coding horror_ a visual explanation of sql joins.pdf

programmingandhumanfactorsbyJeAtwood

AVisualExplanaonofSQLJoinsOctober11,2007

IthoughtLigayaTurmelle'spostonSQLjoinswasagreatprimerfornovicedevelopers.SinceSQLjoinsappeartobesetbased,theuseofVenndiagramstoexplainthemseems,atfirstblush,tobeanaturalfit.However,likethecommenterstoherpost,IfoundthattheVenndiagramsdidn'tquitematchtheSQLjoinsyntaxrealityinmytesng.

Ilovetheconcept,though,solet'sseeifwecanmakeitwork.Assumewehavethefollowingtwotables.TableAisonthele,andTableBisontheright.We'llpopulatethemwithfourrecordseach.

id name id name-- ---- -- ----1 Pirate 1 Rutabaga2 Monkey 2 Pirate3 Ninja 3 Darth Vader4 Spaghetti 4 Ninja

Let'sjointhesetablesbythenamefieldinafewdierentwaysandseeifwecangetaconceptualmatchtothoseniyVenndiagrams.

SELECT * FROM TableAINNER JOIN TableBON TableA.name = TableB.nameid name id name-- ---- -- ----1 Pirate 2 Pirate3 Ninja 4 Ninja

InnerjoinproducesonlythesetofrecordsthatmatchinbothTableAandTableB.

SELECT * FROM TableAFULL OUTER JOIN TableBON TableA.name = TableB.nameid name id name-- ---- -- ----1 Pirate 2 Pirate2 Monkey null null3 Ninja 4 Ninja4 Spaghetti null nullnull null 1 Rutabaga null null 3 Darth Vader

FullouterjoinproducesthesetofallrecordsinTableAandTableB,withmatchingrecordsfrombothsideswhereavailable.Ifthereisnomatch,themissingsidewillcontainnull.

Coding Horror: A Visual Explanation of SQL Joins http://www.codinghorror.com/blog/2007/10/a-visual-explanation-of-sql-j...

1 de 19 31/08/2012 07:18 a.m.

SELECT * FROM TableALEFT OUTER JOIN TableBON TableA.name = TableB.nameid name id name-- ---- -- ----1 Pirate 2 Pirate2 Monkey null null3 Ninja 4 Ninja4 Spaghetti null null

LeouterjoinproducesacompletesetofrecordsfromTableA,withthematchingrecords(whereavailable)inTableB.Ifthereisnomatch,therightsidewillcontainnull.

SELECT * FROM TableALEFT OUTER JOIN TableBON TableA.name = TableB.nameWHERE TableB.id IS nullid name id name-- ---- -- ----2 Monkey null null4 Spaghetti null null

ToproducethesetofrecordsonlyinTableA,butnotinTableB,weperformthesameleouterjoin,thenexcludetherecordswedon'twantfromtherightsideviaawhereclause.

SELECT * FROM TableAFULL OUTER JOIN TableBON TableA.name = TableB.nameWHERE TableA.id IS null OR TableB.id IS nullid name id name-- ---- -- ----2 Monkey null null4 Spaghetti null nullnull null 1 Rutabaganull null 3 Darth Vader

ToproducethesetofrecordsuniquetoTableAandTableB,weperformthesamefullouterjoin,thenexcludetherecordswedon'twantfrombothsidesviaawhereclause.

There'salsoacartesianproductorcrossjoin,whichasfarasIcantell,can'tbeexpressedasaVenndiagram:

SELECT * FROM TableACROSS JOIN TableB

Thisjoins"everythingtoeverything",resulngin4x4=16rows,farmorethanwehadintheoriginalsets.Ifyoudothemath,youcanseewhythisisaverydangerousjointorunagainstlargetables.


2 de 19 31/08/2012 07:18 a.m.

PostedbyJeAtwood

Relatedposts:

AllAbstraconsAreFailedAbstraconsDeadlocked!GivemeparameterizedSQL,orgivemedeathCompiledorBust?StoredProceduresvs.AdHocSQL


3 de 19 31/08/2012 07:18 a.m.

190Comments

Funny,Ijustexplainedthistoacoworkerinthesamemannerearlierintheweek.Iguessitneverdawnedonmethatothersmayhaveneverthoughtaboutjoinsintermsofthesediagrams.Goodpost,Je!

ShawnWheatleyonOctober12,20072:14AM

HeyJe,thanksforthegreatblogIthoroughlyenjoyreadingeverysingleoneofyourposts.

EventhoughIoenamfamiliarwiththeconceptsyoutalkof,Ifindthesimplemannerinwhichyoubreakdowntheissuesisalwaysagreatrefresher.

Keepupthegreatwork.

capdogonOctober12,20072:32AM

linkedfromjoin(SQL)enwikipediatoyourblogentry...yourpostismuchbeer(orsimplisc)thanthetechtalkonthatwikipediaentry;)

hackckonOctober12,20073:00AM

ThepostismuchsimplerthantheWikipediaentrybecauseitomitsallthenontrivialcases.AlloftheseexamplesassumethateachentryintableAwillbematchedbythejoinpredicatetoatmostoneentryintableB,andviceversa.

