Detekcija fragmenata koda raspoloživih na portalu Stack Overflow u aplikacijama na GitHubu

master's thesis

• Title (english): Detecting code fragments available on Stack Overflow in applications on GitHub
• Author: Lea Rački
• Mentor: Stjepan Groš (mentor)
• Committee member: Stjepan Groš (predsjednik povjerenstva)
• Committee member: Ante Đerek (član povjerenstva)
• Committee member: Alan Jović (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2020-07-10, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

Programeri koriste različite portale kao pomoć u svom radu te za vrijeme razvoja programa i njihovog održavanja recikliraju već napisane dijelove programskih komponenata, isječke kodova i programske knjižice. Jedni od najpopularnijih takvih internetskih portala su Stack Overflow i GitHub. Stack Overflow je forum koji se temelji na konceptu pitanja i odgovora, a programeri ga nekad čak preferiraju više od službenih dokumentacija i knjiga kao resursa za traženje programskog rješenja za neki problem. GitHub je jedan od najvećih internetskih portala za pohranu programskog koda.

U ovom je radu istražena pretpostavka da programeri za vrijeme svoga rada kopiraju već napisane fragmente programskog koda u svoje aplikacije koje zatim u određenoj mjeri izmjenjuju i prilagođavaju svojim potrebama. Proces korištenja gotovog koda koji uključuje ponovnu upotrebu odnosno kopiranje izvornog koda naziva se kloniranje koda. Uz to se istražuje koliko je kloniranje kodova često i na koji se način mogu detektirati klonirani fragmenti u izvornim kodovima te u kojoj se mjeri isječci kodova dostupni na Stack Overflow portalu kopiraju u javno dostupnim projektima na GitHub-u.

Prikupljeno je ukupno 1,221,907 Java datoteka s GitHub-a i 349,791 Java datoteka sa portala Stack Overflow. Podaci su obrađeni alatom SourcererCC. SourcererCC je alat za otkrivanje klonova koda koji se temelji na korištenju tokena i može otkrivati klonove u bilo kojem programskom jeziku. Može otkriti klonove tipa T1, T2 pa sve do tipa T3 odnosno može otkriti klonove s dodanim, uklonjenim i/ili izmijenjenim linijama koda, a uz to ima veliku preciznost i odziv pa je konkurentan ostalim alatima za detekciju klonova.

U 2,27 posto datoteka s GitHub-a pronađeno je podudaranje od 80 posto ili više s nekim od fragmenata Java koda objavljenih na Stack Overflow-u, odnosno u njima je detektirano kopiranje izvornog koda. Postotak klonova uvelike ovisi o odabranim setovima podataka za istraživanje te da je moguće da bi neki drugi set podataka prikupljen s GitHub-a rezultirao većim postotkom klonova između Stack Overflow-a i GitHub-a.

Abstract (english)

Developers use various portals to help with their work, and during program development and maintenance they recycle parts of program components, code snippets, and frameworks. One of the most popular such web portals is Stack Overflow and GitHub. Stack Overflow is a forum-based online portal with the concept of questions and answers, and developers sometimes even prefer it more than official documentation and books as a resource for finding a software solution to a problem. GitHub is one of the largest Internet portals for storing program code.

This paper investigates the assumption that developers copy fragments of program code into their applications during their work, which is then modified and adapted to their needs. The process which involves reusing or copying source code is called code cloning. In addition, this paper investigates how common code cloning is and how cloned fragments can be detected in source code and the extent to which code snippets available on the Stack Overflow portal are copied in publicly available projects on GitHub.

A total of 1,221,907 Java files from GitHub and 349,791 Java files from the Stack Overflow were collected. The data was processed with the SourcererCC tool. SourcererC is a token-based code clone detection tool that can detect clones in any programming language. It can detect clones of type-1, type-2 up to type-3, as well as it can detect clones with added, removed and/or modified lines of code, and in addition, it has high precision and response, so it is competitive with other clone detection tools.

2.27 percent of files from GitHub were found to match 80 percent or more of some of the Java code snippets published on Stack Overflow, and source code copying was detected in them. The percentage of clones largely depends on the data sets selected for research, and it is possible that some other data set collected from GitHub would result in a higher percentage of clones between Stack Overflow and GitHub.

• Keywords: Stack Overflow; GitHub; Java; kloniranje kodova; SourcererCC; klonovi tipa T1; klonovi tipa T2; klonovi tipa T3; otkrivanje klonova koda
• Keywords (english): Stack Overflow; GitHub; Java; code clones; SourcererCC; clone type-1; clone type-2; clone type-3; code cloning detection
• Language: croatian
• URN:NBN: urn:nbn:hr:168:877289
• Study programme: Title: Computing; Study programme type: university; Study level: graduate; Academic / professional title: magistar/magistra inženjer/inženjerka računarstva (magistar/magistra inženjer/inženjerka računarstva)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Closed access
• Created on: 2023-01-13 13:37:40