Simple creation of data classes from dictionaries.
Project description
dacite
This module simplifies creation of data classes (PEP 557) from dictionaries.
Installation
To install dacite, simply use pip
(or pipenv
):
$ pip install dacite
Requirements
Minimum Python version supported by dacite
is 3.6.
Quick start
from dataclasses import dataclass
from dacite import from_dict
@dataclass
class User:
name: str
age: int
is_active: bool
data = {
'name': 'john',
'age': 30,
'is_active': True,
}
user = from_dict(data_class=User, data=data)
assert user == User(name='john', age=30, is_active=True)
Features
Dacite supports following features:
- nested structures
- (basic) types checking
- optional fields (i.e.
typing.Optional
) - unions
- forward references
- collections
- values casting and transformation
- remapping of fields names
Motivation
Passing plain dictionaries as a data container between your functions or methods isn't a good practice. Of course you can always create your custom class instead, but this solution is an overkill if you only want to merge a few fields within a single object.
Fortunately Python has a good solution to this problem - data classes.
Thanks to @dataclass
decorator you can easily create a new custom
type with a list of given fields in a declarative manner. Data classes
support type hints by design.
However, even if you are using data classes, you have to create their
instances somehow. In many such cases, your input is a dictionary - it
can be a payload from a HTTP request or a raw data from a database. If
you want to convert those dictionaries into data classes, dacite
is
your best friend.
This library was originally created to simplify creation of type hinted data transfer objects (DTO) which can cross the boundaries in the application architecture.
It's important to mention that dacite
is not a data validation library.
There are dozens of awesome data validation projects and it doesn't make
sense to duplicate this functionality within dacite
. If you want to
validate your data first, you should combine dacite
with one of data
validation library.
Please check Use Case section for a real-life example.
Usage
Dacite is based on a single function - dacite.from_dict
. This function
takes 3 parameters:
data_class
- data class typedata
- dictionary of input dataconfig
(optional) - configuration of the creation process, instance ofdacite.Config
class
Configuration is a (data) class with following fields:
remap
flattened
prefixed
cast
transform
forward references
check_types
The examples below show all features of from_dict
function and usage
of all Config
parameters.
Use a dot-notation path if you want to point a nested data class field in
a configuration, e.g. "a.b"
. It works for all options.
Nested structures
You can pass a data with nested dictionaries and it will create a proper result.
@dataclass
class A:
x: str
y: int
@dataclass
class B:
a: A
data = {
'a': {
'x': 'test',
'y': 1,
}
}
result = from_dict(data_class=B, data=data)
assert result == B(a=A(x='test', y=1))
Optional fields
Whenever your data class has a Optional
field and you will not provide
input data for this field, it will take the None
value.
from typing import Optional
@dataclass
class A:
x: str
y: Optional[int]
data = {
'x': 'test',
}
result = from_dict(data_class=A, data=data)
assert result == A(x='test', y=None)
Unions
If your field can accept multiple types, you should use Union
. Dacite
will try to match data with provided types one by one. If none will
match, it will raise UnionMatchError
exception.
from typing import Union
@dataclass
class A:
x: str
@dataclass
class B:
y: int
@dataclass
class C:
u: Union[A, B]
data = {
'u': {
'y': 1,
},
}
result = from_dict(data_class=C, data=data)
assert result == C(u=B(y=1))
Collections
Dacite supports fields defined as collections. It works for both - basic types and data classes.
@dataclass
class A:
x: str
y: int
@dataclass
class B:
a_list: List[A]
data = {
'a_list': [
{
'x': 'test1',
'y': 1,
},
{
'x': 'test2',
'y': 2,
}
],
}
result = from_dict(data_class=B, data=data)
assert result == B(a_list=[A(x='test1', y=1), A(x='test2', y=2)])
Forward References
Definition of forward references can be passed as a {'name': Type}
mapping to
Config.forward_references
. This dict is passed to typing.get_type_hints()
as the
globalns
param when evaluating each field's type.
@dataclass
class X:
y: "Y"
@dataclass
class Y:
s: str
data = from_dict(X, {"y": {"s": "text"}}, Config(forward_references={"Y": Y}))
assert data == X(Y("text"))
Remapping
If your input data key does not match with a data class field name, you
can use Config.remap
argument to handle this case. You have to pass
dictionary with a following mapping:
{'data_class_field': 'input_field'}
@dataclass
class A:
x: str
data = {
'y': 'test',
}
result = from_dict(data_class=A, data=data, config=Config(remap={'x': 'y'}))
assert result == A(x='test')
Flattened
You often receive a flat structure which you want to convert to
something more sophisticated. In this case you can use
Config.flattened
argument. You have to pass list of flattened fields.
