IFixR

IFixR a patch generation system for user-reported bugs, described in "iFixR: Bug Report driven Program Repair", In Foundations of Software Engineering, 2019.

@inproceedings{arXiv-1907.05620,
 title = {iFixR: Bug Report driven Program Repair},
 author = {Anil Koyuncu and Kui Liu and Tegawendé F. Bissyandé and Dongsun Kim and Martin Monperrus and Jacques Klein and Yves Le Traon},
 booktitle = {Proceedings of the 27th ACM Joint Meeting on European Software
  Engineering Conference and Symposium on the Foundations of Software
  Engineering},
 year = {2019},
 doi = {10.1145/3338906.3338935},
 url = {http://arxiv.org/pdf/1907.05620},
}

• I. Introduction of IFixR
• II. Environment setup
• III. Run Step-by-Step
• IV. Evaluation Result
• V. Generated Patches
• VI. Structure of the project

II. Environment

Manual installation

• OS: macOS Mojave (10.14.3)
• JDK7: Oracle jdk1.7 (important!)
• JDK8: Oracle jdk1.8 (important!)

• Defects4J: Clone and configure defects4j from its original repository

  git clone https://github.com/rjust/defects4j
  cd defects4j
  ./init.sh
  
  export PATH=$PATH:"path2defects4j"/framework/bin
  
  Please add defects4j also to path since compileProjects.sh,createProjects.sh,testProjects.sh uses defects4j

• Download and configure Anaconda
• Create an python environment using the environment file

  conda env create -f environment.yml

Before running

• Update config file with corresponding user paths.

  Defects4J home path requires '/' at the end
  Example: /Users/projects/defects4j/

Docker

• Clone the repository

• Pull image from docker hub

  docker pull anilkoyuncu/ifixr

• Run the docker image sharing the repository folder

  docker run -v <path_of_cloned_repository>/:/data -it anilkoyuncu/ifixr:latest /bin/bash

• Change folder to /data

• Follow Step III to run

III. How to run

• Launch startPy as follows:

  bash startPy.sh [OPTIONS]

Running Options

IFixR needs three input options for running.

  Usage: bash startPy.sh $1 $2 $3
  where $1 is root
        $2 is job
        $3 is subject

• root : The full path of directory where startPy.sh is located (corresponds to pwd in the examples below)

• subject : the project name of buggy program of benchmark. (MATH,LANG are expected values but can be specific for job)

• job : the name of the job in the pipeline , supports multiple steps which needs to executed in order:

  1.clone : Clone target project repository.

    Example: bash startPy.sh pwd clone MATH
    Example: bash startPy.sh pwd clone LANG

  2.collect : Collect all commit from repository.

    Example: bash startPy.sh pwd collect MATH
    Example: bash startPy.sh pwd collect LANG

  3.fix : Collect commits linked to a bug report.

    Example: bash startPy.sh pwd fix MATH
    Example: bash startPy.sh pwd fix LANG

  4.bugPoints : Identify the snapshot of the repository before the bug fixing commit introduced.

    Example: bash startPy.sh pwd bugPoints MATH
    Example: bash startPy.sh pwd bugPoints LANG

  5.brDownload : Download bug reports recovered from commit log

    Example: bash startPy.sh pwd brDownload MATH
    Example: bash startPy.sh pwd brDownload LANG

  6.brParser : Parse bug reports to select the bug report where type labeled as BUG and status as RESOLVED or CLOSED

    Example: bash startPy.sh pwd brParser MATH
    Example: bash startPy.sh pwd brParser LANG

  7.brFeatures : Extract bug report features

    Example: bash startPy.sh pwd brFeatures MATH
    Example: bash startPy.sh pwd brFeatures LANG

  8.verify : Extract source code features

    Example: bash startPy.sh pwd verify MATH
    Example: bash startPy.sh pwd verify LANG

  9.simi : Compute the similarity between bug reports and source code features

    Example: bash startPy.sh pwd simi MATH
    Example: bash startPy.sh pwd simi LANG

  10.features : Compute the features from similarity scores

    Example: bash startPy.sh pwd features MATH
    Example: bash startPy.sh pwd features LANG

  11.predict : Predict file level bug localization

    Example: bash startPy.sh pwd predict ALL

  12.eval : Retrieve predictions for all bug reports and suspiciousness scores of files.

    Example: bash startPy.sh pwd eval ALL

  13.stmt : Compute statement level bug localization

    Example: bash startPy.sh pwd stmt ALL (!!! Clones all defects4j bugs and compile, may take significant amount of time)

  14.gv : Execute generate-validation step to produce patch candidates.

    Execute generate-validation step on a single defects4j bug (eg Lang_15, Math_34 case sensitive)
  
      Example: bash startPy.sh pwd gv Math_34
  
    Execute generate-validation step on all dataset. (!!! Takes significant amount of time !!!)
  
      Example: bash startPy.sh pwd gv ALL

  Generated patches are placed in directory OUTPUT

Data Viewer

The data provided with replication package is listed in directory data The data is stored in different formats. (e.g. pickle, db, csv, etc..)

The see content of the .pickle file the following script could be used.

   import pickle as p
   import gzip
   def load_zipped_pickle(filename):
      with gzip.open(filename, 'rb') as f:
          loaded_object = p.load(f)
          return loaded_object

Usage

  result = load_zipped_pickle('code/LANGbugReportsComplete.pickle')
  # Result is pandas object which can be exported to several formats
  # Details on how to export is listed in offical library documentation
  # https://pandas.pydata.org/pandas-docs/stable/reference/api/pandas.DataFrame.html

IV. Evaluation Result

V. Generated Patches

MIMIC generates genuine/ plausible patches for 21/44 Defects4J bugs with its IR-based bug localizer.

VI. Structure of the project

  |--- README.md                    :  user guidance
  |--- code                         :  code
  |--- data                         :  replication data
  |--- doc                          :  document
  |--- OUTPUT/MIMIC/FixedBugs       :  generated patches