Type Hints and Runtime Validation

Type hints describe expected types for readers, editors, and static analysis tools. A static checker examines code without executing every possible input. Runtime validation inspects the values a program actually receives. These checks address different boundaries: an annotated date parameter helps check callers, while incoming CSV text still needs parsing and validation.

For example, a function can declare that day is a date and amount is a Decimal. Passing a string instead of a date can be detected by a configured type checker. But a Decimal can still be NaN, have an unsupported scale, or exceed the storage range. Those are value constraints that ordinary type hints do not establish.

Python does not automatically enforce function annotations at runtime. A dataclass can organize fields without validating them, and frozen=True restricts ordinary reassignment rather than proving valid input or deep immutability. Validation libraries can parse and constrain values, but coercion settings, allowed currencies, timezone requirements, and rejection policy must be chosen deliberately.

Validate untrusted input before creating the trusted domain records used by transforms. Preserve a useful reason for rejected data and test both accepted boundaries and expected failures. Static checking can miss unannotated code or values treated as Any; runtime tests cover only the cases exercised. Combining these methods reduces different error classes without guaranteeing every future input.

Reference: Python typing documentation. For worked examples, see Python for Data Engineers.