Reptory

This repository provides a dataset along with a framework to assist comparative experimental studies on learning-based automated program repair.

Dataset:

• Dataset is based upon Deepbugs. The following files in the create-dataset/ folder that are in charge of data extraction are borrowed from DeepBugs although we changed them accordingly:

  • extractFromJS.js
  • extractorOfBinOps.js
  • extractorOfCalls.js
  • fileIDs.json
  • jsExtractionUtil.js
  • Util.py

• Dataset comprises of named based bug patterns:

  • Swapped function arguments
  • Wrong binary operator
  • Wrong operand in binary operation

Framework: - A framework where researchers can incorporate additional context and use on the existing dataset.

Prerequisites

• Install conda

  wget https://repo.anaconda.com/miniconda/Miniconda3-latest-MacOSX-x86_64.sh -O ~/miniconda.sh
  bash ~/miniconda.sh -b -p $HOME/miniconda
  conda init
  source ~/miniconda/bin/activate
  source ~/.bash_profile
  

• Setup environment

  • Install conda:

    conda create -n context_ml python=3.6 
    conda activate context_ml
    conda install -y python=3.6    
    

  • Install tensorflow using pip: pip install tensorflow==1.5

  • Install tensorflow using conda: conda install -y -c conda-forge tensorflow=1.5.1

  • Required python packages for embedding generation:

       conda install -c anaconda nltk
       conda install -c anaconda gensim
    

    Detailed:

      conda install -c anaconda nltk
            
      import nltk
      nltk.download('punkt')
      [nltk_data] Downloading package punkt to /Users/UserName/nltk_data...
      [nltk_data]   Unzipping tokenizers/punkt.zip.
    

  • Install nvm and npm:

        curl -o- https://raw.githubusercontent.com/nvm-sh/nvm/v0.35.3/install.sh | bash
        source .bashrc
        nvm install --lts
        nvm use --lts
    

  • Install node dependencies:

      cd src
      npm install
    

Dataset

• Download dataSet

Dataset

Code Representations

ID	Buggy	Fixed	Dataset Generation Script
1a	Word tokenization	Word tokenization	data_word_level_esprima.sh
1b	Word tokenization Enhanced	Word tokenization Enhanced	data_word_level_word_tokenization_enhanced.sh
2	Deepbugs Representation	Deepbugs Representation	data_word_level_deepbugs.sh
3	Deepbugs Representation (with Types Incomplete with variable value)	Deepbugs Representation (with Types Incomplete with variable value)	data_word_level_deepbugs_with_type_and_variable.sh
4	Deepbugs Representation (with Types Incomplete without variable value)	Deepbugs Representation (with Types Incomplete without variable value)	data_word_level_deepbugs_with_type.sh
5	Code Simplification (Signatures)	Code Simplification (Signatures)	data_word_level_synthesized.sh
6	Code Simplification (Signatures with position anchor)	Code Simplification (Signatures with position anchor)	data_word_level_synthesized_with_anchor.sh
7	Code Simplification (Signatures with LIT/ID)	Code Simplification (Signatures with LIT/ID)	data_word_level_synthesized_with_ID_LIT.sh
8	Code Simplification (Signatures with position anchor and LIT/ID)	Code Simplification (Signatures with position anchor and LIT/ID)	data_word_level_synthesized_with_anchor_with_ID_LIT.sh
9	AST (of original code)	AST (of original code)	data_word_level_ast.sh
10	AST (of code simplification -> Type with variable value)	AST (of code simplification -> Type with variable value)	data_word_level_synthesized_with_variable_ast.sh
11	AST (of code simplification -> Types without variable value)	AST (of code simplification -> Types without variable value)	data_word_level_synthesized_without_variable_ast.sh
12	Preorder AST (of original code)	Preorder AST (of original code)	prepare_calls_ast_preorder.sh
13	Abstraction - Tufano	Abstraction - Tufano	prepare_calls_abstraction.sh

Mixed Representations

Buggy	Fixed	Dataset Generation Script
Code Simplification (function signatures with LIT/ID)	AST (of code simplification -> Types without variable value)	data_word_level_synthesized_with_ID_LIT_to_data_word_level_synthesized_without_variable_ast.sh
AST (of code simplification -> Types without variable value)	Code Simplification (function signatures with LIT/ID)	data_word_level_synthesized_without_variable_ast_to_data_word_level_synthesized_with_ID_LIT.sh
Word tokenization	AST	data_word_level_esprima_to_data_word_level_ast.sh
AST	Word tokenization	data_word_level_ast_to_data_word_level_esprima.sh

Buggy	Fixed	Dataset Generation Script
		prepare_calls_tufano_abstraction_to_code_simplification_signatures_with_position_anchor.sh
		prepare_code_simplification_signatures_with_position_anchor_to_calls_tufano_abstraction.sh

Generate Embedding

Embedding	Script
word2Vec-CBOW	getEmbeddings.sh
word2Vec-SkipGram	get-embedding-final-skipgram.sh
fastText	get-embedding-fasttext-final.sh
gloVe	cd glove && make && getEmbeddings_glove.sh

Experiments for Embedding

Embedding	Script
word2Vec-SkipGram	prepare_calls_abstraction_word2vec_skipgram.sh
fastText	prepare_calls_abstraction_fasttext.sh
gloVe	run prepare_calls_abstraction_glove.sh and then cd GloVe && make && getEmbeddings_glove.sh. Finally run ./train-final-save-log.sh

How to calculate Accuracy and Rank?

python calculate_accuracy_and_rank.py test.correct test.buggy model.output