Skip to content

Latest commit

 

History

History
64 lines (50 loc) · 3.19 KB

README.md

File metadata and controls

64 lines (50 loc) · 3.19 KB
Raw

Code Change Sniffer (CCS)

This is a study on predicting code changes with Markov Chains for Java projects based on their commits. To have a better understanding how Markov Chains is used in CCS, please check the Running Example Supplementary Material

How to Run CCS

CCS runs by the class MainCCS.java, and requires two arguments:

  • First argument: Directory to last commit
  • Second argument: Directory last version of the project

Example:

Assume you want to predict the code changes based on the last version of JJWT jjwt-0.6.0, with respect to the changes that made in the commit jjwt-e392524919ecd155c00240533f54e66b6b1eaa82. The arguments shall be given as below:

  • First argument: "../jjwt-e392524919ecd155c00240533f54e66b6b1eaa82"
  • Second argument: "../jjwt-0.6.0"

Setting Inputs for CCS

There are several inputs that needs to be setup for code change prediction:

  • Setting model
  • Setting commit number
  • Setting project commit directory for output

Setting model

CCS has two models implemented Call Graph (CG) and Effect Graph (EG) based models. To set the model to be used, in the MainCCS.java set the MarkovChains.model to:

SelectModel.EFFECT_GRAPH; or SelectModel.CALL_GRAPH;

Setting commit number

To set the commit number that will be analyzed, in the MainCCS.java class, initialize the MarkovChains.commitNum variable to the commit number that will be analyzed. For example, if the 16th commit is going to be analyzed, set the variable as below:

MarkovChains.commitNum=16;

Setting project commit directory for output

To generate the prediction output into the specified directory, in the MainCCS.java class, initialize the predictionOutputFile variable. The prediction output is normally implemtented to generate the output into CCS under the ../data folder. See dataset organization

Dataset

Dataset is organized as the following:

data
└── project
    ├── version + commit_number
    |   └── EIS_Markov_FS_CG.txt (Estimated Impact Set of Call Graph)
    |   └── EIS_Markov_FS_EG.txt (Estimated Impact Set of Effect Graph)
    ├── ...
    ├── README.md (table of precision, recall, f-measure results, and change types for each commit)
    ├── charts
    └── AIS.txt (Actual Impact Set)

How to Generate Ground Truth

To generate ground truth, you need to run the ccs.markov.change.TestNumberOfChanges.java file. The TestNumberOfChanges.java class requires two arguments. The first argument must be the last version of the project, and the second argument must be the previous version of the project. For example, for the versions 0.6.0 and 0.7.0 of project JJWT, the TestNumberOfChanges.java class should be run as below:

"../jjwt-0.7.0" "../jjwt-0.6.0"

Citing

If you have used our work please cite us

@inproceedings{ufuktepe2021ccs,
  title={Code Change Sniffer: Predicting Future Code Changes with Markov Chain},
  author={Ufuktepe, Ekincan and Tuglular, Tugkan},
  booktitle={IEEE Computer Software and Applications Conference},
  year={2021},
}