Unions

Unions are fundamentally different to all other types Pydantic validates - instead of requiring all fields/items/values to be valid, unions require only one member to be valid.

This leads to some nuance around how to validate unions:

• which member(s) of the union should you validate data against, and in which order?
• which errors to raise when validation fails?

Validating unions feels like adding another orthogonal dimension to the validation process.

To solve these problems, Pydantic supports three fundamental approaches to validating unions:

1. left to right mode - the simplest approach, each member of the union is tried in order and the first match is returned
2. smart mode - similar to "left to right mode" members are tried in order; however, validation will proceed past the first match to attempt to find a better match, this is the default mode for most union validation
3. discriminated unions - only one member of the union is tried, based on a discriminator

Tip

In general, we recommend using discriminated unions. They are both more performant and more predictable than untagged unions, as they allow you to control which member of the union to validate against.

For complex cases, if you're using untagged unions, it's recommended to use union_mode='left_to_right' if you need guarantees about the order of validation attempts against the union members.

If you're looking for incredibly specialized behavior, you can use a custom validator.

Union Modes

Left to Right Mode

Note

Because this mode often leads to unexpected validation results, it is not the default in Pydantic >=2, instead union_mode='smart' is the default.

With this approach, validation is attempted against each member of the union in their order they're defined, and the first successful validation is accepted as input.

If validation fails on all members, the validation error includes the errors from all members of the union.

union_mode='left_to_right' must be set as a Field parameter on union fields where you want to use it.

Union with left to right mode

from typing import Union

from pydantic import BaseModel, Field, ValidationError


class User(BaseModel):
    id: Union[str, int] = Field(union_mode='left_to_right')


print(User(id=123))
#> id=123
print(User(id='hello'))
#> id='hello'

try:
    User(id=[])
except ValidationError as e:
    print(e)
    """
    2 validation errors for User
    id.str
      Input should be a valid string [type=string_type, input_value=[], input_type=list]
    id.int
      Input should be a valid integer [type=int_type, input_value=[], input_type=list]
    """

The order of members is very important in this case, as demonstrated by tweak the above example:

Union with left to right - unexpected results

from typing import Union

from pydantic import BaseModel, Field


class User(BaseModel):
    id: Union[int, str] = Field(union_mode='left_to_right')


print(User(id=123))  # (1)
#> id=123
print(User(id='456'))  # (2)
#> id=456

1. As expected the input is validated against the int member and the result is as expected.
2. We're in lax mode and the numeric string '123' is valid as input to the first member of the union, int. Since that is tried first, we get the surprising result of id being an int instead of a str.

Smart Mode

Because of the potentially surprising results of union_mode='left_to_right', in Pydantic >=2 the default mode for Union validation is union_mode='smart'.

In this mode, pydantic attempts to select the best match for the input from the union members. The exact algorithm may change between Pydantic minor releases to allow for improvements in both performance and accuracy.

Note

We reserve the right to tweak the internal smart matching algorithm in future versions of Pydantic. If you rely on very specific matching behavior, it's recommended to use union_mode='left_to_right' or discriminated unions.

Smart Mode Algorithm

The smart mode algorithm uses two metrics to determine the best match for the input:

1. The number of valid fields set (relevant for models, dataclasses, and typed dicts)
2. The exactness of the match (relevant for all types)

Number of valid fields set

Note

This metric was introduced in Pydantic v2.8.0. Prior to this version, only exactness was used to determine the best match.

This metric is currently only relevant for models, dataclasses, and typed dicts.

The greater the number of valid fields set, the better the match. The number of fields set on nested models is also taken into account. These counts bubble up to the top-level union, where the union member with the highest count is considered the best match.

For data types where this metric is relevant, we prioritize this count over exactness. For all other types, we use solely exactness.

Exactness

For exactness, Pydantic scores a match of a union member into one of the following three groups (from highest score to lowest score):

• An exact type match, for example an int input to a float | int union validation is an exact type match for the int member
• Validation would have succeeded in strict mode
• Validation would have succeeded in lax mode

The union match which produced the highest exactness score will be considered the best match.

