art-dupl

art-dupl

art-dupl is a code duplication detection tool for Go projects. It analyzes source code at the AST level to find structural clones — ignoring literal values so it focuses on patterns, not string contents.

A fork of mibk/dupl (via golangci/dupl) with major enhancements: multi-method detection, semantic-aware matching, 7 output formats, templ support, incremental analysis, and a professional CLI.

Why art-dupl?

Installation

go install github.com/LarsArtmann/art-dupl/cmd/art-dupl@latest

Or build from source:

git clone https://github.com/LarsArtmann/art-dupl.git
cd art-dupl
just build    # Creates ./dist/art-dupl

Or with Nix:

nix build     # Builds with Go 1.26, static binary
nix run .#art-dupl -- ./src

Quick Start

# Scan current directory (default: threshold 15 tokens)
art-dupl

# Larger clones only (50+ tokens)
art-dupl -t 50

# HTML report with diff visualization
art-dupl --html --diff side-by-side > report.html

# JSON for CI/CD pipelines
art-dupl --json -t 20

# Structural-only matching (shows more results, including noise)
art-dupl --structural ./src

# Hash-based detection (faster)
art-dupl -m hash ./src

# Both detection methods combined
art-dupl -m "hash,art-dupl" ./src

# All output formats to a directory
art-dupl --all --output-dir ./reports ./src

# Project statistics
art-dupl stats
art-dupl stats --format json

Supported Languages

.templ source files are included by default. Templ-generated *_templ.go files are filtered by default (use --include-templ to include them). Use --only go or --only templ to restrict analysis to a single file type.

Detection Methods

Controlled with -m / --detection-methods:

Semantic vs. Structural Matching

• Default (semantic): Matches by structure AND identifier names. a.String() does NOT match b.Error(). Distinguishes methods by receiver type.
• --structural: Matches by AST shape only. a.String() matches b.Error(). Use for raw structural analysis.

Output Formats

Sort results with -s / --sort: size (default), occurrence, hash, total-tokens.

CLI Reference

art-dupl [flags] [paths...]
art-dupl stats [flags] [paths...]
art-dupl completion [bash|zsh|fish|powershell]

Core Flags

Output Flags

Filtering Flags

Performance Flags

Stats Subcommand

art-dupl stats [flags] [paths...]

Supports all shared flags plus:

Statistics include files scanned, clone groups, duplicate lines/tokens, complexity score, impact score, clone size distribution, health grade (A–F), and top files by duplication.

Configuration

Create dupl.json:

{
  "threshold": 30,
  "outputFormat": "json",
  "paths": ["./src", "./lib"]
}

art-dupl -c dupl.json

CLI flags override config file values. Config file overrides defaults.

Examples

# CI/CD: fail if too many clones
TOTAL=$(art-dupl --json . | jq '.summary.total_clones')
[ "$TOTAL" -gt 100 ] && echo "Too many clones: $TOTAL" && exit 1

# Find test file duplicates
find . -name '*_test.go' | art-dupl -files

# Progressive refactoring: baseline → refactor → measure
art-dupl --json -t 15 > baseline.json
# ... refactor ...
art-dupl --json -t 15 > after.json

# Most widespread clones first
art-dupl --plumbing --sort occurrence ./src

# Filter all generated code (default behavior)
art-dupl ./src

# Override: include sqlc-generated files
art-dupl --include-sqlc ./src

Architecture

Source Files (.go, .templ)
    ↓
Parsing (go/parser + templ parser)
    ↓
AST Serialization → unified syntax.Node[]
    ↓
Suffix Tree (Ukkonen's algorithm, O(1) transitions)
    ↓
Duplicate Search → syntax.Match
    ↓
Clone Classification → domain.ProcessedCloneGroup
    ↓
Output Formatting (text, HTML, JSON, SARIF, plumbing)

Key Packages

Development

just build          # Build binary
just test           # Run all tests
just check          # Run linting
just ci             # Format + lint + test
just coverage       # Generate coverage report
just bench          # Run benchmarks
just test-race      # Run with race detector
just check-coverage # Verify ≥80% coverage

Tests use standard testing + Ginkgo/Gomega BDD. See CONTRIBUTING.md for guidelines.

License

MIT — Based on mibk/dupl (via golangci/dupl).