Migration Guide

Pydantic V2 introduces a number of changes to the API, including some breaking changes.

This page provides a guide highlighting the most important changes to help you migrate your code from Pydantic V1 to Pydantic V2.

Install Pydantic V2¶

Pydantic V2 is now the current production release of Pydantic. You can install Pydantic V2 from PyPI:

pip install -U pydantic

If you encounter any issues, please create an issue in GitHub using the bug V2 label. This will help us to actively monitor and track errors, and to continue to improve the library's performance.

If you need to use latest Pydantic V1 for any reason, see the Continue using Pydantic V1 features section below for details on installation and imports from pydantic.v1.

Code transformation tool¶

We have created a tool to help you migrate your code. This tool is still in beta, but we hope it will help you to migrate your code more quickly.

You can install the tool from PyPI:

pip install bump-pydantic

The usage is simple. If your project structure is:

* repo_folder
    * my_package
        * <python source files> ...

Then you'll want to do:

cd /path/to/repo_folder
bump-pydantic my_package

See more about it on the Bump Pydantic repository.

Continue using Pydantic V1 features¶

Pydantic V1 is still available when you need it, though we recommend migrating to Pydantic V2 for its improvements and new features.

If you need to use latest Pydantic V1, you can install it with:

pip install "pydantic==1.*"

The Pydantic V2 package also continues to provide access to the Pydantic V1 API by importing through pydantic.v1.

For example, you can use the BaseModel class from Pydantic V1 instead of the Pydantic V2 pydantic.BaseModel class:

from pydantic.v1 import BaseModel

You can also import functions that have been removed from Pydantic V2, such as lenient_isinstance:

from pydantic.v1.utils import lenient_isinstance

Pydantic V1 documentation is available at https://docs.pydantic.dev/1.10/.

Using Pydantic v1 features in a v1/v2 environment¶

As of pydantic>=1.10.17, the pydantic.v1 namespace can be used within V1. This makes it easier to migrate to V2, which also supports the pydantic.v1 namespace. In order to unpin a pydantic<2 dependency and continue using V1 features, take the following steps:

Replace pydantic<2 with pydantic>=1.10.17
Find and replace all occurrences of:

from pydantic.<module> import <object>

with:

from pydantic.v1.<module> import <object>

Here's how you can import pydantic's v1 features based on your version of pydantic:

pydantic>=1.10.17,<3pydantic<3

As of v1.10.17 the .v1 namespace is available in V1, allowing imports as below:

from pydantic.v1.fields import ModelField

All versions of Pydantic V1 and V2 support the following import pattern, in case you don't know which version of Pydantic you are using:

try:
    from pydantic.v1.fields import ModelField
except ImportError:
    from pydantic.fields import ModelField

Note

When importing modules using pydantic>=1.10.17,<2 with the .v1 namespace these modules will not be the same module as the same import without the .v1 namespace, but the symbols imported will be. For example pydantic.v1.fields is not pydantic.fields but pydantic.v1.fields.ModelField is pydantic.fields.ModelField. Luckily, this is not likely to be relevant in the vast majority of cases. It's just an unfortunate consequence of providing a smoother migration experience.

Migration guide¶

The following sections provide details on the most important changes in Pydantic V2.

Changes to `pydantic.BaseModel`¶

Various method names have been changed; all non-deprecated BaseModel methods now have names matching either the format model_.* or __.*pydantic.*__. Where possible, we have retained the deprecated methods with their old names to help ease migration, but calling them will emit DeprecationWarnings.

Pydantic V1	Pydantic V2
`__fields__`	`model_fields`
`__private_attributes__`	`__pydantic_private__`
`__validators__`	`__pydantic_validator__`
`construct()`	`model_construct()`
`copy()`	`model_copy()`
`dict()`	`model_dump()`
`json_schema()`	`model_json_schema()`
`json()`	`model_dump_json()`
`parse_obj()`	`model_validate()`
`update_forward_refs()`	`model_rebuild()`

Some of the built-in data-loading functionality has been slated for removal. In particular, parse_raw and parse_file are now deprecated. In Pydantic V2, model_validate_json works like parse_raw. Otherwise, you should load the data and then pass it to model_validate.
The from_orm method has been deprecated; you can now just use model_validate (equivalent to parse_obj from Pydantic V1) to achieve something similar, as long as you've set from_attributes=True in the model config.
The __eq__ method has changed for models.
- Models can only be equal to other BaseModel instances.
- For two model instances to be equal, they must have the same:
  - Type (or, in the case of generic models, non-parametrized generic origin type)
  - Field values
  - Extra values (only relevant when model_config['extra'] == 'allow')
  - Private attribute values; models with different values of private attributes are no longer equal.
  - Models are no longer equal to the dicts containing their data.
  - Non-generic models of different types are never equal.
  - Generic models with different origin types are never equal. We don't require exact type equality so that, for example, instances of MyGenericModel[Any] could be equal to instances of MyGenericModel[int].
We have replaced the use of the __root__ field to specify a "custom root model" with a new type called RootModel which is intended to replace the functionality of using a field called __root__ in Pydantic V1. Note, RootModel types no longer support the arbitrary_types_allowed config setting. See this issue comment for an explanation.
We have significantly expanded Pydantic's capabilities related to customizing serialization. In particular, we have added the @field_serializer, @model_serializer, and @computed_field decorators, which each address various shortcomings from Pydantic V1.
- See Custom serializers for the usage docs of these new decorators.
- Due to performance overhead and implementation complexity, we have now deprecated support for specifying json_encoders in the model config. This functionality was originally added for the purpose of achieving custom serialization logic, and we think the new serialization decorators are a better choice in most common scenarios.
We have changed the behavior related to serializing subclasses of models when they occur as nested fields in a parent model. In V1, we would always include all fields from the subclass instance. In V2, when we dump a model, we only include the fields that are defined on the annotated type of the field. This helps prevent some accidental security bugs. You can read more about this (including how to opt out of this behavior) in the Subclass instances for fields of BaseModel, dataclasses, TypedDict section of the model exporting docs.
GetterDict has been removed as it was just an implementation detail of orm_mode, which has been removed.
In many cases, arguments passed to the constructor will be copied in order to perform validation and, where necessary, coercion. This is notable in the case of passing mutable objects as arguments to a constructor. You can see an example + more detail here.
The .json() method is deprecated, and attempting to use this deprecated method with arguments such as indent or ensure_ascii may lead to confusing errors. For best results, switch to V2's equivalent, model_dump_json(). If you'd still like to use said arguments, you can use this workaround.
JSON serialization of non-string key values is generally done with str(key), leading to some changes in behavior such as the following:

