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
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.
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.
// 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
.
Extending the built-in formatting is possible by creating a custom formatter and adding it to the main formatter:
- 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());
}
}
- 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;
});
Similar to custom formatting, extending the built-in parsing works practically the same way:
- 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";
// use typescript exported by TJS to avoid version conflict
import ts from "ts-json-schema-generator";
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
}
}
- 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;
});
-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)
interface
typesenum
typesunion
,tuple
,type[]
typesDate
,RegExp
typesstring
,boolean
,number
types"value"
,123
,true
,false
,null
,undefined
literals- type aliases
- generics
typeof
keyof
- conditional types
yarn --silent run run --path 'test/valid-data/type-mapped-array/*.ts' --type 'MyObject'
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!
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 withnpm install ts-json-schema-generator@next
- When merging into
next
, please use thesquash and merge
strategy.
- When merging into
- To release a new stable version, open a PR from
next
intostable
using this compare link.- When merging from
next
intostable
, please use thecreate a merge commit
strategy.
- When merging from