WeebleonOctober12,20073:35AM

I'mnotevensophiscatedenoughtousedatabasesanymore(mycareerasadeveloperhasdevolved),butwhenIwas,99%ofthemeIusedinnerjoinsandjustthoughtofitasthedatacommontobothtables.EverynowandthenIwantedsomedatathatdidn'tfitthatkindofjoinandI'ddoaleouterjointogetthedatathatwas"leout".

BradonOctober12,20073:47AM

SpeakingaboutOracle(+)syntaxandLEFT/RIGHT[OUTER]JOINsyntaxthereisadierence.Anditisnicelydescribedherehp://www.oreillynet.com/pub/a/network/2002/10/01/whatsinacondion.html

SpeakingaboutANSIsyntaxwhereeveryonehastoexplicitlysaytomakeCROSSJOINtogetCartesianproductatleastinOracleitisnottrue,usingstupid(atleasttomymind)NATURALJOINclauseonecaneasilygetCROSSJOINandalsoJOINonsuchcolumnsheneverthoughtitreallyis.I'vebloggedaboutitherehp://gplivna.blogspot.com/2007/10/naturaljoinsareevilmooifyou.htmlDon'tknowwehetheritisjustOraclespecificorasperANSISQLstandard.

SpeakingaboutINNNERJOINSasaCartesianproductandtheneliminangallunnecessaryrows,atleasttherealmechanicsinOracleisabsolutelydierent,therearefollowingpossibiliesfordoingthem:NESTEDLOOPS(OUTER)HASHJOIN(OUTER)MERGEJOIN(OUTER)andforCartesianproductitusesMERGEJOINCARTESIAN,andthat'swhenusuallytherealtrouble(fillingtemp)starts;)TherealchoiceamongNestedloops,hashjoinormergejoinatleastdependsondata,stascs,availablememoryforasession,explicitlygivenavailablememoryforhashjoinsormergejoins(btwthereispossibilitytogiveforonebutnotforother),systemworkload,hints,inializaonparametersallowingceratainjointypes(atleastforhashjoins)andprobablysomethingother:)


4 de 19 31/08/2012 07:18 a.m.

GintsPlivnaonOctober12,20073:51AM

HiJe,

IthinkyourillustraonsmakeoneBIGassumpon:TableAandTableBcontainuniquerows.Ifeithertablecontainedduplicatematchingrecords,thenthewholeVenndiagramwillnothold(thecardinalityofthesetswillnotaddup).IamafraidthisvisualexplanaonwillleadsometothinkthatyoucanuseSQLjoinstodofilteroperaonswhichshouldn'tbedonethisway.

Althoughit'sniceaempttoexplainSQLjoins,overallIthinkitismisleading.

dt

dtonOctober12,20073:57AM

OneoftheworstthreemonthsofmyCStrainingwasaclassostensiblyondatabasetheorywhichwasreallyabootcampinSQLsyntax,taughtbyaJohnnyonenoteprofessorwhothoughtSQLandRDMSswerethegreatestinvenonofmankind.

Thisexperienceturnedmeodatabasesfor15yearsunlonedayintheshowerIrealizedthatjoinswereanalogoustosettheory,inaroughway.(Yestheyare,allyounaysayers!"Givemeeverythingthat'sthereANDthere"or"Givemeeverythingthat'sthereORthere"maynothandleallthepossibleinputs,butit'sagoodjumpingopointforexplaining"inner"and"outer."Andwhocameupwiththatstupid,arbitraryterminologyanyway?)

IsllthinkSQLisanawfullanguage,though,andrelaonaldatabasesareanuglyhackmandatedbyhardwarelimitaonstrumpingelegantdesign.OLAP"cubes"aresoclearlyabeersoluonsointuivelyclearandobviousthenaturalgeneralizaontohigherdimensionsoftheflatfiledatabase.

AlexChamberlainonOctober12,20074:09AM

Tome,aCartesianProductwouldbebestillustratedbybothcirclesbeingblank.

Granted,alltheSQLguruswillprobablyspitfeathersatmyoutrageoussuggeson,butforme,it'sabeaufullysimplemethodofexplainingsimplejoinsandI'llbeusingitinmetechnicaldocumentaon!

Thanksforthisarcle,Je.

KevinonOctober12,20074:12AM

Je,youare_the_masterofwellcraedblogpoststhatsomemesdon'tsaymuch,butsomehowgatherscadsofinanecomments!Thisoneandthepreviouspost(onexercise)aregreatexamples.

Now,notallofyourpostsareinthiscategory,thankGod!ButI'mwatchingyou...

SteveonOctober12,20074:25AM

HeyNowJe,Thesediagramsmakemethinkofwhenlearningofclassicbooleanlogic.CodingHorrorFan,Cao

CaoonOctober12,20074:37AM


5 de 19 31/08/2012 07:18 a.m.

Well,thedatabasepraconerrejoice,andthesetrelaonistscringe.

Thereisonlyonejoin,whichisthecrossproduct(eachrowofthefirsttableconcatenatedtoeachrowofthesecondtable).TheINNERJOINexampleonlydiscardssomerowsbythewhereclause,otherwiseitis(orshouldbe)thesameastheCROSSJOIN.Theouterjoinisahackthatisn'tveryfirmlyrootedinrelaonalalgebra.