from typing import Dict, Optional

from pydantic import BaseModel as V2BaseModel
from pydantic.v1 import BaseModel as V1BaseModel


class V1Model(V1BaseModel):
    a: Dict[Optional[str], int]


class V2Model(V2BaseModel):
    a: Dict[Optional[str], int]


v1_model = V1Model(a={None: 123})
v2_model = V2Model(a={None: 123})

# V1
print(v1_model.json())
#> {"a": {"null": 123}}

# V2
print(v2_model.model_dump_json())
#> {"a":{"None":123}}

model_dump_json() results are compacted in order to save space, and don't always exactly match that of json.dumps() output. That being said, you can easily modify the separators used in json.dumps() results in order to align the two outputs:

Python 3.8 and abovePython 3.9 and above

import json
from typing import List

from pydantic import BaseModel as V2BaseModel
from pydantic.v1 import BaseModel as V1BaseModel


class V1Model(V1BaseModel):
    a: List[str]


class V2Model(V2BaseModel):
    a: List[str]


v1_model = V1Model(a=['fancy', 'sushi'])
v2_model = V2Model(a=['fancy', 'sushi'])

# V1
print(v1_model.json())
#> {"a": ["fancy", "sushi"]}

# V2
print(v2_model.model_dump_json())
#> {"a":["fancy","sushi"]}

# Plain json.dumps
print(json.dumps(v2_model.model_dump()))
#> {"a": ["fancy", "sushi"]}

# Modified json.dumps
print(json.dumps(v2_model.model_dump(), separators=(',', ':')))
#> {"a":["fancy","sushi"]}

import json

from pydantic import BaseModel as V2BaseModel
from pydantic.v1 import BaseModel as V1BaseModel


class V1Model(V1BaseModel):
    a: list[str]


class V2Model(V2BaseModel):
    a: list[str]


v1_model = V1Model(a=['fancy', 'sushi'])
v2_model = V2Model(a=['fancy', 'sushi'])

# V1
print(v1_model.json())
#> {"a": ["fancy", "sushi"]}

# V2
print(v2_model.model_dump_json())
#> {"a":["fancy","sushi"]}

# Plain json.dumps
print(json.dumps(v2_model.model_dump()))
#> {"a": ["fancy", "sushi"]}

# Modified json.dumps
print(json.dumps(v2_model.model_dump(), separators=(',', ':')))
#> {"a":["fancy","sushi"]}

Changes to `pydantic.generics.GenericModel`¶

The pydantic.generics.GenericModel class is no longer necessary, and has been removed. Instead, you can now create generic BaseModel subclasses by just adding Generic as a parent class on a BaseModel subclass directly. This looks like class MyGenericModel(BaseModel, Generic[T]): ....

Mixing of V1 and V2 models is not supported which means that type parameters of such generic BaseModel (V2) cannot be V1 models.

While it may not raise an error, we strongly advise against using parametrized generics in isinstance checks.

For example, you should not do isinstance(my_model, MyGenericModel[int]). However, it is fine to do isinstance(my_model, MyGenericModel). (Note that for standard generics, it would raise an error to do a subclass check with a parameterized generic.)
If you need to perform isinstance checks against parametrized generics, you can do this by subclassing the parametrized generic class. This looks like class MyIntModel(MyGenericModel[int]): ... and isinstance(my_model, MyIntModel).

Find more information in the Generic models documentation.

Changes to `pydantic.Field`¶

Field no longer supports arbitrary keyword arguments to be added to the JSON schema. Instead, any extra data you want to add to the JSON schema should be passed as a dictionary to the json_schema_extra keyword argument.

In Pydantic V1, the alias property returns the field's name when no alias is set. In Pydantic V2, this behavior has changed to return None when no alias is set.

The following properties have been removed from or changed in Field:

const
min_items (use min_length instead)
max_items (use max_length instead)
unique_items
allow_mutation (use frozen instead)
regex (use pattern instead)
final (use the typing.Final type hint instead)

Field constraints are no longer automatically pushed down to the parameters of generics. For example, you can no longer validate every element of a list matches a regex by providing my_list: list[str] = Field(pattern=".*"). Instead, use typing.Annotated to provide an annotation on the str itself: my_list: list[Annotated[str, Field(pattern=".*")]]

[TODO: Need to document any other backwards-incompatible changes to pydantic.Field]

Changes to dataclasses¶

Pydantic dataclasses continue to be useful for enabling the data validation on standard dataclasses without having to subclass BaseModel. Pydantic V2 introduces the following changes to this dataclass behavior:

