Looking for the exact opposite? Check out zod-to-json-schema
A runtime package and CLI tool to convert JSON schema (draft 4+) objects or files into Zod schemas in the form of JavaScript code.
Before v2 it used prettier
for formatting and json-refs
to resolve schemas. To replicate the previous behaviour, please use their respective CLI tools.
Since v2 the CLI supports piped JSON.
Just paste your JSON schemas here!
npm i -g json-schema-to-zod
json-schema-to-zod -i mySchema.json -o mySchema.ts
$refs
resolved and output formattednpm i -g json-schema-to-zod json-refs prettier
json-refs resolve mySchema.json | json-schema-to-zod | prettier --parser typescript > mySchema.ts
Flag | Shorthand | Function |
---|---|---|
--input |
-i |
JSON or a source file path. Required if no data is piped. |
--output |
-o |
A file path to write to. If not supplied stdout will be used. |
--name |
-n |
The name of the schema in the output |
--depth |
-d |
Maximum depth of recursion in schema before falling back to z.any() . Defaults to 0. |
--module |
-m |
Module syntax; esm , cjs or none. Defaults to esm in the CLI and none programmaticly. |
--type |
-t |
Export a named type along with the schema. Requires name to be set and module to be esm . |
--noImport |
-ni |
Removes the import { z } from 'zod'; or equivalent from the output. |
import { jsonSchemaToZod } from "json-schema-to-zod";
const myObject = {
type: "object",
properties: {
hello: {
type: "string",
},
},
};
const module = jsonSchemaToZod(myObject, { module: "esm" });
// `type` can be either a string or - outside of the CLI - a boolean. If its `true`, the name of the type will be the name of the schema with a capitalized first letter.
const moduleWithType = jsonSchemaToZod(myObject, { name: "mySchema", module: "esm", type: true });
const cjs = jsonSchemaToZod(myObject, { module: "cjs", name: "mySchema" });
const justTheSchema = jsonSchemaToZod(myObject);
module
import { z } from "zod";
export default z.object({ hello: z.string().optional() });
moduleWithType
import { z } from "zod";
export const mySchema = z.object({ hello: z.string().optional() });
export type MySchema = z.infer<typeof mySchema>;
cjs
const { z } = require("zod");
module.exports = { mySchema: z.object({ hello: z.string().optional() }) };
justTheSchema
z.object({ hello: z.string().optional() });
$refs
resolved and output formattedimport { z } from "zod";
import { resolveRefs } from "json-refs";
import { format } from "prettier";
import jsonSchemaToZod from "json-schema-to-zod";
async function example(jsonSchema: Record<string, unknown>): Promise<string>{
const { resolved } = await resolveRefs(jsonSchema);
const code = jsonSchemaToZod(resolved);
const formatted = await format(code, { parser: "typescript" });
return formatted;
}
You can pass a function to the overrideParser
option, which represents a function that receives the current schema node and the reference object, and should return a string when it wants to replace a default output. If the default output should be used for the node just return void.
The output of this package is not meant to be used at runtime. JSON Schema and Zod does not overlap 100% and the scope of the parsers are purposefully limited in order to help the author avoid a permanent state of chaotic insanity. As this may cause some details of the original schema to be lost in translation, it is instead recommended to use tools such as Ajv to validate your runtime values directly against the original JSON Schema.
That said, it's possible in most cases to use eval
. Here's an example that you shouldn't use:
const zodSchema = eval(jsonSchemaToZod({ type: "string" }, { module: "cjs" }));
zodSchema.safeParse("Please just use Ajv instead");