What Makes a Good Commit Message?

<strong>What-Makes-a-Good-Commit-Message?</strong> This repository contains the main data and scripts used in "What Makes a Good Commit Message" <strong>Notes:</strong> <em>The code will be maintained on GitHub. If you have any questions, please feel free to mention issue.</em> <strong>Dataset:</strong> The folder dataset contains the following files. literature survery.xlsx It contains the data of 46 relevant literature reviewed in this study (Section 3.2). Questionnaire.pdf It is the questionnaire sent to experienced contributors. It contains three questions. It also contains an example of the initialized questionnaire. Frequency.pdf It describes the number and proportion of occurrences of a category/subcategory. posts list.xlsx It contains all posts we studied in Sec. 3.2. sampled messages.csv It contains meta-information of 1649 labeled commit messages. label = 0 means a commit message contains "Why and What". label = 1 means a commit message contains "Neither Why nor What". label = 2 means a commit message contains "No What". label = 3 means a commit message contains "No Why". if_mulit_commit = 1 means a commit is non-atomic. new_message1 means a message after preprocessing. maintenance type and expression way.xlsx It contains the results of our RQ2: the expression ways of Why and What, as well as links to maintenance types. <strong>CommitMessage (Scripts):</strong> The folder contains the following scripts files. Preprocessor It contains the preprocessing of commit messages, including the replacement of token in the message, etc. ModelTraining It contains the code for our model training, that is, the implementation of different classification techniques.