When used as fields, dataclasses (Pydantic or vanilla) no longer accept tuples as validation inputs; dicts should be used instead.
The __post_init__ in Pydantic dataclasses will now be called after validation, rather than before.
- As a result, the __post_init_post_parse__ method would have become redundant, so has been removed.
Pydantic no longer supports extra='allow' for Pydantic dataclasses, where extra fields passed to the initializer would be stored as extra attributes on the dataclass. extra='ignore' is still supported for the purpose of ignoring unexpected fields while parsing data, they just won't be stored on the instance.
Pydantic dataclasses no longer have an attribute __pydantic_model__, and no longer use an underlying BaseModel to perform validation or provide other functionality.
- To perform validation, generate a JSON schema, or make use of any other functionality that may have required __pydantic_model__ in V1, you should now wrap the dataclass with a TypeAdapter (discussed more below) and make use of its methods.
In Pydantic V1, if you used a vanilla (i.e., non-Pydantic) dataclass as a field, the config of the parent type would be used as though it was the config for the dataclass itself as well. In Pydantic V2, this is no longer the case.
- In Pydantic V2, to override the config (like you would with model_config on a BaseModel), you can use the config parameter on the @dataclass decorator. See Dataclass Config for examples.

Changes to config¶

In Pydantic V2, to specify config on a model, you should set a class attribute called model_config to be a dict with the key/value pairs you want to be used as the config. The Pydantic V1 behavior to create a class called Config in the namespace of the parent BaseModel subclass is now deprecated.
When subclassing a model, the model_config attribute is inherited. This is helpful in the case where you'd like to use a base class with a given configuration for many models. Note, if you inherit from multiple BaseModel subclasses, like class MyModel(Model1, Model2), the non-default settings in the model_config attribute from the two models will be merged, and for any settings defined in both, those from Model2 will override those from Model1.
The following config settings have been removed:
- allow_mutation — this has been removed. You should be able to use frozen equivalently (inverse of current use).
- error_msg_templates
- fields — this was the source of various bugs, so has been removed. You should be able to use Annotated on fields to modify them as desired.
- getter_dict — orm_mode has been removed, and this implementation detail is no longer necessary.
- smart_union.
- underscore_attrs_are_private — the Pydantic V2 behavior is now the same as if this was always set to True in Pydantic V1.
- json_loads
- json_dumps
- copy_on_model_validation
- post_init_call
The following config settings have been renamed:
- allow_population_by_field_name → populate_by_name
- anystr_lower → str_to_lower
- anystr_strip_whitespace → str_strip_whitespace
- anystr_upper → str_to_upper
- keep_untouched → ignored_types
- max_anystr_length → str_max_length
- min_anystr_length → str_min_length
- orm_mode → from_attributes
- schema_extra → json_schema_extra
- validate_all → validate_default

See the ConfigDict API reference for more details.

Changes to validators¶

`@validator` and `@root_validator` are deprecated¶

@validator has been deprecated, and should be replaced with @field_validator, which provides various new features and improvements.
- The new @field_validator decorator does not have the each_item keyword argument; validators you want to apply to items within a generic container should be added by annotating the type argument. See validators in Annotated metadata for details. This looks like List[Annotated[int, Field(ge=0)]]
- Even if you keep using the deprecated @validator decorator, you can no longer add the field or config arguments to the signature of validator functions. If you need access to these, you'll need to migrate to @field_validator — see the next section for more details.
- If you use the always=True keyword argument to a validator function, note that standard validators for the annotated type will also be applied even to defaults, not just the custom validators. For example, despite the fact that the validator below will never error, the following code raises a ValidationError:

Note

To avoid this, you can use the validate_default argument in the Field function. When set to True, it mimics the behavior of always=True in Pydantic v1. However, the new way of using validate_default is encouraged as it provides more flexibility and control.

from pydantic import BaseModel, validator


class Model(BaseModel):
    x: str = 1

    @validator('x', always=True)
    @classmethod
    def validate_x(cls, v):
        return v


Model()

@root_validator has been deprecated, and should be replaced with @model_validator, which also provides new features and improvements.
- Under some circumstances (such as assignment when model_config['validate_assignment'] is True), the @model_validator decorator will receive an instance of the model, not a dict of values. You may need to be careful to handle this case.
- Even if you keep using the deprecated @root_validator decorator, due to refactors in validation logic, you can no longer run with skip_on_failure=False (which is the default value of this keyword argument, so must be set explicitly to True).

Changes to `@validator`'s allowed signatures¶

In Pydantic V1, functions wrapped by @validator could receive keyword arguments with metadata about what was being validated. Some of these arguments have been removed from @field_validator in Pydantic V2:

config: Pydantic V2's config is now a dictionary instead of a class, which means this argument is no longer backwards compatible. If you need to access the configuration you should migrate to @field_validator and use info.config.
field: this argument used to be a ModelField object, which was a quasi-internal class that no longer exists in Pydantic V2. Most of this information can still be accessed by using the field name from info.field_name to index into cls.model_fields

from pydantic import BaseModel, ValidationInfo, field_validator


class Model(BaseModel):
    x: int

    @field_validator('x')
    def val_x(cls, v: int, info: ValidationInfo) -> int:
        assert info.config is not None
        print(info.config.get('title'))
        #> Model
        print(cls.model_fields[info.field_name].is_required())
        #> True
        return v


Model(x=1)

`TypeError` is no longer converted to `ValidationError` in validators¶

Previously, when raising a TypeError within a validator function, that error would be wrapped into a ValidationError and, in some cases (such as with FastAPI), these errors might be displayed to end users. This led to a variety of undesirable behavior — for example, calling a function with the wrong signature might produce a user-facing ValidationError.

However, in Pydantic V2, when a TypeError is raised in a validator, it is no longer converted into a ValidationError:

import pytest

from pydantic import BaseModel, field_validator  # or validator


class Model(BaseModel):
    x: int

    @field_validator('x')
    def val_x(cls, v: int) -> int:
        return str.lower(v)  # raises a TypeError


with pytest.raises(TypeError):
    Model(x=1)

This applies to all validation decorators.