AndreasKreyonOctober12,20074:47AM

NiceJe,thankyousomuchforthispost!

IalwaysgotconfusedwithJOINSandhowtheywork,that'soneofthesethingsthatarehardtobeexpainedinabookoramanual.

Butthesediagramsmakeitpreyclear,theyaregreat!

jan.gonOctober12,20075:10AM

Don'tforgetEXCEPT!Whichwas(andaddionallysllis)MINUSinOracle.ForyearsIthinkOraclewastheonlyRDBMSthatimplementeditandIhad,innearly20years,useditallof,oh,once.Unllastweek,whenIuseditthesecondme.

It'sthesameasyourfourthexample:insteadof

SELECT*FROMTableALEFTOUTERJOINTableBONTableA.name=TableB.nameWHERETableB.idISnull

wecansay

SELECT*FROMTableAMINUSSELECT*FROMTableB

(IusedtheoldOracleoperatorit'smoreexpressivetomyeyes)

Noteveryplaormimplementstheoperatoryet:MySQL(whichIjustchecked)doesn't,forexample.MSSQLServer2005does.SQLIte?Hangon...yup,thatworksinv3atleast.

MikeWoodhouseonOctober12,20075:21AM

Verystrange.Inmy"DatabaseManagement"classatArizonaStateUniversityastudentaskedaboutJoin'sandmyprofessorsaid"Ineverusethem.JustsckwiththeWHEREsyntax".

Thiswasamoretheorecalcoursethough.WespentFARmoremeinrelaonalalgebra/calculusandalgorithmsforensuringatomictransaconsthanwedidinmorepraccalstulikedatabasenormalizaon.

Hmm...lotsofpeoplesaythetheorecalstuisawasteofme.Forme,well,IalmostbecameamathmajorinsteadofCS,soIfindthetheorecalstumoreinteresng.

KGonOctober12,20075:21AM

Ifyoudon'tuseranyjoinfeatureandjustdo"select*fromTableA,TableB",youhaveyourcrossjoin.


6 de 19 31/08/2012 07:18 a.m.

joskeonOctober12,20075:26AM

Venndiagrams.

wow,sosimple,soobvious;howcomeIhaven'tseenthisanywhereelse?

davidonOctober12,20075:42AM

HiJethat'sannicediagramandanewwayofpresenngthesqljoinfuncon.

RockonOctober12,20075:47AM

Yetanothercaseoftheblindleadingtheblind(downtheblindalleyofignoranceandarrogance);theoriginal,BTW.Atleastyoumadeandaempttofixitup.Butreadingallthose"nowIgetit"comments,makesitclearerwhyYahooslikeGWBushgetawaywithbaldfacedlies."Peoplebelievewhattheywanttobelieve,whenitmakesnosenseatall"/Mellencamp.

Nowwe'llhaveanothergaggleoffolksrunningaroundfawningonAmblerandFowler.Gad.TheHorror,theHorror.

BuggyFunBunnyonOctober12,20076:22AM

ThankstothisI'mnowskippingmerrilydownthepathofBOMmanagementandstockcontrol,ratherthanflickingfuriouslythroughTeachyourselfbooks,thanks

ChrisonOctober12,20076:36AM

Ilikeitthanksforyourwork.

cocobearonOctober12,20076:42AM

Excellentpost.

Everyonewhohasfollowed(andpassed)acomputersciencebeginnercourseshouldbeabletounderstandthis.

RasmusonOctober12,20076:45AM

It'simportanttonotethatwhenjoiningtwotablesonnonindexedcolumnsmostdatabasesperformtheequivalentofacrossjoinandthenweedouttherowsthatdon'tmatchthejoin/wherecriteria.Whilethisdoesntreallyimpactthequeryresults,ithasabigimpactonqueryperformance.

InMySqlyoucanrun:explainmyquerytodetermineifthequeryisusingindexesandreviewperformanceofallyourqueries.

CRonOctober12,20076:59AM

Perfectming.IjusthadanewprojectthrownonmydeskthatwillrequireworkingwithSQLsomethingIknowlileaboutandhaveavoided.Termsthathaveglazedmyeyesinweekspastsuddenlymakesense.


7 de 19 31/08/2012 07:18 a.m.

WombatonOctober12,20077:13AM

dtis100%right.Youreallyneedadisclaimerthatindicatesthatyoursetdiagramsonlyholdwhenthereisatbestaonetoonecorrespondencebetweenjoincriteriavaluesinthetables.

Thesecondyouhavetwoorthreepiratesinoneofthetables(an_extremely_commonrealworldscenario,probablymuchmorecommonthanthesimplifiedcaseyoushowhere)yourVenndiagramsbreakdown.

MaWigdahlonOctober12,20077:29AM

WhatagreatwaytoexplainJOINs,I'mcertaintherearethousandsofstudentsthatwouldbenefitfromthis...apictureistrulyworthathousandwords(especiallyincaseslikethis).

ThomasBorzeckionOctober12,20077:38AM

Ialwaysuseleouterjoinsandneverreallylikedtherightouterjoin.Isthereanymetheyareuseful?Italwaysseemsbassackwardstousearightouterjoin.

