tiangolo / sqlmodel

SQL databases in Python, designed for simplicity, compatibility, and robustness.
https://sqlmodel.tiangolo.com/
MIT License
13.7k stars 616 forks source link

Flexibly Create Nested Database Entries from Incoming Pydantic/SQLModels #6

Open hay-kot opened 2 years ago

hay-kot commented 2 years ago

First Check

Commit to Help

Example Code

# Sudo Code Based on Examples in Docs

class Team(SQLModel, table=True):
    id: Optional[int] = Field(default=None, primary_key=True)
    name: str
    headquarters: str

    heroes: List["Hero"] = Relationship(back_populates="team")

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

    team_id: Optional[int] = Field(default=None, foreign_key="team.id")

    team: Optional[Team] = Relationship(back_populates="heroes")

payload = {
    "name": "Team Name",
    "headquarters": "Whereever".
    "heroes": [
        "name": "Name 1"
        // Other Requied Fields... 👇
    ]
}

with Session(engine) as session:
    Team.create_all_nested(session, payload) # or something?

Description

I would like to do what is described in FastAPI issue #2194

How to make nested sqlalchemy models from nested pydantic models (or python dicts) in a generic way and write them to the database in "one shot".

In the example above, I'd like to pass in the payload to a method and the following to occur.

Similarly, I'd like the same to happen on update. Effectively making writing to the SQL database akin to writing to MongoDB

I don't believe this is supported or haven't gotten it to work, but my main questions are.

  1. Is this supported?
  2. If no, is this a use-case you've thought of?
  3. Are you interested in a PR to support this either as a utility method or some sort of decorator?

Loving working with this so far, thanks for all your hard work!

Operating System

macOS

Operating System Details

No response

SQLModel Version

0.0.3

Python Version

3.9.6

Additional Context

I have accomplished this with SQLAlchemy in the past by using an auto_init decarator.

from functools import wraps
from typing import Union

from sqlalchemy.orm import MANYTOMANY, MANYTOONE, ONETOMANY

def handle_one_to_many_list(relation_cls, all_elements: list[dict]):
    elems_to_create = []
    updated_elems = []

    for elem in all_elements:
        elem_id = elem.get("id", None)

        existing_elem = relation_cls.get_ref(match_value=elem_id)

        if existing_elem is None:

            elems_to_create.append(elem)

        else:
            for key, value in elem.items():
                setattr(existing_elem, key, value)

            updated_elems.append(existing_elem)

    new_elems = []
    for elem in elems_to_create:
        new_elems = [relation_cls(**elem) for elem in all_elements]

    return new_elems

def auto_init(exclude: Union[set, list] = None):  # sourcery no-metrics
    """Wraps the `__init__` method of a class to automatically set the common
    attributes.

    Args:
        exclude (Union[set, list], optional): [description]. Defaults to None.
    """

    exclude = exclude or set()
    exclude.add("id")

    def decorator(init):
        @wraps(init)
        def wrapper(self, *args, **kwargs):  # sourcery no-metrics
            """
            Custom initializer that allows nested children initialization.
            Only keys that are present as instance's class attributes are allowed.
            These could be, for example, any mapped columns or relationships.

            Code inspired from GitHub.
            Ref: https://github.com/tiangolo/fastapi/issues/2194
            """
            cls = self.__class__
            model_columns = self.__mapper__.columns
            relationships = self.__mapper__.relationships

            session = kwargs.get("session", None)

            for key, val in kwargs.items():
                if key in exclude:
                    continue

                if not hasattr(cls, key):
                    continue
                    # raise TypeError(f"Invalid keyword argument: {key}")

                if key in model_columns:
                    setattr(self, key, val)
                    continue

                if key in relationships:
                    relation_dir = relationships[key].direction.name
                    relation_cls = relationships[key].mapper.entity
                    use_list = relationships[key].uselist

                    if relation_dir == ONETOMANY.name and use_list:
                        instances = handle_one_to_many_list(relation_cls, val)
                        setattr(self, key, instances)

                    if relation_dir == ONETOMANY.name and not use_list:
                        instance = relation_cls(**val)
                        setattr(self, key, instance)

                    elif relation_dir == MANYTOONE.name and not use_list:
                        if isinstance(val, dict):
                            val = val.get("id")

