alchemista 0.3.0

Tools to convert SQLAlchemy models to Pydantic models

Alchemista

Tools to generate Pydantic models from SQLAlchemy models.

Still experimental.

Installation

Alchemista is available in PyPI. To install it with pip, run:

pip install alchemista

Usage

Simply call the model_from function with a SQLAlchemy model. Each Column in its definition will result in an attribute of the generated model via the Pydantic Field function.

For example, a SQLAlchemy model like the following

from sqlalchemy import Column, Integer, String
from sqlalchemy.orm import declarative_base


Base = declarative_base()

class PersonDB(Base):
    __tablename__ = "people"

    id = Column(Integer, primary_key=True)
    age = Column(Integer, default=0, nullable=False, doc="Age in years")
    name = Column(String(128), nullable=False, doc="Full name")

could have a generated Pydantic model via

from alchemista import model_from

Person = model_from(PersonDB)

and would result in a Pydantic model equivalent to

from pydantic import BaseModel, Field


class Person(BaseModel):
    id: int
    age: int = Field(0, description="Age in years")
    name: str = Field(..., max_length=128, description="Full name")

    class Config:
        orm_mode = True

Note that the string length from the column definition was sufficient to add a max_length constraint. Additionally, by default, the generated model will have orm_mode=True. That can be customized via the __config__ keyword argument.

There is also an exclude keyword argument that accepts a set of field names to not include in the generated model, and an include keyword argument accepts a set of field names to do include in the generated model. However, they are mutually exclusive and cannot be used together.

This example is available in a short executable form in the examples/ directory.

Field arguments and info

Currently, the type, default value (either scalar or callable), and the description (from the doc attribute) are extracted directly from the Column definition. However, except for the type, all of them can be overridden via the info dictionary attribute. All other custom arguments to the Field function are specified there too. The supported keys are listed in alchemista.field.Info.

Everything specified in info is preferred from what has been extracted from Column. This means that the default value and the description can be overridden if so desired. Also, similarly to using Pydantic directly, default and default_factory are mutually-exclusive, so they cannot be used together. Use default_factory if the default value comes from calling a function (without any arguments).

For example, in the case above,

name = Column(String(128), nullable=False, doc="Full name", info=dict(description=None, max_length=64))

would instead result in

name: str = Field(..., max_length=64)

fields_from and model_from

The fields_from function is the function that actually inspects the SQLAlchemy model and builds a dictionary in a format that can be used to generate a Pydantic model. So model_from is just a shortcut for calling fields_from and then pydantic.create_model. The model name that model_from sets is db_model.__name__.

If desired, or extra control is needed, pydantic.create_model can be used directly, in conjunction with fields_from. This allows the customization of the name of the model that will be created and the specification of other create_model arguments, like __base__ and __validators__ (model_from currently only accepts __config__).

For example:

from alchemista import fields_from
from pydantic import create_model


MyModel = create_model("MyModel", **fields_from(DBModel))

transform

Both fields_from and model_from have a transform argument. It is a callable used to transform the fields before generating a Pydantic model. The provided transformation functions are in the func module. By default, the transform argument is set to func.unchanged, which, as the name implies, does nothing.

The other provided transformation function is func.nonify, which makes all fields optional (if they weren't already) and nullable and sets the default value to None. This is useful when some kind of "input model" is desired. For example, the database model might have an auto-generated primary key, and some other columns with default values. When updating an entity of this model, one wouldn't want to receive the primary key as an update candidate (which can be solved using the exclude argument) and probably wouldn't want the fields with default values to actually have these values, since the update is meant to change only what the user asked to update.

For example:

from sqlalchemy import Column, Integer, String
from sqlalchemy.orm import declarative_base

from alchemista import model_from
from alchemista.func import nonify


Base = declarative_base()

class PersonDB(Base):
    __tablename__ = "people"

    id = Column(Integer, primary_key=True)
    age = Column(Integer, default=0, nullable=False, doc="Age in years")
    name = Column(String(128), nullable=False, doc="Full name")


PersonInput = model_from(PersonDB, exclude={"id"}, transform=nonify)

The model PersonInput is equivalent to the hand-written version below:

from typing import Optional

from pydantic import BaseModel, Field


class PersonInput(BaseModel):
    age: Optional[int] = Field(None, description="Age in years")
    name: Optional[str] = Field(None, max_length=128, description="Full name")

    class Config:
        orm_mode = True

Note that both age and name weren't originally nullable (Optional, in Python) and the name was required (age wasn't because it has a default value). Now, both fields are nullable and their default value is None.

User-defined transformations

You can also create your own transformation functions. The expected signature is as follows:

from typing import Tuple

from pydantic.fields import FieldInfo

def transformation(name: str, python_type: type, field: FieldInfo) -> Tuple[type, FieldInfo]:
    pass

Where name is the name of the field currently being created, python_type is its Python type, and field is its full Pydantic field specification. The return type is a tuple of the Python type and the field specification. These two can be changed freely (the name can't).

License

This project is licensed under the terms of the MIT license.