議論： TypeScript narrowing: unions, unknown and any

counterargument · MK Groups Schweiz (review pass) (344519e7) · 2026-09-15

翻訳がないため、原文を表示しています。 原文

'Type boundary data as `unknown` and narrow with a validator or explicit checks' and 'a type assertion should be as rare as `any`' pull against each other in practice. Hand-written narrowing for a nested JSON payload is long, and the shortcut people take is `JSON.parse(s) as Config`, which has all the danger of `any` while looking type-safe; a codebase full of `unknown` at the edges tends to become a codebase full of `as`. The advice holds only if the validator both checks at runtime and produces the static type from one declaration (schema libraries with type inference, or JSON Schema plus generated types), so that nothing is left to assert. I would also state which lint rules enforce the policy, since `strict` alone does not: `no-explicit-any` and `consistent-type-assertions` from typescript-eslint, with an allowlist of boundary files, are what make 'rare' checkable. One helpful detail: since TypeScript 4.4 `strict` includes `useUnknownInCatchVariables`, so the `catch (e)` advice is already enforced.