Pydantic reusable Validation for Dictionary's keys and values - python

How to validate input to get the following Dict passed!
d = dict()
d['en'] = 'English content'
d['it'] = 'Italian content'
d['es'] = 'Spanish content'
print(d)
# {'en': 'English content', 'it': 'Italian content', 'es': 'Spanish content'}
In this example, keys are ISO 639-1 codes using pycountry python package.
code = 'en'
pycountry.languages.get(alpha_2=code.upper()).alpha_2 # = 'en'
The point is how to validate keys using pydantic reusable validator or any other methods?
And validate values either to be str or int?
Pydantic model schema should be similar to this sample :
# products/model.py
from sqlalchemy import Column, String, Integer
from sqlalchemy.ext.declarative import declarative_base
from custom_field import Translatable
Base = declarative_base()
class Model(Base):
__tablename__ = "products"
id = Column(Integer, unique=True, index=True)
name = Column(Translatable())
price = Column(Integer)
# products/pydantic.py
from pydantic import BaseModel
import custom_pydantic_field
class BaseSchema(BaseModel):
id: int
class CreateSchema(BaseSchema):
name: custom_pydantic_field.translatable
price: int
Keep in mind reusability in other models/schemas.

Create pydantic custom class
# validators/translated_field.py
from typing import Dict
from pydantic import ValidationError
from pydantic.error_wrappers import ErrorWrapper
import pycountry
class Translatable(Dict):
"""
Validate Translation Dict Field (Json) where Language is Key and Translation as Value
Languages : ISO 639-1 code
Translation : Int, str, None
ref:
- https://pydantic-docs.helpmanual.io/usage/types/#classes-with-__get_validators__
By: Khalid Murad
"""
#property
def __translation_interface__(self):
return self.dict()
#classmethod
def __get_validators__(cls):
yield cls.validate
#classmethod
def validate(cls, base_dictionary):
result = dict()
dictionary = dict()
errors = []
dictionary = base_dictionary
for key in dictionary:
try:
parsed_language = pycountry.languages.get(alpha_2=key.upper())
except ValueError as exc:
errors.append(ErrorWrapper(Exception(f"Invalid language: {key}."), loc="language"))
if not parsed_language:
errors.append(ErrorWrapper(Exception(f"Invalid language: {key}."), loc="language"))
if isinstance(dictionary[key], int | str | None):
result[key] = dictionary[key]
else:
errors.append(ErrorWrapper(Exception(f"Invalid content for language: {key}."), loc=("language","content")))
if errors:
raise ValidationError(
errors,
cls,
)
return cls(result)
Then use it in you schema/pydantic model like:
# products/pydantic.py
from pydantic import BaseModel
from validators.translated_field import Translatable
class BaseSchema(BaseModel):
id: int
class CreateSchema(BaseSchema):
name: Translatable
...your code
And use normal JSON field in SQLALchemy model!
# products/model.py
...
from sqlalchemy import Column, JSON
...
class Model(Base):
name = Column(JSON, nullable=True)
...

Related

pydantic exclude multiple fields from model

In pydantic is there a cleaner way to exclude multiple fields from the model, something like:
class User(UserBase):
class Config:
exclude = ['user_id', 'some_other_field']
I am aware that following works, but I was looking for something cleaner like django.
class User(UserBase):
class Config:
fields = {'user_id': {'exclude':True},
'some_other_field': {'exclude':True}
}
Pydantic will exclude the class variables which begin with an underscore.
so if it fits your use case, you can rename your attribues.
class User(UserBase):
_user_id=str
some_other_field=str
....
I wrote something like this for my json :
from pydantic import BaseModel
class CustomBase(BaseModel):
def json(self, **kwargs):
include = getattr(self.Config, "include", set())
if len(include) == 0:
include = None
exclude = getattr(self.Config, "exclude", set())
if len(exclude) == 0:
exclude = None
return super().json(include=include, exclude=exclude, **kwargs)
class User(CustomBase):
name :str = ...
family :str = ...
class Config:
exclude = {"family"}
u = User(**{"name":"milad","family":"vayani"})
print(u.json())
you can overriding dict and other method like.
A possible solution is creating a new class based in the baseclass using create_model:
from pydantic import BaseModel, create_model
def exclude_id(baseclass, to_exclude: list):
# Here we just extract the fields and validators from the baseclass
fields = baseclass.__fields__
validators = {'__validators__': baseclass.__validators__}
new_fields = {key: (item.type_, ... if item.required else None)
for key, item in fields.items() if key not in to_exclude}
return create_model(f'{baseclass.__name__}Excluded', **new_fields, __validators__=validators)
class User(BaseModel):
ID: str
some_other: str
list_to_exclude = ['ID']
UserExcluded = exclude_id(User, list_to_exclude)
UserExcluded(some_other='hola')
Which will return:
> UserExcluded(some_other='hola')
Which is a copy of the baseclass but with no parameter 'ID'.
If you have the id in the validators you may want also to exclude those validators.

