@amir.best/ts-json-schema-generator

ts-json-schema-generator

Extended version of https://github.com/xiag-ag/typescript-to-json-schema.

Inspired by YousefED/typescript-json-schema. Here's the differences list:

• this implementation avoids the use of typeChecker.getTypeAtLocation() (so probably it keeps correct type aliases)
• processing AST and formatting JSON schema have been split into two independent steps
• not exported types, interfaces, enums are not exposed in the definitions section in the JSON schema

Contributors

This project is made possible by a community of contributors. We welcome contributions of any kind (issues, code, documentation, examples, tests,...). Please read our code of conduct.

CLI Usage

npm install --save ts-json-schema-generator
./node_modules/.bin/ts-json-schema-generator --path 'my/project/**/*.ts' --type 'My.Type.Name'

Note that different platforms (e.g. Windows) may use different path separators so you may have to adjust the command above.

Programmatic Usage

// main.js

const tsj = require("ts-json-schema-generator");
const fs = require("fs");

/** @type {import('ts-json-schema-generator/dist/src/Config').Config} */
const config = {
    path: "path/to/source/file",
    tsconfig: "path/to/tsconfig.json",
    type: "*", // Or <type-name> if you want to generate schema for that one type only
};

const output_path = "path/to/output/file";

const schema = tsj.createGenerator(config).createSchema(config.type);
const schemaString = JSON.stringify(schema, null, 2);
fs.writeFile(output_path, schemaString, (err) => {
    if (err) throw err;
});

Run the schema generator via node main.js.

Custom formatting

Extending the built-in formatting is possible by creating a custom formatter and adding it to the main formatter:

1. First we create a formatter, in this case for formatting function types:

// my-function-formatter.ts
import { BaseType, Definition, FunctionType, SubTypeFormatter } from "ts-json-schema-generator";
import ts from "typescript";

export class MyFunctionTypeFormatter implements SubTypeFormatter {
    // You can skip this line if you don't need childTypeFormatter
    public constructor(private childTypeFormatter: TypeFormatter) {}

    public supportsType(type: FunctionType): boolean {
        return type instanceof FunctionType;
    }

    public getDefinition(type: FunctionType): Definition {
        // Return a custom schema for the function property.
        return {
            type: "object",
            properties: {
                isFunction: {
                    type: "boolean",
                    const: true,
                },
            },
        };
    }

    // If this type does NOT HAVE children, generally all you need is:
    public getChildren(type: FunctionType): BaseType[] {
        return [];
    }

    // However, if children ARE supported, you'll need something similar to
    // this (see src/TypeFormatter/{Array,Definition,etc}.ts for some examples):
    public getChildren(type: FunctionType): BaseType[] {
        return this.childTypeFormatter.getChildren(type.getType());
    }
}

2. Then we add the formatter as a child to the core formatter using the augmentation callback:

import { createProgram, createParser, SchemaGenerator, createFormatter } from "ts-json-schema-generator";
import { MyFunctionTypeFormatter } from "./my-function-formatter.ts";
import fs from "fs";

const config = {
    path: "path/to/source/file",
    tsconfig: "path/to/tsconfig.json",
    type: "*", // Or <type-name> if you want to generate schema for that one type only
};

// We configure the formatter an add our custom formatter to it.
const formatter = createFormatter(config, (fmt, circularReferenceTypeFormatter) => {
    // If your formatter DOES NOT support children, e.g. getChildren() { return [] }:
    fmt.addTypeFormatter(new MyFunctionTypeFormatter());
    // If your formatter DOES support children, you'll need this reference too:
    fmt.addTypeFormatter(new MyFunctionTypeFormatter(circularReferenceTypeFormatter));
});

const program = createProgram(config);
const parser = createParser(program, config);
const generator = new SchemaGenerator(program, parser, formatter, config);
const schema = generator.createSchema(config.type);

const schemaString = JSON.stringify(schema, null, 2);
fs.writeFile(output_path, schemaString, (err) => {
    if (err) throw err;
});

Custom parsing

Similar to custom formatting, extending the built-in parsing works practically the same way:

