Types provide the most leverage where uncertain data enters a program. HTTP requests, webhook payloads, environment variables, and model responses all start as values whose shape cannot be trusted merely because a type annotation was written nearby.
Parse before passing data inward
Treat external data as object (or Any only when unavoidable), validate it
once, and convert it into a small internal representation. After that point,
business logic can work with values whose invariants have already been
checked.
def parse_limit(value: object) -> int:
if isinstance(value, bool) or not isinstance(value, int):
raise ValueError("limit must be an integer")
if value < 1:
raise ValueError("limit must be positive")
return value
The explicit boolean check matters in Python because bool is a subclass of
int. Boundary validation is a good place to make these language-specific
details unambiguous.
Keep internal models boring
Once data is parsed, prefer ordinary dataclasses or clearly defined models over repeated dictionary lookups. This improves autocomplete, refactoring, and static analysis while keeping runtime behavior understandable.
Validate once, then trust the contract
Repeated validation scattered through the codebase makes it hard to know which invariants hold at any point. A narrow parser at each external boundary gives the rest of the system one consistent contract to rely on.