I am trying to understand the way validation works in pydantic.

I create a class and three objects:

import pydantic

class TestClass(pydantic.BaseModel):
    id: int = 'text'
    name: str

obj0 = TestClass(id=1, name="test")
obj1 = TestClass(name="test")
obj2 = TestClass.model_construct(id=2)

One may see that first object is totally valid, second semivalid, and third one definitive not, that's why I omit the validation using model_construct().

Now I want to validate the objects, as pydantic suggests:

vobj0= TestClass.model_validate(obj0, strict=True)
vobj1= TestClass.model_validate(obj1, strict=True)
vobj2= TestClass.model_validate(obj2, strict=True)

The problem is that NONE of the lines rises ValidationError! I found a clumsy way around it, namely transforming objects to dicts:

vobj0= TestClass.model_validate(obj0.model_dump(), strict=True)
vobj1= TestClass.model_validate(obj1.model_dump(), strict=True)
vobj2= TestClass.model_validate(obj2.model_dump(), strict=True)

This code does raise the ValidationError in the second and third lines.

I cannot understand why does it behave in such a funny way? I expected something like obj.validate_me(), but can't find anything like this.

So, my question: if I create an object without validation (with model_construct()), what is the right way to do the validation afterwards?

I am trying to understand the way validation works in pydantic.

I create a class and three objects:

import pydantic

class TestClass(pydantic.BaseModel):
    id: int = 'text'
    name: str

obj0 = TestClass(id=1, name="test")
obj1 = TestClass(name="test")
obj2 = TestClass.model_construct(id=2)

One may see that first object is totally valid, second semivalid, and third one definitive not, that's why I omit the validation using model_construct().

Now I want to validate the objects, as pydantic suggests:

vobj0= TestClass.model_validate(obj0, strict=True)
vobj1= TestClass.model_validate(obj1, strict=True)
vobj2= TestClass.model_validate(obj2, strict=True)

The problem is that NONE of the lines rises ValidationError! I found a clumsy way around it, namely transforming objects to dicts:

vobj0= TestClass.model_validate(obj0.model_dump(), strict=True)
vobj1= TestClass.model_validate(obj1.model_dump(), strict=True)
vobj2= TestClass.model_validate(obj2.model_dump(), strict=True)

This code does raise the ValidationError in the second and third lines.

I cannot understand why does it behave in such a funny way? I expected something like obj.validate_me(), but can't find anything like this.

So, my question: if I create an object without validation (with model_construct()), what is the right way to do the validation afterwards?

• python
• validation
• pydantic

1 Answer 1

By default Pydantic does not re-validate instances, because it assumes they have been validated. This is configurable, the docs is here

In your test program, all you need is to update the TestClass definition:

class TestClass(pydantic.BaseModel, revalidate_instances='always'):
    ...

(There is also another related option you might find interesting: validate_assignment)

Field defaults are also not validated by default. (See the validate_default parameter in Field). Otherwise you would not be able to write:

id: int = 'text'