Ofcourseotherdevelopersusevisualquerydesignerswhichuserightouterjoins...

JoeBeamonOctober12,20077:39AM

IsitjustmeorOracle'ssyntaxismuchsimplerthanthat?Like:

select*fromTableA,TableBwhereTableA.Name=TableB.Name

I'mfamiliarwithboth,bothIalwayslikedthatonebeer.

RicardoonOctober12,20077:48AM

Je,

ThisisgreatforthoseSQLtypesthatdidn'ttakeasettheory,relaonalcalculusorevendiscretemathincollege.

I'dlovetoseeyouvisualizeacrossproduct:)

JayR.WrenonOctober12,20077:57AM

Ricardo,thatsyntaxispreANSI.MostdatabasessllsupportitbuttheadvantagesoftheINNERJOIN,etc.syntaxare:itmakesthequeriesmoreportableacrossdatabases;itallowsseparaonofthejoincriteriafromthefiltercriteria,whichimprovesreadability;andallowsfullouterjoinswithoutsyntacchacks.

AnotherhugebenefitisthatifyouwanttodoaCROSS(Cartesian)JOIN,andusuallyyoudon't,youhavetocallitoutspecifically.Intheoldsyntax,ifyouscreweduptheWHEREclauseyoucouldgetwildlyhugeresultsetsduetoinadvertentCartesianjoins.

MaWigdahlonOctober12,20078:02AM

JoeBeam,


8 de 19 31/08/2012 07:18 a.m.

TheonlyuseIhavefoundforRIGHTOUTERiswheni'mtoolazytoreordermyquery.;)Otherthenthatitmakesmoresensetouseonlyonetype.

IuseLEFTJOINtoo,asdomycompany.

IagreewithotherresponseshereJe,whileit'sareallynicesimplewaytorepresentjoins,itdoesn'thandlemanytomanyoronetomanyverywellandlet'snottalkaboutmalformedjoins.

It'sagoodprimer,butshouldperhapscontainawarning.

EdKennyonOctober12,20078:08AM

Thecommenterspoinngoutthatthediagramsbreakdownincaseofmulpleandorduplicateresults,areabsolutelyright.Iwasactuallythinkingofjoinsalongtheprimarykey,whichtendstobeuniquebydefinion,althoughtheexamplesarenotexpressedthatway.

Likethecartesianorcrossproduct,anythingthatresultsinmorerowsthanyouoriginallystartedwithdoesabsolutelybreaksthewholevenndiagramconcept.Sokeepthatinmind.

JeAtwoodonOctober12,20078:11AM

Exceptthatvenndiagramsexplainsetlogic,andSQLJoinshaveveryliletodowithsetlogic.

Scaryhowmanypeopleareagreeingwiththis.

WyaonOctober12,20078:29AM

Actually,I'vealwaysbeenabitshakyonjoinsmyself.IalwayshavetolookthemupanymeIneedtousingajointhatisn'tastandardfilteredcrossproduct.Butseeingthisvisualguidehasreallyhelpedmetograspitmuchbeer.IjustwentbackandreadtheWikipediaentryforjoinsanditmakescompletesensetomenow.BeforeseeingthisvisualrepresentaonImighthavejustglazedoverwhenreadingtheWikipediaarcle.

Soeveniftheyaren'texactlyrighthavingavisualrepresentaonhasreallyhelpedmetounderstandwhatisgoingon.

Thanks!

MaonOctober12,20078:29AM

AnicetrickisyoucanwritearbitraryfiltersintheONclauseofthejoin.IusethiswhenwringtrulyawfuljoinsorwhenIneedthelejoinwithanalreadyfilteredtable.

JoshuaonOctober12,20079:17AM

FromoneJetoanother,thisisabrilliantessayonasimpletopicsimplebutvexingtonewbies.IamaDB2DBA,andexplainingouterjoinstothenewbiedevelopersIencounterconstantlywillbemucheasierwithmaterialslikethis.Thanks,keepitcoming.

JereyBenneronOctober12,20079:25AM

IhaveneverseensuchagoodillustraonofSQLjoins.Goodwork.


9 de 19 31/08/2012 07:18 a.m.

JeonOctober12,20079:28AM

Imustagreethatwhilethediagramsareafairlygoodillustraonsofwhathappenswithajoin,they'renotVenndiagramsinthestrictsensebecausethat'sjustnotwhat'shappening.WhatthefirstdiagramsaystomeisthatyougetallrecordsthatarebothinsetAandsetB(akaintersecon).That'sjustnottrue.Yougetasetofnewrecords,thatareneitherinsetAnorinsetB,buttheycontainacombinaonoftheaributesofbothsets.

AaronG:joinsareconceptuallyaspecialcaseofthecartesianproduct.Howthedatabaseserveractuallycomputesthemisanothermaerenrely.

Conceptually,inrelaonalalgebra,aninnerjoinisacrossproductfollowedbyaselecononthejoincondion.That'saterriblyinecientwaytocomputeit,butconceptually,that'swhatitis.

SvenGrootonOctober12,200710:09AM

ThemoronsrunningmydatabasefundamentalsclassatmyuniversitythoughtthatteachingtheCartesianproductasamethodofjoining,thenfilteringtheresults,wasavalidalternavetotheinsanelysimpleinnerjoin.Somanypoorstudentsaregoingtohaveahardmeintheindustryasaresult.

