A Log Parsing Algorithm with Source Code

Most existing log parsing algorithms are unable to fully accurately parse logs. This is because these algorithms typically assume that the original log text satisfies features and find the log templates based on these features. However, these characteristics do not always hold. Once certain parts of the log violate these characteristics, it results in the failure of mining the log templates. This paper presents the novel method to obtain the log templates by analyzing the source code which generates the logs. The experiment shows that the log template obtained through our method is correct. However, due to the difficulty in searching for all the source code that generates the logs, this method also faces the problem of incomplete coverage.