Getting Started

PHPCPD is an archived, no-longer-maintained CLI tool for detecting copy-paste duplicates in PHP code. To begin:

Install via PHIVE: phive install phpcpd — never use Composer, per the author’s recommendation.
Run a quick scan on your Laravel app’s app/ directory: ./tools/phpcpd app/
Focus first on the summary metrics: % duplicated lines and Average size of duplication. Values >1–2% duplication or clones >50 lines are strong refactoring candidates.

Implementation Patterns

CI Gatekeeping: Integrate as a non-blocking pre-commit check or a soft CI step with --fail threshold only on new code (e.g., phpcpd --minimum-lines 15 app/ && php phpunit.xml --testdox), or use --dry-run to report without exit code impact.
Targeted scanning: Leverage --exclude to omit tests/, database/, and vendor/, and --suffixes .php to avoid non-PHP noise.
Team feedback loops: Generate XML via --log-pmd cpd.xml and feed into IDE plugins (e.g., PHPStorm’s built-in CPD viewer) or dashboards like SonarQube for recurring visibility.
Adoption ramp-up: Start by scanning new feature modules only; once patterns stabilize, expand gradually—avoid scanning legacy monoliths whole to prevent overwhelming teams.

Archived status = zero support: Since 2022, there are no updates for PHP 8.2/8.3 breaking changes. Always validate with php -v and consider maintaining a fork for critical fixes.
Memory hogs: Large apps (e.g., Laravel apps with 1000+ files) may hit limits; override with php -d memory_limit=2G ./tools/phpcpd app/.
Fuzzy ≠ smart: --fuzzy catches “almost identical” clones (e.g., changed variable names), but may flag legitimate factory patterns. Use --minimum-lines 20 to ignore small boilerplate snippets.
False confidence risk: A clean CPD scan ≠ no duplication; many patterns (e.g., Laravel macros, traits) intentionally duplicate behavior. Treat CPD as one signal among many—not a correctness oracle.
CLI friction: PHPCPD’s outdated sebastian/cli-parser can conflict with other CLI tools using symfony/console 6+. Run via isolated PHAR to avoid version skew in global toolchains.