1. First we create a parser, in this case for parsing construct types:

// my-constructor-parser.ts
import { Context, StringType, ReferenceType, BaseType, SubNodeParser } from "ts-json-schema-generator";
import ts from "typescript";

export class MyConstructorParser implements SubNodeParser {
    supportsNode(node: ts.Node): boolean {
        return node.kind === ts.SyntaxKind.ConstructorType;
    }
    createType(node: ts.Node, context: Context, reference?: ReferenceType): BaseType | undefined {
        return new StringType(); // Treat constructors as strings in this example
    }
}

2. Then we add the parser as a child to the core parser using the augmentation callback:

import { createProgram, createParser, SchemaGenerator, createFormatter } from "ts-json-schema-generator";
import { MyConstructorParser } from "./my-constructor-parser.ts";
import fs from "fs";

const config = {
    path: "path/to/source/file",
    tsconfig: "path/to/tsconfig.json",
    type: "*", // Or <type-name> if you want to generate schema for that one type only
};

const program = createProgram(config);

// We configure the parser an add our custom parser to it.
const parser = createParser(program, config, (prs) => {
    prs.addNodeParser(new MyConstructorParser());
});

const formatter = createFormatter(config);
const generator = new SchemaGenerator(program, parser, formatter, config);
const schema = generator.createSchema(config.type);

const schemaString = JSON.stringify(schema, null, 2);
fs.writeFile(output_path, schemaString, (err) => {
    if (err) throw err;
});

Options

-p, --path 'index.ts'
    The path to the TypeScript source file. If this is not provided, the type will be searched in the project specified in the `.tsconfig`.

-t, --type 'My.Type.Name'
    The type the generated schema will represent. If omitted, the generated schema will contain all
    types found in the files matching path. The same is true if '*' is specified.

-i, --id 'generatedSchemaId'
    The `$id` of the generated schema. If omitted, there will be no `$id`.

-e, --expose <all|none|export>
    all: Create shared $ref definitions for all types.
    none: Do not create shared $ref definitions.
    export (default): Create shared $ref definitions only for exported types (not tagged as `@internal`).

-f, --tsconfig 'my/project/tsconfig.json'
    Use a custom tsconfig file for processing typescript (see https://www.typescriptlang.org/docs/handbook/tsconfig-json.html) instead of the default:
    {
        "compilerOptions": {
            "noEmit": true,
            "emitDecoratorMetadata": true,
            "experimentalDecorators": true,
            "target": "ES5",
            "module": "CommonJS",
            "strictNullChecks": false,
        }
    }

-j, --jsDoc <extended|none|basic>
    none: Do not use JsDoc annotations.
    basic: Read JsDoc annotations to provide schema properties.
    extended (default): Also read @nullable, and @asType annotations.

--unstable
    Do not sort properties.

--strict-tuples
    Do not allow additional items on tuples.

--no-top-ref
    Do not create a top-level $ref definition.

--no-type-check
    Skip type checks for better performance.

--no-ref-encode
    Do not encode references. According to the standard, references must be valid URIs but some tools do not support encoded references.

--validation-keywords
    Provide additional validation keywords to include.

-o, --out
    Specify the output file path. Without this option, the generator logs the response in the console.

--additional-properties <true|false>
    Controls whether or not to allow additional properties for objects that have no index signature.

    true: Additional properties are allowed
    false (default): Additional properties are not allowed

--minify
    Minify generated schema (default: false)

Current state

• interface types
• enum types
• union, tuple, type[] types
• Date, RegExp types
• string, boolean, number types
• "value", 123, true, false, null, undefined literals
• type aliases
• generics
• typeof
• keyof
• conditional types

Run locally

yarn --silent run run --path 'test/valid-data/type-mapped-array/*.ts' --type 'MyObject'

Debug

yarn --silent run debug --path 'test/valid-data/type-mapped-array/*.ts' --type 'MyObject'

And connect via the debugger protocol.

AST Explorer is amazing for developers of this tool!

Publish

Publishing is handled by a 2-branch pre-release process, configured in publish-auto.yml. All changes should be based off the default next branch, and are published automatically.

• PRs made into the default branch are auto-deployed to the next pre-release tag on NPM. The result can be installed with npm install ts-json-schema-generator@next
  • When merging into next, please use the squash and merge strategy.
• To release a new stable version, open a PR from next into stable using this compare link.
  • When merging from next into stable, please use the create a merge commit strategy.