Logging statements' prediction based on source code clones

Abstract

Log files are widely used to record runtime information of software systems, such as the time-stamp of an event, the unique ID of the source of the log, and a part of the state of task execution. The rich information of logs enables system operators to monitor the runtime behaviors of their systems and further track down system problems in production settings. Although logs are useful, there exists a trade-off between their benefit and cost, and it is a crucial problem to optimize the location and content of log messages in the source code, i.e., where and what to log? Prior research has analyzed logging statements in the source code and proposed ways to predict and suggest the location of log statements in order to partially automate log statement addition to the source code. However, there are gaps and unsolved problems in the literature to fully automate the logging process. Thus, in this research, we perform an experimental study on open-source Java projects and apply code-clone detection methods for log statements' prediction. Our work demonstrates the feasibility of logging automation by predicting the location of a log point in a code snippet based on the existence of a logging statement in its corresponding code clone pair. We propose a Log-Aware Code-Clone Detector (LACC) which achieves a higher accuracy of log prediction when compared to state-of-the-art general-purpose clone detectors. Our analysis shows that 98% of clone snippets match in their logging behavior, and LACC can predict the location of logging statements by the accuracy of 90+% for Apache Java projects.