues,and
valuesof127orgreater,maynotproduceconsistentresultsacrossdifferentoperatingsystems.
Next:PatternActionSummary,Previous:Statements,Up:PatternsandActions[Contents][Index]

7.5PredefinedVariables
Mostawkvariablesareavailabletouseforyourownpurposestheyneverchangeunlessyourprogramassignsvaluesto
them,andtheyneveraffectanythingunlessyourprogramexaminesthem.However,afewvariablesinawkhavespecial
builtinmeanings.awkexaminessomeoftheseautomatically,sothattheyenableyoutotellawkhowtodocertain
things.Othersaresetautomaticallybyawk,sothattheycarryinformationfromtheinternalworkingsofawktoyour
program.
Thissectiondocumentsallofgawkspredefinedvariables,mostofwhicharealsodocumentedinthechaptersdescribing
theirareasofactivity.
Usermodified:
Builtinvariablesthatyouchangetocontrolawk.
Autoset:
Builtinvariableswhereawkgivesyouinformation.
ARGCandARGV: WaystouseARGCandARGV.
Next:Autoset,Up:BuiltinVariables[Contents][Index]
7.5.1BuiltinVariablesThatControlawk
Thefollowingisanalphabeticallistofvariablesthatyoucanchangetocontrolhowawkdoescertainthings.
Thevariablesthatarespecifictogawkaremarkedwithapoundsign(#).Thesevariablesaregawkextensions.Inother
awkimplementationsorifgawkisincompatibilitymode(seeOptions),theyarenotspecial.(Anyexceptionsarenotedin
thedescriptionofeachvariable.)
BINMODE#

OnnonPOSIXsystems,thisvariablespecifiesuseofbinarymodeforallI/O.Numericvaluesofone,two,or
threespecifythatinputfiles,outputfiles,orallfiles,respectively,shouldusebinaryI/O.Anumericvalueless
thanzeroistreatedaszero,andanumericvaluegreaterthanthreeistreatedasthree.Alternatively,stringvalues
of"r"or"w"specifythatinputfilesandoutputfiles,respectively,shouldusebinaryI/O.Astringvalueof"rw"or
"wr"indicatesthatallfilesshouldusebinaryI/O.Anyotherstringvalueistreatedthesameas"rw",butcauses
gawktogenerateawarningmessage.BINMODEisdescribedinmoredetailinPCUsing.mawk(seeOtherVersions)
alsosupportsthisvariable,butonlyusingnumericvalues.
CONVFMT

Astringthatcontrolstheconversionofnumberstostrings(seeConversion).Itworksbybeingpassed,ineffect,
asthefirstargumenttothesprintf()function(seeStringFunctions).Itsdefaultvalueis"%.6g".CONVFMTwas
introducedbythePOSIXstandard.
FIELDWIDTHS#

Aspaceseparatedlistofcolumnsthattellsgawkhowtosplitinputwithfixedcolumnarboundaries.Assigninga
valuetoFIELDWIDTHSoverridestheuseofFSandFPATforfieldsplitting.SeeConstantSize,formoreinformation.
FPAT#

Aregularexpression(asastring)thattellsgawktocreatethefieldsbasedontextthatmatchestheregular
expression.AssigningavaluetoFPAToverridestheuseofFSandFIELDWIDTHSforfieldsplitting.SeeSplittingBy
Content,formoreinformation.
https://www.gnu.org/software/gawk/manual/gawk.html

138/479

29/7/2016

TheGNUAwkUsersGuide

FS

Theinputfieldseparator(seeFieldSeparators).Thevalueisasinglecharacterstringoramulticharacterregular
expressionthatmatchestheseparationsbetweenfieldsinaninputrecord.Ifthevalueisthenullstring(""),then
eachcharacterintherecordbecomesaseparatefield.(Thisbehaviorisagawkextension.POSIXawkdoesnot
specifythebehaviorwhenFSisthenullstring.Nonetheless,someotherversionsofawkalsotreat""specially.)
Thedefaultvalueis"",astringconsistingofasinglespace.Asaspecialexception,thisvaluemeansthatany
sequenceofspaces,TABs,and/ornewlinesisasingleseparator.37Italsocausesspaces,TABs,andnewlinesat
thebeginningandendofarecordtobeignored.
YoucansetthevalueofFSonthecommandlineusingtheFoption:
awkF,'program'inputfiles

IfgawkisusingFIELDWIDTHSorFPATforfieldsplitting,assigningavaluetoFScausesgawktoreturntothenormal,
FSbasedfieldsplitting.AneasywaytodothisistosimplysayFS=FS,perhapswithanexplanatorycomment.
IGNORECASE#

IfIGNORECASEisnonzeroornonnull,thenallstringcomparisonsandallregularexpressionmatchingarecase
independent.Thisappliestoregexpmatchingwith~and!~,thegensub(),gsub(),index(),match(),
patsplit(),split(),andsub()functions,recordterminationwithRS,andfieldsplittingwithFSandFPAT.
However,thevalueofIGNORECASEdoesnotaffectarraysubscriptinganditdoesnotaffectfieldsplittingwhen
usingasinglecharacterfieldseparator.SeeCasesensitivity.
LINT#

Whenthisvariableistrue(nonzeroornonnull),gawkbehavesasifthelintcommandlineoptionisineffect
(seeOptions).Withavalueof"fatal",lintwarningsbecomefatalerrors.Withavalueof"invalid",only
warningsaboutthingsthatareactuallyinvalidareissued.(Thisisnotfullyimplementedyet.)Anyothertrue
valueprintsnonfatalwarnings.AssigningafalsevaluetoLINTturnsoffthelintwarnings.
Thisvariableisagawkextension.Itisnotspecialinotherawkimplementations.Unlikewiththeotherspecial
variables,changingLINTdoesaffecttheproductionoflintwarnings,evenifgawkisincompatibilitymode.Much
asthelintandtraditionaloptionsindependentlycontroldifferentaspectsofgawksbehavior,thecontrolof
lintwarningsduringprogramexecutionisindependentoftheflavorofawkbeingexecuted.
OFMT

Astringthatcontrolsconversionofnumberstostrings(seeConversion)forprintingwiththeprintstatement.It
worksbybeingpassedasthefirstargumenttothesprintf()function(seeStringFunctions).Itsdefaultvalueis
"%.6g".EarlierversionsofawkusedOFMTtospecifytheformatforconvertingnumberstostringsingeneral
expressionsthisisnowdonebyCONVFMT.
OFS

Theoutputfieldseparator(seeOutputSeparators).Itisoutputbetweenthefieldsprintedbyaprintstatement.Its
defaultvalueis"",astringconsistingofasinglespace.
ORS

Theoutputrecordseparator.Itisoutputattheendofeveryprintstatement.Itsdefaultvalueis"\n",thenewline
character.(SeeOutputSeparators.)
PREC#

Theworkingprecisionofarbitraryprecisionfloatingpointnumbers,53bitsbydefault(seeSettingprecision).
ROUNDMODE#
https://www.gnu.org/software/gawk/manual/gawk.html

139/479

29/7/2016

TheGNUAwkUsersGuide

Theroundingmodetouseforarbitraryprecisionarithmeticonnumbers,bydefault"N"(roundTiesToEveninthe
IEEE754standardseeSettingtheroundingmode).
RS

Theinputrecordseparator.Itsdefaultvalueisastringcontainingasinglenewlinecharacter,whichmeansthatan
inputrecordconsistsofasinglelineoftext.Itcanalsobethenullstring,inwhichcaserecordsareseparatedby
runsofblanklines.Ifitisaregexp,recordsareseparatedbymatchesoftheregexpintheinputtext.(See
Records.)
TheabilityforRStobearegularexpressionisagawkextension.Inmostotherawkimplementations,orifgawkisin
compatibilitymode(seeOptions),justthefirstcharacterofRSsvalueisused.
SUBSEP

Thesubscriptseparator.Ithasthedefaultvalueof"\034"andisusedtoseparatethepartsoftheindicesofa
multidimensionalarray.Thus,theexpressionfoo["A","B"]reallyaccesses foo["A\034B"](see
Multidimensional).
TEXTDOMAIN#

Usedforinternationalizationofprogramsattheawklevel.Itsetsthedefaulttextdomainforspeciallymarked
stringconstantsinthesourcetext,aswellasforthedcgettext(),dcngettext(),andbindtextdomain()functions
(seeInternationalization).ThedefaultvalueofTEXTDOMAINis"messages".
Next:ARGCandARGV,Previous:Usermodified,Up:BuiltinVariables[Contents][Index]
7.5.2BuiltinVariablesThatConveyInformation
Thefollowingisanalphabeticallistofvariablesthatawksetsautomaticallyoncertainoccasionsinordertoprovide
informationtoyourprogram.
Thevariablesthatarespecifictogawkaremarkedwithapoundsign(#).Thesevariablesaregawkextensions.Inother
awkimplementationsorifgawkisincompatibilitymode(seeOptions),theyarenotspecial:
ARGC,ARGV

ThecommandlineargumentsavailabletoawkprogramsarestoredinanarraycalledARGV.ARGCisthenumberof
commandlineargumentspresent.SeeOtherArguments.Unlikemostawkarrays,ARGVisindexedfrom0toARGC
1.Inthefollowingexample:
$awk'BEGIN{
>for(i=0;i<ARGC;i++)
>printARGV[i]
>}'inventoryshippedmaillist
|awk
|inventoryshipped
|maillist
ARGV[0]containsawk,ARGV[1]containsinventoryshipped,andARGV[2]containsmaillist.Thevalueof
ARGCisthree,onemorethantheindexofthelastelementinARGV,becausetheelementsarenumberedfromzero.

ThenamesARGCandARGV,aswellastheconventionofindexingthearrayfrom0toARGC1,arederivedfromthe
Clanguagesmethodofaccessingcommandlinearguments.
ThevalueofARGV[0]canvaryfromsystemtosystem.Also,youshouldnotethattheprogramtextisnotincluded
inARGV,norareanyofawkscommandlineoptions.SeeARGCandARGV,forinformationabouthowawkuses
thesevariables.(d.c.)
ARGIND#
https://www.gnu.org/software/gawk/manual/gawk.html

140/479

29/7/2016

TheGNUAwkUsersGuide

TheindexinARGVofthecurrentfilebeingprocessed.Everytimegawkopensanewdatafileforprocessing,itsets
ARGINDtotheindexinARGVofthefilename.Whengawkisprocessingtheinputfiles,FILENAME==ARGV[ARGIND]
isalwaystrue.
Thisvariableisusefulinfileprocessingitallowsyoutotellhowfaralongyouareinthelistofdatafilesaswell
astodistinguishbetweensuccessiveinstancesofthesamefilenameonthecommandline.
WhileyoucanchangethevalueofARGINDwithinyourawkprogram,gawkautomaticallysetsittoanewvalue
whenitopensthenextfile.
ENVIRON

Anassociativearraycontainingthevaluesoftheenvironment.Thearrayindicesaretheenvironmentvariable
namestheelementsarethevaluesoftheparticularenvironmentvariables.Forexample,ENVIRON["HOME"]might
be"/home/arnold".Changingthisarraydoesnotaffecttheenvironmentpassedontoanyprogramsthatawkmay
spawnviaredirectionorthesystem()function.(Inafutureversionofgawk,itmaydoso.)
Someoperatingsystemsmaynothaveenvironmentvariables.Onsuchsystems,theENVIRONarrayisempty
(exceptforENVIRON["AWKPATH"]andENVIRON["AWKLIBPATH"]seeAWKPATHVariable,andseeAWKLIBPATH
Variable).
ERRNO#

Ifasystemerroroccursduringaredirectionforgetline,duringareadforgetline,orduringaclose()operation,
thenERRNOcontainsastringdescribingtheerror.
Inaddition,gawkclearsERRNObeforeopeningeachcommandlineinputfile.Thisenablescheckingifthefileis
readableinsideaBEGINFILEpattern(seeBEGINFILE/ENDFILE).
Otherwise,ERRNOworkssimilarlytotheCvariableerrno.Exceptforthecasejustmentioned,gawkneverclearsit
(setsittozeroor"").Thus,youshouldonlyexpectitsvaluetobemeaningfulwhenanI/Ooperationreturnsa
failurevalue,suchasgetlinereturning1.Youare,ofcourse,freetoclearityourselfbeforedoinganI/O
operation.
FILENAME

Thenameofthecurrentinputfile.Whennodatafilesarelistedonthecommandline,awkreadsfromthestandard
inputandFILENAMEissetto"".FILENAMEchangeseachtimeanewfileisread(seeReadingFiles).InsideaBEGIN
rule,thevalueofFILENAMEis"",becausetherearenoinputfilesbeingprocessedyet.38(d.c.)Note,though,that
usinggetline(seeGetline)insideaBEGINrulecangiveFILENAMEavalue.
FNR

Thecurrentrecordnumberinthecurrentfile.awkincrementsFNReachtimeitreadsanewrecord(seeRecords).
awkresetsFNRtozeroeachtimeitstartsanewinputfile.
NF

Thenumberoffieldsinthecurrentinputrecord.NFisseteachtimeanewrecordisread,whenanewfieldis
created,orwhen$0changes(seeFields).
Unlikemostofthevariablesdescribedinthissubsection,assigningavaluetoNFhasthepotentialtoaffectawks
internalworkings.Inparticular,assignmentstoNFcanbeusedtocreatefieldsinorremovefieldsfromthecurrent
record.SeeChangingFields.
FUNCTAB#

Anarraywhoseindicesandcorrespondingvaluesarethenamesofallthebuiltin,userdefined,andextension
functionsintheprogram.
https://www.gnu.org/software/gawk/manual/gawk.html

141/479

29/7/2016

TheGNUAwkUsersGuide

NOTE:AttemptingtousethedeletestatementwiththeFUNCTABarraycausesafatalerror.Any
attempttoassigntoanelementofFUNCTABalsocausesafatalerror.
NR

Thenumberofinputrecordsawkhasprocessedsincethebeginningoftheprogramsexecution(seeRecords).awk
incrementsNReachtimeitreadsanewrecord.
PROCINFO#

Theelementsofthisarrayprovideaccesstoinformationabouttherunningawkprogram.Thefollowingelements
(listedalphabetically)areguaranteedtobeavailable:
PROCINFO["egid"]

Thevalueofthegetegid()systemcall.
PROCINFO["euid"]

Thevalueofthegeteuid()systemcall.
PROCINFO["FS"]

Thisis"FS"iffieldsplittingwithFSisineffect,"FIELDWIDTHS"iffieldsplittingwithFIELDWIDTHSisin
effect,or"FPAT"iffieldmatchingwithFPATisineffect.
PROCINFO["identifiers"]

Asubarray,indexedbythenamesofallidentifiersusedinthetextoftheawkprogram.Anidentifieris
simplythenameofavariable(beitscalarorarray),builtinfunction,userdefinedfunction,orextension
function.Foreachidentifier,thevalueoftheelementisoneofthefollowing:
"array"

Theidentifierisanarray.
"builtin"

Theidentifierisabuiltinfunction.
"extension"

Theidentifierisanextensionfunctionloadedvia@loadorl.
"scalar"

Theidentifierisascalar.
"untyped"

Theidentifierisuntyped(couldbeusedasascalaroranarraygawkdoesntknowyet).
"user"

Theidentifierisauserdefinedfunction.
Thevaluesindicatewhatgawkknowsabouttheidentifiersafterithasfinishedparsingtheprogramtheyare
notupdatedwhiletheprogramruns.
PROCINFO["gid"]

Thevalueofthegetgid()systemcall.
https://www.gnu.org/software/gawk/manual/gawk.html

142/479

29/7/2016

TheGNUAwkUsersGuide

PROCINFO["pgrpid"]

TheprocessgroupIDofthecurrentprocess.
PROCINFO["pid"]

TheprocessIDofthecurrentprocess.
PROCINFO["ppid"]

TheparentprocessIDofthecurrentprocess.
PROCINFO["sorted_in"]

IfthiselementexistsinPROCINFO,itsvaluecontrolstheorderinwhicharrayindiceswillbeprocessedby
for(indxinarray)loops.Thisisanadvancedfeature,sowedeferthefulldescriptionuntillatersee
ScanninganArray.
PROCINFO["strftime"]

Thedefaulttimeformatstringforstrftime().Assigninganewvaluetothiselementchangesthedefault.
SeeTimeFunctions.
PROCINFO["uid"]

Thevalueofthegetuid()systemcall.
PROCINFO["version"]

Theversionofgawk.
ThefollowingadditionalelementsinthearrayareavailabletoprovideinformationabouttheMPFRandGMP
librariesifyourversionofgawksupportsarbitraryprecisionarithmetic(seeArbitraryPrecisionArithmetic):
PROCINFO["mpfr_version"]

TheversionoftheGNUMPFRlibrary.
PROCINFO["gmp_version"]

