[bug] default_factory called twice when parse input by FastAPI/route

[dacrystal]

Current Behavior (example Code):

from typing import Optional
from sqlmodel import Field, SQLModel


def gen_id():
    gen_id.i += 1
    print(f"gen_id called. id={gen_id.i}")
    return gen_id.i


gen_id.i = 0


class Hero(SQLModel):
    id: Optional[int] = Field(primary_key=True, default_factory=gen_id)
    name: str
    secret_name: str
    age: Optional[int] = None


hero_1 = Hero.validate({"name": "Deadpond",
                        "secret_name": "Dive Wilson"})


assert hero_1.id == 1

Expected Behavior

default_factor and validations should run once.

Root cause

The issue is that validation_model called twice directly in L589 and indirectly in L592 (init)
https://github.com/tiangolo/sqlmodel/blob/43a689d369f52b72aac60efd71111aba7d84714d/sqlmodel/main.py#L583-L605

in __init__ :

https://github.com/tiangolo/sqlmodel/blob/43a689d369f52b72aac60efd71111aba7d84714d/sqlmodel/main.py#L498-L498

[antont]

Just a note, that I guess that with the multiple models pattern this does not typically happen, because the *Create model does not include fields with factories, but typically the db model does. https://sqlmodel.tiangolo.com/tutorial/fastapi/multiple-models/

Sure would be nice to fix anyhow.

[dacrystal]

Yes multiple-models-pattern (what I'm using) does fix it for default_factory.
However, validation still run twice for the same data when used with FastAPI(fast/high-performance) or when validating a model (for some hypothetical use case) in general.

[dacrystal]

@antont
I was wrong for multiple-models-pattern, it seems it didn't fix it (I was testing the wrong endpoint earlier).

from_orm and parse_as_obj still cause validations called twice.

Here is a more comprehensive example:

from typing import List, Optional
from sqlmodel import Field, SQLModel
from pydantic import parse_obj_as


def gen_id():
    gen_id.i += 1
    print(f"gen_id called. id={gen_id.i}")
    return gen_id.i


gen_id.i = 0


class HeroBase(SQLModel):
    name: str = Field(index=True)
    secret_name: str
    age: Optional[int] = Field(default=None, index=True)


class Hero(HeroBase, table=True):
    id: int = Field(primary_key=True, default_factory=gen_id)


class HeroCreate(HeroBase):
    pass


hero_create = HeroCreate(name="Deadpond", secret_name="Dive Wilson")
hero_create_list = [
    HeroCreate(name="Spider-Boy", secret_name="Pedro Parqueador"),
    HeroCreate(name="Deadpond", secret_name="Dive Wilson")]


# This work, but...
hero_parse_obj = Hero.parse_obj(hero_create)

gen_id.i = 0
# default_factory/validations called twice
hero_validate = Hero.validate({"name": "Deadpond",
                               "secret_name": "Dive Wilson"})

# default_factory/validations called twice
gen_id.i = 0
hero_from_orm = Hero.from_orm(hero_create)

# default_factory/validations called twice
gen_id.i = 0
hero_list = parse_obj_as(List[Hero], hero_create_list)


assert hero_parse_obj.id == 1
assert hero_validate.id == 1
assert hero_from_orm.id == 1
assert hero_list[0].id == 1

Or I might be doing it wrongly.

[antont]

You are probably right, I guess in our case we just haven't noticed that, with DB generated IDs.