In smart mode, the following steps are taken to try to select the best match for the input:

BaseModel, dataclass, and TypedDictAll other data types

1. Union members are attempted left to right, with any successful matches scored into one of the three exactness categories described above, with the valid fields set count also tallied.
2. After all members have been evaluated, the member with the highest "valid fields set" count is returned.
3. If there's a tie for the highest "valid fields set" count, the exactness score is used as a tiebreaker, and the member with the highest exactness score is returned.
4. If validation failed on all the members, return all the errors.

1. Union members are attempted left to right, with any successful matches scored into one of the three exactness categories described above.
   • If validation succeeds with an exact type match, that member is returned immediately and following members will not be attempted.
2. If validation succeeded on at least one member as a "strict" match, the leftmost of those "strict" matches is returned.
3. If validation succeeded on at least one member in "lax" mode, the leftmost match is returned.
4. Validation failed on all the members, return all the errors.

from typing import Union
from uuid import UUID

from pydantic import BaseModel


class User(BaseModel):
    id: Union[int, str, UUID]
    name: str


user_01 = User(id=123, name='John Doe')
print(user_01)
#> id=123 name='John Doe'
print(user_01.id)
#> 123
user_02 = User(id='1234', name='John Doe')
print(user_02)
#> id='1234' name='John Doe'
print(user_02.id)
#> 1234
user_03_uuid = UUID('cf57432e-809e-4353-adbd-9d5c0d733868')
user_03 = User(id=user_03_uuid, name='John Doe')
print(user_03)
#> id=UUID('cf57432e-809e-4353-adbd-9d5c0d733868') name='John Doe'
print(user_03.id)
#> cf57432e-809e-4353-adbd-9d5c0d733868
print(user_03_uuid.int)
#> 275603287559914445491632874575877060712

Tip

The type Optional[x] is a shorthand for Union[x, None].

See more details in Required fields.

Discriminated Unions

Discriminated unions are sometimes referred to as "Tagged Unions".

We can use discriminated unions to more efficiently validate Union types, by choosing which member of the union to validate against.

This makes validation more efficient and also avoids a proliferation of errors when validation fails.

Adding discriminator to unions also means the generated JSON schema implements the associated OpenAPI specification.

Discriminated Unions with str discriminators

Frequently, in the case of a Union with multiple models, there is a common field to all members of the union that can be used to distinguish which union case the data should be validated against; this is referred to as the "discriminator" in OpenAPI.

To validate models based on that information you can set the same field - let's call it my_discriminator - in each of the models with a discriminated value, which is one (or many) Literal value(s). For your Union, you can set the discriminator in its value: Field(discriminator='my_discriminator').

from typing import Literal, Union

from pydantic import BaseModel, Field, ValidationError


class Cat(BaseModel):
    pet_type: Literal['cat']
    meows: int


class Dog(BaseModel):
    pet_type: Literal['dog']
    barks: float


class Lizard(BaseModel):
    pet_type: Literal['reptile', 'lizard']
    scales: bool


class Model(BaseModel):
    pet: Union[Cat, Dog, Lizard] = Field(..., discriminator='pet_type')
    n: int


print(Model(pet={'pet_type': 'dog', 'barks': 3.14}, n=1))
#> pet=Dog(pet_type='dog', barks=3.14) n=1
try:
    Model(pet={'pet_type': 'dog'}, n=1)
except ValidationError as e:
    print(e)
    """
    1 validation error for Model
    pet.dog.barks
      Field required [type=missing, input_value={'pet_type': 'dog'}, input_type=dict]
    """

Discriminated Unions with callable Discriminator

API Documentation

pydantic.types.Discriminator

In the case of a Union with multiple models, sometimes there isn't a single uniform field across all models that you can use as a discriminator. This is the perfect use case for a callable Discriminator.