@dataclass
class A:
x: str
y: int
@dataclass
class B:
a: A
z: float
data = {
'x': 'test',
'y': 1,
'z': 2.0,
}
result = from_dict(data_class=B, data=data, config=Config(flattened=['a']))
assert result == B(a=A(x='test', y=1), z=2.0)
Prefixed
Sometimes your data is prefixed rather than nested. To handle this case,
you have to use Config.prefixed
argument, just pass a following
mapping: {'data_class_field': 'prefix'}
@dataclass
class A:
x: str
y: int
@dataclass
class B:
a: A
z: float
data = {
'a_x': 'test',
'a_y': 1,
'z': 2.0,
}
result = from_dict(data_class=B, data=data, config=Config(prefixed={'a': 'a_'}))
assert result == B(a=A(x='test', y=1), z=2.0)
Casting
Input values are not casted by default. If you want to use field type
information to transform input value from one type to another, you have
to pass given field name as an element of the Config.cast
argument
list.
@dataclass
class A:
x: str
data = {
'x': 1,
}
result = from_dict(data_class=A, data=data, config=Config(cast=['x']))
assert result == A(x='1')
Transformation
You can use Config.transform
argument if you want to transform the
input data into the new value. You have to pass a following mapping:
{'data_class_field': callable}
, where callable
is a
Callable[[Any], Any]
.
@dataclass
class A:
x: str
data = {
'x': 'TEST',
}
result = from_dict(data_class=A, data=data, config=Config(transform={'x': str.lower}))
assert result == A(x='test')
Types checking
There are rare cases when dacite
built-in type checker can not validate
your types (e.g. custom generic class) or you have such functionality
covered by other library and you don't want to validate your types twice.
In such case you can disable type checking with Config(check_types=False)
.
By default types checking is enabled.
T = TypeVar('T')
class X(Generic[T]):
pass
@dataclass
class A:
x: X[str]
x = X[str]()
assert from_dict(A, {'x': x}, config=Config(check_types=False)) == A(x=x)
Exceptions
Whenever something goes wrong, from_dict
will raise adequate
exception. There are a few of them:
WrongTypeError
- raised when a type of a input value does not match with a type of a data class fieldMissingValueError
- raised when you don't provide a value for a required fieldInvalidConfigurationError
- raised when you provide a invalid value (a field name or a input data key) for a configurationUnionMatchError
- raised when provided data does not match any type ofUnion
ForwardReferenceError
- raised when undefined forward reference encountered in dataclass
Development
First of all - if you want to submit your pull request, thank you very much! I really appreciate your support.
Please remember that every new feature, bug fix or improvement should be tested. 100% code coverage is a must have.
We are using a few static code analysis tools to increase the code quality
(black
, mypy
, pylint
). Please make sure that you are not generating any
errors/warnings before you submit your PR. You can find current configuration
in .travis.yml
file.
Last but not least, if you want to introduce new feature, please discuss it first within an issue.
How to start
Clone dacite
repository:
$ git clone git@github.com:konradhalas/dacite.git
Create and activate virtualenv in the way you like:
$ python3 -m venv dacite-env
$ source dacite-env/bin/activate
Install all dacite
dependencies:
$ pip install -e .[dev]
To run tests you just have to fire:
$ pytest
Use case
There are many cases when we receive "raw" data (Python dicts) as a input to our system. HTTP request payload is a very common use case. In most web frameworks we receive request data as a simple dictionary. Instead of passing this dict down to your "business" code, it's a good idea to create something more "robust".
Following example is a simple flask
app - it has single /products
endpoint.
You can use this endpoint to "create" product in your system. Our core
create_product
function expects data class as a parameter. Thanks to dacite
we can easily build such data class from POST
request payload.
from dataclasses import dataclass
from typing import List
from flask import Flask, request, Response
import dacite
app = Flask(__name__)
@dataclass
class ProductVariantData:
code: str
description: str = ''
stock: int = 0
@dataclass
class ProductData:
name: str
price: float
variants: List[ProductVariantData]
def create_product(product_data: ProductData) -> None:
pass # your business logic here
@app.route("/products", methods=['POST'])
def products():
product_data = dacite.from_dict(
data_class=ProductData,
data=request.get_json(),
)
create_product(product_data=product_data)
return Response(status=201)
What if we want to validate our data (e.g. check if code
has 6 characters) or
use something different than simple built-in types (e.g. we want to use
Decimal
as a type for price
field)? Such features are out of scope of
dacite
but we can easily combine it with one of data validation library.
Let's try with marshmallow.
First of all we have to define our data validation schemas:
from marshmallow import Schema, fields, ValidationError
def validate_code(code):
if len(code) != 6:
raise ValidationError('Code must have 6 characters.')
class ProductVariantDataSchema(Schema):
code = fields.Str(required=True, validate=validate_code)
description = fields.Str(required=False)
stock = fields.Int(required=False)
class ProductDataSchema(Schema):
name = fields.Str(required=True)
price = fields.Decimal(required=True)
variants = fields.Nested(ProductVariantDataSchema(many=True))
And use them within our endpoint:
@app.route("/products", methods=['POST'])
def products():
schema = ProductDataSchema()
result, errors = schema.load(request.get_json())
if errors:
return Response(
response=json.dumps(errors),
status=400,
mimetype='application/json',
)
product_data = dacite.from_dict(
data_class=ProductData,
data=result,
)
create_product(product_data=product_data)
return Response(status=201)
Still dacite
helps us to create data class from "raw" dict with validated data.
Authors
Created by Konrad Hałas.
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.