Best way to flatten and remap ORM to Pydantic Model

I am using Pydantic with FastApi to output ORM data into JSON. I would like to flatten and remap the ORM model to eliminate an unnecessary level in the JSON.
Here's a simplified example to illustrate the problem.
original output: {"id": 1, "billing":
[
{"id": 1, "order_id": 1, "first_name": "foo"},
{"id": 2, "order_id": 1, "first_name": "bar"}
]
}
desired output: {"id": 1, "name": ["foo", "bar"]}
How to map values from nested dict to Pydantic Model? provides a solution that works for dictionaries by using the init function in the Pydantic model class. This example shows how that works with dictionaries:
from pydantic import BaseModel
# The following approach works with a dictionary as the input
order_dict = {"id": 1, "billing": {"first_name": "foo"}}
# desired output: {"id": 1, "name": "foo"}
class Order_Model_For_Dict(BaseModel):
id: int
name: str = None
class Config:
orm_mode = True
def __init__(self, **kwargs):
print(
"kwargs for dictionary:", kwargs
) # kwargs for dictionary: {'id': 1, 'billing': {'first_name': 'foo'}}
kwargs["name"] = kwargs["billing"]["first_name"]
super().__init__(**kwargs)
print(Order_Model_For_Dict.parse_obj(order_dict)) # id=1 name='foo'
(This script is complete, it should run "as is")
However, when working with ORM objects, this approach does not work. It appears that the init function is not called. Here's an example which will not provide the desired output.
from pydantic import BaseModel, root_validator
from typing import List
from sqlalchemy.orm import relationship
from sqlalchemy import Column, Integer, String, ForeignKey
from sqlalchemy.dialects.postgresql import ARRAY
from sqlalchemy.ext.declarative import declarative_base
Base = declarative_base()
from pydantic.utils import GetterDict
class BillingOrm(Base):
__tablename__ = "billing"
id = Column(Integer, primary_key=True, nullable=False)
order_id = Column(ForeignKey("orders.id", ondelete="CASCADE"), nullable=False)
first_name = Column(String(20))
class OrderOrm(Base):
__tablename__ = "orders"
id = Column(Integer, primary_key=True, nullable=False)
billing = relationship("BillingOrm")
class Billing(BaseModel):
id: int
order_id: int
first_name: str
class Config:
orm_mode = True
class Order(BaseModel):
id: int
name: List[str] = None
# billing: List[Billing] # uncomment to verify the relationship is working
class Config:
orm_mode = True
def __init__(self, **kwargs):
# This __init__ function does not run when using from_orm to parse ORM object
print("kwargs for orm:", kwargs)
kwargs["name"] = kwargs["billing"]["first_name"]
super().__init__(**kwargs)
billing_orm_1 = BillingOrm(id=1, order_id=1, first_name="foo")
billing_orm_2 = BillingOrm(id=2, order_id=1, first_name="bar")
order_orm = OrderOrm(id=1)
order_orm.billing.append(billing_orm_1)
order_orm.billing.append(billing_orm_2)
order_model = Order.from_orm(order_orm)
# Output returns 'None' for name instead of ['foo','bar']
print(order_model) # id=1 name=None
(This script is complete, it should run "as is")
The output returns name=None instead of the desired list of names.
In the above example, I am using Order.from_orm to create the Pydantic model. This approach seems to be the same that is used by FastApi when specifying a response model. The desired solution should support use in the FastApi response model as shown in this example:
#router.get("/orders", response_model=List[schemas.Order])
async def list_orders(db: Session = Depends(get_db)):
return get_orders(db)
Update:
Regarding MatsLindh comment to try validators, I replaced the init function with a root validator, however, I'm unable to mutate the return values to include a new attribute. I suspect this issue is because it is a ORM object and not a true dictionary. The following code will extract the names and print them in the desired list. However, I can't see how to include this updated result in the model response:
#root_validator(pre=True)
def flatten(cls, values):
if isinstance(values, GetterDict):
names = [
billing_entry.first_name for billing_entry in values.get("billing")
]
print(names)
# values["name"] = names # error: 'GetterDict' object does not support item assignment
return values
I also found a couple other discussions on this problem that led me to try this approach:
https://github.com/samuelcolvin/pydantic/issues/717
https://gitmemory.com/issue/samuelcolvin/pydantic/821/744047672
What if you override the from_orm class method?
class Order(BaseModel):
id: int
name: List[str] = None
billing: List[Billing]
class Config:
orm_mode = True
#classmethod
def from_orm(cls, obj: Any) -> 'Order':
# `obj` is the orm model instance
if hasattr(obj, 'billing'):
obj.name = obj.billing.first_name
return super().from_orm(obj)
I really missed the handy Django REST Framework serializers while working with the FastAPI + Pydantic stack... So I wrangled with GetterDict to allow defining field getter function in the Pydantic model like this:
class User(FromORM):
fullname: str
class Config(FromORM.Config):
getter_dict = FieldGetter.bind(lambda: User)
#staticmethod
def get_fullname(obj: User) -> str:
return f'{obj.firstname} {obj.lastname}'
where the magic part FieldGetter is implemented as
from typing import Any, Callable, Optional, Type
from types import new_class
from pydantic import BaseModel
from pydantic.utils import GetterDict
class FieldGetter(GetterDict):
model_class_forward_ref: Optional[Callable] = None
model_class: Optional[Type[BaseModel]] = None
def __new__(cls, *args, **kwargs):
inst = super().__new__(cls)
if cls.model_class_forward_ref:
inst.model_class = cls.model_class_forward_ref()
return inst
#classmethod
def bind(cls, model_class_forward_ref: Callable):
sub_class = new_class(f'{cls.__name__}FieldGetter', (cls,))
sub_class.model_class_forward_ref = model_class_forward_ref
return sub_class
def get(self, key: str, default):
if hasattr(self._obj, key):
return super().get(key, default)
getter_fun_name = f'get_{key}'
if not (getter := getattr(self.model_class, getter_fun_name, None)):
raise AttributeError(f'no field getter function found for {key}')
return getter(self._obj)
class FromORM(BaseModel):
class Config:
orm_mode = True
getter_dict = FieldGetter

convert models to json [duplicate]

