TreeNose - A Language-Independent Code Smell Detector

Table of Contents

• TreeNose
  • Table of Contents
  • Introduction
  • Associated Thesis
  • Run Code Smell Detector
    • Dependencies
    • Environment Setup
    • Main Feature: Code Smell Detection
    • Side Feature: Tree Structure Log
  • Technical Details
  • Related Work
  • Development Note

Introduction

If code smells are language-independent, code smell detectors could also be.

TreeNose provides language-independent code smell detections across multiple programming languages. TreeNose currently supports the following code smells:

• Complex Conditional
• Long Class
• Long Method
• Long Parameter List
• Message Chain

Associated Thesis

Here is a related senior thesis

Run Code Smell Detector

Dependencies

To run the example program, Node runtime environment and npm package manager are required. Here are some documentations on how to install them.

• https://www.npmjs.com/package/npm
• https://nodejs.org/en/download

Environment Setup

• Go to the directory src on your terminal
• Run npm install

Main Feature: Code Smell Detection

main.js provides CLI to run TreeNose. It requires three flags:

• input: Path to a directory or to a single file
• lang: The target programming language that the target file is written in
• out: Path to a directory the reports will reside
• smell (optional): The smell types to detect, you are able to pass several code smells separated by space

Note:

1. lang flag accepts: 1. java 2. python 3. javascript
2. smell flag accepts: 1. long-class 2. long-param 3. long-method 4. long-message-chain 5. complex-conditional
3. if a directory is provided to flag input, TreeNose will fetch all the target files under both the current directory and all of its subdirectories in any level.
4. By default TreeNose checks all kinds of code smells when no smell flag needs to be set up

Here is an example:

node main.js --input ../example_codes/python_codes  --lang python --out ../reports --smell long-param long-class

Side Feature: Tree Structure Log

Note: you don't need to understand anything in this section to do code smell detection. However, the tool introduced in section helps you to understand the [tree-sitter code structure](https://tree-sitter.github.io/tree-sitter/using-parsers#syntax-nodes), which is necessary to build a new language interface.

tree_printer.js provides the features to display the Abstract Syntax Tree of a file parsed from tree-sitter.

tree_printer.js expects two required arguments:

1. lang
2. file

Here is an example:

node tree_printer.js --lang javascript --input ../example_codes/js_codes/conditional.js

Technical Details

TreeNose currently supports the following languages:

• Java
• JavaScript
• Python 2 and Python 3

TreeNose supports the detection of 5 kinds of code smells with default thresholds

• Complex Conditional
• Long Class
• Long Method
• Long Message Chain
• Long Parameter List

Here is a table about the metrics and thresholds TreeNose by default used to detect those smells:

Code Smell	Thresholds
Long Class	LOC > 200, NOM > 20
Long Method	LOM > 100
Long Parameter List	PAR > 5
Complex Conditional	NOC > 3
Long Message Chain	LMC > 4

• LOC: lines of class
• NOM: number of sub methods in class
• LOM: length of method
• PAR: number of parameters
• NOC: number of conditional cases
• LMC: length of message chains

Note: To fit your need, you can change the thresholds in the detection configuration file

Related Work

• Pysmell
• Jscent
• DesigniteJava

Development Note

• Currently god class code smell has only method amounts as the metric
• Generator function in Python tree-sitter is treated as expression statement but not function definition