Untangling Composite Commits Using Program Slicing

Ward Muylaert; Coen De Roover

doi:10.1109/SCAM.2018.00030

Untangling Composite Commits Using Program Slicing

Ward Muylaert, Coen De Roover

Research output: Chapter in Book/Report/Conference proceeding › Conference paper

3 Citations (Scopus)

92 Downloads (Pure)

Abstract

Composite commits are a common mistake in the use of version control software.
A composite commit groups many unrelated tasks,
rendering the commit difficult for developers to understand, revert, or integrate
and for empirical researchers to analyse.
We propose an algorithmic foundation for tool support to identify such composite commits.
Our algorithm computes both a program dependence graph and
the changes to the abstract syntax tree for the files that have been changed in a commit.
Our algorithm then groups these fine-grained changes according to the slices through the dependence graph they belong to.
To evaluate our technique, we analyse and refine an established dataset of Java commits, the results of which we also make available.
We find that our algorithm can determine whether or not a commit is composite.
For the majority of commits, this analysis takes but a few seconds.
The parts of a commit that our algorithm identifies do not map directly to the commit's tasks.
The parts tend to be smaller, but stay within their respective tasks.

Original language	English
Title of host publication	Proceedings - 18th IEEE International Working Conference on Source Code Analysis and Manipulation, SCAM 2018
Publisher	IEEE
Pages	193-202
Number of pages	10
ISBN (Electronic)	978-1-5386-8290-6
ISBN (Print)	978-1-5386-8291-3
DOIs	https://doi.org/10.1109/SCAM.2018.00030
Publication status	Published - 9 Nov 2018
Event	18th IEEE International Working Conference on Source Code Analysis and Manipulation - Madrid, Spain, Madrid, Spain Duration: 23 Sep 2018 → 24 Sep 2018 Conference number: 18 http://www.ieee-scam.org/2018/

Conference

Conference	18th IEEE International Working Conference on Source Code Analysis and Manipulation
Abbreviated title	SCAM
Country	Spain
City	Madrid
Period	23/09/18 → 24/09/18
Internet address	http://www.ieee-scam.org/2018/

Keywords

Changedistiller
Changenodes
Class dependence graph
Commit
Inter procedural
Intra procedural
Program dependence graph
Program slicing
Slicing
Static analysis
System dependence graph
Tangled
Tinypdg
Untangling

Access to Document

10.1109/SCAM.2018.00030

camera readyAccepted author manuscript, 353 KB

Link to publication in Scopus

Fingerprint Dive into the research topics of 'Untangling Composite Commits Using Program Slicing'. Together they form a unique fingerprint.

1 Talk or presentation at a conference

Untangling Composite Commits Using Program Slicing
Ward Muylaert (Speaker)
23 Sep 2018 → 24 Sep 2018
Activity: Talk or presentation › Talk or presentation at a conference

Cite this

@inproceedings{7170f6293b314982b7735722964d196e,

title = "Untangling Composite Commits Using Program Slicing",

abstract = "Composite commits are a common mistake in the use of version control software. A composite commit groups many unrelated tasks, rendering the commit difficult for developers to understand, revert, or integrate and for empirical researchers to analyse. We propose an algorithmic foundation for tool support to identify such composite commits. Our algorithm computes both a program dependence graph and the changes to the abstract syntax tree for the files that have been changed in a commit. Our algorithm then groups these fine-grained changes according to the slices through the dependence graph they belong to. To evaluate our technique, we analyse and refine an established dataset of Java commits, the results of which we also make available. We find that our algorithm can determine whether or not a commit is composite. For the majority of commits, this analysis takes but a few seconds. The parts of a commit that our algorithm identifies do not map directly to the commit's tasks. The parts tend to be smaller, but stay within their respective tasks.",

keywords = "Changedistiller, Changenodes, Class dependence graph, Commit, Inter procedural, Intra procedural, Program dependence graph, Program slicing, Slicing, Static analysis, System dependence graph, Tangled, Tinypdg, Untangling",

author = "Ward Muylaert and {De Roover}, Coen",

year = "2018",

month = "11",

day = "9",

doi = "10.1109/SCAM.2018.00030",

language = "English",

isbn = "978-1-5386-8291-3",

pages = "193--202",

booktitle = "Proceedings - 18th IEEE International Working Conference on Source Code Analysis and Manipulation, SCAM 2018",

publisher = "IEEE",

}

Muylaert, W & De Roover, C 2018, Untangling Composite Commits Using Program Slicing. in Proceedings - 18th IEEE International Working Conference on Source Code Analysis and Manipulation, SCAM 2018., 8530734, IEEE, pp. 193-202, 18th IEEE International Working Conference on Source Code Analysis and Manipulation, Madrid, Spain, 23/09/18. https://doi.org/10.1109/SCAM.2018.00030

TY - GEN

T1 - Untangling Composite Commits Using Program Slicing

AU - Muylaert, Ward

AU - De Roover, Coen

PY - 2018/11/9

Y1 - 2018/11/9

N2 - Composite commits are a common mistake in the use of version control software. A composite commit groups many unrelated tasks, rendering the commit difficult for developers to understand, revert, or integrate and for empirical researchers to analyse. We propose an algorithmic foundation for tool support to identify such composite commits. Our algorithm computes both a program dependence graph and the changes to the abstract syntax tree for the files that have been changed in a commit. Our algorithm then groups these fine-grained changes according to the slices through the dependence graph they belong to. To evaluate our technique, we analyse and refine an established dataset of Java commits, the results of which we also make available. We find that our algorithm can determine whether or not a commit is composite. For the majority of commits, this analysis takes but a few seconds. The parts of a commit that our algorithm identifies do not map directly to the commit's tasks. The parts tend to be smaller, but stay within their respective tasks.

AB - Composite commits are a common mistake in the use of version control software. A composite commit groups many unrelated tasks, rendering the commit difficult for developers to understand, revert, or integrate and for empirical researchers to analyse. We propose an algorithmic foundation for tool support to identify such composite commits. Our algorithm computes both a program dependence graph and the changes to the abstract syntax tree for the files that have been changed in a commit. Our algorithm then groups these fine-grained changes according to the slices through the dependence graph they belong to. To evaluate our technique, we analyse and refine an established dataset of Java commits, the results of which we also make available. We find that our algorithm can determine whether or not a commit is composite. For the majority of commits, this analysis takes but a few seconds. The parts of a commit that our algorithm identifies do not map directly to the commit's tasks. The parts tend to be smaller, but stay within their respective tasks.

KW - Changedistiller

KW - Changenodes

KW - Class dependence graph

KW - Commit

KW - Inter procedural

KW - Intra procedural

KW - Program dependence graph

KW - Program slicing

KW - Slicing

KW - Static analysis

KW - System dependence graph

KW - Tangled

KW - Tinypdg

KW - Untangling

UR - http://www.scopus.com/inward/record.url?scp=85058303499&partnerID=8YFLogxK

U2 - 10.1109/SCAM.2018.00030

DO - 10.1109/SCAM.2018.00030

M3 - Conference paper

SN - 978-1-5386-8291-3

SP - 193

EP - 202

BT - Proceedings - 18th IEEE International Working Conference on Source Code Analysis and Manipulation, SCAM 2018

PB - IEEE

ER -