ValueError: object has no field when using monkeypatch.setattr on a pydantic.BaseModel's method

Question

I usually can patch methods of normal objects using pytest monkeypatch. However when I try with a pydantic.BaseModel it fails.

from pydantic import BaseModel
class Person(BaseModel):
 name: str
 age: int
 def intro(self) -> str:
 return f"I am {self.name}, {self.age} years old."
def test_person_intro(monkeypatch):
 p = Person(name='Joe', age=20)
 monkeypatch.setattr(p, 'intro', lambda: 'patched intro')
 assert p.intro() == 'patched intro'

Raises:

monkeypatch = <_pytest.monkeypatch.MonkeyPatch object at 0x10f0d3040>
 def test_intro(monkeypatch):
 p = Person(name='Joe', age=20)
> monkeypatch.setattr(p, 'intro', lambda: 'patched intro')
- - - - - - - - - - - - - - - - - - - - - 
> ???
E ValueError: "Person" object has no field "intro"
pydantic/main.py:422: ValueError

Answer 1

You could patch the type:

def test_person_intro(monkeypatch):
 p = Person(name='Joe', age=20)
 monkeypatch.setattr(Person, 'intro', lambda self: 'patched intro')
 assert p.intro() == 'patched intro'

Or you could set item in the instance dict:

def test_person_intro(monkeypatch):
 p = Person(name='Joe', age=20)
 monkeypatch.setitem(p.__dict__, 'intro', lambda: 'patched intro')
 assert p.intro() == 'patched intro'

The reason an instance setattr isn't working as usual is that pydantic.BaseModel overrides __setattr__ to only allow a restricted set of names. You will see the same error doing person.intro = "other" as well, it is unrelated to pytest's monkeypatch fixture.

Answer 2

To simplify from the answer by wim, If wanting to patch manually, this worked for me:

p.__dict__['myattr'] = 'myval'

After this, p.myattr also worked.