Laravel Factory Generator Laravel Package

Getting Started

composer require thedoctor0/laravel-factory-generator --dev
php artisan vendor:publish --provider="TheDoctor0\LaravelFactoryGenerator\FactoryGeneratorServiceProvider"

Then run the generator against a live database connection:

php artisan generate:factory

Where to look first: the README now ships a version table and modern (Laravel 8+) factory syntax examples — use these as the reference output. Confirm the generated database/factories/ files match the class-based public function definition(): array style, not the legacy $factory->define() closure style.

First use case: point it at an existing or legacy schema (or a freshly-migrated local DB) and generate factories in bulk so you can immediately write feature tests and seeders without hand-writing every definition(). Verify your Laravel/PHP version against the new README table before assuming generated code will match your project's factory conventions.

Implementation Patterns

• Generate → curate → commit. Treat output as a scaffold. Run the generator, then hand-edit the result: tighten enums to realistic values, add state() methods, and wire up relationships the generator can't infer. Commit the curated factories, not the raw output.
• Preview with --dry-run (new in 1.5.0). Use it in CI or on unfamiliar schemas to see what would be produced/overwritten before touching database/factories/. This is the safest way to run the tool in an existing codebase.
• Regenerate per-table after schema changes. Rather than regenerating everything, run against the connection/table you changed and diff the result against your curated version.
• Enum columns as states. The hardened enum driver (1.5.0) produces better candidate values. Follow up by converting meaningful enums into explicit state() methods (->state(UserState::Suspended)) instead of leaving raw enum samples in definition().
• Modern Faker output. Generated factories now use current Faker idioms (fake()->... / fake()->unique() style). If your project still pins older Faker conventions, normalize the output or align your Faker version.
• Test-suite-first workflow. Run the generator once, then use Model::factory()->create() in Pest/PHPUnit tests immediately. Curate incrementally as tests surface missing columns or bad sample data.
• Connection-scoped generation. Point it at a specific connection (e.g. a migrated test DB) rather than your production-linked local DB so generated samples reflect the schema you actually test against.

Gotchas and Tips

• Missing database/factories/ directory. Previously crashed if the directory didn't exist; 1.5.0 fixes this. If you're on <1.5.0, mkdir -p database/factories first, or upgrade.
• It cannot infer relationships. Foreign keys become plain integer/ID faker values, not Model::factory() references. Always replace *_id columns with real factory relationships in definition(), and add configure()/afterCreating() where necessary.
• Overwrites destroy curation. Anything that writes files (--force-style options) will clobber your hand-edited factories. Gate destructive regeneration behind --dry-run review and keep curated files in version control so diffs are visible.
• Schema must be live. The generator introspects the actual DB, so run migrations first. Enum handling is driver-sensitive — 1.5.0 hardened the enum driver, but verify output on MySQL/MariaDB/Postgres enums rather than assuming parity.
• Requires a working connection at generation time. Don't wire it into build steps that run before the DB is available; keep it as an on-demand dev command (--dev dependency).
• Version table matters for syntax. The new README maps package versions to generated factory syntax. If you're on an older Laravel, pin an older package version rather than post-processing legacy syntax by hand.
• Extension points: the config file controls output paths and naming; adjust there rather than moving generated files. Custom column-type mappings are the place to improve enum/JSON/set handling for non-standard schemas.
• First run on a large schema: generate into a temporary output path (or via --dry-run), cherry-pick the tables you need, and avoid dumping dozens of low-quality factories you'll never maintain.