Sub-clones: Considering the Part Rather than the Whole Robert Tairas Department of Computer and Information Sciences University of Alabama at Birmingham Jeff Gray Department of Computer Science University of Alabama This research is supported by University of Alabama at Birmingham Software Composition and Modeling Lab University of Alabama NSF grant CPA-0702764
2 Cloning in Software Code Clones: A section of code that is duplicated in multiple locations in a program Source Code Different granularity levels: Statements, Block, Method, Class, Program Clone Group: Clones of the same duplication Cloned Code
3 Maintaining Clones After a period of time A new programmer
4 Removing Clones through Refactoring Modularizing the code represented by clones through appropriate abstractions may improve code quality Less duplicated code to maintain Modularized Clone Ease of future maintenance efforts Clone 1 Clone 2 Refactoring is one means of improving the quality of code The goal of refactoring is to preserve the external behavior of code while improving its internal structure
5 Refactoring Clones publ i c c l a s s A { publ i c c l a s s A { publ i c voi d m e t hod( ) { publ i c voi d m e t hod( ) { {cloned statements} Extract-Method ne wM e t hod( ) ; . . . {cloned statements} Refactoring ne wM e t hod( ) ; {cloned statements} } . . . {cloned statements} publ i c voi d ne wM e t hod( ) { {cloned statements} {cloned statements} {cloned statements} {cloned statements} } {cloned statements} } } }
6 Clone Refactoring Process Manually Determine Clones Refactoring Detect Clones For Refactoring Clones Changes between two versions First version contains original code Second version contains refactored code
7 Clone Refactoring Process Manually Determine Clones Refactoring Detect Clones For Refactoring Clones Automated Clone Detection Tool What are the refactoring characteristics of clones detected by a clone detection tool, if such a tool was used in the clone maintenance process?
8 Approach: Observing Refactorings Observing actual clone-related refactorings in multiple release versions of JBoss (v2.2.0–4.2.3) Used Simian clone detection tool Compare Source Code Source Code Clone Code Clones (Version 1) (Version 2) Detection @ @ - 2471, 13 +2469, 7 @ @ s c a n_pos i t i on. c ur r e nt _s l ot = Pa ge . I NVALI D_SLOT_NUM BER; / / r e l e a s e t he s c a n l oc k now t ha t we ha ve s a ve d a wa y t he r ow. - - i f ( s c a n_pos i t i on. c ur r e nt _s c a n_pa ge no ! = 0) - { - t hi s . ge t Loc ki ngPol i c y( ) . unl oc kSc a n( - s c a n_pos i t i on. c ur r e nt _s c a n_pa ge no) ; - s c a n_pos i t i on. c ur r e nt _s c a n_pa ge no = 0; - } + unl oc kCur r e nt Sc a n( s c a n_pos i t i on) ; } V ersion 1 V ersion 2 V ersion 1 V ersion 2 V ersion 1 V ersion 2 } } diff region C lone c C lone c C lone c C lone c C lone c diff region
9 Refactoring of Simian Clones Observations 21 Extract Method -type Refactorings Range of refactored code not equal to the range reported as a clone Type T otal Extract Method 14 Extract Method with Pull-up Method 1 Extract Method to utility class 6 Total 21
10 Observing with Other Tools Consider clones reported by other tools CCFinder, CloneDR, Deckard, and SimScan Run these tools on source files associated with the 21 Extract Method -type refactorings from Simian clones Source Code Simian (Version 1) Compare Source Code Code Clones Code Clones CCFinder (Version 2) Compare Source Code Code Clones CloneDR (Version 2) Source Files Compare Source Code Code Clones Deckard (Version 2) Compare Source Code Code Clones SimScan (Version 2)
11 Evaluation: Tool Coverage Coverage of 21 Extract Method -type refactorings in JBoss Initially detected by using Simian clones Reported clones that exactly covered the refactored code were less than half for all the tools T ool Exact Larger Coverage Coverage 1. CCFinder 4 (19%) 8 (38%) 2. CloneDR 6 (28%) 9 (42%) 3. Deckard 8 (38%) 3 (14%) 4. Simian 2 (9%) 0 (0%) 5. Simscan 6 (28%) 12 (57%)
