Prevent duplicate items

[Chris-May-WS]

Our system has a flaky test because occasionally polyfactory creates two identical Phone models for a class, and we have a Pydantic validator that raises a value error when this happens.

class Contact(BaseModel, frozen=True, extra=Extra.forbid):
    email: NonEmptyStr
    first_name: NonEmptyStr
    last_name: NonEmptyStr
    phones: List[Phone] = Field(min_items=1)

    @validator('phones')
    @classmethod
    def no_duplicate_phones(cls, value):
        counter = Counter(p.type for p in value)
        for _, count in counter.items():
            if count > 1:
                raise ValueError('phone of same type cannot be entered twice')
        return value

What would you suggest I do to prevent this from happening?

[guacs]

You'll probably want to override the generation of a field in Phone using a classmethod to make sure it satisfies your constraint. Could you show the Phone model??

[Chris-May-WS]

Sure thing!

class PhoneType(str, Enum):
    HOME = 'home'
    WORK = 'work'
    MOBILE = 'mobile'
    FAX = 'fax'
    ALT = 'alt'

class Phone(BaseModel, frozen=True, extra=Extra.forbid):
    type: Union[PhoneType, NonEmptyStr] = Field(default=None)
    number: Optional[NonEmptyStr] = Field(default=None)
    ext: Optional[str] = Field(default=None)

[Chris-May-WS]

I've solved it by catching ValidationErrors and generating new instances, but I wonder if there's a better way:

class ContactFactory(ModelFactory[Contact]):
    __set_as_default_factory_for_type__ = True

    @classmethod
    def build(cls, **kwargs):
        instance = None
        # prevent duplicate phones
        while instance is None:
            try:
                instance = super().build(**kwargs)
            except ValidationError:
                pass
        return instance

[guacs]

Looking at your validator, it seems you need to ensure that the type is unique for each Phone instance right? If so, then I would consider doing something like the following:

class PhoneFactory(ModelFactory[Phone]):
    __set_as_default_factory_for_type__ = True
    
    @classmethod
    def type(cls) -> str:
        try:
            return cls.__faker__.unique.word()
        except UniquenessException:
            cls.__faker__.unique.clear()
            return cls.__faker__.unique.word()

Now unless you're creating a really large list of phones for a single contact, then this should ensure that you get a list of unique Phone instances for a single Contact.

Also, if you use the Faker instance provided as a fixture in pytest by the faker library, then you don't have to do the manual handling of UniquenessException since faker will call the unique.clear method before every test.

[Chris-May-WS]

Thank you very much!