Lib/typing_extensions.py (part 2)

Source:

cpython 3.14 @ ab2d84fe1023/Lib/typing_extensions.py ↗

typing_extensions backports newer typing features to older Python versions. In gopy (targeting 3.14), most of these are already in typing, but typing_extensions is imported by many popular packages and must be importable.

Map

Lines	Symbol	Role
1-200	TypeAlias	Explicit type alias annotation (3.10+, X: TypeAlias = int)
201-400	TypeGuard	Narrowing predicate return type
401-600	Self	Reference to the current class
601-800	LiteralString, Never	String literals, bottom type
801-1100	ParamSpec, Concatenate	Callable parameter typing
1101-1400	TypeVarTuple, Unpack	Variadic generics
1401-1700	override	Decorator marking method overrides
1701-2000	dataclass_transform	PEP 681: decorator metadata for type checkers
2001-3000	Backports of stdlib typing features	Various re-exports with version guards

Reading

TypeAlias

# CPython: Lib/typing_extensions.py:180 TypeAlias
# In Python 3.14 TypeAlias is a SpecialForm.
# In Python 3.10 it's just a sentinel object:
class _TypeAliasMeta(type):
    def __instancecheck__(cls, obj):
        return False
TypeAlias = _TypeAliasForm('TypeAlias', doc="...")

x: TypeAlias = int is treated as a type alias by type checkers; at runtime TypeAlias is just an annotation marker.

override

# CPython: Lib/typing_extensions.py:1480 override
def override(method):
    """Indicate that a method is intended to override a base class method."""
    method.__override__ = True
    return method

@override is a no-op at runtime (just sets __override__ = True). Type checkers verify the base class has the method.

dataclass_transform

# CPython: Lib/typing_extensions.py:1720 dataclass_transform
def dataclass_transform(
    *, eq_default=True, order_default=False, frozen_default=False,
    field_specifiers=(), **kwargs
):
    def decorator(cls_or_fn):
        cls_or_fn.__dataclass_transform__ = {
            "eq_default": eq_default,
            ...
        }
        return cls_or_fn
    return decorator

@dataclass_transform marks a class or function as a dataclass-like decorator. Type checkers use __dataclass_transform__ metadata to understand the generated __init__ signature.

gopy notes

typing_extensions is importable in gopy when the basic typing module is available. The module consists almost entirely of pure Python with sys.version_info guards. No C extension is required. The key requirement is that typing.TypeVar, typing.ParamSpec, typing.Protocol, and typing.get_type_hints all work so that the version guards resolve correctly.