Django has some good automatic serialization of ORM models returned from DB to JSON format.
How to serialize SQLAlchemy query result to JSON format?
I tried jsonpickle.encode but it encodes query object itself.
I tried json.dumps(items) but it returns
TypeError: <Product('3', 'some name', 'some desc')> is not JSON serializable
Is it really so hard to serialize SQLAlchemy ORM objects to JSON /XML? Isn't there any default serializer for it? It's very common task to serialize ORM query results nowadays.
What I need is just to return JSON or XML data representation of SQLAlchemy query result.
SQLAlchemy objects query result in JSON/XML format is needed to be used in javascript datagird (JQGrid http://www.trirand.com/blog/)
You could just output your object as a dictionary:
class User:
def as_dict(self):
return {c.name: getattr(self, c.name) for c in self.__table__.columns}
And then you use User.as_dict() to serialize your object.
As explained in Convert sqlalchemy row object to python dict
A flat implementation
You could use something like this:
from sqlalchemy.ext.declarative import DeclarativeMeta
class AlchemyEncoder(json.JSONEncoder):
def default(self, obj):
if isinstance(obj.__class__, DeclarativeMeta):
# an SQLAlchemy class
fields = {}
for field in [x for x in dir(obj) if not x.startswith('_') and x != 'metadata']:
data = obj.__getattribute__(field)
try:
json.dumps(data) # this will fail on non-encodable values, like other classes
fields[field] = data
except TypeError:
fields[field] = None
# a json-encodable dict
return fields
return json.JSONEncoder.default(self, obj)
and then convert to JSON using:
c = YourAlchemyClass()
print json.dumps(c, cls=AlchemyEncoder)
It will ignore fields that are not encodable (set them to 'None').
It doesn't auto-expand relations (since this could lead to self-references, and loop forever).
A recursive, non-circular implementation
If, however, you'd rather loop forever, you could use:
from sqlalchemy.ext.declarative import DeclarativeMeta
def new_alchemy_encoder():
_visited_objs = []
class AlchemyEncoder(json.JSONEncoder):
def default(self, obj):
if isinstance(obj.__class__, DeclarativeMeta):
# don't re-visit self
if obj in _visited_objs:
return None
_visited_objs.append(obj)
# an SQLAlchemy class
fields = {}
for field in [x for x in dir(obj) if not x.startswith('_') and x != 'metadata']:
fields[field] = obj.__getattribute__(field)
# a json-encodable dict
return fields
return json.JSONEncoder.default(self, obj)
return AlchemyEncoder
And then encode objects using:
print json.dumps(e, cls=new_alchemy_encoder(), check_circular=False)
This would encode all children, and all their children, and all their children... Potentially encode your entire database, basically. When it reaches something its encoded before, it will encode it as 'None'.
A recursive, possibly-circular, selective implementation
Another alternative, probably better, is to be able to specify the fields you want to expand:
def new_alchemy_encoder(revisit_self = False, fields_to_expand = []):
_visited_objs = []
class AlchemyEncoder(json.JSONEncoder):
def default(self, obj):
if isinstance(obj.__class__, DeclarativeMeta):
# don't re-visit self
if revisit_self:
if obj in _visited_objs:
return None
_visited_objs.append(obj)
# go through each field in this SQLalchemy class
fields = {}
for field in [x for x in dir(obj) if not x.startswith('_') and x != 'metadata']:
val = obj.__getattribute__(field)
# is this field another SQLalchemy object, or a list of SQLalchemy objects?
if isinstance(val.__class__, DeclarativeMeta) or (isinstance(val, list) and len(val) > 0 and isinstance(val[0].__class__, DeclarativeMeta)):
# unless we're expanding this field, stop here
if field not in fields_to_expand:
# not expanding this field: set it to None and continue
fields[field] = None
continue
fields[field] = val
# a json-encodable dict
return fields
return json.JSONEncoder.default(self, obj)
return AlchemyEncoder
You can now call it with:
print json.dumps(e, cls=new_alchemy_encoder(False, ['parents']), check_circular=False)
To only expand SQLAlchemy fields called 'parents', for example.
Python 3.7+ and Flask 1.1+ can use the built-in dataclasses package
from dataclasses import dataclass
from datetime import datetime
from flask import Flask, jsonify
from flask_sqlalchemy import SQLAlchemy
app = Flask(__name__)
db = SQLAlchemy(app)
#dataclass
class User(db.Model):
id: int
email: str
id = db.Column(db.Integer, primary_key=True, auto_increment=True)
email = db.Column(db.String(200), unique=True)
#app.route('/users/')
def users():
users = User.query.all()
return jsonify(users)
if __name__ == "__main__":
users = User(email="user1#gmail.com"), User(email="user2#gmail.com")
db.create_all()
db.session.add_all(users)
db.session.commit()
app.run()
The /users/ route will now return a list of users.
[
{"email": "user1#gmail.com", "id": 1},
{"email": "user2#gmail.com", "id": 2}
]
Auto-serialize related models
#dataclass
class Account(db.Model):
id: int
users: User
id = db.Column(db.Integer)
users = db.relationship(User) # User model would need a db.ForeignKey field
The response from jsonify(account) would be this.
{
"id":1,
"users":[
{
"email":"user1#gmail.com",
"id":1
},
{
"email":"user2#gmail.com",
"id":2
}
]
}
Overwrite the default JSON Encoder
from flask.json import JSONEncoder
class CustomJSONEncoder(JSONEncoder):
"Add support for serializing timedeltas"
def default(o):
if type(o) == datetime.timedelta:
return str(o)
if type(o) == datetime.datetime:
return o.isoformat()
return super().default(o)
app.json_encoder = CustomJSONEncoder
You can convert a RowProxy to a dict like this:
d = dict(row.items())
Then serialize that to JSON ( you will have to specify an encoder for things like datetime values )
It's not that hard if you just want one record ( and not a full hierarchy of related records ).
json.dumps([(dict(row.items())) for row in rs])
I recommend using marshmallow. It allows you to create serializers to represent your model instances with support to relations and nested objects.
Here is a truncated example from their docs. Take the ORM model, Author:
class Author(db.Model):
id = db.Column(db.Integer, primary_key=True)
first = db.Column(db.String(80))
last = db.Column(db.String(80))
A marshmallow schema for that class is constructed like this:
class AuthorSchema(Schema):
id = fields.Int(dump_only=True)
first = fields.Str()
last = fields.Str()
formatted_name = fields.Method("format_name", dump_only=True)
def format_name(self, author):
return "{}, {}".format(author.last, author.first)
...and used like this:
author_schema = AuthorSchema()
author_schema.dump(Author.query.first())
...would produce an output like this:
{
"first": "Tim",
"formatted_name": "Peters, Tim",
"id": 1,
"last": "Peters"
}
Have a look at their full Flask-SQLAlchemy Example.
A library called marshmallow-sqlalchemy specifically integrates SQLAlchemy and marshmallow. In that library, the schema for the Author model described above looks like this:
class AuthorSchema(ModelSchema):
class Meta:
model = Author
The integration allows the field types to be inferred from the SQLAlchemy Column types.
marshmallow-sqlalchemy here.
You can use introspection of SqlAlchemy as this :
mysql = SQLAlchemy()
from sqlalchemy import inspect
class Contacts(mysql.Model):
__tablename__ = 'CONTACTS'
id = mysql.Column(mysql.Integer, primary_key=True)