                            if val is None:
                                raise ValueError(f"Expected 'id' to be provided for {key}")

                        if isinstance(val, (str, int)):
                            instance = relation_cls.get_ref(match_value=val, session=session)
                            setattr(self, key, instance)

                    elif relation_dir == MANYTOMANY.name:

                        if not isinstance(val, list):
                            raise ValueError(f"Expected many to many input to be of type list for {key}")

                        if len(val) > 0 and isinstance(val[0], dict):
                            val = [elem.get("id") for elem in val]

                        instances = [relation_cls.get_ref(elem, session=session) for elem in val]
                        setattr(self, key, instances)

            return init(self, *args, **kwargs)

        return wrapper

    return decorator

Usage



class AdminModel(SqlAlchemyBase, BaseMixins):
    name = Column(String, index=True)
    email = Column(String, unique=True, index=True)
    password = Column(String)
    is_superuser = Column(Boolean(), default=False)

    @auto_init(exclude={'is_superuser'})
    def __init__(self, **_):
        this.is_superuser = false

    @classmethod
    def get_ref(cls, match_value: str, match_attr: str = "id"):
        with SessionLocal() as session:
            eff_ref = getattr(cls, match_attr)
            return session.query(cls).filter(eff_ref == match_value).one_or_none()

```decorator
khialb32 commented 2 years ago

Hello @hay-kot , i have tried your solution on nested models, and it works just fine, I will try with more nested pedantic models and see if it works, yet I would greatly appreciate if you could provide some explanation(documentation) of what you did because although I read it many times, I am still afraid to use it for production.

hobbsAU commented 2 years ago

SQLModel is fantastic as it cuts so much time out managing pydantic schemas and sqlalchemy models.

I too would like to know how to create database entries with incoming nested SQLModels.

Does anyone know how to do this yet?

woprandi commented 2 years ago

It's even worse than use separate pydantic and sqlalchemy models because the nested list on the payload received from client are always empty. So it's not even possible to create each element separarely

scd75 commented 2 years ago

Agree with @hay-kot : it would be great to have the ability to instantiate nested objects from a nested SQLModel instance. in the same vein as @hay-kot : i used with the past a custom constructor for SQLAlchemy Base class, and also proposed that to SQLAlchemey repo: https://github.com/tiangolo/fastapi/issues/2194 https://github.com/sqlalchemy/sqlalchemy/discussions/6322 @tiangolo : any idea on how this could be achieved? Thanks!

Fanna1119 commented 2 years ago

This would such an amazing feature. Any update on this?

Sancho66 commented 2 years ago

Any update on this ?

Sillocan commented 1 year ago

This is a major feature that is missing from sqlmodel that is preventing myself from migrating a larger project from marshmallow. There is a lot of boilerplate code that needs to be added to handle nested objects

zrajna commented 1 year ago

I would also love to see this addition to the awesome SQLModel library.

50Bytes-dev commented 1 year ago

Started working with this wonderful library and encountered the same problem. Any updates?

micoloth commented 1 year ago

+1 on this.

First of all, this library looks amazing!

Last year, I was using Sqlalchemy and Pydantic models separately, but this library would be such a quality of life improvement..

Unfortunately, the fact that this feature is missing is a deal breaker, so i cannot start using it.

Last year, I also built a custom recursive solution to do this in Sqlalchemy+Pydantic, just like OP did.

But, I don't think it would be easy/a good idea to port it to SqlModels...

Any updates would be greatly appreciated!

s-froghyar commented 6 months ago

Any update on this? Would love to use hydra as a configuration management the way it is meant to be used (hierarchical and out of the box configurable with separation of concerns) rather than making a single flat yaml for all configs

whugarra commented 1 week ago

There any temporary guidance from best practices to avoid this issue?