archibate/dotfiles-opencode/code-quality-metrics

Code Quality Metrics Guide

Overview

Code quality metrics provide quantitative measures of code characteristics. This skill covers complexity metrics, maintainability indices, and their practical thresholds across different languages.

Complexity Metrics

Cyclomatic Complexity (CC)

Measures the number of linearly independent paths through code.

Calculation: CC = E - N + 2P

• E = edges in control flow graph
• N = nodes in control flow graph
• P = connected components (usually 1)

Simplified: Count decision points + 1

• Each if, elif, else, for, while, case, catch, &&, ||, ?: adds 1

Thresholds:

Cognitive Complexity

Measures how difficult code is to understand (introduced by SonarQube).

Key differences from CC:

• Penalizes nested structures more heavily
• Doesn't count all decision points equally
• Focuses on human comprehension

Calculation rules:

1. +1 for each break in linear flow (if, for, while, switch, catch, ?:)
2. +1 for each nesting level inside these structures
3. +1 for binary logical operators (&&, ||)
4. Recursion adds +1

Thresholds:

Example:

def process(items):           # 0
    for item in items:        # +1 (loop)
        if item.active:       # +1 (if) +1 (nesting=1)
            if item.valid:    # +1 (if) +2 (nesting=2)
                handle(item)
    return items              # Total: 6

Halstead Metrics

Based on counting operators and operands:

Maintainability Metrics

Maintainability Index (MI)

Composite metric combining complexity, size, and comments.

Formula: MI = 171 - 5.2 × ln(V) - 0.23 × CC - 16.2 × ln(LOC)

Thresholds:

Lines of Code (LOC)

Coupling Metrics

Afferent Coupling (Ca)

Number of classes that depend on this class.

• High Ca = Many dependents, changes are risky

Efferent Coupling (Ce)

Number of classes this class depends on.

• High Ce = Many dependencies, class is fragile

Instability (I)

I = Ce / (Ca + Ce)

• 0 = Stable (many dependents, few dependencies)
• 1 = Unstable (few dependents, many dependencies)

Cohesion Metrics

Lack of Cohesion of Methods (LCOM)

Measures how related methods are within a class.

LCOM1: Number of method pairs without shared instance variables

• 0 = Perfect cohesion
• High = Low cohesion, consider splitting

LCOM4: Number of connected components in method-field graph

• 1 = Cohesive class
• >1 = Multiple responsibilities

Quick Reference Table

Measurement Commands

ESLint Complexity

npx eslint --rule 'complexity: ["error", 10]' src/

SonarQube Analysis

sonar-scanner -Dsonar.projectKey=myproject

Python Radon

radon cc src/ -a -s  # Cyclomatic complexity
radon mi src/ -s     # Maintainability index

Additional Resources

Reference Files

For detailed measurement techniques:

• `references/measurement-tools.md` - Tool configurations and usage
• `references/thresholds-by-language.md` - Language-specific thresholds

Integration with Other Skills

Combine with:

• solid-principles for structural analysis
• refactoring-patterns for improvement strategies