Validator behavior changes¶

Pydantic V2 includes some changes to type coercion. For example:

coercing int, float, and Decimal values to strings is now optional and disabled by default, see Coerce Numbers to Strings.
iterable of pairs is no longer coerced to a dict.

See the Conversion table for details on Pydantic V2 type coercion defaults.

The `allow_reuse` keyword argument is no longer necessary¶

Previously, Pydantic tracked "reused" functions in decorators as this was a common source of mistakes. We did this by comparing the function's fully qualified name (module name + function name), which could result in false positives. The allow_reuse keyword argument could be used to disable this when it was intentional.

Our approach to detecting repeatedly defined functions has been overhauled to only error for redefinition within a single class, reducing false positives and bringing the behavior more in line with the errors that type checkers and linters would give for defining a method with the same name multiple times in a single class definition.

In nearly all cases, if you were using allow_reuse=True, you should be able to simply delete that keyword argument and have things keep working as expected.

`@validate_arguments` has been renamed to `@validate_call`¶

In Pydantic V2, the @validate_arguments decorator has been renamed to @validate_call.

In Pydantic V1, the decorated function had various attributes added, such as raw_function, and validate (which could be used to validate arguments without actually calling the decorated function). Due to limited use of these attributes, and performance-oriented changes in implementation, we have not preserved this functionality in @validate_call.

Input types are not preserved¶

In Pydantic V1 we made great efforts to preserve the types of all field inputs for generic collections when they were proper subtypes of the field annotations. For example, given the annotation Mapping[str, int] if you passed in a collection.Counter() you'd get a collection.Counter() as the value.

Supporting this behavior in V2 would have negative performance implications for the general case (we'd have to check types every time) and would add a lot of complexity to validation. Further, even in V1 this behavior was inconsistent and partially broken: it did not work for many types (str, UUID, etc.), and for generic collections it's impossible to re-build the original input correctly without a lot of special casing (consider ChainMap; rebuilding the input is necessary because we need to replace values after validation, e.g. if coercing strings to ints).

In Pydantic V2 we no longer attempt to preserve the input type in all cases; instead, we only promise that the output type will match the type annotations.

Going back to the Mapping example, we promise the output will be a valid Mapping, and in practice it will be a plain dict:

Python 3.8 and abovePython 3.9 and above

from typing import Mapping

from pydantic import TypeAdapter


class MyDict(dict):
    pass


ta = TypeAdapter(Mapping[str, int])
v = ta.validate_python(MyDict())
print(type(v))
#> <class 'dict'>

from collections.abc import Mapping

from pydantic import TypeAdapter


class MyDict(dict):
    pass


ta = TypeAdapter(Mapping[str, int])
v = ta.validate_python(MyDict())
print(type(v))
#> <class 'dict'>

If you want the output type to be a specific type, consider annotating it as such or implementing a custom validator:

Python 3.8 and abovePython 3.9 and above

from typing import Any, Mapping, TypeVar

from typing_extensions import Annotated

from pydantic import (
    TypeAdapter,
    ValidationInfo,
    ValidatorFunctionWrapHandler,
    WrapValidator,
)


def restore_input_type(
    value: Any, handler: ValidatorFunctionWrapHandler, _info: ValidationInfo
) -> Any:
    return type(value)(handler(value))


T = TypeVar('T')
PreserveType = Annotated[T, WrapValidator(restore_input_type)]


ta = TypeAdapter(PreserveType[Mapping[str, int]])


class MyDict(dict):
    pass


v = ta.validate_python(MyDict())
assert type(v) is MyDict

from typing import Any, TypeVar
from collections.abc import Mapping

from typing import Annotated

from pydantic import (
    TypeAdapter,
    ValidationInfo,
    ValidatorFunctionWrapHandler,
    WrapValidator,
)


def restore_input_type(
    value: Any, handler: ValidatorFunctionWrapHandler, _info: ValidationInfo
) -> Any:
    return type(value)(handler(value))


T = TypeVar('T')
PreserveType = Annotated[T, WrapValidator(restore_input_type)]


ta = TypeAdapter(PreserveType[Mapping[str, int]])


class MyDict(dict):
    pass


v = ta.validate_python(MyDict())
assert type(v) is MyDict

While we don't promise to preserve input types everywhere, we do preserve them for subclasses of BaseModel, and for dataclasses:

import pydantic.dataclasses
from pydantic import BaseModel


class InnerModel(BaseModel):
    x: int


class OuterModel(BaseModel):
    inner: InnerModel


class SubInnerModel(InnerModel):
    y: int


m = OuterModel(inner=SubInnerModel(x=1, y=2))
print(m)
#> inner=SubInnerModel(x=1, y=2)


@pydantic.dataclasses.dataclass
class InnerDataclass:
    x: int


@pydantic.dataclasses.dataclass
class SubInnerDataclass(InnerDataclass):
    y: int


@pydantic.dataclasses.dataclass
class OuterDataclass:
    inner: InnerDataclass


d = OuterDataclass(inner=SubInnerDataclass(x=1, y=2))
print(d)
#> OuterDataclass(inner=SubInnerDataclass(x=1, y=2))

Changes to Handling of Standard Types¶

Dicts¶

Iterables of pairs (which include empty iterables) no longer pass validation for fields of type dict.

