Towards the Automation of Metamorphic Testing in Model Transformations

281 XXI Jo nadas de Ingenie ía del So wa e y Bases de Da os, pp. 281-284 © Ediciones Uni e sidad de Salamanca Towa ds he Au oma ion o Me amo phic Tes ing in Model T ans o ma ions Ja ie T oya, Se gio Segu a, and An onio Ruiz-Co ´ es Depa men o Compu e Languages and Sys ems Uni e sidad de Se illa, Spain {j oya, se giosegu a, a uiz}@us.es Abs ac . Model ans o ma ions a e he co ne s one o Model-D i en Engi- nee ing, and p o ide he essen ial mechanisms o manipula ing and ans o m- ing models. Checking whe he he ou pu o a model ans o ma ion is co ec is a manual and e o -p one ask, his is e e ed o as he o acle p oblem in he so wa e es ing li e a u e. The co ec ness o he model ans o ma ion p og am is c ucial o he p ope gene a ion o i s ou pu , so i should be es ed. Me amo - phic es ing is a es ing echnique o alle ia e he o acle p oblem consis ing on exploi ing he ela ions be ween di e en inpu s and ou pu s o he p og am un- de es , so-called me amo phic ela ions. In his pape we gi e an insigh in o ou app oach o gene ically de ine me amo phic ela ions o model ans o ma ions, which can be au oma ically ins an ia ed gi en any speci ic model ans o ma ion. Keywo ds: Me amo phic Tes ing, Model T ans o ma ion, Au oma ion, Gene ic 1 In oduc ion Model T ans o ma ions (MTs) a e he co ne s one o Model-D i en Enginee ing (MDE). They p o ide he essen ial mechanisms o manipula ing and ans o ming models. Checking whe he he ou pu o a model ans o ma ion is co ec is a manual and e o -p one ask, his is e e eed o as he o acle p oblem in he so wa e es ing li - e a u e. Indeed, he quali y o he gene a ed so wa e a i ac s is highly a ec ed by he co ec ness o he de eloped model ans o ma ions. Fo his eason, se e al app oaches ha e been p oposed ha e i y he co ec beha io o he ans o ma ions using o mal me hods [6,1] o ce i y hei beha io o a selec ed se o es models mainly o iden i y bugs in a cos -e ec i e way [2,7]. Me amo phic Tes ing (MT’) [5] is a me hodology designed o alle ia e he o a- cle p oblem. Di e en om con en ional es ing s a egies, MT’ consis s on exploi ing he ela ions be ween di e en inpu s and ou pu s o he p og am unde es , so-called Me amo phic Rela ions (MRs). In p ac ice, MRs de ine possible modi ica ions o a es inpu and how hose changes a e p opaga ed o he p og am ou pu . A basic example o MR can be de ined o he p og am ha compu es he sine unc ion. Le us suppose we wan o know he exac alue o sin(5). Is an obse ed ou pu o 0.091 co ec ? A ma hema ical p ope y o he sine unc ion s a es ha sin(x) = sin(π−x), and we can use his o es whe he sin(5) = sin(π−5) wi hou knowing he conc e e alues o ei he sine calcula ion. Cu en ly, he bigges limi a ion o MT’ has o do wi h he 282 © Ediciones Uni e sidad de Salamanca ja ie oya, se gio segu a, an onio uiz-co és owa ds he au oma ion o me amo phic es ing in model ans o ma ions XXI Jo nadas de Ingenie ía del So wa e y Bases de Da os, pp. 281-284 2 Towa ds he Au oma ion o Me amo phic Tes ing in Model T ans o ma ions C1: es model (sou ce model) T1: esul model ( a ge model) MTexecu ion C2: ollow‐up es model ( ollow‐upsou ce model) T2: ollow‐up esul model ( ollow‐up a ge model) con olled modi ica ion MTexecu ion Fig. 1: Me amo phic Tes ing in Model T ans o ma ions. de ini ion o he MRs. In ac , he au oma ic gene a ion o MRs has been acknowledged as one o he big challenges in MT’ [5]. Figu e 1 displays he scena io o MT’ in MTs. We ha e a es model (C1), which is ypically he sou ce model. I we apply he MT, we ob ain he esul model (T1), i.e., he a ge model. By applying a con olled modi ica ion in he es model, we ob ain he ollow-up es model (C2). I we now apply he same MT o C2, we ge he ollow-up esul model (T2). In his con ex , a MR conside s he modi ica ion done in C2 wi h espec o C1, and he consequences ha his has in T2 wi h espec o