pydantic-i18n
pydantic-i18n is an extension to support an i18n for the pydantic error messages.
Documentation: https://pydantic-i18n.boardpack.org
Source Code: https://github.com/boardpack/pydantic-i18n
Requirements¶
Python 3.8+
pydantic-i18n has the next dependencies:
Installation¶
$ pip install pydantic-i18n
---> 100%
First steps¶
To start to work with pydantic-i18n, you can just create a dictionary (or
create any needed translations storage and then convert it into dictionary)
and pass to the main PydanticI18n
class.
To translate messages, you need to pass result of exception.errors()
call to
the translate
method:
from pydantic import BaseModel, ValidationError
from pydantic_i18n import PydanticI18n
translations = {
"en_US": {
"Field required": "field required",
},
"de_DE": {
"Field required": "Feld erforderlich",
},
}
tr = PydanticI18n(translations)
class User(BaseModel):
name: str
try:
User()
except ValidationError as e:
translated_errors = tr.translate(e.errors(), locale="de_DE")
print(translated_errors)
# [
# {
# 'type': 'missing',
# 'loc': ('name',),
# 'msg': 'Feld erforderlich',
# 'input': {
#
# },
# 'url': 'https://errors.pydantic.dev/2.6/v/missing'
# }
# ]
In the next chapters, you will see current available loaders and how to implement your own loader.
Usage with FastAPI¶
Here is a simple example usage with FastAPI.
Create it¶
Let's create a tr.py
file:
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 |
|
13-22
: As you see, we selected the simplest variant to store translations,
you can use any that you need.
25-26
: To not include locale
query parameter into every handler, we
created a simple function get_locale
, which we will include as a global
dependency with Depends
.
29-36
: An example of overridden function to return translated messages of the
validation exception.
Now we are ready to create a FastAPI application:
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 |
|
8
: Add get_locale
function as a global dependency.
Note
If you need to use i18n only for specific part of your
application, you can add this get_locale
function to the specific
APIRouter
. More information about APIRouter
you can find
here.
10
: Override default request validation error handler with
validation_exception_handler
.
Run it¶
Run the server with:
$ uvicorn main:app --reload
INFO: Uvicorn running on http://127.0.0.1:8000 (Press CTRL+C to quit)
INFO: Started reloader process [28720]
INFO: Started server process [28722]
INFO: Waiting for application startup.
INFO: Application startup complete.
About the command uvicorn main:app --reload
...
The command uvicorn main:app
refers to:
main
: the filemain.py
(the Python "module").app
: the object created inside ofmain.py
with the lineapp = FastAPI()
.--reload
: make the server restart after code changes. Only do this for development.
Send it¶
Open your browser at http://127.0.0.1:8000/docs#/default/create_user_user_post.
Send POST-request with empty body and de_DE
locale query param via swagger UI
or curl
:
$ curl -X 'POST' \
'http://127.0.0.1:8000/user?locale=de_DE' \
-H 'accept: application/json' \
-H 'Content-Type: application/json' \
-d '{
}'
Check it¶
As a result, you will get the next response body:
{
"detail": [
{
"loc": [
"body",
"name"
],
"msg": "Feld erforderlich",
"type": "value_error.missing"
}
]
}
If you don't mention the locale
param, English locale will be used by
default.
Use placeholder in error strings¶
You can use placeholders in error strings, but you must mark every placeholder with {}
.
from decimal import Decimal
from pydantic import BaseModel, ValidationError, Field
from pydantic_i18n import PydanticI18n
translations = {
"en_US": {
"Decimal input should have no more than {} in total":
"Decimal input should have no more than {} in total",
},
"es_AR": {
"Decimal input should have no more than {} in total":
"La entrada decimal no debe tener más de {} en total",
},
}
tr = PydanticI18n(translations)
class CoolSchema(BaseModel):
my_field: Decimal = Field(max_digits=3)
try:
CoolSchema(my_field=1111)
except ValidationError as e:
translated_errors = tr.translate(e.errors(), locale="es_AR")
print(translated_errors)
# [
# {
# 'type': 'decimal_max_digits',
# 'loc': ('my_field',),
# 'msg': 'La entrada decimal no debe tener más de 3 digits en total',
# 'input': 1111,
# 'ctx': {
# 'max_digits': 3
# },
# 'url': 'https://errors.pydantic.dev/2.6/v/decimal_max_digits'
# }
# ]
Get current error strings from Pydantic¶
pydantic-i18n doesn't provide prepared translations of all current error
messages from pydantic, but you can use a special class method
PydanticI18n.get_pydantic_messages
to load original messages in English. By
default, it returns a dict
object:
from pydantic_i18n import PydanticI18n
print(PydanticI18n.get_pydantic_messages())
# {
# "Object has no attribute '{}'": "Object has no attribute '{}'",
# "Invalid JSON: {}": "Invalid JSON: {}",
# "JSON input should be string, bytes or bytearray": "JSON input should be string, bytes or bytearray",
# "Recursion error - cyclic reference detected": "Recursion error - cyclic reference detected",
# "Field required": "Field required",
# "Field is frozen": "Field is frozen",
# .....
# }
You can also choose JSON string or Babel format with output
parameter values
"json"
and "babel"
:
from pydantic_i18n import PydanticI18n
print(PydanticI18n.get_pydantic_messages(output="json"))
# {
# "Field required": "Field required",
# "Field is frozen": "Field is frozen",
# "Error extracting attribute: {}": "Error extracting attribute: {}",
# .....
# }
print(PydanticI18n.get_pydantic_messages(output="babel"))
# msgid "Field required"
# msgstr "Field required"
#
# msgid "Field is frozen"
# msgstr "Field is frozen"
#
# msgid "Error extracting attribute: {}"
# msgstr "Error extracting attribute: {}"
# ....
Loaders¶
pydantic-i18n provides a list of loaders to use translations.
