get_type_hints() vs __annotations__ with future annotations

Question

Accepted Answer

When 'from __future__ import annotations' is active, Python stores all annotations as their literal source text - strings. Reading __annotations__ gives you {'x': 'int', 'return': 'str'} - raw strings, not type objects. Code that expects type objects (like some metaclasses, older versions of Pydantic, or custom introspection logic) will break silently or raise errors when it tries to use these strings as types. get_type_hints() does the resolution step: it looks up each string in the function's global and local namespaces and returns the actual type object. For 'int' that's trivial, but for forward references like 'MyClass' defined later in the file, get_type_hints() resolves correctly because by the time you call it, all names are defined. The practical consequence: if you write library code that introspects annotations, always use get_type_hints() rather than __annotations__ directly. The two are equivalent only when 'from __future__ import annotations' is NOT active and no annotations are strings.