GUI Bug Localization Data

1. Download source code repositories (~4.29GB) using the following commands:

pip install gdown
gdown "https://drive.google.com/uc?export=download&id=1zluoAWnPLPx1nCK7WSb0EQ5ctCwwisU3"

2. Unzip the source code repositories file
3. Download augmented source code repositories (~1.45GB) using the following command:

gdown "https://drive.google.com/uc?export=download&id=1s9dpzEXdT_zEVPanhsa5rf-etEGm18xL"

4. Unzip the augmented source code repositories file

Directory Structure (Dataset)

├── UI Bug Localization Dataset
│   ├──  BugLocalizationGroundTruth --> contains the ground truth for buggy commit ids and relevant filenames in JSON format for all 80 bug reports
│   ├──  BugReports --> contains bug reports in text format concating titles and contents
│   ├──  BugReportsTitles --> contains bug report titles in text format
│   ├──  BugReportsContents --> contains bug report contents excluding titles in text format
│   ├──  TraceReplayer-Data --> GUI metadata from bug reproduction scenarios having app screenshots, relevant XMLs, and app execution information in JSON format
│   ├──  Augmentation-Info --> necessary GUI information and potential filenames based on the number of screens, re-ranking criteria used in the configurations
│   ├──  Preprocessed-Data --> preprocessed queries with original content, query replacement, query expansion, and preprocessed source code 
|   |   ├── PreprocessedTitles -> preprocessed queries containing only bug report titles 
|   |   ├── PreprocessedContents -> preprocessed queries containing bug reports excluding titles 
|   |   ├── PreprocessedBugReports -> preprocessed queries containing bug reports including titles 
|   |   ├── BugLocatorQueries -> preprocessed BugLocator queries 
|   |   ├── PreprocessedCode -> preprocessed source code

All the bug reports that we used in our experiments are taken from the AndroR2 [1][2] dataset. The bug ids used in the filenames are the same bug ids from the AndroR2 dataset. We provide the detailed contents for each directory below.

BugLocalizationGroundTruth

This directory contains 80 json files with <bug-id>.json file format. Each file contains the following information:

- id: Bug Id
- bug_location: 
    - commit_ids: the commit id where a bug occurs
    - file_name: the file where a bug exists

BugReports

This directory contains 80 txt files containing bug reports containing titles and contents in bug_report_<bug-id>.txt file format.

BugReportsTitles

This directory contains 80 txt files containing only bug report titles in bug_title_<bug-id>.txt file format.

BugReportsContents

This directory contains 80 txt files containing only bug report contents, excluding titles in bug_report_<bug-id>.txt file format.

TraceReplayer-Data

This directory contains 80 folders with TR<bug-id> file format each having the following information:

- Execution-1.json: execution information having activities, actions, GUI component information.
- .xml files: GUI hierarchy information for the screens that were traversed in the app
- screenshots: app screenshots, screenshots with marked exercised GUI components and GUI component screenshots.

Augmentation-Info

Here, the subdirectory is designed as follows:

├── Screen (2/3/4 screens)
│   ├──  Corpus (all files or files if filtered based on GUI types): Files_In_Corpus.csv file contains bug report id and necessary file paths and Queries.csv file contains bug report ids, activities, fragments, exercised GUI components and all GUI components on the selected screens
|   |   ├── Boosting (files if boosted based on GUI types): This directory contains files that are boosted in Match_Query_File_List.csv and the remaining files are in Not_Match_Query_File_List.csv

GUI Types acronyms used in the paper and the same names in this repository:

• GS = GUI_States
• EGC = Interacted_GUI_Component_IDs
• GS+EGC = GUI_State_and_Interacted_GUI_Component_IDs
• SC = All_GUI_Component_IDs
• GS+SC = GUI_State_and_All_GUI_Component_IDs

References

1. Tyler Wendland, Jingyang Sun, Junayed Mahmud, SMH Mansur, Steven Huang, Kevin Moran, Julia Rubin, and Mattia Fazzini. 2021. Andror2: A Dataset of Manually-Reproduced Bug Reports for Android apps. In MSR’21. 600–604.

2. Jack Johnson, Junayed Mahmud, Tyler Wendland, Kevin Moran, Julia Rubin, and Mattia Fazzini. 2022. An Empirical Investigation into the Reproduction of Bug Reports for Android Apps. SANER’22 (2022).