Pydantic Dataclasses

Provide an enhanced dataclass that performs validation.

dataclass

dataclass(
    _cls: type[_T] | None = None,
    *,
    init: Literal[False] = False,
    repr: bool = True,
    eq: bool = True,
    order: bool = False,
    unsafe_hash: bool = False,
    frozen: bool | None = None,
    config: ConfigDict | type[object] | None = None,
    validate_on_init: bool | None = None,
    kw_only: bool = False,
    slots: bool = False
) -> (
    Callable[[type[_T]], type[PydanticDataclass]]
    | type[PydanticDataclass]
)

Usage Documentation

Dataclasses

A decorator used to create a Pydantic-enhanced dataclass, similar to the standard Python dataclass, but with added validation.

This function should be used similarly to dataclasses.dataclass.

Parameters:

Name	Type	Description	Default
_cls	type[_T] | None	The target dataclass.	None
init	Literal[False]	Included for signature compatibility with dataclasses.dataclass, and is passed through to dataclasses.dataclass when appropriate. If specified, must be set to False, as pydantic inserts its own __init__ function.	False
repr	bool	A boolean indicating whether to include the field in the __repr__ output.	True
eq	bool	Determines if a __eq__ method should be generated for the class.	True
order	bool	Determines if comparison magic methods should be generated, such as __lt__, but not __eq__.	False
unsafe_hash	bool	Determines if a __hash__ method should be included in the class, as in dataclasses.dataclass.	False
frozen	bool | None	Determines if the generated class should be a 'frozen' dataclass, which does not allow its attributes to be modified after it has been initialized. If not set, the value from the provided config argument will be used (and will default to False otherwise).	None
config	ConfigDict | type[object] | None	The Pydantic config to use for the dataclass.	None
validate_on_init	bool | None	A deprecated parameter included for backwards compatibility; in V2, all Pydantic dataclasses are validated on init.	None
kw_only	bool	Determines if __init__ method parameters must be specified by keyword only. Defaults to False.	False
slots	bool	Determines if the generated class should be a 'slots' dataclass, which does not allow the addition of new attributes after instantiation.	False

Returns:

Type	Description
Callable[[type[_T]], type[PydanticDataclass]] | type[PydanticDataclass]	A decorator that accepts a class as its argument and returns a Pydantic dataclass.

Raises:

Type	Description
AssertionError	Raised if init is not False or validate_on_init is False.

Source code in pydantic/dataclasses.py

@dataclass_transform(field_specifiers=(dataclasses.field, Field, PrivateAttr))
def dataclass(
    _cls: type[_T] | None = None,
    *,
    init: Literal[False] = False,
    repr: bool = True,
    eq: bool = True,
    order: bool = False,
    unsafe_hash: bool = False,
    frozen: bool | None = None,
    config: ConfigDict | type[object] | None = None,
    validate_on_init: bool | None = None,
    kw_only: bool = False,
    slots: bool = False,
) -> Callable[[type[_T]], type[PydanticDataclass]] | type[PydanticDataclass]:
    """Usage docs: https://docs.pydantic.dev/2.10/concepts/dataclasses/

    A decorator used to create a Pydantic-enhanced dataclass, similar to the standard Python `dataclass`,
    but with added validation.

    This function should be used similarly to `dataclasses.dataclass`.

    Args:
        _cls: The target `dataclass`.
        init: Included for signature compatibility with `dataclasses.dataclass`, and is passed through to
            `dataclasses.dataclass` when appropriate. If specified, must be set to `False`, as pydantic inserts its
            own  `__init__` function.
        repr: A boolean indicating whether to include the field in the `__repr__` output.
        eq: Determines if a `__eq__` method should be generated for the class.
        order: Determines if comparison magic methods should be generated, such as `__lt__`, but not `__eq__`.
        unsafe_hash: Determines if a `__hash__` method should be included in the class, as in `dataclasses.dataclass`.
        frozen: Determines if the generated class should be a 'frozen' `dataclass`, which does not allow its
            attributes to be modified after it has been initialized. If not set, the value from the provided `config` argument will be used (and will default to `False` otherwise).
        config: The Pydantic config to use for the `dataclass`.
        validate_on_init: A deprecated parameter included for backwards compatibility; in V2, all Pydantic dataclasses
            are validated on init.
        kw_only: Determines if `__init__` method parameters must be specified by keyword only. Defaults to `False`.
        slots: Determines if the generated class should be a 'slots' `dataclass`, which does not allow the addition of
            new attributes after instantiation.

    Returns:
        A decorator that accepts a class as its argument and returns a Pydantic `dataclass`.