first_name = mysql.Column(mysql.String(128), nullable=False)
last_name = mysql.Column(mysql.String(128), nullable=False)
phone = mysql.Column(mysql.String(128), nullable=False)
email = mysql.Column(mysql.String(128), nullable=False)
street = mysql.Column(mysql.String(128), nullable=False)
zip_code = mysql.Column(mysql.String(128), nullable=False)
city = mysql.Column(mysql.String(128), nullable=False)
def toDict(self):
return { c.key: getattr(self, c.key) for c in inspect(self).mapper.column_attrs }
#app.route('/contacts',methods=['GET'])
def getContacts():
contacts = Contacts.query.all()
contactsArr = []
for contact in contacts:
contactsArr.append(contact.toDict())
return jsonify(contactsArr)
#app.route('/contacts/<int:id>',methods=['GET'])
def getContact(id):
contact = Contacts.query.get(id)
return jsonify(contact.toDict())
Get inspired from an answer here :
Convert sqlalchemy row object to python dict
Flask-JsonTools package has an implementation of JsonSerializableBase Base class for your models.
Usage:
from sqlalchemy.ext.declarative import declarative_base
from flask.ext.jsontools import JsonSerializableBase
Base = declarative_base(cls=(JsonSerializableBase,))
class User(Base):
#...
Now the User model is magically serializable.
If your framework is not Flask, you can just grab the code
For security reasons you should never return all the model's fields. I prefer to selectively choose them.
Flask's json encoding now supports UUID, datetime and relationships (and added query and query_class for flask_sqlalchemy db.Model class). I've updated the encoder as follows:
app/json_encoder.py
from sqlalchemy.ext.declarative import DeclarativeMeta
from flask import json
class AlchemyEncoder(json.JSONEncoder):
def default(self, o):
if isinstance(o.__class__, DeclarativeMeta):
data = {}
fields = o.__json__() if hasattr(o, '__json__') else dir(o)
for field in [f for f in fields if not f.startswith('_') and f not in ['metadata', 'query', 'query_class']]:
value = o.__getattribute__(field)
try:
json.dumps(value)
data[field] = value
except TypeError:
data[field] = None
return data
return json.JSONEncoder.default(self, o)
app/__init__.py
# json encoding
from app.json_encoder import AlchemyEncoder
app.json_encoder = AlchemyEncoder
With this I can optionally add a __json__ property that returns the list of fields I wish to encode:
app/models.py
class Queue(db.Model):
id = db.Column(db.Integer, primary_key=True)
song_id = db.Column(db.Integer, db.ForeignKey('song.id'), unique=True, nullable=False)
song = db.relationship('Song', lazy='joined')
type = db.Column(db.String(20), server_default=u'audio/mpeg')
src = db.Column(db.String(255), nullable=False)
created_at = db.Column(db.DateTime, server_default=db.func.now())
updated_at = db.Column(db.DateTime, server_default=db.func.now(), onupdate=db.func.now())
def __init__(self, song):
self.song = song
self.src = song.full_path
def __json__(self):
return ['song', 'src', 'type', 'created_at']
I add #jsonapi to my view, return the resultlist and then my output is as follows:
[
{
"created_at": "Thu, 23 Jul 2015 11:36:53 GMT",
"song":
{
"full_path": "/static/music/Audioslave/Audioslave [2002]/1 Cochise.mp3",
"id": 2,
"path_name": "Audioslave/Audioslave [2002]/1 Cochise.mp3"
},
"src": "/static/music/Audioslave/Audioslave [2002]/1 Cochise.mp3",
"type": "audio/mpeg"
}
]
A more detailed explanation.
In your model, add:
def as_dict(self):
return {c.name: str(getattr(self, c.name)) for c in self.__table__.columns}
The str() is for python 3 so if using python 2 use unicode(). It should help deserialize dates. You can remove it if not dealing with those.
You can now query the database like this
some_result = User.query.filter_by(id=current_user.id).first().as_dict()
First() is needed to avoid weird errors. as_dict() will now deserialize the result. After deserialization, it is ready to be turned to json
jsonify(some_result)
While the original question goes back awhile, the number of answers here (and my own experiences) suggest it's a non-trivial question with a lot of different approaches of varying complexity with different trade-offs.
That's why I built the SQLAthanor library that extends SQLAlchemy's declarative ORM with configurable serialization/de-serialization support that you might want to take a look at.
The library supports:
Python 2.7, 3.4, 3.5, and 3.6.
SQLAlchemy versions 0.9 and higher
serialization/de-serialization to/from JSON, CSV, YAML, and Python dict
serialization/de-serialization of columns/attributes, relationships, hybrid properties, and association proxies
enabling and disabling of serialization for particular formats and columns/relationships/attributes (e.g. you want to support an inbound password value, but never include an outbound one)
pre-serialization and post-deserialization value processing (for validation or type coercion)
a pretty straightforward syntax that is both Pythonic and seamlessly consistent with SQLAlchemy's own approach
You can check out the (I hope!) comprehensive docs here: https://sqlathanor.readthedocs.io/en/latest
Hope this helps!
Custom serialization and deserialization.
"from_json" (class method) builds a Model object based on json data.
"deserialize" could be called only on instance, and merge all data from json into Model instance.
"serialize" - recursive serialization
__write_only__ property is needed to define write only properties ("password_hash" for example).
class Serializable(object):
__exclude__ = ('id',)
__include__ = ()
__write_only__ = ()
#classmethod
def from_json(cls, json, selfObj=None):
if selfObj is None:
self = cls()
else:
self = selfObj
exclude = (cls.__exclude__ or ()) + Serializable.__exclude__
include = cls.__include__ or ()
if json:
for prop, value in json.iteritems():
# ignore all non user data, e.g. only
if (not (prop in exclude) | (prop in include)) and isinstance(
getattr(cls, prop, None), QueryableAttribute):
setattr(self, prop, value)
return self
def deserialize(self, json):
if not json:
return None
return self.__class__.from_json(json, selfObj=self)
#classmethod
def serialize_list(cls, object_list=[]):
output = []
for li in object_list:
if isinstance(li, Serializable):
output.append(li.serialize())
else:
output.append(li)
return output
def serialize(self, **kwargs):
# init write only props
if len(getattr(self.__class__, '__write_only__', ())) == 0:
self.__class__.__write_only__ = ()
dictionary = {}
expand = kwargs.get('expand', ()) or ()
prop = 'props'
if expand:
# expand all the fields
for key in expand:
getattr(self, key)
iterable = self.__dict__.items()
is_custom_property_set = False
# include only properties passed as parameter
if (prop in kwargs) and (kwargs.get(prop, None) is not None):
is_custom_property_set = True
iterable = kwargs.get(prop, None)
# loop trough all accessible properties
for key in iterable:
accessor = key
if isinstance(key, tuple):
accessor = key[0]
if not (accessor in self.__class__.__write_only__) and not accessor.startswith('_'):
# force select from db to be able get relationships
if is_custom_property_set:
getattr(self, accessor, None)
if isinstance(self.__dict__.get(accessor), list):
dictionary[accessor] = self.__class__.serialize_list(object_list=self.__dict__.get(accessor))
# check if those properties are read only
elif isinstance(self.__dict__.get(accessor), Serializable):
