Stage 3 · L07

Understand Pydantic types, conversion and validation

Core · original Session 3

Pydantic validates data at a boundary and produces typed values. EquipmentIn accepts a name, integer quantity and Decimal daily_rate. quantity='3' converts to integer 3 in the default policy; quantity='three' cannot convert and raises validation error. Conversion is deliberate behavior, not proof that clients originally sent the correct JSON type.

Constraints are separate from types. An integer quantity of zero has the right Python type but violates ge=1. A rate of 12.505 violates decimal_places=2. An empty name violates min_length. extra='forbid' rejects hidden keys such as role. These request errors occur before the handler creates a database row; the baseline checks a fresh row count after every rejected payload.

Required and nullable describe different things. nickname: str | None permits null but remains required without a default. nickname: str | None = None permits omission too. The scoped test uses the first form and confirms that omission fails. For a future PATCH, model_dump(exclude_unset=True) preserves omission; an explicitly supplied null remains supplied and needs a domain decision.

Field(strict=True) on an integer rejects a numeric string that default conversion would accept. Use strictness where the contract needs it, and test that boundary. Strictness does not replace business rules, foreign keys or a transaction. Pydantic is not permission checking, email deliverability, stock enforcement or SQL injection protection.

Errors include locations and kinds so a client can identify the invalid field. Do not echo credential values or raw database exceptions in an API error. FastAPI maps request validation to 422 for these routes. A valid request can still encounter a uniqueness conflict at PostgreSQL; that later failure is 409 and must leave no partial write.

from pydantic import BaseModel, ConfigDict, Field
class QuantityIn(BaseModel):
    model_config = ConfigDict(extra='forbid')
    quantity: int = Field(ge=1)
    note: str | None  # Nullable, but still required without a default.
valid = QuantityIn(quantity='2', note=None)
assert valid.quantity == 2

The non-strict integer accepts '2'; zero violates the bound; note may be None but omission is invalid because it has no default; role is an extra forbidden key. Field(strict=True) rejects the numeric string. None of these schema checks reserves stock or commits a database row.