jetbrains-research/kotlin-code-anomaly

Code anomaly detection in Kotlin compiler

kotlin-code-anomaly

This is a tool for detecting anomalies in kotlin code. Anomalies are found in two ways: by classical machine learning algorithms on code metrics and by variational autoencoder on n-grams.

Requirements:

• java 8
  • java fx for visualizer.
• JDK 1.6, 1.7 for PSI extraction
• python 3 with packages listed in requiremets.txt

Usage:

• ./build.sh - to build the feture extraction and visualization kotlin modules
  • --no-visualizer to build only feature extractor
• ./run-all.sh <path to the data set> - run the pipline. The extracted features are placed in the data/ directory and the resulting anomalies data is put in the out-data/ directory.
• java -jar visualizer-cli.jar helps analyze the anomalies found by classical algorithms with metrics
  • --seek to extract code with anomalies
  • --binary-mark to manualy mark anomalies and useful or not
  • --copy to copy kotlin files
  • --categ-copy to sort anomalies by category (useful or useless)

The following parts contain external code:

• detection/ngram/anomaly_detection
  • License: Apache 2
  • origin: https://github.com/PetukhovVictor/code-anomaly-detection
• feature_extraction/ngram/ast_set2matrix
  • License: Apache 2
  • origin: https://github.com/PetukhovVictor/tree-set2matrix/
• feature_extraction/ngram/bytecode-parser
  • License: Apache 2
  • origin: https://github.com/PetukhovVictor/bytecode-parser/
• feature_extraction/ngram/ngram-generator
  • License: Apache 2
  • origin: https://github.com/PetukhovVictor/ngram-generator/
• code_mapping/bytecode-to-source-mapper
  • License: Apache 2
  • origin: https://github.com/PetukhovVictor/bytecode-to-source-mapper
• code_mapping/bytecode-anomalies-source-finder
  • License: Apache 2
  • origin: https://github.com/PetukhovVictor/bytecode-anomalies-source-finder