    Raises:
        AssertionError: Raised if `init` is not `False` or `validate_on_init` is `False`.
    """
    assert init is False, 'pydantic.dataclasses.dataclass only supports init=False'
    assert validate_on_init is not False, 'validate_on_init=False is no longer supported'

    if sys.version_info >= (3, 10):
        kwargs = {'kw_only': kw_only, 'slots': slots}
    else:
        kwargs = {}

    def make_pydantic_fields_compatible(cls: type[Any]) -> None:
        """Make sure that stdlib `dataclasses` understands `Field` kwargs like `kw_only`
        To do that, we simply change
          `x: int = pydantic.Field(..., kw_only=True)`
        into
          `x: int = dataclasses.field(default=pydantic.Field(..., kw_only=True), kw_only=True)`
        """
        for annotation_cls in cls.__mro__:
            # In Python < 3.9, `__annotations__` might not be present if there are no fields.
            # we therefore need to use `getattr` to avoid an `AttributeError`.
            annotations = getattr(annotation_cls, '__annotations__', [])
            for field_name in annotations:
                field_value = getattr(cls, field_name, None)
                # Process only if this is an instance of `FieldInfo`.
                if not isinstance(field_value, FieldInfo):
                    continue

                # Initialize arguments for the standard `dataclasses.field`.
                field_args: dict = {'default': field_value}

                # Handle `kw_only` for Python 3.10+
                if sys.version_info >= (3, 10) and field_value.kw_only:
                    field_args['kw_only'] = True

                # Set `repr` attribute if it's explicitly specified to be not `True`.
                if field_value.repr is not True:
                    field_args['repr'] = field_value.repr

                setattr(cls, field_name, dataclasses.field(**field_args))
                # In Python 3.8, dataclasses checks cls.__dict__['__annotations__'] for annotations,
                # so we must make sure it's initialized before we add to it.
                if cls.__dict__.get('__annotations__') is None:
                    cls.__annotations__ = {}
                cls.__annotations__[field_name] = annotations[field_name]

    def create_dataclass(cls: type[Any]) -> type[PydanticDataclass]:
        """Create a Pydantic dataclass from a regular dataclass.

        Args:
            cls: The class to create the Pydantic dataclass from.

        Returns:
            A Pydantic dataclass.
        """
        from ._internal._utils import is_model_class

        if is_model_class(cls):
            raise PydanticUserError(
                f'Cannot create a Pydantic dataclass from {cls.__name__} as it is already a Pydantic model',
                code='dataclass-on-model',
            )

        original_cls = cls

        # we warn on conflicting config specifications, but only if the class doesn't have a dataclass base
        # because a dataclass base might provide a __pydantic_config__ attribute that we don't want to warn about
        has_dataclass_base = any(dataclasses.is_dataclass(base) for base in cls.__bases__)
        if not has_dataclass_base and config is not None and hasattr(cls, '__pydantic_config__'):
            warn(
                f'`config` is set via both the `dataclass` decorator and `__pydantic_config__` for dataclass {cls.__name__}. '
                f'The `config` specification from `dataclass` decorator will take priority.',
                category=UserWarning,
                stacklevel=2,
            )

        # if config is not explicitly provided, try to read it from the type
        config_dict = config if config is not None else getattr(cls, '__pydantic_config__', None)
        config_wrapper = _config.ConfigWrapper(config_dict)
        decorators = _decorators.DecoratorInfos.build(cls)

        # Keep track of the original __doc__ so that we can restore it after applying the dataclasses decorator
        # Otherwise, classes with no __doc__ will have their signature added into the JSON schema description,
        # since dataclasses.dataclass will set this as the __doc__
        original_doc = cls.__doc__

        if _pydantic_dataclasses.is_builtin_dataclass(cls):
            # Don't preserve the docstring for vanilla dataclasses, as it may include the signature
            # This matches v1 behavior, and there was an explicit test for it
            original_doc = None

            # We don't want to add validation to the existing std lib dataclass, so we will subclass it
            #   If the class is generic, we need to make sure the subclass also inherits from Generic
            #   with all the same parameters.
            bases = (cls,)
            if issubclass(cls, Generic):
                generic_base = Generic[cls.__parameters__]  # type: ignore
                bases = bases + (generic_base,)
            cls = types.new_class(cls.__name__, bases)

        make_pydantic_fields_compatible(cls)

        # Respect frozen setting from dataclass constructor and fallback to config setting if not provided
        if frozen is not None:
            frozen_ = frozen
            if config_wrapper.frozen:
                # It's not recommended to define both, as the setting from the dataclass decorator will take priority.
                warn(
                    f'`frozen` is set via both the `dataclass` decorator and `config` for dataclass {cls.__name__!r}.'
                    'This is not recommended. The `frozen` specification on `dataclass` will take priority.',
                    category=UserWarning,
                    stacklevel=2,
                )
        else:
            frozen_ = config_wrapper.frozen or False

        cls = dataclasses.dataclass(  # type: ignore[call-overload]
            cls,
            # the value of init here doesn't affect anything except that it makes it easier to generate a signature
            init=True,
            repr=repr,
            eq=eq,
            order=order,
            unsafe_hash=unsafe_hash,
            frozen=frozen_,
            **kwargs,
        )

