Deploy Containers Close to Your Users

This Engineering Education (EngEd) Program is supported by Section.

Instantly deploy containers globally. Section is affordable, simple and powerful.

Get Started for Free.

Strict Type Validation With Pydantic

November 21, 2021

Type validation is the process of making sure what you get is what you are expecting. If an endpoint is supposed to get an integer, you use type validation to ensure the input is an integer and not a string. It could be time-consuming to write your validation logic.

Many API frameworks have type validation out of the box, whereas lightweight programming frameworks like Flask do not. Type validation can be made more accessible with frameworks like Pydantic.

In this post, we’ll look at various Pydantic features and examples on how to use them.

Prerequisites

For the reader to follow along, they must have the following:

  • Have Python (>= 3.6) installed.
  • Install Pydantic with:
pip install pydantic

Data models

Objects in Pydantic are defined using models. A model class inherits from the BaseModel class. All of the fields and custom validation logic sit in the data model class.

A simple example is a model that defines a user profile and the fields it contains.

from pydantic import BaseModel

class Profile(BaseModel):
    firstname: str
    lastname: str
    location: str
    bio: str

In the above snippet, the name of the data model is Profile. It inherits Pydantic’s BaseModel class; it also defines some fields in a profile like firstname, lastname, and their types. They are all expected to be strings in this case.

# create a new profile with some fields
new_profile = {
    "firstname": "Tomi",
    "lastname": "Bamimore",
}

# validate the new_profile with the Profile data model
profile = Profile(**new_profile)
print(profile)

Passing the new_profile dictionary, which contains information about a new profile, into the Profile model will validate the new_profile.

If you run the code snippet above, you will get this error:

pydantic.error_wrappers.ValidationError: 2 validation errors for Profile
location
 field required (type=value_error.missing)
bio
 field required (type=value_error.missing)

Missing fields in the new_profile dictionary caused this error. Pydantic makes all the fields defined in the data model to be “required” by default.

Alternatively, you can use Optional defined by the typing module in Python’s standard library to make a field optional.

from typing import Optional
from pydantic import BaseModel

# create a pydantic data model
class Profile(BaseModel):
    firstname: str
    lastname: str
    location: Optional[str]
    bio: Optional[str]

# create a new profile with some fields
new_profile = {
    "firstname": "Tomi",
    "lastname": "Bamimore",
}

# validate the new_profile with the Profile data model
profile = Profile(**new_profile)
print(profile.json())

Output:

{"firstname": "Tomi", "lastname": "Bamimore", "location": null, "bio": null}

This time, the output is a JSON.

A JSON output is useful when working with APIs. You can also access results like attributes of an object in Python.

new_profile = {
    "firstname": "Jane",
    "lastname": "Doe",
}
profile = Profile(**new_profile)
print(profile.firstname, profile.lastname)

Output:

Jane Doe

Recursive models

When dealing with nested fields, using a data model as a data type in another model arises.

A data model can be declared as a type in another data model. You need recursive models when a field in one model has other child fields related to it. This is the concept of recursive models.

In the example below, Bio is defined as a data model. Bio is also a type in the Profile model.

from typing import Optional

from pydantic import BaseModel

class Bio(BaseModel):
    age: Optional[int]
    profession: str
    school: str

class Profile(BaseModel):
    firstname: str
    lastname: str
    location: Optional[str]
    # Model Bio is now the type of a field in the Profile model
    bio: Bio

new_profile = {
    "firstname": "Jane",
    "lastname": "Doe",
    "bio": {"age": 38, "profession": "Nurse", "school": "MIT"},
}
profile = Profile(**new_profile)
print(profile.dict())

Output:

{
    "firstname": "Jane",
    "lastname": "Doe",
    "location": None,
    "bio": {"age": 38, "profession": "Nurse", "school": "MIT"},
}

Pydantic field types

Pydantic supports an extensive range of field types from Python’s standard library. The list is limitless and can’t be exhausted in this article.

Pydantic also has custom types like PaymentCardNumber.

See how it works in the snippet below:

from pydantic import BaseModel

from pydantic.types import PaymentCardNumber, ConstrainedInt

# Define the Payment model
class Payment(BaseModel):
    # card_number is defined as a field with PaymentCardNumber type
    card_number: PaymentCardNumber

# A valid credit card number
new_card = 4238721116652766
new_payment = Payment(card_number=new_card)
print(new_payment)

Output:

card_number='4238721116652766'

Credit card numbers are validated using the Luhn algorithm. Pydantic runs the validation under the hood to validate any input to the card_number field.

If the input is invalid:

from pydantic import BaseModel, conint
from pydantic.types import PaymentCardNumber

class Payment(BaseModel):
    # constrain amount to greater than or equal to 300
    amount: conint(ge=300)
    card_number: PaymentCardNumber

new_card = 7618972848548894
new_payment = Payment(card_number=new_card, amount=300)
print(new_payment)

Output:

pydantic.error_wrappers.ValidationError: 1 validation error for Payment
card_number
 card number is not luhn valid (type=value_error.payment_card_number.luhn_check)

Custom validators

Pydantic also allows writing custom validation methods. It is useful when working with generic data types that need custom validation.

In the example below, we will validate an employee ID. It is a string with four integers, a hyphen and two alphabets.

For example, 2345-HG.

from pydantic import BaseModel, validator

class Employee(BaseModel):
    employee_id: str
    # validator decorator is used to wrap custom validation fuction for a field
    @validator("employee_id")
    def employee_id_validator(cls, value):
        splitted = value.split("-")
        if len(splitted) != 2:
            raise ValueError("Invalid id")
        if len(splitted[0]) != 4:
            raise ValueError("Invalid id")
        if len(splitted[1]) != 2:
            raise ValueError("Invalid id")
        return value

# validate a new employee id
new_employee = Employee(employee_id="234-HG")
print(new_employee)

Output:

pydantic.error_wrappers.ValidationError: 1 validation error for Employee
employee_id
 Invalid id (type=value_error)

In the code snippet above, an incorrect employee_id is passed into the model. Pydantic runs the custom validator and returns an error if any of the checks fail.

With the correct input, it runs successfully:

new_employee = Employee(employee_id="2345-HG")
print(new_employee.dict())

Output:

{'employee_id': '2345-HG'}

Your validation logic can be as complex as you want. It is a good practice to use a try/except block when working with Pydantic models.

try:
    new_employee = Employee(employee_id="2345-HG")
    print(new_employee.dict())

except:
    # Error handling logic
    print("ERROR")

Generating JSON schemas

Pydantic models can generate JSON schema complaints with the OpenAPI specifications. You can use the Field object to populate the schema with information.

Schemas help define the structure of a JSON document. Schemas are needed for generating API documentation.

from typing import Optional
from pydantic import BaseModel, Field

class Bio(BaseModel):
    age: Optional[int]
    profession: str
    school: str

class Profile(BaseModel):
    firstname: str = Field("Jane", title="Firstname", description="User's firstname")
    lastname: str = Field("Doe", title="Lastname", description="User's lastname")
    location: Optional[str] = Field(
        None, title="Location", description="User's location"
    )
    bio: Optional[Bio] = Field(None, title="Bio", description="Short bio of user")

# create a new profile
new_profile = {"firstname": "Tomi", "lastname": "Bamimore"}
profile = Profile(**new_profile)
# generate json schema
print(profile.schema_json())

Output:

{
    "title": "Profile",
    "type": "object",
    "properties": {
        "firstname": {
            "title": "Firstname",
            "description": "User's firstname",
            "default": "Jane",
            "type": "string",
        },
        "lastname": {
            "title": "Lastname",
            "description": "User's lastname",
            "default": "Doe",
            "type": "string",
        },
        "location": {
            "title": "Location",
            "description": "User's location",
            "type": "string",
        },
        "bio": {
            "title": "Bio",
            "description": "Short bio of user",
            "allOf": [{"$ref": "#/definitions/Bio"}],
        },
    },
    "definitions": {
        "Bio": {
            "title": "Bio",
            "type": "object",
            "properties": {
                "age": {"title": "Age", "type": "integer"},
                "profession": {"title": "Profession", "type": "string"},
                "school": {"title": "School", "type": "string"},
            },
            "required": ["profession", "school"],
        }
    },
}

The first argument to the Field object is the default value of the field. You should set it to None if you don’t want any default value. The other keyword arguments in the Field are for optional properties in the schema.

Conclusion

Pydantic is built in a way that allows room for flexibility. You can use Pydantic with any development framework, and it works just fine.

Frameworks like FastAPI support Pydantic out of the box. Other loosely coupled frameworks like Flask do not come bundled with Pydantic but allow room for integration.

From examples in the article, Pydantic enables you to control input types custom validation, because input validation is a significant step towards securing your application.

Happy learning!

Further reading


Peer Review Contributions by: Srishilesh P S