LachlanMcDonOctober12,200710:21AM

JonRaynor:Outofcuriosity,butwhatdatabasesotherthanOraclesupport+forjoins?

Anyway,Ipreferthe"TableAaINNERJOINTableBbONa.something=b.somethingelse"syntaxover"TableAa,TableBbWHEREa.something=b.somethingelse"

(MostDBsalsosupport"USING(something)"insteadof"ONa.something=b.something"ifthecolumnshavethesamename)

ManyDBswillthrowanerrorifIaccidentlyomittheONstatementinsteadofgivingmeacrossjointhatIdidn'twant.Iimaginethatthequeryparserwouldalsoparseitfaster,asIspecifyupfrontwhichtypeofjoinI'mdoing;thequeryparserdoesn'thavetofigureitout.

PowerlordonOctober12,200710:31AM

Je,

Ilikehowyoucanregurgitateanotherauthor'scontent,andgetawaywithit.

PhilihponOctober12,200710:45AM

Bigdeal,SQLfinallycaughtupwiththe1980's.UsedaveryfineDBMSandlanguage,NOMAD2,onmainframesthatdidthisstubutonlymuchbeerandmorethoroughly.


nicelydone.ThoughIdohavetoadmittofindingithumorousthatwhatstartedoutasafast,dirtywaytotrytoclarifytosomeone(onanIRCchannel)whytheyweregengthewronginformaonforajoinnowhasotherscorrecngandlinkingtome.

LigayaTurmelleonOctober12,200710:59AM


10 de 19 31/08/2012 07:18 a.m.

Bindun!

hp://www.khankennels.com/blog/index.php/archives/2007/04/20/gengjoins/

Nicepostthough;)

BarryonOctober12,200711:04AM

"IwasundertheimpressionthatINNERJOINsuseahashmatchinternally(infact,I'mquitecertainofthis).CROSSJOINsobviouslydon't.ThereforeIdon'tseehowit'sreasonabletosaythatanIJisjustaspecialcaseoftheCJ.Isupposeinbreadandbuersettheory,wherethere'snonoonofahashoranindex,thiscouldwork,butthisistheworldofCS,wherethecleanacademicsoluonisalmostneverthemostpraccalone.TheIJyouspeakofisthenaivestpossiblealgorithm,onlyusedwhentherearenoindexes(andifyou'redesigningdatabaseswithoutanyindexes,well...stopdesigningdatabases)."AaronG

AnIJisdefinedasbeingtheequivalentofafilteredCJ.Thefactthatthiswouldnotbeareasonableimplementaondoesnotmakeadierence.

IJsareABSOLUTELYNOTimplementedexclusivelywithhashalgorithms.Sortmergealgorithmsandnestedloopalgorithmsareusedfrequently.

RevMikeonOctober12,200711:10AM

Je:

Niceillustraon.IthinkyourexamplewouldbeminutelymorereadableifthelehandtablehadanIDcolumncontainingvalues1,2,3,4(asitdoes)buttherighthandtablehadanIDcolumncontainingvaluesA,B,C,D(insteadofnumbers).Itjustmakesthatnybiteasiertotellwhat'swhatbyavoidingthetwosimilarlookingbutdierentcolumns.

MichaelChermsideonOctober12,200712:04PM

Verynicearcle.IuseaSqlitedatabase(viathecommandline)tomanagemydatabaseofcomicsandthiswouldhavebeenaverynicethingtohavewhenIstartedout.Hopefully,thiswillhelppeopleingraspingSQLjoins.

JeFlowersonOctober12,200712:07PM

Iloveit.Thegrumpycomments,thatis.Plentyofcomplaintsthatthisistoosimpliscaviewtobeuseful,naryalinktoanythingmeaer.

AndSQLaddictswonderwhyeveryoneelsewhohastodealwithRDBSjumpsatthefirstlibrarythatletsthemtreatthemasanythingotherthanRDBS...

(justfinishedreworkingabigfatchunkofcodeoriginallywrientopullalltablesintomemoryanditeravelyquery,join,andsortthemevenanultrasimpliscunderstandingwouldhavedonesomeoneatonofgood)

Shog9onOctober12,200712:11PM

TheOracle(+),actuallypredatesOracle'susage,wasbecausewhenimplementedtherewasANSInoouterjoinsyntax.InOracle8,theyaddedtheANSIsyntax.Nooneshouldbeusingtheoldsyntaxsohopfullyyouwillnotbeingseeingitoutinthewide.

willdieterichonOctober12,200712:17PM


11 de 19 31/08/2012 07:18 a.m.

Veryhelpful,thanks!!!

SpeedmasteronOctober12,20071:04PM

Whilethisisaninteresngexampleofthesekindsofjoin,IhavefoundthatinpracceIalmostneverneedanythingotherthanle/innerjoin.10yearsofpracce,andit'snottheworldsmostcomplicatedsqlmostoftheme,butthereitis.ThankGODthatstupidthingdidn'tillustrate"RIGHT"joinstoo!