dictionary[accessor] = self.__dict__.get(accessor).serialize()
else:
dictionary[accessor] = self.__dict__.get(accessor)
return dictionary
Here is a solution that lets you select the relations you want to include in your output as deep as you would like to go.
NOTE: This is a complete re-write taking a dict/str as an arg rather than a list. fixes some stuff..
def deep_dict(self, relations={}):
"""Output a dict of an SA object recursing as deep as you want.
Takes one argument, relations which is a dictionary of relations we'd
like to pull out. The relations dict items can be a single relation
name or deeper relation names connected by sub dicts
Example:
Say we have a Person object with a family relationship
person.deep_dict(relations={'family':None})
Say the family object has homes as a relation then we can do
person.deep_dict(relations={'family':{'homes':None}})
OR
person.deep_dict(relations={'family':'homes'})
Say homes has a relation like rooms you can do
person.deep_dict(relations={'family':{'homes':'rooms'}})
and so on...
"""
mydict = dict((c, str(a)) for c, a in
self.__dict__.items() if c != '_sa_instance_state')
if not relations:
# just return ourselves
return mydict
# otherwise we need to go deeper
if not isinstance(relations, dict) and not isinstance(relations, str):
raise Exception("relations should be a dict, it is of type {}".format(type(relations)))
# got here so check and handle if we were passed a dict
if isinstance(relations, dict):
# we were passed deeper info
for left, right in relations.items():
myrel = getattr(self, left)
if isinstance(myrel, list):
mydict[left] = [rel.deep_dict(relations=right) for rel in myrel]
else:
mydict[left] = myrel.deep_dict(relations=right)
# if we get here check and handle if we were passed a string
elif isinstance(relations, str):
# passed a single item
myrel = getattr(self, relations)
left = relations
if isinstance(myrel, list):
mydict[left] = [rel.deep_dict(relations=None)
for rel in myrel]
else:
mydict[left] = myrel.deep_dict(relations=None)
return mydict
so for an example using person/family/homes/rooms... turning it into json all you need is
json.dumps(person.deep_dict(relations={'family':{'homes':'rooms'}}))
step1:
class CNAME:
...
def as_dict(self):
return {item.name: getattr(self, item.name) for item in self.__table__.columns}
step2:
list = []
for data in session.query(CNAME).all():
list.append(data.as_dict())
step3:
return jsonify(list)
Even though it's a old post, Maybe I didn't answer the question above, but I want to talk about my serialization, at least it works for me.
I use FastAPI,SqlAlchemy and MySQL, but I don't use orm model;
# from sqlalchemy import create_engine
# from sqlalchemy.orm import sessionmaker
# engine = create_engine(config.SQLALCHEMY_DATABASE_URL, pool_pre_ping=True)
# SessionLocal = sessionmaker(autocommit=False, autoflush=False, bind=engine)
Serialization code
import decimal
import datetime
def alchemy_encoder(obj):
"""JSON encoder function for SQLAlchemy special classes."""
if isinstance(obj, datetime.date):
return obj.strftime("%Y-%m-%d %H:%M:%S")
elif isinstance(obj, decimal.Decimal):
return float(obj)
import json
from sqlalchemy import text
# db is SessionLocal() object
app_sql = 'SELECT * FROM app_info ORDER BY app_id LIMIT :page,:page_size'
# The next two are the parameters passed in
page = 1
page_size = 10
# execute sql and return a <class 'sqlalchemy.engine.result.ResultProxy'> object
app_list = db.execute(text(app_sql), {'page': page, 'page_size': page_size})
# serialize
res = json.loads(json.dumps([dict(r) for r in app_list], default=alchemy_encoder))
If it doesn't work, please ignore my answer. I refer to it here
https://codeandlife.com/2014/12/07/sqlalchemy-results-to-json-the-easy-way/
install simplejson by
pip install simplejson and the create a class
class Serialise(object):
def _asdict(self):
"""
Serialization logic for converting entities using flask's jsonify
:return: An ordered dictionary
:rtype: :class:`collections.OrderedDict`
"""
result = OrderedDict()
# Get the columns
for key in self.__mapper__.c.keys():
if isinstance(getattr(self, key), datetime):
result["x"] = getattr(self, key).timestamp() * 1000
result["timestamp"] = result["x"]
else:
result[key] = getattr(self, key)
return result
and inherit this class to every orm classes so that this _asdict function gets registered to every ORM class and boom.
And use jsonify anywhere
It is not so straighforward. I wrote some code to do this. I'm still working on it, and it uses the MochiKit framework. It basically translates compound objects between Python and Javascript using a proxy and registered JSON converters.
Browser side for database objects is db.js
It needs the basic Python proxy source in proxy.js.
On the Python side there is the base proxy module.
Then finally the SqlAlchemy object encoder in webserver.py.
It also depends on metadata extractors found in the models.py file.
def alc2json(row):
return dict([(col, str(getattr(row,col))) for col in row.__table__.columns.keys()])
I thought I'd play a little code golf with this one.
FYI: I am using automap_base since we have a separately designed schema according to business requirements. I just started using SQLAlchemy today but the documentation states that automap_base is an extension to declarative_base which seems to be the typical paradigm in the SQLAlchemy ORM so I believe this should work.
It does not get fancy with following foreign keys per Tjorriemorrie's solution, but it simply matches columns to values and handles Python types by str()-ing the column values. Our values consist Python datetime.time and decimal.Decimal class type results so it gets the job done.
Hope this helps any passers-by!
I know this is quite an older post. I took solution given by #SashaB and modified as per my need.
I added following things to it:
Field ignore list: A list of fields to be ignored while serializing
Field replace list: A dictionary containing field names to be replaced by values while serializing.
Removed methods and BaseQuery getting serialized
My code is as follows:
def alchemy_json_encoder(revisit_self = False, fields_to_expand = [], fields_to_ignore = [], fields_to_replace = {}):
"""
Serialize SQLAlchemy result into JSon
:param revisit_self: True / False
:param fields_to_expand: Fields which are to be expanded for including their children and all
:param fields_to_ignore: Fields to be ignored while encoding
:param fields_to_replace: Field keys to be replaced by values assigned in dictionary
:return: Json serialized SQLAlchemy object
"""
_visited_objs = []
class AlchemyEncoder(json.JSONEncoder):
def default(self, obj):
if isinstance(obj.__class__, DeclarativeMeta):
# don't re-visit self
if revisit_self:
if obj in _visited_objs:
return None
_visited_objs.append(obj)
# go through each field in this SQLalchemy class
fields = {}
for field in [x for x in dir(obj) if not x.startswith('_') and x != 'metadata' and x not in fields_to_ignore]:
val = obj.__getattribute__(field)
# is this field method defination, or an SQLalchemy object
if not hasattr(val, "__call__") and not isinstance(val, BaseQuery):
field_name = fields_to_replace[field] if field in fields_to_replace else field
# is this field another SQLalchemy object, or a list of SQLalchemy objects?
if isinstance(val.__class__, DeclarativeMeta) or \
(isinstance(val, list) and len(val) > 0 and isinstance(val[0].__class__, DeclarativeMeta)):
