schemas not compatible with openapi in litestar

[YuriFontella]

I'm trying to define the correct response in the api documentation
the schema:

class Jobs(Table):
    id = Serial
    name = Varchar()
    lines = Integer(default=0)
    status = Varchar(default='pending')
    uuid = UUID(default=uuid4)
    progress = Integer(default=0)
    date = Timestamptz()

In the route function directly assigns the class

@get(path='/', summary='Retornar todos os trabalhos')
async def jobs(limit: int = 100, offset: int = 0) ->list[ Jobs]:
    ...

But it did not generate a response in the documentation
I tried using pydantic

JobsModel = create_pydantic_model(Jobs)

@get(path='/', summary='Retornar todos os trabalhos')
async def jobs(limit: int = 100, offset: int = 0) -> list[JobsModel]:
    ...

generated the data in the documentation, but it didn't come with the data type, I didn't understand where the {} comes from

I was waiting for this result:

[
  {
    "name": "string",
    "lines": "integer",
    "status": "boolean",
    "uuid": "uuid",
    "progress": "integer",
    "date": "string"
  }
]

Does anyone know the correct way to do this?

piccolo 1.5.0
litestar 2.7.1
python 3.12.2

[dantownsend]

I think you've got a typo in your table definition, it should be:

    id = Serial(primary_key=True)

You're right about needing to return list[JobsModel], instead of list[Jobs].

Which UI are you using to view the documentation? Is it the Swagger UI?

[YuriFontella]

@sinisaos I can open a discussion on some litestar channel about this, since they are inegrated with piccolo.

Another test I did was generating a dataclass, and it worked showing the types.

from dataclasses import dataclass

@dataclass
class JobsCustomResponse:
    name: str
    status: bool
    lines: int
    uuid: str
    progress: int
    date: str


@get(path='/', summary='Retornar todos os trabalhos')
async def jobs(limit: int = 100, offset: int = 0) -> list[JobsCustomResponse]:
    ...

[sinisaos]

@YuriFontella Litestar and FastAPI seem to treat optional fields differently. If you try to put optional fields in a dataclass or a Pydantic model like this

import typing as t
from dataclasses import dataclass
from pydantic import BaseModel

@dataclass
class JobsCustomResponse:
    name: t.Optional[str] = None
    status: t.Optional[bool] = None
    lines: t.Optional[int] = None
    uuid: t.Optional[str] = None
    progress: t.Optional[int] = None
    date: t.Optional[str] = None
# or 
class JobsCustomResponse(BaseModel):
    name: t.Optional[str] = None
    status: t.Optional[bool] = None
    lines: t.Optional[int] = None
    uuid: t.Optional[str] = None
    progress: t.Optional[int] = None
    date: t.Optional[str] = None

Litestar won't generate the types in the openapi docs (I don't know if this is intentional in Litestar or ?), but FastAPI will. You can do one thing and make all columns required by adding required=True (docs) in the Piccolo table definition like this

class Jobs(Table):
    id = Serial(primary_key=True, required=True)
    name = Varchar(required=True)
    lines = Integer(default=0, required=True)
    status = Varchar(default="pending", required=True)
    uuid = UUID(default=uuid.uuid4, required=True)
    progress = Integer(default=0, required=True)
    date = Timestamptz(required=True)
# create Pydantic model
JobsModel = create_pydantic_model(Jobs)

and then Litestar will consider them as non-optional (required) fields and they will be types displayed in Redoc

[YuriFontella]

This really worked, it's not so bad to put required in the schema fields.
Do you think this is a problem and we should notify the litestar team?

[sinisaos]

I don't know, as you wish. Regarding Litestar, I've only made one starter template to use Piccolo with Litestar and I don't have enough knowledge about Litestar. Maybe you will get an answer there whether such behavior is intentional or not.

[YuriFontella]

@sinisaos thanks a lot for the help