Unions¶

While union types will still attempt validation of each choice from left to right, they now preserve the type of the input whenever possible, even if the correct type is not the first choice for which the input would pass validation. As a demonstration, consider the following example:

Python 3.8 and abovePython 3.10 and above

from typing import Union

from pydantic import BaseModel


class Model(BaseModel):
    x: Union[int, str]


print(Model(x='1'))
#> x='1'

from pydantic import BaseModel


class Model(BaseModel):
    x: int | str


print(Model(x='1'))
#> x='1'

In Pydantic V1, the printed result would have been x=1, since the value would pass validation as an int. In Pydantic V2, we recognize that the value is an instance of one of the cases and short-circuit the standard union validation.

To revert to the non-short-circuiting left-to-right behavior of V1, annotate the union with Field(union_mode='left_to_right'). See Union Mode for more details.

Required, optional, and nullable fields¶

Pydantic V2 changes some of the logic for specifying whether a field annotated as Optional is required (i.e., has no default value) or not (i.e., has a default value of None or any other value of the corresponding type), and now more closely matches the behavior of dataclasses. Similarly, fields annotated as Any no longer have a default value of None.

The following table describes the behavior of field annotations in V2:

State	Field Definition
Required, cannot be `None`	`f1: str`
Not required, cannot be `None`, is `'abc'` by default	`f2: str = 'abc'`
Required, can be `None`	`f3: Optional[str]`
Not required, can be `None`, is `None` by default	`f4: Optional[str] = None`
Not required, can be `None`, is `'abc'` by default	`f5: Optional[str] = 'abc'`
Required, can be any type (including `None`)	`f6: Any`
Not required, can be any type (including `None`)	`f7: Any = None`

Note

A field annotated as typing.Optional[T] will be required, and will allow for a value of None. It does not mean that the field has a default value of None. (This is a breaking change from V1.)

Note

Any default value if provided makes a field not required.

Here is a code example demonstrating the above:

Python 3.8 and abovePython 3.10 and above

from typing import Optional

from pydantic import BaseModel, ValidationError


class Foo(BaseModel):
    f1: str  # required, cannot be None
    f2: Optional[str]  # required, can be None - same as str | None
    f3: Optional[str] = None  # not required, can be None
    f4: str = 'Foobar'  # not required, but cannot be None


try:
    Foo(f1=None, f2=None, f4='b')
except ValidationError as e:
    print(e)
    """
    1 validation error for Foo
    f1
      Input should be a valid string [type=string_type, input_value=None, input_type=NoneType]
    """

from pydantic import BaseModel, ValidationError


class Foo(BaseModel):
    f1: str  # required, cannot be None
    f2: str | None  # required, can be None - same as str | None
    f3: str | None = None  # not required, can be None
    f4: str = 'Foobar'  # not required, but cannot be None


try:
    Foo(f1=None, f2=None, f4='b')
except ValidationError as e:
    print(e)
    """
    1 validation error for Foo
    f1
      Input should be a valid string [type=string_type, input_value=None, input_type=NoneType]
    """

Patterns / regex on strings¶

Pydantic V1 used Python's regex library. Pydantic V2 uses the Rust regex crate. This crate is not just a "Rust version of regular expressions", it's a completely different approach to regular expressions. In particular, it promises linear time searching of strings in exchange for dropping a couple of features (namely look arounds and backreferences). We believe this is a tradeoff worth making, in particular because Pydantic is used to validate untrusted input where ensuring things don't accidentally run in exponential time depending on the untrusted input is important. On the flipside, for anyone not using these features complex regex validation should be orders of magnitude faster because it's done in Rust and in linear time.

If you still want to use Python's regex library, you can use the regex_engine config setting.

Introduction of `TypeAdapter`¶

Pydantic V1 had weak support for validating or serializing non-BaseModel types.

To work with them, you had to either create a "root" model or use the utility functions in pydantic.tools (namely, parse_obj_as and schema_of).

In Pydantic V2 this is a lot easier: the TypeAdapter class lets you create an object with methods for validating, serializing, and producing JSON schemas for arbitrary types. This serves as a complete replacement for parse_obj_as and schema_of (which are now deprecated), and also covers some of the use cases of "root" models. (RootModel, discussed above, covers the others.)

Python 3.8 and abovePython 3.9 and above

from typing import List

from pydantic import TypeAdapter

adapter = TypeAdapter(List[int])
assert adapter.validate_python(['1', '2', '3']) == [1, 2, 3]
print(adapter.json_schema())
#> {'items': {'type': 'integer'}, 'type': 'array'}

from pydantic import TypeAdapter

adapter = TypeAdapter(list[int])
assert adapter.validate_python(['1', '2', '3']) == [1, 2, 3]
print(adapter.json_schema())
#> {'items': {'type': 'integer'}, 'type': 'array'}

Due to limitations of inferring generic types with common type checkers, to get proper typing in some scenarios, you may need to explicitly specify the generic parameter:

from pydantic import TypeAdapter

adapter = TypeAdapter[str | int](str | int)
...

See Type Adapter for more information.

Defining custom types¶

We have completely overhauled the way custom types are defined in pydantic.

We have exposed hooks for generating both pydantic-core and JSON schemas, allowing you to get all the performance benefits of Pydantic V2 even when using your own custom types.

We have also introduced ways to use typing.Annotated to add custom validation to your own types.

The main changes are:

__get_validators__ should be replaced with __get_pydantic_core_schema__. See Custom Data Types for more information.
__modify_schema__ becomes __get_pydantic_json_schema__. See JSON Schema Customization for more information.

Additionally, you can use typing.Annotated to modify or provide the __get_pydantic_core_schema__ and __get_pydantic_json_schema__ functions of a type by annotating it, rather than modifying the type itself. This provides a powerful and flexible mechanism for integrating third-party types with Pydantic, and in some cases may help you remove hacks from Pydantic V1 introduced to work around the limitations for custom types.

See Custom Data Types for more information.