12 Refactoring in Clone Ranges 1 2 4 5 pr ot e c t e d St r i ng ge t Va l ue ( St r i ng na m e , St r i ng va l ue ) { 1 2 4 5 i f ( va l ue . s t a r t s W i t h( " ${" ) & & va l ue . e nds W i t h( " }" ) ) { 1 2 3 4 5 - t r y { 1 2 3 4 5 - St r i ng pr ope r t yNa m e = va l ue . s ubs t r i ng( 2, va l ue . l e ngt h( ) - 1) ; 1 2 3 4 5 - Obj e c t Na m e pr ope r t ySe r vi c e ON = ne w Obj e c t Na m e ( “ . . . " ) ; 1 2 3 4 5 - Ke r ne l Abs t r a c t i on ke r ne l Abs t r a c t i on = Ke r ne l Abs t r a c t i onFa c t or y. ge t I ns t a nc e ( ) ; 1 2 3 4 5 - St r i ng pr ope r t yVa l ue = ( St r i ng) ke r ne l Abs t r a c t i on. i nvoke ( . . . ) ; 1 2 3 5 - l og. de bug( " Re pl a c e d e j b- j a r . xm l e l e m e nt " + na m e + " wi t h va l ue " + pr ope r t yVa l ue ) ; 1 2 3 5 - r e t ur n pr ope r t yVa l ue ; 1 2 3 5 - } c a t c h ( Exc e pt i on e ) { 1 2 3 5 - l og. wa r n( " Una bl e t o l ook up pr ope r t y s e r vi c e f or e j b- j a r . xm l e l e m e nt " + . . . ) ; 1 2 3 5 - } + St r i ng r e pl a c e m e nt = St r i ngPr ope r t yRe pl a c e r . r e pl a c e Pr ope r t i e s ( va l ue ) ; + i f ( r e pl a c e m e nt ! = nul l ) i f ( e dge i ns t a nc e of M Tr a ns i t i on) { + va l ue = r e pl a c e m e nt ; M Tr a ns i t i on t r = ( M Tr a ns i t i on) e dge ; 1 2 5 } - Fi gTr a ns t r Fi g = ne w Fi gTr a ns ( t r ) ; 1 2 5 r e t ur n va l ue ; - / / s e t s our c e a nd de s t 1 2 5 } - / / s e t a ny a r r owhe a ds , l a be l s , or c ol or s - M St a t e Ve r t e x s our c e SV = t r . ge t Sour c e ( ) ; Refactoring performed on - M St a t e Ve r t e x de s t SV = t r . ge t Ta r ge t ( ) ; - Fi gNode s our c e FN = ( Fi gNode ) l a y. . . only part of the reported - Fi gNode de s t FN = ( Fi gNode ) l a y. . . - t r Fi g. s e t Sour c e Por t Fi g( s our c e FN) ; clone range - t r Fi g. s e t Sour c e Fi gNode ( s our c e FN) ; - t r Fi g. s e t De s t Por t Fi g( de s t FN) ; - t r Fi g. s e t De s t Fi gNode ( de s t FN) ; Sub-clone refactoring + Fi gTr a ns t r Fi g = ne w Fi gTr a ns ( t r , l a y) ; r e t ur n t r Fi g; }
13 Evaluation: Focus on Deckard Deckard selected due to tree-based tool performance JBoss re-evaluated Additional artifacts: ArgoUML (v0.10.1–0.26) and Apache Derby (v10.1.1.0–10.5.3.0) Property JBoss ArgoUML Derby Refactoring Exact coverage 19 17 12 Coverage Sub-clone coverage 14 9 15 Coverage Same level 4 4 6 Levels 1 level above 9 2 8 > 1 level above 1 3 1 Clone Refactorable 7 4 8 Differences Not Refactorable 7 5 7
14 Evaluation: Focus on Deckard Reported clone range mainly the same level or one syntactic level above the actual refactored code Possibly to keep some logic in the original location Property JBoss ArgoUML Derby Refactoring Exact coverage 19 17 12 Coverage Sub-clone coverage 14 9 15 Coverage Same level 4 4 6 Levels 1 level above 9 2 8 > 1 level above 1 3 1 Clone Refactorable 7 4 8 Differences Not Refactorable 7 5 7 Programmers only refactored a sub-clone even when the entire clone was refactorable
15 Conclusion We observed the actual refactoring of clones by evaluating source code changes between multiple versions In various instances only part of the reported clone (i.e., sub-clone) was refactored We conclude that sub-clone refactoring should be included in the clone maintenance process Future Work Individual evaluation of other clone detection tools Provide support for sub-clone refactoring in an IDE
16 CeDAR plug-in
17 Sub-clones in CeDAR
18 Thank you Personal: http://www.cis.uab.edu/tairasr Code Clones Literature: http://www.cis.uab.edu/tairasr/clones/literature SoftCom Laboratory: http://www.cis.uab.edu/softcom
Recommend
More recommend
