Roadmap

[pmdevita]

Here's my current in progress plan for the roadmap:

Short Term

• Get builds up on PyPI
• Fix building docs from GitHub Actions
• Fix aliases for foreign keys in ModelSchema (Ninja 1111, 828, 469)
• Fix primary keys being unconditionally marked as nullable in ModelSchema (Ninja 1249, 1227)
• Fix blanks being unconditionally marked as nullable in ModelSchema (Ninja 1374, 882)
• Merge Ksauder's ModelSchema rewrite (Ninja 1281)
  • Might need some additional work, I'm less inclined to consider making an option for nullable primary keys without an example use case.
• Merge the django_getter function cache performance improvements (Ninja 1150) Superseded by Improve performance by removing DjangoGetter #28
• Fix uploading files with PUT/PATCH (Ninja 397) Solved in 1423!
• Fix mypy q problem for FilterSchema (Ninja 1037)
• Separate validation and serialization schemas (Ninja 1162, 1139, 1132)
• Implement Field .choices as a literal for Pydantic (Ninja 1078)
• Fix missing dev dependencies and related issues
• Add marking by_alias at Router and API level (may be partially fixed with Ninja 1268)
• Determine if/what to do about default factories making fields not required (Ninja 366)

Medium Term

• Proper hybrid sync/async support in Ninja auth
• Add Rapidocs as an OpenAPI docs option
• Cursor-based pagination (There are multiple community implementations now and I'd like to try and merge the best ideas from them)
• OpenAPI schema generation caching
• Solve @paginate with multiple responses (either error on start up or make it work)
• Make add_router idempotent
• Expand options for custom field typing

Long Term

• Decide what to do about Pydantic v1 hold outs like class Config and .dict(), and determine whether to follow Pydantic closer or keep the current strategy of adapting to Django patterns
  • Update, Ninja has decided to remove class Config in favor of the new config dict setting
• AsyncAPI support for Channels (Websockets)
• Investigate performance against FastAPI https://github.com/oscarychen/building-efficient-api/

Longer Term (Breaking changes, not sure if we'll do them)

• Full pass on the source and docs to improve grammar
• Fully type for mypy --strict
• Clean up and streamline Readme

Release Plan

Shinobi will follow Ninja's minor release versions, so 1.3.x releases will be based off of Ninja's 1.3.x releases, but we'll increment the patch version on our own.

[dpgraham4401]

Would the maintainers of shinobi consider adding Authorization/permission checks? Similar to what's proposed in ninja 789?

I know there's a couple packages that are trying to do something in this space like django-ninja-auth and Django ninja extras but there's not well-adopted option that makes use of django.contrib.auth like DRF does. I know this has been a roadblock to adopting ninja/shinobi for anything more than an internal API on teams I've worked on.

[pmdevita]

Saw this late, yes I think that would be a good idea to have. My team ended up making our own decorators that do this but it would be nice to see at least some simple ones in the library.