Changes to JSON schema generation¶

We received many requests over the years to make changes to the JSON schemas that pydantic generates.

In Pydantic V2, we have tried to address many of the common requests:

The JSON schema for Optional fields now indicates that the value null is allowed.
The Decimal type is now exposed in JSON schema (and serialized) as a string.
The JSON schema no longer preserves namedtuples as namedtuples.
The JSON schema we generate by default now targets draft 2020-12 (with some OpenAPI extensions).
When they differ, you can now specify if you want the JSON schema representing the inputs to validation, or the outputs from serialization.

However, there have been many reasonable requests over the years for changes which we have not chosen to implement.

In Pydantic V1, even if you were willing to implement changes yourself, it was very difficult because the JSON schema generation process involved various recursive function calls; to override one, you'd have to copy and modify the whole implementation.

In Pydantic V2, one of our design goals was to make it easier to customize JSON schema generation. To this end, we have introduced the class GenerateJsonSchema, which implements the translation of a type's pydantic-core schema into a JSON schema. By design, this class breaks the JSON schema generation process into smaller methods that can be easily overridden in subclasses to modify the "global" approach to generating JSON schema.

The various methods that can be used to produce JSON schema (such as BaseModel.model_json_schema or TypeAdapter.json_schema) accept a keyword argument schema_generator: type[GenerateJsonSchema] = GenerateJsonSchema, and you can pass your custom subclass to these methods in order to use your own approach to generating JSON schema.

Hopefully this means that if you disagree with any of the choices we've made, or if you are reliant on behaviors in Pydantic V1 that have changed in Pydantic V2, you can use a custom schema_generator, modifying the GenerateJsonSchema class as necessary for your application.

`BaseSettings` has moved to `pydantic-settings`¶

BaseSettings, the base object for Pydantic settings management, has been moved to a separate package, pydantic-settings.

Also, the parse_env_var classmethod has been removed. So, you need to customise settings sources to have your own parsing function.

Color and Payment Card Numbers moved to `pydantic-extra-types`¶

The following special-use types have been moved to the Pydantic Extra Types package, which may be installed separately if needed.

Url and Dsn types in `pydantic.networks` no longer inherit from `str`¶

In Pydantic V1 the AnyUrl type inherited from str, and all the other Url and Dsn types inherited from these. In Pydantic V2 these types are built on two new Url and MultiHostUrl classes using Annotated.

Inheriting from str had upsides and downsides, and for V2 we decided it would be better to remove this. To use these types in APIs which expect str you'll now need to convert them (with str(url)).

Pydantic V2 uses Rust's Url crate for URL validation. Some of the URL validation differs slightly from the previous behavior in V1. One notable difference is that the new Url types append slashes to the validated version if no path is included, even if a slash is not specified in the argument to a Url type constructor. See the example below for this behavior:

from pydantic import AnyUrl

assert str(AnyUrl(url='https://google.com')) == 'https://google.com/'
assert str(AnyUrl(url='https://google.com/')) == 'https://google.com/'
assert str(AnyUrl(url='https://google.com/api')) == 'https://google.com/api'
assert str(AnyUrl(url='https://google.com/api/')) == 'https://google.com/api/'

If you still want to use the old behavior without the appended slash, take a look at this solution.

Constrained types¶

The Constrained* classes were removed, and you should replace them by Annotated[<type>, Field(...)], for example:

from pydantic import BaseModel, ConstrainedInt


class MyInt(ConstrainedInt):
    ge = 0


class Model(BaseModel):
    x: MyInt

...becomes:

Python 3.8 and abovePython 3.9 and above

from typing_extensions import Annotated

from pydantic import BaseModel, Field

MyInt = Annotated[int, Field(ge=0)]


class Model(BaseModel):
    x: MyInt

from typing import Annotated

from pydantic import BaseModel, Field

MyInt = Annotated[int, Field(ge=0)]


class Model(BaseModel):
    x: MyInt

Read more about it in the Composing types via Annotated docs.

For ConstrainedStr you can use StringConstraints instead.

Mypy Plugins¶

Pydantic V2 contains a mypy plugin in pydantic.mypy.

When using V1 features the pydantic.v1.mypy plugin might need to also be enabled.

To configure the mypy plugins:

=== mypy.ini

```ini
[mypy]
plugins = pydantic.mypy, pydantic.v1.mypy # include `.v1.mypy` if required.
```

=== pyproject.toml

```toml
[tool.mypy]
plugins = [
    "pydantic.mypy",
    "pydantic.v1.mypy",
]
```

Other changes¶

Dropped support for email-validator<2.0.0. Make sure to update using pip install -U email-validator.

Moved in Pydantic V2¶

Pydantic V1	Pydantic V2
`pydantic.BaseSettings`	`pydantic_settings.BaseSettings`
`pydantic.color`	`pydantic_extra_types.color`
`pydantic.types.PaymentCardBrand`	`pydantic_extra_types.PaymentCardBrand`
`pydantic.types.PaymentCardNumber`	`pydantic_extra_types.PaymentCardNumber`
`pydantic.utils.version_info`	`pydantic.version.version_info`
`pydantic.error_wrappers.ValidationError`	`pydantic.ValidationError`
`pydantic.utils.to_camel`	`pydantic.alias_generators.to_pascal`
`pydantic.utils.to_lower_camel`	`pydantic.alias_generators.to_camel`
`pydantic.PyObject`	`pydantic.ImportString`

