Using discriminative feature in software entities for relevance identification of code changes

Abstract

Developers often bundle unrelated changes (eg, bug fix and feature addition) in a single commit and then submit a “poor cohesive” commit to version control system. Such a commit consists of multiple independent code changes and makes review of code changes harder. If the code changes before commit can be identified as related and unrelated ones, the “cohesiveness” of a commit can be guaranteed. Inspired by the effectiveness of machine learning techniques in classification field, we model the relevance identification of code changes as a binary classification problem (ie, related and unrelated changes) and propose discriminative feature in software entities to characterize the relevance of code changes. In particular, to quantify the discriminative feature, 21 coupling rules and 4 cochanged type relationships are elaborately extracted from software entities to construct related changes vector (RCV). Twenty‐one coupling rules at granularities of class, attribute, and method can capture the relevance of code changes from structural coupling dimension, and 4 cochanged type relationships are defined to capture the change type combinations of software entities that may cause related changes. Based on RCV, machine learning algorithms are applied to identify the relevance of code changes. The experiment results show that probabilistic neural network and general regression neural network provide statistically significant improvements in accuracy of relevance identification of code changes over the other 4 machine learning algorithms. Related changes vector with 72 dimensions (RCV72) outperforms other 2RCVs with less dimensions.

Change-Patterns Mapping: A Boosting Way for Change Impact Analysis

“…Then, our method will extract the changepatterns between each salient class and the rest of the classes in the commit, and apply the change-patterns to the rank boosting step. We found in previous study [39] that if a commit contains 2 starting entities, it contains unrelated changes. Then, one starting entity builds change-patterns between parts of the classes in the commit, while another starting entity builds change-patterns between the other parts of the classes in the commit.…”

“…Since there can be various coupling dependencies between the starting entity and impacted entities [23], there are many types of change-patterns. According to the study [23] as well as our observation from the objectoriented programming paradigm in previous study [13], we get 21 common instances of coupling dependencies from the source code, and these coupling dependencies may exist at the class level, method level, and variable level. Coupling Dependency 1.…”

“…We conduct change impact analysis for 7 Java opensource projects in our evaluation. The selected projects are commonly evaluated projects in most of the software engineering related research [13], [14], [15], i.e., FreeCol [28], HSQLDB [29], JAMWiki [30], jEdit [31], JHotDraw [32], Makagiga [33], OmegaT [34]. Since these 7 projects have more than ten years evolution history, they have a relatively high number of commits.…”

“…After that, we boost the classes in the initial impact set when they involve similar change-patterns as comparing with the ones betweenc init and its impacted classes. We chose 7 open-source projects for CIA tasks, which are considered to be high-quality, well evolved and widely used in related studies [13], [14], [15]. We found that the combination of traditional CIA methods and our boosting method is superior to the standalone CIA methods, i.e., the maximum relative improvements of precision for ImpactMiner, JRipples and ROSE are 21.98%, 6.52% and 11.81%, respectively.…”

Does your code need comment?

“…According to the previous experience in mining data sets from software repository, there is a small portion of instances in the training set might represent incorrect commenting instances. Therefore, it is reasonable to assume that some instances that required header comments are not actually commented, while others are unnecessarily commented.…”