TheversionoftheGNUMPlibrary.
PROCINFO["prec_max"]

ThemaximumprecisionsupportedbyMPFR.
PROCINFO["prec_min"]

TheminimumprecisionrequiredbyMPFR.
Thefollowingadditionalelementsinthearrayareavailabletoprovideinformationabouttheversionofthe
extensionAPI,ifyourversionofgawksupportsdynamicloadingofextensionfunctions(seeDynamic
Extensions):
PROCINFO["api_major"]

ThemajorversionoftheextensionAPI.
PROCINFO["api_minor"]

TheminorversionoftheextensionAPI.
Onsomesystems,theremaybeelementsinthearray,"group1"through"groupN"forsomeN.Nisthenumberof
https://www.gnu.org/software/gawk/manual/gawk.html

143/479

29/7/2016

TheGNUAwkUsersGuide

supplementarygroupsthattheprocesshas.Usetheinoperatortotestfortheseelements(seeReferenceto
Elements).
ThePROCINFOarrayhasthefollowingadditionaluses:
Itmaybeusedtoprovideatimeoutwhenreadingfromanyopeninputfile,pipe,orcoprocess.SeeRead
Timeout,formoreinformation.
Itmaybeusedtocausecoprocessestocommunicateoverpseudottysinsteadofthroughtwowaypipes
thisisdiscussedfurtherinTwowayI/O.
RLENGTH

Thelengthofthesubstringmatchedbythematch()function(seeStringFunctions).RLENGTHissetbyinvokingthe
match()function.Itsvalueisthelengthofthematchedstring,or1ifnomatchisfound.
RSTART

Thestartindexincharactersofthesubstringthatismatchedbythematch()function(seeStringFunctions).
RSTARTissetbyinvokingthematch()function.Itsvalueisthepositionofthestringwherethematchedsubstring
starts,orzeroifnomatchwasfound.
RT#

TheinputtextthatmatchedthetextdenotedbyRS,therecordseparator.Itisseteverytimearecordisread.
SYMTAB#

Anarraywhoseindicesarethenamesofalldefinedglobalvariablesandarraysintheprogram.SYMTABmakes
gawkssymboltablevisibletotheawkprogrammer.Itisbuiltasgawkparsestheprogramandiscompletebefore
theprogramstartstorun.
Thearraymaybeusedforindirectaccesstoreadorwritethevalueofavariable:
foo=5
SYMTAB["foo"]=4
printfoo#prints4

Theisarray()function(seeTypeFunctions)maybeusedtotestifanelementinSYMTABisanarray.Also,you
maynotusethedeletestatementwiththeSYMTABarray.
YoumayuseanindexforSYMTABthatisnotapredefinedidentifier:
SYMTAB["xxx"]=5
printSYMTAB["xxx"]

Thisworksasexpected:inthiscaseSYMTABactsjustlikearegulararray.Theonlydifferenceisthatyoucantthen
deleteSYMTAB["xxx"].
TheSYMTABarrayismoreinterestingthanitlooks.AndrewSchorrpointsoutthatiteffectivelygivesawkdata
pointers.Considerhisexample:
#Indirectmultiplyofanyvariablebyamount,returnresult
functionmultiply(variable,amount)
{
returnSYMTAB[variable]*=amount
}

NOTE:Inordertoavoidseveretimetravelparadoxes,39neitherFUNCTABnorSYMTABisavailableas
anelementwithintheSYMTABarray.
ChangingNRandFNR
https://www.gnu.org/software/gawk/manual/gawk.html

144/479

29/7/2016

TheGNUAwkUsersGuide

awkincrementsNRandFNReachtimeitreadsarecord,insteadofsettingthemtotheabsolutevalueofthenumberof

recordsread.Thismeansthataprogramcanchangethesevariablesandtheirnewvaluesareincrementedforeach
record.(d.c.)Thefollowingexampleshowsthis:
$echo'1
>2
>3
>4'|awk'NR==2{NR=17}
>{printNR}'
|1
|17
|18
|19

BeforeFNRwasaddedtotheawklanguage(seeV7/SVR3.1),manyawkprogramsusedthisfeaturetotrackthenumber
ofrecordsinafilebyresettingNRtozerowhenFILENAMEchanged.

