Schema Validator III

Requirements

• Export schema as named and default, plus ValidationError. Keep string(), number(), boolean(), optional(), object(shape), parse(input), and safeParse(input) from Schema Validator. safeParse returns { success: true, data } or { success: false, errors: [{ path, message }] }. parse returns validated data or throws ValidationError with an issues array.
• Add chainable min(limit) and max(limit) to string and number schemas, including optional variants. String limits are nonnegative integers and measure UTF-16 length. Number limits are finite numbers. Bounds are inclusive; invalid limits throw RangeError. Boolean schemas do not expose min/max.
• Builders are immutable. Every min/max call appends a constraint without changing its source schema. Check every constraint in chain order and collect all failed bounds. Contradictory limits are allowed and may yield multiple errors. Skip bounds when the primitive type is wrong or optional undefined is accepted.
• Use root paths [] for primitive issues. Type messages are Expected string, Expected finite number, and Expected boolean. String-bound messages are Expected at least N characters or Expected at most N characters. Number-bound messages are Expected number >= N or Expected number <= N. null never satisfies optional().
• object(shape) now accepts any schema factory result, including nested objects and arrays. Add array(elementSchema) for dense arrays. optional() must work on primitive, bounded, object, and array schemas. Optional undefined succeeds; null fails the underlying type check.
• Infer recursive output types, including required versus optional object keys and arrays of the element output. object strips unknown own fields at every nesting level and builds fresh output containers; array builds a fresh array. Missing optional object fields are omitted, while explicitly present undefined is retained. Read only own input fields.
• Collect all errors in depth-first shape order, then ascending array index order. Issue paths are (string | number)[]; prepend each traversed field name or numeric index. Primitive/root issues start at []. Wrong object type reports Expected plain object; wrong array type reports Expected array; sparse arrays report Expected dense array at their array path without visiting elements.
• Shape and element schemas must expose optional and safeParse; reject invalid schema definitions with TypeError when constructing a container schema. Snapshot object shape entries so changing the shape object later cannot alter validation. Safely define special field names in outputs.
• Keep parse, safeParse, ValidationError, immutable min/max builders, inclusive bounds, constraint order, and messages from version II. Do not mutate schemas or inputs. No coercion, defaults, async checks, unions, recursion through self-referential schemas, cyclic inputs, or external validator library is required.

Examples

Input: schema.object({user:schema.object({name:schema.string().min(2)}),scores:schema.array(schema.number())}).parse({user:{name:"Ada",extra:1},scores:[1,2],extra:true})
Output: {user:{name:"Ada"},scores:[1,2]}

Input: schema.array(schema.object({age:schema.number().min(0)})).safeParse([{age:-1},{age:"bad"}])
Output: Errors at [0,"age"] and [1,"age"], in that order.