Thefirst"leouterjoin"isthesameasjust"lejoin".Thesecondoneisthesameaswell,justput"b.idisnull"inthewhere.Andpleasedon'tanyonetellmethattheopmizerisn'tgoingtofigurethatout(ordothesamethingwiththeouter.)Well,maybeitwon'tifitisMySQL,butthatswhyyoudon'tusethat.

Die"outer",die!

AndrewonOctober12,20071:12PM

Excellentvisualrepresentaon.Thisshouldhelpthenoviceseeaconcretepictureintheirmindofthedierentjoins.WhileLINQisn'tquitesetbased(butclose)thesesamevisualizaonswouldworktheretoo.

ShawnWildermuthonOctober13,20073:40AM

Excellentarcle.

Thanksalot.

canyouwritesomethingsimilarforindexes.

mpunaskaratgmaildotcom

MandaronOctober13,20075:30AM

Somethingyoumayormaynotknow....

Duringthelate70'sMathswasrepackagedtoinclude

VennDiagramsMatrices

andsoon,withtheideaofpreparingforthecomingofcomputers.AtleastinEnglandthathappened...

Whattheyteachnowishardformetodetermine,asmostvicmsoftheeducaonalsystemcannotcountoutthecorrectchangeinashop.

DavidGingeronOctober13,20077:08AM

Wowman,yourdesignisstellar.Contentisgreattoo...you'vegotanewreader.

WillHarrisonOctober13,20078:40AM


12 de 19 31/08/2012 07:18 a.m.

HolycrapthisisWAYoverdue!I'vereadTONSofSQLbooksandNONEofthemhadthissimplediagram!

TerrySmithonOctober13,20079:33AM

IthinktheCartesianequaonorCrossProductinVennDiagramformwouldbebothcirclescombiningtoformasphere.

TokesonOctober13,20079:35AM

Interesng.Ifindjoinstobepreyobviouspersonally.Fundamentally,youhavetwosetsofrows,andyou'rematchingthemupbysomecriteria.Inaninnerjoin,youonlytakerowsthatmatchup.Inaleouterjoin,youtakealltheonesfromthele,plustheonesfromtherightthatmatch.Inarightouterjoin,theopposite.Ifyouhavemulplematches,youtakeallpossiblecombinaonsofthem.

smackfuonOctober13,200710:00AM

Thankyou!I'vebeenusingSQLforsolongandhavemanybooksonSQLandIhaveneverseensuchsimpleexamples.Iloveit!

MitchellHashimotoonOctober13,200710:36AM

ThereasonmanypeoplewhodoDBdevelopmenthavesuchainadequateunderstandingofthissimpletopicisbecausetheSQLlanguageisaninadequatetooltotrytoabsorbitfrom.So,ifyouneverweretaughtthisinschooloronthejob,it'skindofasurprise.

SomeDBMS'hassyntaxwithmoredescripveverbs(e.g.,'Merge','Subset','Reject'),andVenndiagramsaswellinthevendordocumentaon.

SQL2005isanimprovementover2000.ButthesadfactisismanypeoplelearnaboutrelaonalalgebrabywringSQLstatements.

AlesspreyversionofmuchofwhatJeissayingcanbefoundhere:

hp://db.grussell.org/secon010.html


WithallrespecttoJeAtwood,bewareifyou'vereadthispostandthink,"wow,that'seasywhydidnooneexplainitlikethatbefore?"Well,there'sareasonwhysomanySQLtutorialsandbooksdon'tusethesediagramsthey'reaninadequateexplanaonofwhatdatabasejoinsareallabout.

Joinsareindeedanalogoustosetoperaons,butitisonlyananalogy.Understandwhat'sreallygoingon,oryouwillgetburnedwhenyouactuallyhavetousethethings.

MaonOctober13,200711:48AM

Verynicefirstpostonthesubject.IthinkmanyofthecommentsarevalidandIthinkitwouldbegoodtoconnuethisasaseriesthatbuildsupthemorecomplexissuesofSQLjoins.

I,wholearnedthisstuthehardway,reallyapprecieatetheseprimers!

RobCromaronOctober13,20071:18PM


13 de 19 31/08/2012 07:18 a.m.

IfonlysomeonehadexplainedSQLtomelikethiswhenIstarted...greatblog.

SteveonOctober13,20071:38PM

Greatstu,

thanks

VladonOctober14,20073:05AM

verycoolvisualisaon,pleasedomoreofthat

ravemanonOctober14,20077:44AM

theproblemwithvenndiagramsisthatardbmstableisnotthesamethingasaset.

WesleyTanakaonOctober14,20078:37AM

Greatvisualexplanaon.I'veaddedthattomylistofreferencebookmarksforwhenI(oen)forgetthenatureofeachSQLjointype.

AndyonOctober15,20072:22AM

You'rerightJe.

Thisisausefulexplanaon,asfarasgengpeopletounderstandjoinsintermsofsomethingtheyalreadyunderstand(rudimentarysettheory).However,itshouldbemadeclearthatthisisnotwhatismeantbythestatementthat"therelaonaldatabasemodelisbasedonsettheory".Foralongme,IthoughtIunderstoodthatstatementbecauseIhadthisVenndiagramunderstandingofjoins.

