Code for the TSE article: 'Change-Patterns Mapping: A Boosting Way for Change Impact Analysis'
Author: Yuan Huang, Jinyu Jiang, Xiapu Luo, Xiangping Chen, Member, IEEE, Zibin Zheng, Senior Member, IEEE, Nan Jia, and Gang Huang Senior Member, IEEE
If you try to reproduce the same experiment results in this paper, you should follow the following steps:
STEP 1: Prepare data (dataset and testdata)
STEP 2: Generate coreclass of each commit
STEP 3: Generate coupling relationship of each commit
STEP 4: Generate corpus of each commit
STEP 5: On the testdata, use JRipples, Rose, Impactminer to generate the original result
STEP 6: Run Impactanalysis code to boost the result.
Sample dataset can be found at ./dataset
, these files are used to store change-patterns which may later be used to boost the results.
Sample testdata can be found at ./testdata
, these files are used to do experiments.
Notice: You'd better use the same file structure to store data if you want to do this experiment with ease. Otherwise, you should change the code in order to adapt to your file structure.
See ./coreclass_generation_project
for details
See ./CouplingRel_generation_project
for details
See ./Corpus_generation_project
for details
The output of these tools should organize as follow:
This file include orginal sorted(according to their probability of being impacted) predicted impacted classes.
See ./ImpactAnalysis_project
for details