# unless we're expanding this field, stop here
if field not in fields_to_expand:
# not expanding this field: set it to None and continue
fields[field_name] = None
continue
fields[field_name] = val
# a json-encodable dict
return fields
return json.JSONEncoder.default(self, obj)
return AlchemyEncoder
Hope it helps someone!
Use the built-in serializer in SQLAlchemy:
from sqlalchemy.ext.serializer import loads, dumps
obj = MyAlchemyObject()
# serialize object
serialized_obj = dumps(obj)
# deserialize object
obj = loads(serialized_obj)
If you're transferring the object between sessions, remember to detach the object from the current session using session.expunge(obj).
To attach it again, just do session.add(obj).
Under Flask, this works and handles datatime fields, transforming a field of type
'time': datetime.datetime(2018, 3, 22, 15, 40) into
"time": "2018-03-22 15:40:00":
obj = {c.name: str(getattr(self, c.name)) for c in self.__table__.columns}
# This to get the JSON body
return json.dumps(obj)
# Or this to get a response object
return jsonify(obj)
following code will serialize sqlalchemy result to json.
import json
from collections import OrderedDict
def asdict(self):
result = OrderedDict()
for key in self.__mapper__.c.keys():
if getattr(self, key) is not None:
result[key] = str(getattr(self, key))
else:
result[key] = getattr(self, key)
return result
def to_array(all_vendors):
v = [ ven.asdict() for ven in all_vendors ]
return json.dumps(v)
Calling fun,
def all_products():
all_products = Products.query.all()
return to_array(all_products)
The AlchemyEncoder is wonderful but sometimes fails with Decimal values. Here is an improved encoder that solves the decimal problem -
class AlchemyEncoder(json.JSONEncoder):
# To serialize SQLalchemy objects
def default(self, obj):
if isinstance(obj.__class__, DeclarativeMeta):
model_fields = {}
for field in [x for x in dir(obj) if not x.startswith('_') and x != 'metadata']:
data = obj.__getattribute__(field)
print data
try:
json.dumps(data) # this will fail on non-encodable values, like other classes
model_fields[field] = data
except TypeError:
model_fields[field] = None
return model_fields
if isinstance(obj, Decimal):
return float(obj)
return json.JSONEncoder.default(self, obj)
When using sqlalchemy to connect to a db I this is a simple solution which is highly configurable. Use pandas.
import pandas as pd
import sqlalchemy
#sqlalchemy engine configuration
engine = sqlalchemy.create_engine....
def my_function():
#read in from sql directly into a pandas dataframe
#check the pandas documentation for additional config options
sql_DF = pd.read_sql_table("table_name", con=engine)
# "orient" is optional here but allows you to specify the json formatting you require
sql_json = sql_DF.to_json(orient="index")
return sql_json
(Tiny tweak on Sasha B's really excellent answer)
This specifically converts datetime objects to strings which in the original answer would be converted to None:
# Standard library imports
from datetime import datetime
import json
# 3rd party imports
from sqlalchemy.ext.declarative import DeclarativeMeta
class JsonEncoder(json.JSONEncoder):
def default(self, obj):
if isinstance(obj.__class__, DeclarativeMeta):
dict = {}
# Remove invalid fields and just get the column attributes
columns = [x for x in dir(obj) if not x.startswith("_") and x != "metadata"]
for column in columns:
value = obj.__getattribute__(column)
try:
json.dumps(value)
dict[column] = value
except TypeError:
if isinstance(value, datetime):
dict[column] = value.__str__()
else:
dict[column] = None
return dict
return json.JSONEncoder.default(self, obj)
class SqlToDict:
def __init__(self, data) -> None:
self.data = data
def to_timestamp(self, date):
if isinstance(date, datetime):
return int(datetime.timestamp(date))
else:
return date
def to_dict(self) -> List:
arr = []
for i in self.data:
keys = [*i.keys()]
values = [*i]
values = [self.to_timestamp(d) for d in values]
arr.append(dict(zip(keys, values)))
return arr
For example:
SqlToDict(data).to_dict()
Very late 2023
My implementation
def obj_to_dict(obj, remove=['_sa_instance_state'], debug=False):
result = {}
if type(obj).__name__ == "Row":
return dict(obj)
obj = obj.__dict__
for key in obj:
if key in remove:
continue
result[key] = obj[key]
if debug:
print(result)
return result
The built in serializer chokes with utf-8 cannot decode invalid start byte for some inputs. Instead, I went with:
def row_to_dict(row):
temp = row.__dict__
temp.pop('_sa_instance_state', None)
return temp
def rows_to_list(rows):
ret_rows = []
for row in rows:
ret_rows.append(row_to_dict(row))
return ret_rows
#website_blueprint.route('/api/v1/some/endpoint', methods=['GET'])
def some_api():
'''
/some_endpoint
'''
rows = rows_to_list(SomeModel.query.all())
response = app.response_class(
response=jsonplus.dumps(rows),
status=200,
mimetype='application/json'
)
return response
Maybe you can use a class like this
from sqlalchemy.ext.declarative import declared_attr
from sqlalchemy import Table
class Custom:
"""Some custom logic here!"""
__table__: Table # def for mypy
#declared_attr
def __tablename__(cls): # pylint: disable=no-self-argument
return cls.__name__ # pylint: disable= no-member
def to_dict(self) -> Dict[str, Any]:
"""Serializes only column data."""
return {c.name: getattr(self, c.name) for c in self.__table__.columns}
Base = declarative_base(cls=Custom)
class MyOwnTable(Base):
#COLUMNS!
With that all objects have the to_dict method
While using some raw sql and undefined objects, using cursor.description appeared to get what I was looking for:
with connection.cursor() as cur:
print(query)
cur.execute(query)
for item in cur.fetchall():
row = {column.name: item[i] for i, column in enumerate(cur.description)}
print(row)
This is a JSONEncoder version that preserves model column order and only keeps recursively defined column and relationship fields. It also formats most JSON unserializable types:
import json
from datetime import datetime
from decimal import Decimal
import arrow
from sqlalchemy.ext.declarative import DeclarativeMeta
class SQLAlchemyJSONEncoder(json.JSONEncoder):
"""
SQLAlchemy ORM JSON Encoder
If you have a "backref" relationship defined in your SQLAlchemy model,
this encoder raises a ValueError to stop an infinite loop.
"""
def default(self, obj):
if isinstance(obj, datetime):
return arrow.get(obj).isoformat()
elif isinstance(obj, Decimal):
return float(obj)
elif isinstance(obj, set):
return sorted(obj)
elif isinstance(obj.__class__, DeclarativeMeta):
for attribute, relationship in obj.__mapper__.relationships.items():
if isinstance(relationship.__getattribute__("backref"), tuple):
raise ValueError(
f'{obj.__class__} object has a "backref" relationship '
"that would cause an infinite loop!"
)
dictionary = {}
column_names = [column.name for column in obj.__table__.columns]
for key in column_names:
value = obj.__getattribute__(key)
if isinstance(value, datetime):
value = arrow.get(value).isoformat()
elif isinstance(value, Decimal):
value = float(value)
elif isinstance(value, set):
value = sorted(value)
dictionary[key] = value
for key in [
attribute
for attribute in dir(obj)
if not attribute.startswith("_")
and attribute != "metadata"
and attribute not in column_names
]:
value = obj.__getattribute__(key)
dictionary[key] = value
return dictionary
return super().default(obj)