T1. We show an example in nex sec ion. As a as we a e conce ned, he e is only one app oach ha applies MT’ in MTs [3]. The au ho s demons a e he e ec i eness and easibili y o i s applica ion, al hough hey apply i manually in a speci ic scena io, o which hey de ine he MRs. In ou app oach, we p opose o au oma ically gene a e MRs o any model ans o ma ion, as we explain in he nex sec ion. Then, in Sec ion 3 we desc ibe ou nex s eps. 2 App oach The goal o ou app oach is o au oma e he p ocess o me amo phic es ing (MT’) in model ans o ma ions (MTs). The eby, we p opose he au oma ic gene a ion o me a- mo phic ela ions (MRs) o any model ans o ma ion. We wo k wi h ans o ma ions w i en in he ATL T ans o ma ion Language due o i s impo ance bo h in academia and indus y. In o de o au oma ically ex ac in o ma ion ou o a ans o ma ion, we make use o explici ace models. A ace model can be au oma ically ob ained om a ans o - ma ion execu ion, e.g., by using Jouaul ’s T aceAdde [4], and is composed o a se o aces, one o each ule execu ion. A ace cap u es he name o he applied ule and he elemen s o he sou ce model (sou ceElems ela ionship) ha a e used o c ea e new elemen s in he a ge model ( a ge Elems ela ionship). The e o e, by na iga ing he ace model, we know which a ge elemen (s) ha e been c ea ed om which sou ce el- emen (s) and by which ule. A simple example o a gene ic ace is shown in Figu e 2(a). Please no e ha mo e han one elemen may appea as sou ceElems and a ge Elems. We conside his ace as gene ic because each o he h ee elemen s appea ing in i (Sou ceElemen ,T ace and Ta ge Elemen ) can be ins an ia ed in a pa icula scena io. The idea o ou app oach is o de ine gene ic MRs o gene ic aces. These MRs can hen be ins an ia ed oge he wi h he gene ic aces. Fo ins ance, conside ing he gene ic ace o Figu e 2(a), we know ha i we ha e a es model (C1, Figu e 1) and we add an elemen o ype Sou ceElemen in he ollow-up es model (C2), hen an 283 ja ie oya, se gio segu a, an onio uiz-co és owa ds he au oma ion o me amo phic es ing in model ans o ma ions XXI Jo nadas de Ingenie ía del So wa e y Bases de Da os, pp. 281-284 © Ediciones Uni e sidad de Salamanca Towa ds he Au oma ion o Me amo phic Tes ing in Model T ans o ma ions 3 T ace uleName:Da aType2Type Da aType name:S ing Type name:S ing sou ceElems a ge Elems T ace uleName Sou ceElemen sou ceElems a ge Elems Ta ge Elemen T ace uleName:Class2Table Class name:School Table name:School sou ceElems a ge Elems Column name:objec Id key ype (a) Simple Gene ic T ace. T ace uleName:Da aType2Type Da aType name:S ing Type name:S ing sou ceElems a ge Elems T ace uleName Sou ceElemen sou ceElems a ge Elems Ta ge Elemen T ace uleName:Class2Table Class name:School Table name:School sou ceElems a ge Elems Column name:objec Id key ype (b) T aces o he ans o ma ion exce p o Lis ing 1.2. Fig. 2: Gene ic and ins an ia ed aces elemen o ype Ta ge Elemen is c ea ed o i in he ollow-up esul model (T2). This means ha T2 has one mo e elemen o his ype han T1. Ha ing his in o ma ion in o accoun , we can de ine he i s gene ic MR shown in Lis ing 1.1, w i en in he OCL language. Besides, he numbe o elemen s o any o he ype should emain he same in T1 and T2, so u he MRs can be de ined, such as he second one in he same lis ing. Lis ing 1.1: Gene ic MRs o he addi ion o a Sou ceElemen 1T1_Ta ge Elemen .allIns ances()->size()=T2_Ta ge Elemen .allIns ances()->size()-1 2T1_AnyO he Type.allIns ances()->size()=T2_AnyO he Type.allIns ances()->size() In o de o show an example o ins an ia ion o he gene ic MRs shown be o e, we choose he well-known Class2Rela ional case s udy. We will ocus on he exce p o he ans o ma ion, which has been sligh ly modi ied o simplici y in he explana ion, shown in Lis ing 1.2. I we ha e a sou ce model wi h a Da aType and a Class and we apply he ans o ma ion, he esul ing ace model is depic ed in Figu e 2(b). We