Skip to content

Latest commit

 

History

History
106 lines (69 loc) · 5 KB

README.md

File metadata and controls

106 lines (69 loc) · 5 KB
Raw

Features

Coming soon.

See our paper(accepted by ICPC2022) in https://arxiv.org/pdf/2204.01028.pdf

Docker Image

You can use the provideded docker image to avoiding environment dependence setting

Link: https://drive.google.com/file/d/17zsCf-5FnKbE1iPw6Ca4onW5ckQX69eQ/view?usp=sharing

MSCCD is in /root/MSCCD

Remember to update to the newest MSCCD by git pull

Environment dependence

We have tested MSCCD on Ubuntu 18.04LTS / MacOS Monterey.

MSCCD mainly depends on these environments:

  • Python v3.6.9
  • Java 11 (Newer than Java9) (Remember to set version by editting modules/msccd_tokenizers/pom.xml when using a different version)
  • Maven v3.8.5
  • jinja2 (pip3)
  • ujson (pip3)

We added some interfaces and methods to ANTLR4.8 and packaged a .jar file for MSCCD. Please install the provided antlr-4.8-modified.jar to your local maven repository.

mvn install:install-file -Dfile=./lib/antlr-4.8-modified.jar -DgroupId=org.nagoya_u.ertl.sa -DartifactId=antlr-v4.8-modified -Dversion=4.8 -Dpackaging=jar

Generate a tokenizer for the target language

First, edit ./parserConfig.json :

  • parser: The path of the grammar folder, including g4 files and sometimes java programs.
  • grammarName: The grammar name defined in the g4 file. It can also be checked in pom.xml (for grammars from grammarsv4)
  • startSymbol: Can be easily checked in pom.xml or the g4 file.

Then, generate the tokenizer by:

python3 tokenizerGeneration.py

Configure the tool

We can configure the tool by config.json. Here are the items:

  • inputProject: A list of paths. Each path presents a project you want to detect.
  • keywordsList: The path of the keywordslist.
  • languageExtensionName: A list of the extension names of the target language.
  • minTokens: The minimum size of the token bag in clone detection.
  • minTokensForBagGeneration: The minimum size of the token bag in tokenization. A smaller value will provide a larger range of token bag sizes in clone detection; a bigger one will make the tokenizer faster when you don’t want small bags.
  • detectionThreshold: The similarity threshold with a number in the range(0,1). If the overlapping similarity of a code pair is higher than the threshold, they will be seen as clones. A higher threshold will increase accuracy and reduce recall, and vice versa.
  • maxRound: The max granularity value to detect.
  • tokenizer: The name of generated tokenizer. It is the same as “grammarName” in parserConfig.json
  • threadNum_tokenizer
  • threadNum_detection

Execute MSCCD

Users may always need to do several detections for the same project. So we can save the necessary data in a task object to save time for the execution next time.

Execute for the first time

By this part, we will execute the tool by generating a new task from the configuration file.

1 Edit the config.json file, and check the grammar file, keyword list file, and your input file.

2 Run by python3 controller.py, and just wait for the result.

3 Check the information in tasks/task[taskId]/, for each execution, there will be a folder named detection* to save the result files

Execute from a generated task

By this part, we will execute the tool from a generated task. We can easily change the detection granularity(required) and threshold(optional) by command.

Just run it by python3 controller.py [taskId] ([statementThreshold]).

For example, python3 controller.py 1 means excute from tasks/task1. python3 controller.py 2 0.9 means excute from tasks/task2, and set the detectionThreshold to 0.9

Check the detection results.

For each task, all the data is saved in the tasks/task* folder, including configurations, file list, token bags. Here is the description:

file description
fileList.txt Each line represents a source file, formatting with (projectId, file Path). The index of each file in each project is defined as fileId.
tokenBags Each line represents a token bag and uses '@ @' to separate each data field: projectId @ @ fileId @ @ bagId @ @ granularity value @ @ number of keywords @ @ symbol number @@ token number @@ start line in original file -- end line in original file@@ tokens(token text :: frequency)
taskData.obj Configurations

Results of each detection is saved in tasks/task*/detection* folder.

file description
pairs.file Reported clones in [[projectId,fileId,bagId],[projectId,fileId,bagId]]
info.obj Exection times...

Scripts:

  • scripts/blockPairOutput.py : generate a output file in csv format: [file1Path,startLine,endLine,file2Path,startLine,endline]
    • python3 scripts/blockPairOutput.py taskId detectionId outputFile
  • scripts/filePairOutput.py : generate a output file in csv format: [file1Path,file2Path]
    • It's useful when MSCCD is executed as a file-level clone detector. (When setting maxRound in config.json as 1 or 0)
    • python3 scripts/filePairOutput.py taskId detectionId outputFile

Comming soon

  • Speed up
  • Analysis scripts to make the detection results easier to read and use