Check on duplicate validators breaks validators in jupyter notebooks

[amogorkon]

Check on duplicate validators breaks validators in jupyter notebooks if a cell is executed more than once.

from pydantic import BaseModel, validator
import yaml
from ipaddress import ip_address

class Config(BaseModel):
    ip: str
    name: str
    port: int
        
    @validator("ip")
    def check_ip(cls, value):
        return ip_address(value)            
            

with open("config") as f:
    d = yaml.safe_load(f)
    c = Config(**d)
    print(c)

Returns

Config ip=IPv4Address('123.0.0.200') name='asdf' port=2392
when the cell is called first and raises

ConfigError Traceback (most recent call last)
in ()
3 from ipaddress import ip_address
4
----> 5 class Config(BaseModel):
6 ip: str
7 name: str

in Config()
8 port: int
9
---> 10 @validator("ip")
11 def check_ip(cls, value):
12 return ip_address(value)

~/anaconda3/lib/python3.7/site-packages/pydantic/main.py in dec(f)
449 ref = f.module + '.' + f.qualname
450 if ref in _FUNCS:
--> 451 raise ConfigError(f'duplicate validator function "{ref}"')
452 _FUNCS.add(ref)
453 f_cls = classmethod(f)

ConfigError: duplicate validator function "main.Config.check_ip"

The second time the same cell is called.

I was somewhat surprised by this behaviour and when I checked the code, I didn't really get why this exception is raised since we're dealing with a set here, so behaviour shouldn't change.
I would suggest to replace this exception with a logger message, warning level, which won't mess with people's code and is well supported by jupyter.

[samuelcolvin]

Would be great if you could format code or output like this to make it more readable.

I see what you mean. What about an environment variable and/or some check if we're in the jupyter before doing the duplicate check?

[samuelcolvin]

I understand what you mean about warnings, but this catches a lot of potential problems for me alone, and probably for others too.

I'd rather keep it as an error but allow a workaround.

[Abraxos]

I am having this problem as well, and I do not really understand what kind of behavior this is supposed to prevent. Could you please explain and perhaps I can re-write my code better?

[samuelcolvin]

The test for this behavior demonstrates what it's there to avoid:

https://github.com/samuelcolvin/pydantic/blob/2512f25df7e423de2da87718427d4a0d8ee33855/tests/test_validators.py#L181-L191

If you setup multiple validators with the same name on a class you may get unexpected behaviour when the second validator overwrites the first. The error is there to avoid this behavour.

I often get this myself when I have two similar validators, and create the second by copy and pasting the first and forget to change the function name.

I'll try to fix for jupyter now.