Deprecated and moved in Pydantic V2¶

Pydantic V1	Pydantic V2
`pydantic.tools.schema_of`	`pydantic.deprecated.tools.schema_of`
`pydantic.tools.parse_obj_as`	`pydantic.deprecated.tools.parse_obj_as`
`pydantic.tools.schema_json_of`	`pydantic.deprecated.tools.schema_json_of`
`pydantic.json.pydantic_encoder`	`pydantic.deprecated.json.pydantic_encoder`
`pydantic.validate_arguments`	`pydantic.deprecated.decorator.validate_arguments`
`pydantic.json.custom_pydantic_encoder`	`pydantic.deprecated.json.custom_pydantic_encoder`
`pydantic.json.ENCODERS_BY_TYPE`	`pydantic.deprecated.json.ENCODERS_BY_TYPE`
`pydantic.json.timedelta_isoformat`	`pydantic.deprecated.json.timedelta_isoformat`
`pydantic.decorator.validate_arguments`	`pydantic.deprecated.decorator.validate_arguments`
`pydantic.class_validators.validator`	`pydantic.deprecated.class_validators.validator`
`pydantic.class_validators.root_validator`	`pydantic.deprecated.class_validators.root_validator`
`pydantic.utils.deep_update`	`pydantic.v1.utils.deep_update`
`pydantic.utils.GetterDict`	`pydantic.v1.utils.GetterDict`
`pydantic.utils.lenient_issubclass`	`pydantic.v1.utils.lenient_issubclass`
`pydantic.utils.lenient_isinstance`	`pydantic.v1.utils.lenient_isinstance`
`pydantic.utils.is_valid_field`	`pydantic.v1.utils.is_valid_field`
`pydantic.utils.update_not_none`	`pydantic.v1.utils.update_not_none`
`pydantic.utils.import_string`	`pydantic.v1.utils.import_string`
`pydantic.utils.Representation`	`pydantic.v1.utils.Representation`
`pydantic.utils.ROOT_KEY`	`pydantic.v1.utils.ROOT_KEY`
`pydantic.utils.smart_deepcopy`	`pydantic.v1.utils.smart_deepcopy`
`pydantic.utils.sequence_like`	`pydantic.v1.utils.sequence_like`

Removed in Pydantic V2¶

pydantic.ConstrainedBytes
pydantic.ConstrainedDate
pydantic.ConstrainedDecimal
pydantic.ConstrainedFloat
pydantic.ConstrainedFrozenSet
pydantic.ConstrainedInt
pydantic.ConstrainedList
pydantic.ConstrainedSet
pydantic.ConstrainedStr
pydantic.JsonWrapper
pydantic.NoneBytes
- This was an alias to None | bytes.
pydantic.NoneStr
- This was an alias to None | str.
pydantic.NoneStrBytes
- This was an alias to None | str | bytes.
pydantic.Protocol
pydantic.Required
pydantic.StrBytes
- This was an alias to str | bytes.
pydantic.compiled
pydantic.config.get_config
pydantic.config.inherit_config
pydantic.config.prepare_config
pydantic.create_model_from_namedtuple
pydantic.create_model_from_typeddict
pydantic.dataclasses.create_pydantic_model_from_dataclass
pydantic.dataclasses.make_dataclass_validator
pydantic.dataclasses.set_validation
pydantic.datetime_parse.parse_date
pydantic.datetime_parse.parse_time
pydantic.datetime_parse.parse_datetime
pydantic.datetime_parse.parse_duration
pydantic.error_wrappers.ErrorWrapper
pydantic.errors.AnyStrMaxLengthError
pydantic.errors.AnyStrMinLengthError
pydantic.errors.ArbitraryTypeError
pydantic.errors.BoolError
pydantic.errors.BytesError
pydantic.errors.CallableError
pydantic.errors.ClassError
pydantic.errors.ColorError
pydantic.errors.ConfigError
pydantic.errors.DataclassTypeError
pydantic.errors.DateError
pydantic.errors.DateNotInTheFutureError
pydantic.errors.DateNotInThePastError
pydantic.errors.DateTimeError
pydantic.errors.DecimalError
pydantic.errors.DecimalIsNotFiniteError
pydantic.errors.DecimalMaxDigitsError
pydantic.errors.DecimalMaxPlacesError
pydantic.errors.DecimalWholeDigitsError
pydantic.errors.DictError
pydantic.errors.DurationError
pydantic.errors.EmailError
pydantic.errors.EnumError
pydantic.errors.EnumMemberError
pydantic.errors.ExtraError
pydantic.errors.FloatError
pydantic.errors.FrozenSetError
pydantic.errors.FrozenSetMaxLengthError
pydantic.errors.FrozenSetMinLengthError
pydantic.errors.HashableError
pydantic.errors.IPv4AddressError
pydantic.errors.IPv4InterfaceError
pydantic.errors.IPv4NetworkError
pydantic.errors.IPv6AddressError
pydantic.errors.IPv6InterfaceError
pydantic.errors.IPv6NetworkError
pydantic.errors.IPvAnyAddressError
pydantic.errors.IPvAnyInterfaceError
pydantic.errors.IPvAnyNetworkError
pydantic.errors.IntEnumError
pydantic.errors.IntegerError
pydantic.errors.InvalidByteSize
pydantic.errors.InvalidByteSizeUnit
pydantic.errors.InvalidDiscriminator
pydantic.errors.InvalidLengthForBrand
pydantic.errors.JsonError
pydantic.errors.JsonTypeError
pydantic.errors.ListError
pydantic.errors.ListMaxLengthError
pydantic.errors.ListMinLengthError
pydantic.errors.ListUniqueItemsError
pydantic.errors.LuhnValidationError
pydantic.errors.MissingDiscriminator
pydantic.errors.MissingError
pydantic.errors.NoneIsAllowedError
pydantic.errors.NoneIsNotAllowedError
pydantic.errors.NotDigitError
pydantic.errors.NotNoneError
pydantic.errors.NumberNotGeError
pydantic.errors.NumberNotGtError
pydantic.errors.NumberNotLeError
pydantic.errors.NumberNotLtError
pydantic.errors.NumberNotMultipleError
pydantic.errors.PathError
pydantic.errors.PathNotADirectoryError
pydantic.errors.PathNotAFileError
pydantic.errors.PathNotExistsError
pydantic.errors.PatternError
pydantic.errors.PyObjectError
pydantic.errors.PydanticTypeError
pydantic.errors.PydanticValueError
pydantic.errors.SequenceError
pydantic.errors.SetError
pydantic.errors.SetMaxLengthError
pydantic.errors.SetMinLengthError
pydantic.errors.StrError
pydantic.errors.StrRegexError
pydantic.errors.StrictBoolError
pydantic.errors.SubclassError
pydantic.errors.TimeError
pydantic.errors.TupleError
pydantic.errors.TupleLengthError
pydantic.errors.UUIDError
pydantic.errors.UUIDVersionError
pydantic.errors.UrlError
pydantic.errors.UrlExtraError
pydantic.errors.UrlHostError
pydantic.errors.UrlHostTldError
pydantic.errors.UrlPortError
pydantic.errors.UrlSchemeError
pydantic.errors.UrlSchemePermittedError
pydantic.errors.UrlUserInfoError
pydantic.errors.WrongConstantError
pydantic.main.validate_model
pydantic.networks.stricturl
pydantic.parse_file_as
pydantic.parse_raw_as
pydantic.stricturl
pydantic.tools.parse_file_as
pydantic.tools.parse_raw_as
pydantic.types.JsonWrapper
pydantic.types.NoneBytes
pydantic.types.NoneStr
pydantic.types.NoneStrBytes
pydantic.types.PyObject
pydantic.types.StrBytes
pydantic.typing.evaluate_forwardref
pydantic.typing.AbstractSetIntStr
pydantic.typing.AnyCallable
pydantic.typing.AnyClassMethod
pydantic.typing.CallableGenerator
pydantic.typing.DictAny
pydantic.typing.DictIntStrAny
pydantic.typing.DictStrAny
pydantic.typing.IntStr
pydantic.typing.ListStr
pydantic.typing.MappingIntStrAny
pydantic.typing.NoArgAnyCallable
pydantic.typing.NoneType
pydantic.typing.ReprArgs
pydantic.typing.SetStr
pydantic.typing.StrPath
pydantic.typing.TupleGenerator
pydantic.typing.WithArgsTypes
pydantic.typing.all_literal_values
pydantic.typing.display_as_type
pydantic.typing.get_all_type_hints
pydantic.typing.get_args
pydantic.typing.get_origin
pydantic.typing.get_sub_types
pydantic.typing.is_callable_type
pydantic.typing.is_classvar
pydantic.typing.is_finalvar
pydantic.typing.is_literal_type
pydantic.typing.is_namedtuple
pydantic.typing.is_new_type
pydantic.typing.is_none_type
pydantic.typing.is_typeddict
pydantic.typing.is_typeddict_special
pydantic.typing.is_union
pydantic.typing.new_type_supertype
pydantic.typing.resolve_annotations
pydantic.typing.typing_base
pydantic.typing.update_field_forward_refs
pydantic.typing.update_model_forward_refs
pydantic.utils.ClassAttribute
pydantic.utils.DUNDER_ATTRIBUTES
pydantic.utils.PyObjectStr
pydantic.utils.ValueItems
pydantic.utils.almost_equal_floats
pydantic.utils.get_discriminator_alias_and_values
pydantic.utils.get_model
pydantic.utils.get_unique_discriminator_alias
pydantic.utils.in_ipython
pydantic.utils.is_valid_identifier
pydantic.utils.path_type
pydantic.utils.validate_field_name
pydantic.validate_model

