How-to: the decimal module and Unicode digit input

Python's Decimal accepts Unicode full-width digits

decimal.Decimal() interprets Unicode full-width digits (U+FF10–U+FF19: ０１２３４５６７８９) as numbers as-is. When a user submits full-width digits through an HTTP API, they can pass straight through a Pydantic str field.

from decimal import Decimal

Decimal("１２３")   # → Decimal('123')  ← converted normally
Decimal("１.５")   # → Decimal('1.5')

Problem: unexpected input may slip through

In a financial-calculation API that declares price: str = Field(...), if a client sends "１０００" it is processed as Decimal("１０００") → Decimal('1000'). That is not an error in itself, but when you need normalized input (logging, DB storage, etc.) perform Unicode normalization before processing.

import unicodedata
from decimal import Decimal

def parse_decimal_safe(value: str) -> Decimal:
    """Normalize (NFKC) then convert to Decimal."""
    normalized = unicodedata.normalize("NFKC", value.strip())
    return Decimal(normalized)

unicodedata.normalize("NFKC", ...) converts full-width digits to half-width.

unicodedata.normalize("NFKC", "１２３")  # → "123"
unicodedata.normalize("NFKC", "１.５")  # → "1.5"

Validation with Pydantic

We recommend forcing input normalization with a Pydantic model_validator.

from pydantic import BaseModel, Field, model_validator

class PriceRequest(BaseModel):
    price: str = Field(max_length=20, description="Price (half-width digits)")

    @model_validator(mode="before")
    @classmethod
    def normalize_unicode(cls, values: dict) -> dict:
        import unicodedata
        if "price" in values and isinstance(values["price"], str):
            values["price"] = unicodedata.normalize("NFKC", values["price"])
        return values

Related issue

• [FT176] #500: document the full-width Unicode digit acceptance behavior of parse_decimal_safe()