djm 0.0.1a0

Generate goose-style SQL migrations from Django migrations, plus base model mixins

DJM

CLI and helpers to go from Django models → goose-style SQL migrations (up/down). One app (djm): write your models there, then generate SQL for goose or golang-migrate.

Quick start (CLI)

1. Install and create a project:

   pip install djm
   djm init myapp
   cd myapp
   

2. Write your models in the djm app (edit djm/models/), then run migrations:

   uv sync   # or: pip install -e .
   uv run python manage.py makemigrations djm
   uv run python manage.py migrate
   

3. Generate goose SQL (up/down) into goose_migrations/:

   uv run python manage.py generate_goose
   

   From a project that has the djm CLI installed you can instead run:

   djm goose
   

4. Use the generated .sql files with goose or golang-migrate.

Requirements

• Python ≥3.13 (or adjust in pyproject.toml)
• Django ≥6.0
• uv (recommended) or pip

CLI

Command	Description
djm init [path]	Create a new DJM project (default: current directory). Preconfigured with djm app and generate_goose command.
djm goose [-o DIR]	Generate goose SQL migrations (run from project root; uses manage.py generate_goose).

Project layout (after djm init)

Path	Purpose
djm/	The only app: write your models here. Includes the generate_goose management command.
base/	Reusable abstract model mixins: UUIDPrimaryKeyMixin, TimestampsMixin, SoftDeleteMixin, SortNumberMixin.
core/	Main Django project (settings, urls).

Commands (inside a project)

• Generate goose migrations:

  uv run python manage.py generate_goose [--output-dir DIR] [--no-sql-extension]
  

  Uses the djm app. Default output dir is goose_migrations. Files are named {migration_name}.sql.

Base mixins

Use these in your models (see djm.models and base.models):

• UUIDPrimaryKeyMixin – UUID primary key (via uuid6).
• TimestampsMixin – created, updated (auto_now_add / auto_now).
• SoftDeleteMixin – deleted nullable datetime for soft deletes.
• SortNumberMixin – order_no and default ordering.

Import from base.models; the mixins are abstract and do not require base in INSTALLED_APPS unless you add migrations in base.

Shipping as a package

The package ships djm (models + generate_goose command) and base (mixins). One app only; users write models in djm.

• Install: pip install -e /path/to/djm or pip install djm
• In the consuming project: Add djm to INSTALLED_APPS, write models in djm, run python manage.py generate_goose -o your_goose_dir.
• Build: make build or uv build

Publishing to PyPI (GitHub Actions)

The repo includes a workflow that builds and publishes to PyPI when you create a GitHub Release (or run it manually from the Actions tab).

1. One-time: Add a trusted publisher on PyPI:

   • PyPI project name: djm
   • Repository: OWNER/djm
   • Workflow name: publish.yml

2. To release: Create a new release on GitHub (tag, e.g. v0.1.0). The workflow will build and publish the package to PyPI. No API token is needed (OIDC).

• .github/workflows/publish.yml – build + publish to PyPI on release
• .github/workflows/ci.yml – build only on push/PR to main

Makefile

• make help – show targets
• make migrate – run Django migrations
• make goose – generate goose SQL into goose_migrations
• make goose OUT=./sql – generate into ./sql
• make clean – remove generated goose files and __pycache__
• make install – uv sync
• make build – build distribution artifacts

Notes

• Django’s sqlmigrate supports --backwards (with an “s”) to print SQL that unapplies a migration; the generator uses this for the “Down” section.
• Generated SQL is for the database configured in settings.DATABASES["default"] (e.g. SQLite). For PostgreSQL/MySQL, run the generator with that database configured.