Tip

When you're designing callable discriminators, remember that you might have to account for both dict and model type inputs. This pattern is similar to that of mode='before' validators, where you have to anticipate various forms of input.

But wait! You ask, I only anticipate passing in dict types, why do I need to account for models? Pydantic uses callable discriminators for serialization as well, at which point the input to your callable is very likely to be a model instance.

In the following examples, you'll see that the callable discriminators are designed to handle both dict and model inputs. If you don't follow this practice, it's likely that you'll, in the best case, get warnings during serialization, and in the worst case, get runtime errors during validation.

from typing import Any, Literal, Union

from typing_extensions import Annotated

from pydantic import BaseModel, Discriminator, Tag


class Pie(BaseModel):
    time_to_cook: int
    num_ingredients: int


class ApplePie(Pie):
    fruit: Literal['apple'] = 'apple'


class PumpkinPie(Pie):
    filling: Literal['pumpkin'] = 'pumpkin'


def get_discriminator_value(v: Any) -> str:
    if isinstance(v, dict):
        return v.get('fruit', v.get('filling'))
    return getattr(v, 'fruit', getattr(v, 'filling', None))


class ThanksgivingDinner(BaseModel):
    dessert: Annotated[
        Union[
            Annotated[ApplePie, Tag('apple')],
            Annotated[PumpkinPie, Tag('pumpkin')],
        ],
        Discriminator(get_discriminator_value),
    ]


apple_variation = ThanksgivingDinner.model_validate(
    {'dessert': {'fruit': 'apple', 'time_to_cook': 60, 'num_ingredients': 8}}
)
print(repr(apple_variation))
"""
ThanksgivingDinner(dessert=ApplePie(time_to_cook=60, num_ingredients=8, fruit='apple'))
"""

pumpkin_variation = ThanksgivingDinner.model_validate(
    {
        'dessert': {
            'filling': 'pumpkin',
            'time_to_cook': 40,
            'num_ingredients': 6,
        }
    }
)
print(repr(pumpkin_variation))
"""
ThanksgivingDinner(dessert=PumpkinPie(time_to_cook=40, num_ingredients=6, filling='pumpkin'))
"""

Discriminators can also be used to validate Union types with combinations of models and primitive types.

For example:

from typing import Any, Union

from typing_extensions import Annotated

from pydantic import BaseModel, Discriminator, Tag, ValidationError


def model_x_discriminator(v: Any) -> str:
    if isinstance(v, int):
        return 'int'
    if isinstance(v, (dict, BaseModel)):
        return 'model'
    else:
        # return None if the discriminator value isn't found
        return None


class SpecialValue(BaseModel):
    value: int


class DiscriminatedModel(BaseModel):
    value: Annotated[
        Union[
            Annotated[int, Tag('int')],
            Annotated['SpecialValue', Tag('model')],
        ],
        Discriminator(model_x_discriminator),
    ]


model_data = {'value': {'value': 1}}
m = DiscriminatedModel.model_validate(model_data)
print(m)
#> value=SpecialValue(value=1)

int_data = {'value': 123}
m = DiscriminatedModel.model_validate(int_data)
print(m)
#> value=123

try:
    DiscriminatedModel.model_validate({'value': 'not an int or a model'})
except ValidationError as e:
    print(e)  # (1)!
    """
    1 validation error for DiscriminatedModel
    value
      Unable to extract tag using discriminator model_x_discriminator() [type=union_tag_not_found, input_value='not an int or a model', input_type=str]
    """

1. Notice the callable discriminator function returns None if a discriminator value is not found. When None is returned, this union_tag_not_found error is raised.

Note

Using the typing.Annotated fields syntax can be handy to regroup the Union and discriminator information. See the next example for more details.

There are a few ways to set a discriminator for a field, all varying slightly in syntax.

