A code clone oracle

Abstract

Code clones are functionally equivalent code segments. Detecting code clones is important for determining bugs, fixes and software reuse. Code clone detection is also essential for developing fast and precise code search algorithms. However, the challenge of such research is to evaluate that the clones detected are indeed functionally equivalent, considering the majority of clones are not textual or even syntactically identical. The goal of this work is to generate a set of method level code clones with a high confidence to help to evaluate future code clone detection and code search tools to evaluate their techniques. We selected three open source programs, Apache, Python and PostgreSQL, and randomly sampled a total of 1536 function pairs. To confirm whether or not these function pairs indicate a clone and what types of clones they belong to, we recruited three experts who have experience in code clone research and four students who have experience in programming for manual inspection. For confidence of the data, the experts consulted multiple code clone detection tools to make the consensus. To assist manual inspection, we built a tool to automatically load function pairs of interest and record the manual inspection results. We found that none of the 66 pairs are textual identical type-1 clones, and 9 pairs are type-4 clones. Our data is available at

Examining the effectiveness of using concolic analysis to detect code clones

“…Figure 2 displays the results of various Levenshtein values in discovering clones in a single class as defined by Krawitz and Roy et al Figure 3 shows a similar analysis the results from the Krutz and Le [14] oracle. A higher Levenshtein threshold score is likely to aid in the discovery of more clones, but will also lead to more false positives, creating lower precision but higher recall.…”

“…In the following sections, concolic analysis for clone detection will be evaluated against two small oracles created by Krawitz [12] and Roy et al [21] and a larger oracle built by Krutz and Le [14] to determine what types of clones concolic analysis is capable of finding, along with its accuracy, precision, recall, and F-score.…”

“…Our first step in determining this as the most appropriate Levenshtein value was to produce concolic output from the oracles by Krawitz [12], Roy et al [21] and Krutz and Le [14] using Levenshtein scores of 0-40 with 10 point increments as a basis for determining clones. To obtain the optimal number, we compared the precision, recall, F-score, and accuracy scores of each increment and found that for all of the code bases, the Levenshtein value of 30 produced the best rates.…”

“…In order to evaluate the effectiveness of concolic analysis for code clone detection, we used a function level clone oracle created by Krutz and Le [14] since it is the only known oracle to contain all four types of code clones explicitly defined. We ran concolic analysis for code clone detection and measured the accuracy, precision, recall, and F-score of this technique against this clone oracle.…”

“…We measured the precision, recall, accuracy, and F-score of concolic analysis for code clone detection against two small existing oracles created by Krawitz [12] and Roy et al [21] and a larger one built by Krutz and Le [14]. We found that concolic analysis was able to discover clones with an average precision of .8, recall of .91, F-score of .85, and an accuracy of .99.…”