Joinsdonotmapdirectlytothebasicaddiveoperaonsofsettheory(union,intersect,dierence),whichcreatenewsetsbywayofsimplyincludingorexcludingelementsofothersets.TherearewaystogetthoseinSQL,butitinvolveschainingUNIONandMINUSstatements.

Joins,rather,aremappingoperaons,whichcreatenewsetsbytakinganelementfromeachset,accordingtoasetofrules,andlumpingthemtogetherinanewelementforthenewsetthatisactuallyitselfasmallsetcontainingtheoriginaltwoelementsfromtheothersets.

IfAisanaddivesetoperaon(representablebyaVenndiagram),thentheresultofAonsetsXandYisasetofmembersofXandY.

A(X,Y)={x1,y1,x2,x3,y4,y5,...}

Indatabaseterms,eachelementoftheresulngsetissimplyarowfromeitherXorY.

Inrealworldterms,it'slikesorngjellybeans.Throwthemallintoabucket,throughasieve.Thesieveeliminatesallbuttheappropriateshapedones,butwhatgetsthroughissllwhatyoustartedwith:anassortmentofjellybeans.

IfBisamappingsetoperaon(representablebyalinedrawingmatchingqueson),thentheresultofBonsetsXandYisasetofsetseachcontainingamemberfromXandamemberofY.

B(X,Y)={{x1,y1},{x2,y2},{x3,y3},...}

Indatabaseterms,eachelementoftheresulngsetisanewtypeofrowthatisalumpingtogetherofarowfromXandarowfromY.


14 de 19 31/08/2012 07:18 a.m.

Inrealworldterms,anexamplewouldbetakingabucketoforangesandabucketofapples,andpullingoutpairsofoneofeachthatarethesamesize,pungthepairstogetherinlilebags,andthenpungthosebagsintoathirdbucket.Youdon'tgetabucketofapplesandoranges.Yougetabucketofpairsofanappleandanorange.

Lookingatitthisway,itshouldbereasonablyeasytoseethatthereisafundamentaldierencebetweenjoinsandthesimpleunion/intersect/dierencesetoperaons.

WaterBreathonOctober15,20075:40AM

"Exceptthatvenndiagramsexplainsetlogic,andSQLJoinshaveveryliletodowithsetlogic.

Scaryhowmanypeopleareagreeingwiththis.WyaonOctober12,200707:29AM"

Mythoughtsexactly!

cloud9ineonOctober15,20075:58AM

Je,thisisgreat!HowaboutvisuallyexplainingwhyyouhaveSELECTallthecolumnsyouareGROUPingby?Ialwayshaveahardmeexplainingthatonetopeople!

JFonOctober15,20078:39AM

I'mafraidthisconceptismisleadingonthereasonsmenonedbySvenGroot.

TheresultofajoinisnotthoseitemsintableAandthoseitemsintableB,itisa"joined"combinaonofthetwo.ThediagramsyouhavedrawnaresetdiagramsandarecreatedinSQLusingthesetoperatorsUNIONandINTERSECT.

JoinsareacompletelydierentconceptwhichIbelievearebestexplainedusingjustthetableoutputsyouhaveabovewithoutaddingtheincorrectvenndiagrams.

JF,theexplanaonbehindSELECTandGROUPBYcanbedescribedsimply.

SELECTCOUNT(*)FROMPeopleThiswillreturnacountof100people

SELECTEyeColour,COUNT(*)FROMPeopleGROUPBYEyeColourThiswillreturnBlue20,Green50,Brown30

SELECTCOUNT(*)FROMPeopleGROUPBYEyeColourThiswillreturn20,50,30whilstthiscontainsthecountsthatyouarelookingfor,theyareuselessasthereisnorelaontoeachcountandwhattherepresent.

RobinDayonOctober15,200710:53AM

AssumingPirateandNinjaareidencalrecordsinbothtables,theveryfirstthingthatcametomindaswhatresultIwouldwantwhenIjoinTablesABwas

PirateMonkeyNinjaSpagheRutabaga


15 de 19 31/08/2012 07:18 a.m.

DarthVader

Thisseemstobetheintuivemeaningofjoin.Venn'sdon'twork.

jamesmaidaonOctober15,200712:09PM

Thisisaverylikablearcle,Ipersonallyreallylikethewayyou'vedonethis...

DanielonOctober16,200711:01AM

Itneverceasestoamazemehowmanyfolksbecome"experts"onatechnicaltopiconcesomebodytakesthemetowriteaboutone.

Igivepropstotheauthorofthearclepresentedhere.Ineededtounderstandthisinformaon,anditwasputtomevisuallyinawayIcouldparsequiteeasily.

Asforthefolksdissecngthisarcle,nitpickingeventhepremise,Iaskthatyoupleasewriteabeerone,andpostthelinkinsteadofcomplaining.Iwillthendropbyandpickitapartoutofprofessionaldiscourtesy.

Thanks.

DadonOctober18,200711:48AM

TheCartesianproductof2setsisthesetofallcombinaonsoforderedpairswhichcanbeproducedfromtheelementsofbothsets.SotheVennrepresentaonwouldlookliketwopilesofpokerchipsnexttoeachother,eachpilenumberingthenumberofdatapointsintheotherboomchip.

