tashikame

Lightweight data validation for JavaScript and TypeScript

Validata data against schemas at runtime and infer TypeScript types from schemas. Lightweight alternative to and inspired by Zod, Valibot and ArkType.

Usage

Schemas are simply string keywords for primitives (and some other values) or functions for complex data. Schema functions return a validation report (see Feedback). tashikame ships with built-in schemas for the most common use cases.

Schema keywords

tashikame recognizes the following strings as schema keywords:

• unknown
• string
• number
• boolean
• bigint
• symbol
• null
• undefined
• function

Validating data

Input data is validated with the parse function. The function takes a schema as first argument, the input data as second argument and returns the input data or throws an error if it doesn't conform to the given schema.

⚠️ Input data is only validated and not cloned, i.e. parse returns the input as is if it conforms to the given schema.

Additionally to the parse function there are also parse.safe which returns a ValidationReport instead of throwing an error and parse.is which is a type predicate function.

Feedback

parse.safe returns a ValidationReport which can be one of two shapes:

type DataValidReport<T> = {
	valid: true;
	data: T;
};

type DataInvalidReport = {
	valid: false;
	description?: string;
	expected?: string;
	received?: string | Record<string, DataInvalidReport>;
}

type ValidationReport<T> =  DataValidReport<T> | DataInvalidReport;

parse throws with an instance of ValidationError when the given input doesn't match the schema. The error contains a detailed DataInvalidReport of the issue. Currently built-in schemas stop validation at the first issue encountered.

TypeScript

In TypeScript it is not necessary to write a type that matches a schema separately. Instead the type helper Infer can be used to produce a type from a schema. Some schemas can be configured to produce readonly variants by setting the inferReadonly option.

import { array, type Infer } from "tashikame";

const schema = array("number", { inferReadonly: true });

type S = Infer<typeof schema>; // readonly number[]

The readonly flag is only considered at the type level and not actually checked at runtime.

Note

Infer automatically converts any types to unknown

Custom schemas

Custom schemas are simply functions that return a ValidationReport. TypeScript inference works by providing the generic type argument for the data property when the input is valid.

The tashikame/core entry point provides a few utilities for writing custom schemas (it's not mandatory to use them though). See tashikame/core in the API section for details.

• registerSchemaName
  A function that registers a name for a given schema function that is used in ValidationReports.

• formatSchema
  Retrieves the registered name for a given schema.

• formatValue
  Generates a string represantion of an input value.

Recursive schemas

Self-referencing schemas can be created with the lazy function. It takes a getter function as argument that produces the actual schema which makes it possible to reference the created schema in itself. Recursive schema types cannot be inferred in TypeScript and have to be written separately.

import { record, union, lazy, type Schema } from "tashikame";

type Tree = {
	[key: string]: number | Tree;
};

const treeSchema: Schema<Tree> = lazy(() => record(union(["number", treeSchema])));

Literals and unions

Literal and union schemas can be created with the literal and union functions respectively. The literal function creates a schema for an exact value and is restricted to the same types as TypeScript literals, i.e. a value of type string, number, bigint or boolean. The union function takes an array of schemas and validates if the input matches one of the given schemas. union and literal can be combined to match against a list of allowed values.

Arrays and tuples

Array schemas are created with the array function which takes a schema as argument to match items against.

Tuple schemas can be created with the tuple function which takes an array of schemas for each item position. Item schemas can additionally be a schema created with the special spread function which wraps a spreadable schema, i.e. an array schema or another tuple schema. Spreadables in tuple schemas have the same restrictions as spreadables in tuple types in TypeScript, that is only one item schema can be a spreadable of arbitrary length.

import { array, tuple, type Infer } from "tashikame";
import { spread } from "tashikame/tuple";

const schema = tuple([spread(array("number")), "string"]);

type S = Infer<typeof schema>; // [...number[], string]

Both array and tuple can take a second argument of the type { inferReadonly: boolean } for inferring a readonly array in TypeScript. Addtional constraints like array length can be imposed with the refine function (see Refining schemas).

Objects and records

Object schemas are created with the object function. It takes as first argument an object that describes the shape of the object by providing a schema for each property. Instead of only a schema a property can be described by the ObjectSchemaProperty type. The object function can additionally take a second argument of the type ObjectSchemaConfig.

type ObjectSchemaProperty = {
	value: Schema;
	optional?: boolean;
	inferReadonly?: boolean;
}

type ObjectSchemaConfig = {
	name?: string;
	additionalProperties?: boolean | Schema | { value?: Schema; inferReadonly?: boolean };
}

When optional properties are used with TypeScript it is recommended to enable the exactOptionalPropertyTypes setting in tsconfig.json as that's how optional properties are treated in tashikame.

The name option in ObjectSchemaConfig controls what to display in ValidationReports and defaults to a formatted representation of the property schemas. The additionalProperties option configures whether unspecified keys are allowed and if they should conform to a schema. If the option is set to true additional properties are inferred as type unknown. Currently it's not possible to validate properties with symbol keys.

Record schemas are created with the record function. It is simply a shorthand for object({}, { additionalProperties: schema }). In contrast to how TypeScript's Record is defined, the record function only uses a schema for the values and expects all keys to be strings.

Refining schemas

The refine function creates a wrapped schema with additional constraints.

import { parse, refine } from "tashikame";

// a number greater or equal to 0 and lower than 10
const schema = refine("number", (n) => n >= 0, (n) => n < 10);

parse(schema, 11); // 💥

Alternatively, the constraints can be given as an object, with the property names providing additional information for the ValidationReport.

import { parse, refine } from "tashikame";

const schema = refine("number", { Int: Number.isInteger });

console.log(parse.safe(schema, 0.1).expected); // number with constraint "Int"

Constraint functions are only called when the input matches the wrapped schema.

API

tashikame is very modular and provides several entry points. The most common functions and types are also exported from the top-level package. The overview below only lists the identifiers exported by each entry point and which are also exported from the top-level package. For details see the respective sources.

tashikame/core

// also exported from "tashikame"
export type SchemaKeyword<T>;
export type SchemaFunction<T>;
export type Schema<T>;
export type DataValidReport<T>;
export type DataInvalidReport;
export type ValidationReport<T>;
export type Infer<S extends Schema>;

export class ValidationError;

export function parse<S extends Schema>(schema: S, input: unknown): Infer<S>;
parse.safe = function <S extends Schema>(schema: S, input: unknown): ValidationReport<Infer<S>>;
parse.is = function <S extends Schema>(schema: S, input: unknown): input is Infer<S>;

// only exported from "tashikame/core"
export function registerSchemaName<T extends SchemaFunction>(name: string, schema: T): T;
export function formatSchema(schema: Schema): string;
export function formatValue(value: unknown): string;

tashikame/lazy

// also exported from "tashikame"
export function lazy<LazySchema extends Schema>(getSchema: () => LazySchema): SchemaFunction<Infer<LazySchema>>;

tashikame/literal

// also exported from "tashikame"
export type Literal;
export function literal<T extends Literal>(value: T): SchemaFunction<T>;

tashikame/union

// also exported from "tashikame"
export function union<UnionSchema extends readonly [Schema, ...Schema[]]>(schema: UnionSchema): SchemaFunction<InferUnion<UnionSchema>>;

// only exported from "tashikame/union"
export type InferUnion<S extends readonly Schema[]>;

tashikame/iterable

// only exported from "tashikame/iterable"
export type IterableSchema<T extends Iterable<unknown>>;
export function makeIterable<T extends Iterable<unknown>>(size: number, schema: SchemaFunction<T>): IterableSchema<T>;

tashikame/array

// also exported from "tashikame"
export type ArraySchemaConfig;
export function array<
	ItemSchema extends Schema,
	Config extends ArraySchemaConfig
>(schema: ItemSchema, config?: Config): IterableSchema<InferArray<ItemSchema, Config>>;

// only exported from "tashikame/array"
export type InferArray<ItemSchema extends Schema, Config extends ArraySchemaConfig>;

tashikame/tuple

// also exported from "tashikame"
export type TupleSchemaConfig;
export function tuple<
	const TupleSchema extends TupleSchemaBase,
	const Config extends TupleSchemaConfig
>(schema: TupleSchema, config?: Config): IterableSchema<InferTuple<TupleSchema, Config>>;

// only exported from "tashikame/tuple"
export type SpreadableSchema<T extends readonly unknown[]>;
export type TupleSchemaBase = ReadonlyArray<Schema | SpreadableSchema>;
export type InferTuple<TupleSchema extends TupleSchemaBase, Config extends TupleSchemaConfig>;
export function spread<SpreadSchema extends IterableSchema<readonly unknown[]>>(schema: SpreadSchema): SpreadableSchema<Infer<SpreadSchema>>;

tashikame/object

// also exported from "tashikame"
export type ObjectSchemaProperty;
export type AdditionalPropertiesConfig;
export type ObjectPropertiesSchema;
export type ObjectSchemaConfig;
export function object<
	PropertySchema extends ObjectPropertiesSchema,
	Config extends ObjectSchemaConfig
>(properties: PropertySchema, config?: Config): Schema<InferObject<PropertySchema, Config>>;

export type RecordSchemaConfig;
export function record<
	RecordSchema extends Schema,
	Config extends RecordSchemaConfig
>(schema: RecordSchema, config?: Config): Schema<InferRecord<RecordSchema, Config>>;

// only exported from "tashikame/object"
export type InferObject<S extends ObjectPropertiesSchema, C extends ObjectSchemaConfig>;
export type InferRecord<S extends Schema, C extends RecordSchemaConfig>;

tashikame/refine

// also exported from "tashikame"
export function refine<S extends Schema>(schema: S, ...constraints: ConstraintsArray<S>): SchemaFunction<Infer<S>>;
export function refine<S extends Schema>(schema: S, constraints: NamedConstraints<S>): SchemaFunction<Infer<S>>;

// only exported from "tashikame/refine"
export type Constraint<S extends Schema>;
export type ConstraintsArray<S extends Schema>;
export type NamedConstraints<S extends Schema>;