Question: How to handle custom input validation errors in @field_validator #825
Comments
|
I'm running into this same issue on [{'type': 'value_error', 'loc': ('body', 'payload', 'query'), 'msg': "Value error, Couldn't parse search query.", 'ctx': {'error': ValueError("Couldn't parse search query.")}}]It seems if the |
|
Yeah I dug into it a bit as well. Either the JSON encoder or in I guess doing a Also, wouldn't it be a bit more safe to be explicit and pick out relevant keys instead of doing Source: https://github.com/vitalik/django-ninja/blob/master/ninja/operation.py#L226-L237 |
|
This is how I decided to implement this for now: import uuid
from django.contrib.auth.password_validation import validate_password
from django.core.validators import validate_email
from ninja import Router, Schema
from pydantic import constr, field_validator
from django.core.exceptions import ValidationError as DjangoValidationError
from pydantic_core import PydanticCustomError
from users import models
from django.utils.translation import gettext as _
router = Router()
class UserMeInSchema(Schema):
email: constr(to_lower=True) = None
first_name: str
last_name: str
password: str
@field_validator("email")
@classmethod
def email_validator(cls, v: str):
try:
validate_email(v)
except DjangoValidationError as e:
raise PydanticCustomError(
e.code,
e.message,
)
if models.User.objects.filter(email=v).exists():
raise PydanticCustomError(
"unique",
_("A user with this email already exists"),
)
return v
@field_validator("password")
@classmethod
def password_validator(cls, v: str):
try:
validate_password(v)
except DjangoValidationError as e:
raise PydanticCustomError(
"too_easy_password",
_("This password is too easy to guess. Please create a new one."),
{"errors": e.messages},
)
return v
class UserMeOutSchema(Schema):
id: uuid.UUID
email: str
first_name: str
last_name: str
@router.post("/register", response=UserMeOutSchema)
def register(request, data: UserMeInSchema):
user = models.User.objects.create_user(
first_name=data.first_name,
last_name=data.last_name,
email=data.email,
password=data.password,
)
return userSample output: {
"detail": [
{
"type": "unique",
"loc": [
"body",
"data",
"email"
],
"msg": "A user with this email already exists"
},
{
"type": "string_type",
"loc": [
"body",
"data",
"last_name"
],
"msg": "Input should be a valid string"
},
{
"type": "too_easy_password",
"loc": [
"body",
"data",
"password"
],
"msg": "This password is too easy to guess. Please create a new one.",
"ctx": {
"errors": [
"This password is too short. It must contain at least 8 characters.",
"This password is too common.",
"This password is entirely numeric."
]
}
}
]
}For clarity: I think a single error in The discussion about handling ValueError is still relevant though I guess. |
|
could you share all your versions (django, python, pydantic, ninja) - cannot reproduce the error on my enviroment generally speaking you throw either keep in mind that you should pass string and not list to value error : class UserMeInSchema(Schema):
email: constr(to_lower=True) = None
first_name: str
last_name: str
password: str
@field_validator("email")
@classmethod
def email_validator(cls, v: str):
try:
validate_email(v)
except DjangoValidationError as e:
raise ValueError(', '.join(e.messages)). # <--------------- join
if xxxx:
raise ValueError("A user with this email already exists")
return v
@field_validator("password")
@classmethod
def password_validator(cls, v: str):
try:
validate_password(v)
except DjangoValidationError as e:
raise ValueError(', '.join(e.messages)) # <--------------- join
return v |
Python 3.10.4 In my case, I saw the same error when passing just a string e.g. Thanks, I will try @SRautila 's suggestion for now. |
|
Python 3.11.4 |
|
@vitalik I'm not sure when this changed, but this behavior of pydantic is documented here: https://docs.pydantic.dev/latest/errors/errors/#custom-errors Note the |
This should go away when vitalik/django-ninja#825 is resolved.
|
this ctx behaviour introduced in fixed try with |
|
That was fast! Thanks! Regarding best practise, I'm till considering sticking to from django.utils.translation import gettext as _
...
@field_validator("email")
@classmethod
def email_validator(cls, v: str):
if models.User.objects.filter(email=v).exists():
raise ValueError(_("A user with this email already exists"))
return vSwedish output: {
"type": "value_error",
"loc": [
"body",
"data",
"email"
],
"msg": "Value error, En användare med den här eposten finns redan",
"ctx": {
"error": "En användare med den här eposten finns redan"
}
},So now frontend either needs to use But issue closed. Thanks! |
Hey. I'm a long-time DRF user trying out django-ninja 1.0a2 with pydantic 2.1 and I'm curious about some best practises.
What I'm doing now is writing registration. Code example:
with this I get exception
TypeError: Object of type ValueError is not JSON serializabletraceback:
I noticed I can raise
from ninja.errors import ValidationErrorinstead which works but it feels wrong since pydantic docs says "validators should either return the parsed value or raise a ValueError or AssertionError (assert statements may be used)." . Also this leads to errors having a different structure than the other pydantic errors. https://docs.pydantic.dev/2.1/usage/validators/#field-validatorsHow do I do this in the right way? Thanks!
The text was updated successfully, but these errors were encountered: