diedeadcode

Conservative dead code detection for TypeScript and JavaScript.

ddd finds unused exports, unreachable functions, and dead code in your codebase with confidence scoring to minimize false positives.

Features

• Fast - Built on oxc for blazing-fast parsing
• Conservative - Confidence scoring reduces false positives
• Transitive analysis - Detects code that's only called by other dead code
• Framework-aware - Understands Next.js, Express, Jest, Vitest patterns
• Configurable - Ignore patterns, entry points, and more

Installation

npm (recommended)

npm install -D @dean0x/diedeadcode

Or run directly with npx:

npx @dean0x/diedeadcode .

Cargo

cargo install diedeadcode

Homebrew (macOS)

brew install dean0x/tap/diedeadcode

Build from source

git clone https://github.com/dean0x/diedeadcode.git
cd diedeadcode
cargo build --release

Quick Start

# Analyze current directory
ddd .

# Analyze with verbose output
ddd . --verbose

# Initialize config file
ddd init

# Output as JSON
ddd analyze . --format json

# Check mode (exit code 1 if dead code found, useful for CI)
ddd analyze . --check

Configuration

Run ddd init to generate a ddd.toml in your project root, or create one manually:

# Files to analyze
include = ["**/*.ts", "**/*.tsx", "**/*.js", "**/*.jsx"]
exclude = ["**/node_modules/**", "**/dist/**", "**/*.test.*"]

# Entry points
[entry]
files = ["src/index.ts", "src/main.ts"]
patterns = ["**/pages/**/*.tsx"]  # e.g., Next.js pages
autoDetect = true  # Detect from package.json

# Output settings
[output]
format = "table"  # table, json, or compact
minConfidence = "high"  # high, medium, or low
showChains = true

# Analysis settings
[analysis]
ignoreSymbols = ["logger", "debug"]
ignorePatterns = ["^_"]  # Ignore symbols starting with _

How It Works

1. Parse - Uses oxc to build an AST for each file
2. Extract - Identifies all exports, functions, classes, and variables
3. Graph - Builds a call graph tracking all references
4. Propagate - BFS from entry points to find unreachable code
5. Score - Assigns confidence based on patterns and heuristics

Confidence Scoring

Not all dead code warnings are equal. ddd scores each finding:

Factors that reduce confidence:

• Dynamic imports (import())
• Reflection (eval, Reflect.*)
• Decorated symbols (decorators often indicate framework usage)
• String property access patterns
• Exported symbols (may be used externally)

Output Formats

# Table (default)
ddd .

# JSON for tooling integration
ddd analyze . --format json

# Compact single-line per issue
ddd analyze . --format compact

Exit Codes

Performance

ddd is designed for large codebases:

• Parallel file parsing with rayon
• Memory-efficient AST traversal

Contributing

See CONTRIBUTING.md for guidelines.

License

MIT - see LICENSE