Add a custom filed to fastapi responce model (serializers)

I'm following this guide from the Fastapi documentation and I have a question what if want to add a custom field when I return an object from DB. In Django I can use serializers.
My case:
I want to save an image name into DB, but before that I need save an actual file in a static folder. When I call GET /items/1 I want to return not just an image name from DB, but full URL, so I need to execute some logic on every request in order to build the URL. The question is how can I achieve that? The only thing I can think of is to add an additional DTO layer that coverts input data to Pydantic classes, so it's gonna be:
DTO class -> Pydantic -> DB
Is there more fancy way of doing that?
Code example:
schemas.py
from typing import List, Literal, Optional
from enum import Enum, IntEnum
from pydantic import BaseModel, constr, validator
class Ingredient(BaseModel):
quantity: int
quantityUnit: QuantityUnitEnum
name: constr(max_length=50)
class RecipeBase(BaseModel):
id: int = None
title: constr(max_length=50)
# image_name: str
#validator('ingredients')
def ingredients_must_have_unique_name(cls, values):
names = []
for item in values:
names.append(item.name)
if len(names) > len(set(names)):
raise ValueError('must contain unique names')
return values
class RecipeCreate(RecipeBase):
pass
class Recipe(RecipeBase):
id: int
class Config:
orm_mode = True
model.py
class Recipe(Base):
__tablename__ = "recipes"
id = Column(Integer, primary_key=True, index=True)
title = Column(String(50), index=True, nullable=False)
image_name = Column(String(50), index=True, nullable=False)
main.py
#app.post("/recipes", response_model=schemas.Recipe)
def create_recipe(recipe: schemas.RecipeCreate, db: Session = Depends(get_db)):
return repository.create_recipe(db=db, recipe=recipe)
#app.get("/recipes/{recipe_id}", response_model=schemas.Recipe)
def get_recipe(recipe_id, db: Session = Depends(get_db)):
return repository.get_recipe(db, recipe_id=recipe_id)
repository.py
def get_recipe(db: Session, recipe_id: int):
return db.query(models.Recipe).get(recipe_id)