can see a ace ha e lec s he c ea ion o a Type om a Da aType and ano he one ha s o es he c ea ion o a Table and a Column om a Class. Lis ing 1.2: Exce p o Class2Rela ional ans o ma ion 1 ule Da aType2Type { 2 om 3d : Class!Da aType 4 o 5ou : Rela ional!Type ( 6name <- d .name 7) 8} 9 10 11 ule Class2Table { 12 om c : Class!Class 13 o 14 ou : Rela ional!Table ( 15 name <- c.name, 16 key <- key), 17 key : Rela ional!Column ( 18 name <- ’objec Id’, 19 ype <- hisModule.objec IdType) 20 } Since hese wo aces a e ins an ia ions o he gene ic one shown in Figu e 2(a), we can also ins an ia e he MRs shown in Lis ing 1.1. In pa icula , we ha e wo scena ios. The i s one consis s o adding an elemen o ype Da aType in C2, wha yields he 284 © Ediciones Uni e sidad de Salamanca ja ie oya, se gio segu a, an onio uiz-co és owa ds he au oma ion o me amo phic es ing in model ans o ma ions XXI Jo nadas de Ingenie ía del So wa e y Bases de Da os, pp. 281-284 4 Towa ds he Au oma ion o Me amo phic Tes ing in Model T ans o ma ions MRs shown in Lis ing 1.3. In he second scena io we add an elemen o ype Class in C2, ob aining he MRs shown in Lis ing 1.4. Lis ing 1.3: MRs o he addi ion o a Da aType in C2 1T1_Type.allIns ances()->size()=T2_Type.allIns ances()->size()-1 2T1_Column.allIns ances()->size()=T2_Column.allIns ances()->size() 3T1_Table.allIns ances()->size()=T2_Table.allIns ances()->size() Lis ing 1.4: MRs o he addi ion o a Class in C2 1T1_Column.allIns ances()->size()=T2_Column.allIns ances()->size()-1 2T1_Table.allIns ances()->size()=T2_Table.allIns ances()->size()-1 3T1_Type.allIns ances()->size()=T2_Type.allIns ances()->size() 3 Nex S eps and Obse a ions In his pape we ha e gi en an insigh in o ou app oach o au oma e he gene a ion o MRs o MTs. We iden i y gene ic pa e ns in he aces, om which we de ine gene ic MRs o ganized as well in pa e ns. Fo ins ance, one pa e n is he gene ic ace and MRs we ha e shown in his pape . Despi e i s simplici y, we a e pe o ming ongoing wo ks de ining mo e pa e ns whe e elemen s, a ibu es and ela ionships a e aken in o accoun , so ha we end up wi h a la ge se o MRs. One o he pu poses o he gene a ed MRs is o au oma e eg ession es s, since hey can be checked as o whe he hey hold in di e en e sions o he model ans o ma ion p og am and o any es model. As men ioned, ou app oach akes as inpu one o mo e execu ions o a MT, i.e., he esul ing ace models. The numbe o execu ions o he MT ecei ed and hei size in luence he comple eness o he MRs gene a ed. Fo ins ance, i a ule is ne e applied in any o he execu ions ecei ed as inpu , no MRs will conside i s beha io . Acknowledgmen s. This wo k has been pa ially unded by he Eu opean Commission (FEDER) and Spanish Go . unde CICYT p ojec BELI (TIN2015-70560-R), and by he Andalusian Go . p ojec s THEOS (TIC-5906) and COPAS (P12- TIC-1867). Re e ences 1. Am ani, M., Lucio, L., Selim, G.M.K., Combemale, B., Dingel, J., Vangheluwe, H., T aon, Y.L., Co dy, J.R.: A idimensional app oach o s udying he o mal e i ica ion o model ans o ma ions. In: P oc. o ICST. pp. 921–928. IEEE (2012) 2. Gogolla, M., Vallecillo, A.: T ac able Model T ans o ma ion Tes ing. In: ECMFA. LNCS, ol. 6698, pp. 221–235. Sp inge (2011) 3. Jiang, M., Chen, T.Y., Kuo, F., Zhou, Z., Ding, Z.: Tes ing Model T ans o ma ion P og ams using Me amo phic Tes ing. In: SEKE’14. pp. 94–99 (2014) 4. Jouaul , F.: Loosely Coupled T aceabili y o ATL. In: Wo kshop P oc. o ECMDA (2005) 5. Segu a, S., F ase , G., Sanchez, A., Ruiz-Co es, A.: A su ey on me amo phic es ing. IEEE T ansac ions on So wa e Enginee ing (99), 1–1 (2016) 6. T oya, J., Vallecillo, A.: A Rew i ing Logic Seman ics o ATL. Jou nal o Objec Technology 10, 5:1–29 (2011) 7. Vallecillo, A., Gogolla, M., Bu gue˜ no, L., Wimme , M., Hamann, L.: Fo mal speci ica ion and es ing o model ans o ma ions. In: P oc. o SFM. LNCS, ol. 7320, pp. 399–437. Sp inge (2012)