Previous:Autoset,Up:BuiltinVariables[Contents][Index]
7.5.3UsingARGCandARGV
Autoset,presentedthefollowingprogramdescribingtheinformationcontainedinARGCandARGV:
$awk'BEGIN{
>for(i=0;i<ARGC;i++)
>printARGV[i]
>}'inventoryshippedmaillist
|awk
|inventoryshipped
|maillist

Inthisexample,ARGV[0]containsawk,ARGV[1]containsinventoryshipped,andARGV[2]containsmaillist.
NoticethattheawkprogramisnotenteredinARGV.Theothercommandlineoptions,withtheirarguments,arealsonot
entered.Thisincludesvariableassignmentsdonewiththevoption(seeOptions).Normalvariableassignmentsonthe
commandlinearetreatedasargumentsanddoshowupintheARGVarray.Giventhefollowingprograminafilenamed
showargs.awk:
BEGIN{
printf"A=%d,B=%d\n",A,B
for(i=0;i<ARGC;i++)
printf"\tARGV[%d]=%s\n",i,ARGV[i]
}
END{printf"A=%d,B=%d\n",A,B}

Runningitproducesthefollowing:
$awkvA=1fshowargs.awkB=2/dev/null
|A=1,B=0
|ARGV[0]=awk
|ARGV[1]=B=2
|ARGV[2]=/dev/null
|A=1,B=2

AprogramcanalterARGCandtheelementsofARGV.Eachtimeawkreachestheendofaninputfile,itusesthenext
elementofARGVasthenameofthenextinputfile.Bystoringadifferentstringthere,aprogramcanchangewhichfiles
areread.Use""torepresentthestandardinput.StoringadditionalelementsandincrementingARGCcausesadditional
filestoberead.
IfthevalueofARGCisdecreased,thateliminatesinputfilesfromtheendofthelist.ByrecordingtheoldvalueofARGC
elsewhere,aprogramcantreattheeliminatedargumentsassomethingotherthanfilenames.
https://www.gnu.org/software/gawk/manual/gawk.html

145/479

29/7/2016

TheGNUAwkUsersGuide

Toeliminateafilefromthemiddleofthelist,storethenullstring("")intoARGVinplaceofthefilesname.Asaspecial
feature,awkignoresfilenamesthathavebeenreplacedwiththenullstring.Anotheroptionistousethedeletestatement
toremoveelementsfromARGV(seeDelete).
AlloftheseactionsaretypicallydoneintheBEGINrule,beforeactualprocessingoftheinputbegins.SeeSplitProgram,
andseeTeeProgram,forexamplesofeachwayofremovingelementsfromARGV.
Toactuallygetoptionsintoanawkprogram,endtheawkoptionswithandthensupplytheawkprogramsoptions,in
thefollowingmanner:
awkfmyprog.awkvqfile1file2

ThefollowingfragmentprocessesARGVinordertoexamine,andthenremove,thepreviouslymentionedcommandline
options:
BEGIN{
for(i=1;i<ARGC;i++){
if(ARGV[i]=="v")
verbose=1
elseif(ARGV[i]=="q")
debug=1
elseif(ARGV[i]~/^./){
e=sprintf("%s:unrecognizedoption%c",
ARGV[0],substr(ARGV[i],2,1))
printe>"/dev/stderr"
}else
break
deleteARGV[i]
}
}

Endingtheawkoptionswithisntnecessaryingawk.Unlessposixhasbeenspecified,gawksilentlyputsany
unrecognizedoptionsintoARGVfortheawkprogramtodealwith.Assoonasitseesanunknownoption,gawkstops
lookingforotheroptionsthatitmightotherwiserecognize.Thepreviouscommandlinewithgawkwouldbe:
gawkfmyprog.awkqvfile1file2

Becauseqisnotavalidgawkoption,itandthefollowingvarepassedontotheawkprogram.(SeeGetoptFunction,
foranawklibraryfunctionthatparsescommandlineoptions.)
Whendesigningyourprogram,youshouldchooseoptionsthatdontconflictwithgawks,becauseitwillprocessany
optionsthatitacceptsbeforepassingtherestofthecommandlineontoyourprogram.Using#!withtheEoption
mayhelp(seeExecutableScripts,andseeOptions,).
Previous:BuiltinVariables,Up:PatternsandActions[Contents][Index]

