Skip to main content

parse vs typed

parse and typed solve different runtime tasks.

parse(unknown) for external-input validation​

Use parse when data comes from outside your code and is not trusted.

import {number, object, string} from '@livon/schema';

const User = object({
  name: 'User',
  shape: {
    id: string(),
    age: number().int().min(18),
  },
});

const unknownInput: unknown = JSON.parse(raw);
const user = User.parse(unknownInput);

Parameters in this example​

User.parse(value):

  • value (unknown): untrusted incoming data.

typed(validInput) for invariant entrypoints​

Use typed when data is already local and should be structurally aligned at compile time before entering runtime flow. typed is also useful for editor autocomplete, because the schema-derived type guides field suggestions while building payloads.

import {type Infer} from '@livon/schema';

type UserType = Infer<typeof User>;

const invalid: UserType = {id: 'u-1', age: '21'};
// TypeScript error: age must be number

const valid = User.typed({id: 'u-1', age: 21});

Autocomplete example:

const draft = User.typed({
  id: 'u-1',
  // IDE suggests: age
  age: 21,
});

Parameters in this example​

User.typed(value):

  • value (Infer<typeof User>): compile-time aligned value that is still checked against runtime constraints.

Compile-time alignment plus runtime constraints​

typed shifts structural mismatches (field type/shape) to compile time, but constraint checks still run at runtime.

const maybeAdult = User.typed({id: 'u-1', age: 16});
// Compiles structurally (number), but runtime validation fails on min(18)

Practical rule​

  • Use parse for external input.
  • Use typed for internal invariants before emitting or executing.
  • Both reinforce deterministic runtime checks.