        cls.__pydantic_decorators__ = decorators  # type: ignore
        cls.__doc__ = original_doc
        cls.__module__ = original_cls.__module__
        cls.__qualname__ = original_cls.__qualname__
        cls.__pydantic_complete__ = False  # `complete_dataclass` will set it to `True` if successful.
        # TODO `parent_namespace` is currently None, but we could do the same thing as Pydantic models:
        # fetch the parent ns using `parent_frame_namespace` (if the dataclass was defined in a function),
        # and possibly cache it (see the `__pydantic_parent_namespace__` logic for models).
        _pydantic_dataclasses.complete_dataclass(cls, config_wrapper, raise_errors=False)
        return cls

    return create_dataclass if _cls is None else create_dataclass(_cls)

rebuild_dataclass

rebuild_dataclass(
    cls: type[PydanticDataclass],
    *,
    force: bool = False,
    raise_errors: bool = True,
    _parent_namespace_depth: int = 2,
    _types_namespace: MappingNamespace | None = None
) -> bool | None

Try to rebuild the pydantic-core schema for the dataclass.

This may be necessary when one of the annotations is a ForwardRef which could not be resolved during the initial attempt to build the schema, and automatic rebuilding fails.

This is analogous to BaseModel.model_rebuild.

Parameters:

Name	Type	Description	Default
cls	type[PydanticDataclass]	The class to rebuild the pydantic-core schema for.	required
force	bool	Whether to force the rebuilding of the schema, defaults to False.	False
raise_errors	bool	Whether to raise errors, defaults to True.	True
_parent_namespace_depth	int	The depth level of the parent namespace, defaults to 2.	2
_types_namespace	MappingNamespace | None	The types namespace, defaults to None.	None

Returns:

Type	Description
bool | None	Returns None if the schema is already "complete" and rebuilding was not required.
bool | None	If rebuilding was required, returns True if rebuilding was successful, otherwise False.

Source code in pydantic/dataclasses.py

def rebuild_dataclass(
    cls: type[PydanticDataclass],
    *,
    force: bool = False,
    raise_errors: bool = True,
    _parent_namespace_depth: int = 2,
    _types_namespace: MappingNamespace | None = None,
) -> bool | None:
    """Try to rebuild the pydantic-core schema for the dataclass.

    This may be necessary when one of the annotations is a ForwardRef which could not be resolved during
    the initial attempt to build the schema, and automatic rebuilding fails.

    This is analogous to `BaseModel.model_rebuild`.

    Args:
        cls: The class to rebuild the pydantic-core schema for.
        force: Whether to force the rebuilding of the schema, defaults to `False`.
        raise_errors: Whether to raise errors, defaults to `True`.
        _parent_namespace_depth: The depth level of the parent namespace, defaults to 2.
        _types_namespace: The types namespace, defaults to `None`.

    Returns:
        Returns `None` if the schema is already "complete" and rebuilding was not required.
        If rebuilding _was_ required, returns `True` if rebuilding was successful, otherwise `False`.
    """
    if not force and cls.__pydantic_complete__:
        return None

    if '__pydantic_core_schema__' in cls.__dict__:
        delattr(cls, '__pydantic_core_schema__')  # delete cached value to ensure full rebuild happens

    if _types_namespace is not None:
        rebuild_ns = _types_namespace
    elif _parent_namespace_depth > 0:
        rebuild_ns = _typing_extra.parent_frame_namespace(parent_depth=_parent_namespace_depth, force=True) or {}
    else:
        rebuild_ns = {}

    ns_resolver = _namespace_utils.NsResolver(
        parent_namespace=rebuild_ns,
    )

    return _pydantic_dataclasses.complete_dataclass(
        cls,
        _config.ConfigWrapper(cls.__pydantic_config__, check=False),
        raise_errors=raise_errors,
        ns_resolver=ns_resolver,
        # We could provide a different config instead (with `'defer_build'` set to `True`)
        # of this explicit `_force_build` argument, but because config can come from the
        # decorator parameter or the `__pydantic_config__` attribute, `complete_dataclass`
        # will overwrite `__pydantic_config__` with the provided config above:
        _force_build=True,
    )

is_pydantic_dataclass

is_pydantic_dataclass(
    class_: type[Any],
) -> TypeGuard[type[PydanticDataclass]]

Whether a class is a pydantic dataclass.

Parameters:

Name	Type	Description	Default
class_	type[Any]	The class.	required

Returns:

Type	Description
TypeGuard[type[PydanticDataclass]]	True if the class is a pydantic dataclass, False otherwise.

Source code in pydantic/dataclasses.py

def is_pydantic_dataclass(class_: type[Any], /) -> TypeGuard[type[PydanticDataclass]]:
    """Whether a class is a pydantic dataclass.

    Args:
        class_: The class.

    Returns:
        `True` if the class is a pydantic dataclass, `False` otherwise.
    """
    try:
        return '__pydantic_validator__' in class_.__dict__ and dataclasses.is_dataclass(class_)
    except AttributeError:
        return False