Migration Guide

Install Pydantic V2¶

Code transformation tool¶

Continue using Pydantic V1 features¶

Using Pydantic v1 features in a v1/v2 environment¶

Migration guide¶

Changes to pydantic.BaseModel¶

Changes to pydantic.generics.GenericModel¶

Changes to pydantic.Field¶

Changes to dataclasses¶

Changes to config¶

Changes to validators¶

@validator and @root_validator are deprecated¶

Changes to @validator's allowed signatures¶

TypeError is no longer converted to ValidationError in validators¶

Validator behavior changes¶

The allow_reuse keyword argument is no longer necessary¶

@validate_arguments has been renamed to @validate_call¶

Input types are not preserved¶

Changes to Handling of Standard Types¶

Dicts¶

Unions¶

Required, optional, and nullable fields¶

Patterns / regex on strings¶

Introduction of TypeAdapter¶

Defining custom types¶

Changes to JSON schema generation¶

BaseSettings has moved to pydantic-settings¶

Color and Payment Card Numbers moved to pydantic-extra-types¶

Url and Dsn types in pydantic.networks no longer inherit from str¶

Constrained types¶

Mypy Plugins¶

Other changes¶

Moved in Pydantic V2¶

Deprecated and moved in Pydantic V2¶

Removed in Pydantic V2¶

Changes to `pydantic.BaseModel`¶

Changes to `pydantic.generics.GenericModel`¶

Changes to `pydantic.Field`¶

`@validator` and `@root_validator` are deprecated¶

Changes to `@validator`'s allowed signatures¶

`TypeError` is no longer converted to `ValidationError` in validators¶

The `allow_reuse` keyword argument is no longer necessary¶

`@validate_arguments` has been renamed to `@validate_call`¶

Introduction of `TypeAdapter`¶

`BaseSettings` has moved to `pydantic-settings`¶

Color and Payment Card Numbers moved to `pydantic-extra-types`¶

Url and Dsn types in `pydantic.networks` no longer inherit from `str`¶