1	import { z } from "zod";
2	import { createMirror } from "exact-mirror-zod";
3
4	const schema = z.object({ name: z.string(), age: z.number() });
5	const mirror = createMirror(schema);
6
7	mirror({ name: "test", age: 25, password: "leaked" });
8	// => { name: "test", age: 25 }

createMirror(schema) compiles a Zod schema into a function that strips unknown properties from objects to match the schema's shape. It does not validate -- it assumes input is already valid and just projects known keys.

Why?

Validate on input (z.parse()), strip on output (createMirror()). Prevents leaking internal fields from API responses without the overhead of full re-parsing. Typically 5-15x faster than z.parse() for stripping.

Supported Types

Type	Behavior
Object	Keeps only declared keys
Array	Per-element mirroring (passes through primitive arrays)
Tuple	Per-index mirroring
Record	`Object.keys` iteration with value mirroring
Discriminated Union	`switch` on discriminator key
Union	Cheap `typeof` / `Array.isArray` structural checks
Intersection	Merges object shapes
Recursive (`z.lazy`)	Memoized helpers with cycle detection
Wrappers	`optional`, `nullable`, `default`, `readonly`, `branded`, `pipe` -- unwraps and recurses

Options

typescript

1	createMirror(schema, {
2	// Sanitization functions applied to every string value
3	sanitize: (v) => v.trim(),
4
5	// Max recursion depth for z.lazy() schemas (default: 8)
6	recursionLimit: 8,
7
8	// Remove values that don't match any union variant (default: false)
9	removeUnknownUnionType: false,
10	});

String Sanitization

Pass one or more functions to transform every string value in the output:

typescript

1	const mirror = createMirror(schema, {
2	sanitize: [(v) => v.trim(), (v) => v.replace(/</g, "<")],
3	});

Development

bash

1	bun test # run tests
2	bun run bench # run benchmarks
3	bun run bench:small # benchmark small schemas
4	bun run bench:compare # compare against z.parse()
5	bun run typecheck # type check
6	bun run fix # lint + format

License

MIT