DictLoader¶
DictLoader is the simplest loader and default in PydanticI18n. So you can just pass your translations dictionary without any other preparation steps.
from pydantic import BaseModel, ValidationError
from pydantic_i18n import PydanticI18n
translations = {
"en_US": {
"Field required": "field required",
},
"de_DE": {
"Field required": "Feld erforderlich",
},
}
tr = PydanticI18n(translations)
class User(BaseModel):
name: str
try:
User()
except ValidationError as e:
translated_errors = tr.translate(e.errors(), locale="de_DE")
print(translated_errors)
# [
# {
# 'type': 'missing',
# 'loc': ('name',),
# 'msg': 'Feld erforderlich',
# 'input': {
#
# },
# 'url': 'https://errors.pydantic.dev/2.6/v/missing'
# }
# ]
JsonLoader¶
JsonLoader needs to get the path to some directory with the next structure:
|-- translations
|-- en_US.json
|-- de_DE.json
|-- ...
where e.g. en_US.json
looks like:
{
"Field required": "field required"
}
and de_DE.json
:
{
"Field required": "Feld erforderlich"
}
Then we can use JsonLoader
to load our translations:
from pydantic import BaseModel, ValidationError
from pydantic_i18n import PydanticI18n, JsonLoader
loader = JsonLoader("./translations")
tr = PydanticI18n(loader)
class User(BaseModel):
name: str
try:
User()
except ValidationError as e:
translated_errors = tr.translate(e.errors(), locale="de_DE")
print(translated_errors)
# [
# {
# 'type': 'missing',
# 'loc': ('name',
# ),
# 'msg': 'Feld erforderlich',
# 'input': {
#
# },
# 'url': 'https://errors.pydantic.dev/2.6/v/missing'
# }
# ]
BabelLoader¶
To use this loader, you need to install babel
first:
(venv) $ pip install babel
BabelLoader works in the similar way as JsonLoader. It also needs a translations directory with the next structure:
|-- translations
|-- en_US
|-- LC_MESSAGES
|-- messages.mo
|-- messages.po
|-- de_DE
|-- LC_MESSAGES
|-- messages.mo
|-- messages.po
|-- ...
Information about translations preparation you can find on the Babel docs pages and e.g. from this article.
Here is an example of the BabelLoader
usage:
from pydantic import BaseModel, ValidationError
from pydantic_i18n import PydanticI18n, BabelLoader
loader = BabelLoader("./translations")
tr = PydanticI18n(loader)
class User(BaseModel):
name: str
try:
User()
except ValidationError as e:
translated_errors = tr.translate(e.errors(), locale="de_DE")
print(translated_errors)
# [
# {
# 'type': 'missing',
# 'loc': ('name',),
# 'msg': 'Feld erforderlich',
# 'input': {
#
# },
# 'url': 'https://errors.pydantic.dev/2.6/v/missing'
# }
# ]
Write your own loader¶
If current loaders aren't suitable for you, it's possible to write your own
loader and use it with pydantic-i18n. To do it, you need to import
BaseLoader
and implement the next items:
- property
locales
to get a list of locales; - method
get_translations
to get content for the specific locale.
In some cases you will also need to change implementation of the gettext
method.
Here is an example of the loader to get translations from CSV files:
|-- translations
|-- en_US.csv
|-- de_DE.csv
|-- ...
en_US.csv
content:
Field required,field required
de_DE.csv
content:
Field required,Feld erforderlich
import os
from typing import List, Dict
from pydantic import BaseModel, ValidationError
from pydantic_i18n import PydanticI18n, BaseLoader
class CsvLoader(BaseLoader):
def __init__(self, directory: str):
self.directory = directory
@property
def locales(self) -> List[str]:
return [
filename[:-4]
for filename in os.listdir(self.directory)
if filename.endswith(".csv")
]
def get_translations(self, locale: str) -> Dict[str, str]:
with open(os.path.join(self.directory, f"{locale}.csv")) as fp:
data = dict(line.strip().split(",") for line in fp)
return data
class User(BaseModel):
name: str
if __name__ == '__main__':
loader = CsvLoader("./translations")
tr = PydanticI18n(loader)
try:
User()
except ValidationError as e:
translated_errors = tr.translate(e.errors(), locale="de_DE")
print(translated_errors)
# [
# {
# 'type': 'missing',
# 'loc': ('name',),
# 'msg': 'Feld erforderlich',
# 'input': {
#
# },
# 'url': 'https://errors.pydantic.dev/2.6/v/missing'
# }
# ]
Acknowledgments¶
Thanks to Samuel Colvin and his pydantic library.
Also, thanks to Sebastián Ramírez and his FastAPI project, some scripts and documentation structure and parts were used from there.
License¶
This project is licensed under the terms of the MIT license.