For str discriminators:

some_field: Union[...] = Field(discriminator='my_discriminator'
some_field: Annotated[Union[...], Field(discriminator='my_discriminator')]

For callable Discriminators:

some_field: Union[...] = Field(discriminator=Discriminator(...))
some_field: Annotated[Union[...], Discriminator(...)]
some_field: Annotated[Union[...], Field(discriminator=Discriminator(...))]

Warning

Discriminated unions cannot be used with only a single variant, such as Union[Cat].

Python changes Union[T] into T at interpretation time, so it is not possible for pydantic to distinguish fields of Union[T] from T.

Nested Discriminated Unions

Only one discriminator can be set for a field but sometimes you want to combine multiple discriminators. You can do it by creating nested Annotated types, e.g.:

from typing import Literal, Union

from typing_extensions import Annotated

from pydantic import BaseModel, Field, ValidationError


class BlackCat(BaseModel):
    pet_type: Literal['cat']
    color: Literal['black']
    black_name: str


class WhiteCat(BaseModel):
    pet_type: Literal['cat']
    color: Literal['white']
    white_name: str


Cat = Annotated[Union[BlackCat, WhiteCat], Field(discriminator='color')]


class Dog(BaseModel):
    pet_type: Literal['dog']
    name: str


Pet = Annotated[Union[Cat, Dog], Field(discriminator='pet_type')]


class Model(BaseModel):
    pet: Pet
    n: int


m = Model(pet={'pet_type': 'cat', 'color': 'black', 'black_name': 'felix'}, n=1)
print(m)
#> pet=BlackCat(pet_type='cat', color='black', black_name='felix') n=1
try:
    Model(pet={'pet_type': 'cat', 'color': 'red'}, n='1')
except ValidationError as e:
    print(e)
    """
    1 validation error for Model
    pet.cat
      Input tag 'red' found using 'color' does not match any of the expected tags: 'black', 'white' [type=union_tag_invalid, input_value={'pet_type': 'cat', 'color': 'red'}, input_type=dict]
    """
try:
    Model(pet={'pet_type': 'cat', 'color': 'black'}, n='1')
except ValidationError as e:
    print(e)
    """
    1 validation error for Model
    pet.cat.black.black_name
      Field required [type=missing, input_value={'pet_type': 'cat', 'color': 'black'}, input_type=dict]
    """

Tip

If you want to validate data against a union, and solely a union, you can use pydantic's TypeAdapter construct instead of inheriting from the standard BaseModel.

In the context of the previous example, we have the following:

type_adapter = TypeAdapter(Pet)

pet = type_adapter.validate_python(
    {'pet_type': 'cat', 'color': 'black', 'black_name': 'felix'}
)
print(repr(pet))
#> BlackCat(pet_type='cat', color='black', black_name='felix')

Union Validation Errors

When Union validation fails, error messages can be quite verbose, as they will produce validation errors for each case in the union. This is especially noticeable when dealing with recursive models, where reasons may be generated at each level of recursion. Discriminated unions help to simplify error messages in this case, as validation errors are only produced for the case with a matching discriminator value.

You can also customize the error type, message, and context for a Discriminator by passing these specifications as parameters to the Discriminator constructor, as seen in the example below.

from typing import Union

from typing_extensions import Annotated

from pydantic import BaseModel, Discriminator, Tag, ValidationError


# Errors are quite verbose with a normal Union:
class Model(BaseModel):
    x: Union[str, 'Model']


try:
    Model.model_validate({'x': {'x': {'x': 1}}})
except ValidationError as e:
    print(e)
    """
    4 validation errors for Model
    x.str
      Input should be a valid string [type=string_type, input_value={'x': {'x': 1}}, input_type=dict]
    x.Model.x.str
      Input should be a valid string [type=string_type, input_value={'x': 1}, input_type=dict]
    x.Model.x.Model.x.str
      Input should be a valid string [type=string_type, input_value=1, input_type=int]
    x.Model.x.Model.x.Model
      Input should be a valid dictionary or instance of Model [type=model_type, input_value=1, input_type=int]
    """

