Mapped Types

When you build CRUD features you often want variations of one base DTO: the same fields, all optional, for an update; just two of them for a summary. Mapped types derive them so the rules stay in one place.

from pydantic import BaseModel, Field
 
from austial import IntersectionType, OmitType, PartialType, PickType
 
 
class CreateCatDto(BaseModel):
    name: str = Field(min_length=2)
    age: int
    breed: str
 
 
class UpdateCatDto(PartialType(CreateCatDto)):               # every field optional
    pass
 
 
class CatNameDto(PickType(CreateCatDto, ["name"])):           # only `name`
    pass
 
 
class CatSummaryDto(OmitType(CreateCatDto, ["breed"])):       # everything but `breed`
    pass
 
 
class CatWithTimestampsDto(IntersectionType(CreateCatDto, TimestampsDto)):   # fields of both
    pass

Function	Result
`PartialType(Base)`	every field becomes optional, default `None`
`PickType(Base, keys)`	only `keys`
`OmitType(Base, keys)`	all but `keys`
`IntersectionType(A, B, ...)`	the union of every field; later models win a name clash

Field types, defaults and Field(...) constraints are carried over, so FastAPI validation and the OpenAPI schema stay accurate:

UpdateCatDto(age=3)            # ok
UpdateCatDto(name="x")         # ValidationError: min_length=2 still applies

They return a new pydantic model class that you subclass (as above) to give it a name. Naming an unknown field raises KeyError. Validators (@field_validator, @model_validator) are not carried over: re-declare them on the subclass.