7.6Summary
Patternactionpairsmakeupthebasicelementsofanawkprogram.Patternsareeithernormalexpressions,range
expressions,orregexpconstantsoneofthespecialkeywordsBEGIN,END,BEGINFILE,orENDFILEorempty.The
actionexecutesifthecurrentrecordmatchesthepattern.Empty(missing)patternsmatchallrecords.
I/OfromBEGINandENDruleshascertainconstraints.Thisisalsotrue,onlymoreso,forBEGINFILEandENDFILE
rules.Thelattertwogiveyouhooksintogawksfileprocessing,allowingyoutorecoverfromafilethat
otherwisewouldcauseafatalerror(suchasafilethatcannotbeopened).
Shellvariablescanbeusedinawkprogramsbycarefuluseofshellquoting.Itiseasiertopassashellvariableinto
awkbyusingthevoptionandanawkvariable.
Actionsconsistofstatementsenclosedincurlybraces.Statementsarebuiltupfromexpressions,control
statements,compoundstatements,inputandoutputstatements,anddeletionstatements.
Thecontrolstatementsinawkareifelse,while,for,anddowhile.gawkaddstheswitchstatement.Therearetwo
flavorsofforstatement:oneforperforminggenerallooping,andtheotherforiteratingthroughanarray.
https://www.gnu.org/software/gawk/manual/gawk.html

146/479

29/7/2016

TheGNUAwkUsersGuide

breakandcontinueletyouexitearlyorstartthenextiterationofaloop(orgetoutofaswitch).
nextandnextfileletyoureadthenextrecordandstartoveratthetopofyourprogramorskiptothenextinput

fileandstartover,respectively.
Theexitstatementterminatesyourprogram.Whenexecutedfromanaction(orfunctionbody),ittransfers
controltotheENDstatements.FromanENDstatementbody,itexitsimmediately.Youmaypassanoptional
numericvaluetobeusedasawksexitstatus.
Somepredefinedvariablesprovidecontroloverawk,mainlyforI/O.Othervariablesconveyinformationfromawk
toyourprogram.
ARGCandARGVmakethecommandlineargumentsavailabletoyourprogram.ManipulatingthemfromaBEGINrule
letsyoucontrolhowawkwillprocesstheprovideddatafiles.
Next:Functions,Previous:PatternsandActions,Up:Top[Contents][Index]

8Arraysinawk
Anarrayisatableofvaluescalledelements.Theelementsofanarrayaredistinguishedbytheirindices.Indicesmaybe
eithernumbersorstrings.
Thischapterdescribeshowarraysworkinawk,howtousearrayelements,howtoscanthrougheveryelementinan
array,andhowtoremovearrayelements.Italsodescribeshowawksimulatesmultidimensionalarrays,aswellassome
ofthelessobviouspointsaboutarrayusage.Thechaptermovesontodiscussgawksfacilityforsortingarrays,andends
withabriefdescriptionofgawksabilitytosupporttruearraysofarrays.
ArrayBasics:
Thebasicsofarrays.
NumericArraySubscripts: Howtousenumbersassubscriptsinawk.
UninitializedSubscripts: UsingUninitializedvariablesassubscripts.
Delete:
Thedeletestatementremovesanelementfromanarray.
Multidimensional:
Emulatingmultidimensionalarraysinawk.
ArraysofArrays:
Truemultidimensionalarrays.
ArraysSummary:
Summaryofarrays.
Next:NumericArraySubscripts,Up:Arrays[Contents][Index]

8.1TheBasicsofArrays
Thissectionpresentsthebasics:workingwithelementsinarraysoneatatime,andtraversingalloftheelementsinan
array.
ArrayIntro:
IntroductiontoArrays
ReferencetoElements: Howtoexamineoneelementofanarray.
AssigningElements: Howtochangeanelementofanarray.
ArrayExample:
BasicExampleofanArray
ScanninganArray:
Avariationoftheforstatement.Itloopsthroughtheindicesofanarraysexistingelements.
ControllingScanning: Controllingtheorderinwhicharraysarescanned.
Next:ReferencetoElements,Up:ArrayBasics[Contents][Index]
8.1.1IntroductiontoArrays
DoinglinearscansoveranassociativearrayisliketryingtoclubsomeonetodeathwithaloadedUzi.
LarryWall
https://www.gnu.org/software/gawk/manual/gawk.html