try:
    Model.model_validate({'x': {'x': {'x': {}}}})
except ValidationError as e:
    print(e)
    """
    4 validation errors for Model
    x.str
      Input should be a valid string [type=string_type, input_value={'x': {'x': {}}}, input_type=dict]
    x.Model.x.str
      Input should be a valid string [type=string_type, input_value={'x': {}}, input_type=dict]
    x.Model.x.Model.x.str
      Input should be a valid string [type=string_type, input_value={}, input_type=dict]
    x.Model.x.Model.x.Model.x
      Field required [type=missing, input_value={}, input_type=dict]
    """


# Errors are much simpler with a discriminated union:
def model_x_discriminator(v):
    if isinstance(v, str):
        return 'str'
    if isinstance(v, (dict, BaseModel)):
        return 'model'


class DiscriminatedModel(BaseModel):
    x: Annotated[
        Union[
            Annotated[str, Tag('str')],
            Annotated['DiscriminatedModel', Tag('model')],
        ],
        Discriminator(
            model_x_discriminator,
            custom_error_type='invalid_union_member',  # (1)!
            custom_error_message='Invalid union member',  # (2)!
            custom_error_context={'discriminator': 'str_or_model'},  # (3)!
        ),
    ]


try:
    DiscriminatedModel.model_validate({'x': {'x': {'x': 1}}})
except ValidationError as e:
    print(e)
    """
    1 validation error for DiscriminatedModel
    x.model.x.model.x
      Invalid union member [type=invalid_union_member, input_value=1, input_type=int]
    """

try:
    DiscriminatedModel.model_validate({'x': {'x': {'x': {}}}})
except ValidationError as e:
    print(e)
    """
    1 validation error for DiscriminatedModel
    x.model.x.model.x.model.x
      Field required [type=missing, input_value={}, input_type=dict]
    """

# The data is still handled properly when valid:
data = {'x': {'x': {'x': 'a'}}}
m = DiscriminatedModel.model_validate(data)
print(m.model_dump())
#> {'x': {'x': {'x': 'a'}}}

1. custom_error_type is the type attribute of the ValidationError raised when validation fails.
2. custom_error_message is the msg attribute of the ValidationError raised when validation fails.
3. custom_error_context is the ctx attribute of the ValidationError raised when validation fails.

You can also simplify error messages by labeling each case with a Tag. This is especially useful when you have complex types like those in this example:

from typing import Dict, List, Union

from typing_extensions import Annotated

from pydantic import AfterValidator, Tag, TypeAdapter, ValidationError

DoubledList = Annotated[List[int], AfterValidator(lambda x: x * 2)]
StringsMap = Dict[str, str]


# Not using any `Tag`s for each union case, the errors are not so nice to look at
adapter = TypeAdapter(Union[DoubledList, StringsMap])

try:
    adapter.validate_python(['a'])
except ValidationError as exc_info:
    print(exc_info)
    """
    2 validation errors for union[function-after[<lambda>(), list[int]],dict[str,str]]
    function-after[<lambda>(), list[int]].0
      Input should be a valid integer, unable to parse string as an integer [type=int_parsing, input_value='a', input_type=str]
    dict[str,str]
      Input should be a valid dictionary [type=dict_type, input_value=['a'], input_type=list]
    """

tag_adapter = TypeAdapter(
    Union[
        Annotated[DoubledList, Tag('DoubledList')],
        Annotated[StringsMap, Tag('StringsMap')],
    ]
)

try:
    tag_adapter.validate_python(['a'])
except ValidationError as exc_info:
    print(exc_info)
    """
    2 validation errors for union[DoubledList,StringsMap]
    DoubledList.0
      Input should be a valid integer, unable to parse string as an integer [type=int_parsing, input_value='a', input_type=str]
    StringsMap
      Input should be a valid dictionary [type=dict_type, input_value=['a'], input_type=list]
    """