Support "Field()" on dataclasses

[mariusvniekerk]

Bug

When using pydantic.Schema in combination with optional

import pydantic

from pydantic.dataclasses import dataclass

@dataclass
class User:
    id: int
    name: pydantic.NoneStr = pydantic.Schema(None)

User(4)

---------------------------------------------------------------------------
ValidationError                           Traceback (most recent call last)
<ipython-input-24-c6a72a265df5> in <module>
      8     name: pydantic.NoneStr = pydantic.Schema(None)
      9 
---> 10 User(4)

<string> in __init__(self, id, name)

~/miniconda3/lib/python3.6/site-packages/pydantic/dataclasses.py in _pydantic_post_init(self)
     25 
     26 def _pydantic_post_init(self: 'DataclassType') -> None:
---> 27     d = validate_model(self.__pydantic_model__, self.__dict__)
     28     object.__setattr__(self, '__dict__', d)
     29     object.__setattr__(self, '__initialised__', True)

~/miniconda3/lib/python3.6/site-packages/pydantic/main.py in validate_model(model, input_data, raise_exc)
    625 
    626     if errors:
--> 627         raise ValidationError(errors)
    628     return values

ValidationError: 2 validation errors
name
  str type expected (type=type_error.str)
name
  value is not none (type=type_error.none.allowed)

When calling it with all the arguments filled it works as expected

In  [2]: User(4, None)
Out [2]: User(id=4, name=None)

When not using the pydantic.Schema this also behaves as expected

[samuelcolvin]

I think pydantic.dataclasses isn't setup to work with Schema, shouldn't be that hard to fix. PR welcome.

[Benoss]

Probably related to this, you can't export a schema with dataclasses.
I have to use BaseModel wich is fine but I am loosing the nice autocomplete and liinting from PyCharm by not using @dataclass.I tried to use both BaseModel and @DataClass but you get a recursion error.

from pydantic.schema import schema
from pydantic.dataclasses import dataclass

@dataclass
class Foo:
    a: str = None

@dataclass
class Bar:
    b: Foo
    name: str

top_level_schema = schema([Foo, Bar])
print(top_level_schema)

AttributeError: type object 'Foo' has no attribute '__fields__'

[Midnighter]

You should use the underlying pydantic model:

top_level_schema = schema([Foo.__pydantic_model__, Bar.__pydantic_model__])

it's not yet in the documentation but I found it here in a few issues.

[therefromhere]

A related (maybe the same?) issue on using Schema with dataclasses if anyone's working on it: default value isn't used, instead you get the Schema instance.

import pydantic
import pydantic.dataclasses

@pydantic.dataclasses.dataclass 
class UserDc: 
    name: pydantic.Any = pydantic.Schema(None) 

class User(pydantic.BaseModel): 
    name: pydantic.Any = pydantic.Schema(None) 

In [2]: repr(UserDc().name)                                                                                    
Out[2]: 'Schema(extra: {})'

In [3]: repr(User().name)                                                                                      
Out[3]: 'None'

[samuelcolvin]

Similar issue, again shouldn't be that hard to fix if someone wants to work on it.