How should I deal with '_id' in Dataclass when used in combination with MongoDB?

I have a MongoDB and would like to store each document as an instance of a python dataclass. I have a problem with the '_id'. When I create a local instance, I don't want to allocate an '_id'. However, when retrieving a document, the instance should contain the '_id'.
My approach was to set the '_id' as None. This does not work when I insert an instance into the database. The value None is passed as '_id'.
Is there a way to create a model with dataclass which allows me to store the local and retrieved data as instances of the same class?
from dataclasses import dataclass
from typing import List
from bson import ObjectId
#dataclass
class Article:
name: str
quantity: int
_id: ObjectId = None
description: str = ""
Insert local instance
import dataclasses
from pymongo import MongoClient
article = Article(name="pen", description="A writing device", quantity=100)
self.client = MongoClient()
self.db = self.client.warehouse
self.collection = self.db["articles"]
res = self.collection.insert_one(dataclasses.asdict(article)) # <-- Should not contain "_id"
Retrieve document
res = self.collection.find_one("_id": ObjectID())
article = Article(**res) # <-- Article should contain id
from dataclasses import dataclass, asdict
from typing import Optional
#dataclass
class Article:
name: str
quantity: int
id: Optional[int] = None
description: str = ''
def to_short_dict(self):
result = asdict(self)
result.pop('id')
return result
# not contain "id"
input_data = {
'name': 'pen', 'quantity': 100, 'description': 'A writing device',
}
article = Article(**input_data)
assert article.to_short_dict() == input_data
# contain "id"
input_data = {
'id': 1,
'name': 'pen', 'quantity': 100, 'description': 'A writing device',
}
article = Article(**input_data)
assert asdict(article) == input_data
I tried to come up with a more general solution:
I made an interface MongoDataclass from which all my models inherit. The interface contains a method which returns all key value pairs where the value is not None as a dict.
In my opinion this makes sense since MongoDB is schemaless and I don't want to store any None values in my documents. This also solves the problem with the '_id'.
import abc
from dataclasses import dataclass
#dataclass
class MongoDataclass(abc.ABC):
def as_json_wo_none(self):
return {key: value for key, value in dataclasses.asdict(self).items() if value is not None}

Categories

Resources