Python Daily Group

Django ORM Lens — read your models, migrations and relations without booting Django I kept hitting the same small problem: I'd open an unfamiliar Django codebase and want a straight answer to "what does this schema actually look like, and what breaks if I touch this model" — without setting up a database, resolving the settings module, or getting the app to import at all. So django-orm-lens reads the source instead of the runtime. It parses models.py and the migration files directly, so it works on a checkout you cannot run: no DJANGO_SETTINGS_MODULE, no database, no credentials, no django.setup(). That constraint is the point rather than a limitation I'm apologising for — it means it also works in CI on a repo with no services, on a colleague's branch, or on a project whose dependencies you have not installed. What it does, concretely: Schema drift — replays the migration graph and compares it against what the models declare. Fields declared but never migrated, and columns migrated but no longer declared. Missing indexes — flags lookups that would table-scan, and it knows which indexes Django already gives you: primary keys, db_index, unique, foreign keys, unique_together, UniqueConstraint. Blast radius — what a change to one model reaches through FK / M2M / O2O, `on_delete` behaviour included, so "can I drop this field" has an answer before you try it. ER diagrams — Mermaid, DBML, D2, PlantUML, Graphviz. N+1 heuristics and a signal graph, because signals are where the surprises live. There is a CLI, a VS Code extension, and an MCP server for anyone who points an AI agent at a codebase and would rather it read the schema than guess at it. On correctness, which is the part I actually care about: it is checked against golden snapshots of six real projects — Zulip, Saleor, Wagtail, django-CMS, Mezzanine and Read the Docs — currently 75 models and 538 fields of other people's Django. That suite is also how I keep the Python and TypeScript parsers answering identically. It earns its keep the same way. Running it over real checkouts of django-oscar, django-guardian, django-allauth and django-cms turned up three genuine bugs my green test suite had not: model classes defined inside an `if` block were skipped entirely, `abstract_models.py` was never walked, and `drift` reported a false failure on a project with two apps sharing a directory name — which is the worst thing a CI gate can do. What it is not: it does not execute your code, so anything decided at runtime is invisible to it — dynamically constructed models, fields assigned in `__init__`, anything behind a factory. It reads Django's idioms, not Python's full semantics. If your models are unusual it will tell you less than you want, and I would rather say so up front than have you find out. MIT, free, and it stays that way — there is no paid tier planned and there never was. Repo: https://github.com/FROWNINGdev/django-orm-lens Docs: [https://frowningdev.github.io/django-orm-lens/](https://frowningdev.github.io/django-orm-lens/) pip install django-orm-lens What I would genuinely like from this forum: the drift and index checks are the parts most likely to be wrong in ways I cannot see from my own projects. If you run it on something real and it tells you something false, that is the most useful thing you could send me — open an issue with the model that broke it. Several of the fixes above arrived exactly that way, and two of the people who did it ended up sending patches. https://preview.redd.it/dwnu68d850hh1.png?width=1280&format=png&auto=webp&s=ab9cb354b3362618fcf2f87549b775ae929d85cc https://redd.it/1vdozfg @pythondaily