BarfoRamaonOctober19,20071:25PM

ThanksforthisgreatTutorial:)

LeononOctober22,200710:58AM

Nicewaytoexplainthingsinjoins.Byseeingtheexamplenobodycannotforgetjoins.

Keepitup.Hopeyouwillpublishmanymorearclewhichwillbeusefultoall.

Thanks,Satheesh.

SatheeshonOctober27,200711:49AM

Ireallyhateyou.IhavenorecolleconofyourapingmywifebutIsllhateyou.

CheesecakesLikeMeonOctober28,20079:45AM

thanksforthis,IaminaclassforcrystalreportsandIshowedmanyandithelpedthem

BrianonOctober30,20079:28AM


16 de 19 31/08/2012 07:18 a.m.

nicequicklessononjoins.Thisisexactlywhatiwanted.

JaxonOctober31,200711:25AM

Greatexplanaonofthebasicsofinnerandouterjoins.I'llbereferringitoenandpassingiton.Alsoverybravetoraiseanythinghereasnot100%perfectequalsrubbish

Marn

MarnonNovember2,20075:21AM

Verysimple,I'amalwaysconfusedupwithsqloperaons,butIunderstandbeerwithsimplediagramslikeyours.Thksagainforthisnicejob.

GhislainonNovember7,20073:36AM

ThankgodforVenndiagrams!Goodjobmate.

LukeonNovember13,20078:19AM

Wow!Terrifictutorial.I'veneverunderstoodJOINSbeyondtheconceptofthem.Thankyoufordemysfying!

Clay

ClayonNovember13,20078:52AM

Thankyousomuchformakingthisavailable.CheersfromtheUK

JamesPrinteronNovember14,20072:54AM

VERYGOODEXAMPLE,[email protected]

sureshonNovember15,20077:09AM

Onehelpfulsteptounderstandcomplexqueriestoindentthesubqueries.ThiscanbedoneautomacallywiththisfreeSQLFormaerathp://www.sqlinform.com

GuidoMarcelonNovember22,200711:45AM

thisisawesome..

ArunaonDecember2,20073:00AM


17 de 19 31/08/2012 07:18 a.m.

Thisarcleisgreat!Apictureisworthathousandwords,andyourpictureshavehelpedmetounderstandSQLjoins.Thanks.

DerekonDecember7,200711:46AM

HeythatwasaCLEARexplanaon...juststartedupwithsqliteanddidntgetwhyiwasnotgengtheseleconiwanted.

Nowido!

HennieonDecember8,20079:05AM

howdoyoujointwocolumnstogetherandgetonecolumninaquery.ForEX:ihavelastnameandfirstnamecolumnsinatable.iwanttojointhesetwocolumnstogetherandgetlastnameandfirstnameinonecolumn??

InquisitoronDecember13,200712:15PM

Goodjob.

GurmeetonJanuary2,20089:45AM

AlthoughI'mfamiliarwiththetopic,Islllikeyourniceandeasywayofexplainingthings,nicework.

A.NETDeveloper'sBlogonJanuary6,20089:35AM

ThiswasaGREATexplanaon.IamworkingwithajuniorDBAandthishashelpedtremendously.Iwillhavetogobackandreadsomeofyourotherblogs.

AWESOMEJOB!!!!!

SHALLonJanuary8,20088:58AM

Thereisaquerybuildertoolthatcomeswithoracleapex.Checkitout.

OracleTube.comonJanuary16,200810:44AM

Assomeonesuggestedtocreatedsomethingthemselvesandthencricize;)I'vetriedtodescriberelaonshipsamongvariousjointypesusingERdiagrammhere:hp://gplivna.blogspot.com/2008/01/sqljointypesimstudyingbitsql.htmlAllcommentswelcome!

GintsPlivnaonJanuary25,20082:43AM

It'sagreatwaytoexplainthings....thanks

PrajeeshonFebruary9,20085:05AM


18 de 19 31/08/2012 07:18 a.m.

ThanksforthisJe,wasstrugglingabitwiththeconceptbutyou'vemadeitclearenoughthatevenIcangetmyheadaroundit:)

AndyonFebruary18,20085:29AM

Awesome!

JulieMcBrienonMarch4,200810:09AM

IreallyhavetohandittoyouJe,thewayyoubreakthingsdownisjustgreat.Ireadtheblogseverymeyoupostone.Thisoneinparcularcameinhandyjusttheotherday.OurQAteamwashavingtroubleunderstandingwhatwasgoingoninthedatabaseandwedeveloperswouldsendthemqueriestoruntofindthedatatheyneeded.TheQAteamhasarudimentaryunderstandingofSQL,butcouldn'tquitegraspjoins.IsentalinktothisarcletoourleadQA,andnowyourarcleisfirmlypostedonhercubewallforreference.Justwantedtosaythankyoufortheblogs!

ArcondonMarch11,20088:56AM

Morecomments

Thecommentstothisentryareclosed.

Content2012JeAtwood.Logoimageusedwithpermissionoftheauthor.1993StevenC.McConnell.AllRightsReserved.


19 de 19 31/08/2012 07:18 a.m.

coding horror_ a visual explanation of sql joins.pdf

Documents