dryft

Schema in. Migration out.

Go CLI tool for PostgreSQL schema management via Prisma schemas and Goose migrations.

Features

• PostgreSQL introspection → Generate Prisma schemas from existing databases
• Prisma schema parsing → Understand your data models
• Schema diff engine → Detect changes between schema versions
• Goose migration generation → Create SQL migrations automatically
• Destructive change detection → Warn about data loss risks

Installation

Build from source:

git clone https://github.com/phathdt/dryft.git
cd dryft
make build
# or: go build -o dryft ./cmd/dryft

Or install to $GOPATH/bin:

make install

Quick Start

Initialize a new dryft project:

dryft init

Pull schema from an existing PostgreSQL database:

dryft db pull

Create a migration from schema changes:

dryft migration create "add user table"

Commands

• dryft init - Initialize configuration and state directory
• dryft db pull - Introspect PostgreSQL and generate Prisma schema
• dryft migration create <name> - Generate Goose migration from schema diff
• dryft schema diff - Preview changes between current and previous schemas
• dryft validate - Validate Prisma schema syntax and semantics

Configuration

Create a .dryft.yaml file in your project root (or run dryft init):

database:
  provider: postgresql
  url: ${DATABASE_URL}

schema:
  file: prisma/schema.prisma
  naming:
    fields: camelCase
    tables: PascalCase

migration:
  directory: migrations
  format: goose
  naming: timestamp

goose:
  version_table: goose_db_version

See .dryft.yaml.example for full configuration options.

How It Works

dryft generates incremental migrations by comparing your Prisma schema against migration history (100% offline):

Offline Migration Generation

Migration generation is 100% offline and does NOT require database connection:

# First migration (empty migrations/ directory)
dryft migration create initial
# → Creates full schema (CREATE TABLE)

# Incremental migrations (existing migrations/ files)
dryft migration create add_user_bio
# → Creates incremental changes (ALTER TABLE)

Workflow:

1. Parse existing migrations in migrations/ directory
2. Build virtual schema from migration history
3. Compare with current schema.prisma
4. Generate incremental SQL (ALTER instead of CREATE)

When Database Connection is Needed

Database connection is ONLY needed for db pull:

# Introspect existing database → generate schema.prisma
dryft db pull

After db pull, all future migrations are generated offline from migration history.

Development

# Run tests (skip integration tests)
make test-short

# Run all tests including integration
make test

# Format code
make fmt

# Run static analysis
make vet

# Run linter
make lint

# Development workflow (fmt + vet + test + build)
make dev

Contributing

We welcome contributions! Before submitting a pull request, please ensure:

• All tests pass: make test
• Code is formatted: make fmt
• No lint errors: make lint
• Test coverage ≥85%

See CONTRIBUTING.md for detailed guidelines.

Pull Request Requirements

All PRs must pass these automated checks:

✅ Format - Code properly formatted with gofmt
✅ Lint - No linting errors (golangci-lint)
✅ Vet - Pass go vet analysis
✅ Build - Code compiles successfully
✅ Unit Tests - All unit tests pass
✅ Integration Tests - All integration tests pass
✅ Coverage - Maintain ≥85% test coverage

Development Status

✅ v0.1 MVP Complete - All 8 phases implemented and tested

• ✅ CLI framework and configuration
• ✅ Internal schema model (normalized IR)
• ✅ PostgreSQL introspection
• ✅ Prisma schema writer with relation fields
• ✅ Prisma schema parser
• ✅ Schema diff engine with dependency ordering
• ✅ PostgreSQL SQL generator
• ✅ Goose migration formatter
• ✅ 57 integration tests passing (testcontainers)

See MVP_COMPLETE.md for detailed summary.

Known Limitations

Current MVP has the following limitations:

High Priority Issues

• @@index not generated in migrations (#21) - Index definitions are parsed but not included in SQL output
• Prisma relation fields skipped (#10) - @relation syntax is parsed with warnings but not converted to FK constraints

Medium Priority

• No automatic column rename detection (#15) - Renames appear as DROP+ADD (potential data loss)
• Enum drops disabled (#12) - Safety feature pending dependency checks
• Single schema only (#14) - Only public schema introspected
• Views not supported (#13) - Database views are ignored

Future Features

• Advanced env var expansion (#19) - Only basic ${VAR} supported
• go-migrate format (#18) - Only Goose format currently
• Deferrable foreign keys (#16) - Advanced PostgreSQL feature
• MySQL support (#17) - PostgreSQL only for now

See GitHub Issues for complete list and progress.

Roadmap

v0.2 (In Progress)

• Fix @@index generation in migrations (#21)
• Support Prisma @relation fields and FK constraints (#10)
• Add end-to-end migration tests with PostgreSQL (#23)
• Implement automatic column rename detection (#15)

v0.3 (Planned)

• Enum drop operations with safety checks (#12)
• Multi-schema PostgreSQL support (#14)
• PostgreSQL views introspection and migration (#13)

Future

• MySQL database provider support (#17)
• go-migrate format output (#18)
• Advanced constraint features (deferrable FKs, etc.)

License

MIT