Skip to content

Latest commit

 

History

History
223 lines (153 loc) · 8.55 KB

README.md

File metadata and controls

223 lines (153 loc) · 8.55 KB
Raw

openapi-io-ts

openapi-io-ts is a code generation tool capable of generating io-ts decoders from an OpenAPI document. It can also generate the code needed to perform the request and decode/parse the response returned by the server.

It is composed of a CLI for code generation and a runtime used to perform the request and decode the response.

Project status

WARNING The project is still an alpha version. The generation of the decoders is working (apart from some edge cases like recursive decoders), but the API of the runtime can be changed in future versions with breaking changes.

CLI

Installation

You can install @openapi-io-ts/cli as a devDependency:

# npm
npm i -D @openapi-io-ts/cli

# yarn
yarn add -D @openapi-io-ts/cli

Usage

You can launch the openapi-io-ts binary by adding a script to the package.json.

{
  "scripts": {
    "generate-api": "openapi-io-ts -i petstore.yaml -o src/api"
  }
}
# npm
npm run generate-api

# yarn
yarn generate-api

The -i parameter is used to pass an OpenAPI document, and the -o parameter is the output directory.

Note: At the moment the only OpenAPI version supported is 3.0. If you have a 2.0 file you have to convert it to 3.0 before. Version 3.1 is not supported at the moment.

Note: Each operation defined in the paths section of the OpenAPI document must have an operationId.

Generated code

The code generated by the CLI contains the io-ts decoders for the schemas that are defined in the OpenAPI document and other objects that are used by the runtime.

The folder structure of the generated code is similar to the structure of the OpenAPI document, with a components folder that contains common objects, and an operations folder containing the various operations.

You can see an example in the examples/react-query-petstore/src/api folder.

Note: The generated code is formatted using the default TypeScript formatter. If you want to format it with another tool like Prettier you can run it after the code generation:

{
  "scripts": {
    "generate": "openapi-io-ts -i petstore.yaml -o src/api && prettier --write ./src/api"
  }
}

Generated client

The CLI will generate code for performing requests defined in the OpenAPI document and decode the returned response. For using the generated code you will need to install the runtime, that is a package that contains functions for preparing the request and decoding the response

Installation

You can install @openapi-io-ts/runtime as a dependency of your project:

# npm
npm i @openapi-io-ts/runtime

# yarn
yarn add @openapi-io-ts/runtime

The runtime has fp-ts, io-ts, and io-ts-types as peer dependencies, so you have to install them as well if they are not already installed in your project:

# npm
npm i fp-ts io-ts io-ts-types

# yarn
yarn add fp-ts io-ts io-ts-types

Generated request functions

For each operation described in the OpenAPI document, the code generator will generate a function for calling the API and decoding the result. The return type of the function is TaskEither<ApiError, ApiResponse<T>>, where T is the type returned by the API on a successful response.

When the TaskEither is executed, the API will be called and the result returned in the promise will be either a Right containing the response or a Left containing an error.

The result is Right when the API returns a successful HTTP code and the decoding of the returned value succeeds. The ApiResponse object will contain the decoded data and a Response object that can be useful for example for reading response headers.

Otherwise, the result will be a Left. The ApiError is a tagged union with one of these types:

  • RequestError when the function using for calling the API threw an exception
  • HttpError when the response HTTP code is not a 2XX code
  • DecodeError when the decoding of the response failed
  • ContentParseError when there was an error during the parse of the returned JSON.

HttpRequestAdapter

The sequence of operation for calling an API are:

  1. Preparing the request: replacing path parameters, encoding query parameters in the query string, creating headers, encoding the body
  2. Performing the request using an HTTP client
  3. Parsing the response and returning either an ApiError or an ApiResponse.

The code generated uses the runtime for steps 1 and 3, and requires a function supplied by the user for step 2. This function takes in input an URL and a RequestInit object created by the first step and returns a promise containing a Response object used by the following step:

type HttpRequestAdapter = (url: string, req: RequestInit) => Promise<Response>;

The URL passed to the function is a relative URL, and you need to add the base URL. In your function, you can also modify the RequestInit object, for example adding authorization headers.

As you can see, the API is identical to the fetch API. In fact, the easiest way to define the HttpRequestAdapter is using the fetch API:

const fetchRequestAdapter: HttpRequestAdapter = (url, init) =>
  fetch(`http://example.com/api${url}`, init);

You can also use a different HTTP client, but you need to convert the RequestInit to the request object used by the client and create a Response object from the response returned by the client. Like fetch, the promise should be always resolved with any HTTP code. Keep also in mind that since Request and Response are part of the fetch API, you should polyfill them if fetch is not available in your execution environment.

Using the request functions

The CLI will generate a request function for each operation defined in the OpenAPI document. There is a function builder (requestFunctionsBuilder) that accepts an HttpRequestAdapter and returns request functions for all operations.

For convenience, the client will also generate a service builder for each tag defined in the OpenAPI document. This service builder takes the result of requestFunctionsBuilder as a parameter and returns an object containing all of the request functions for that tag.

For example, if your OpenAPI document contains a updateUser operation that takes a username as a path parameter, a User object in the body, and returns an User on success, the request functions builder will have an updateUser key that return a function with this signature:

export type UpdateUserRequestParameters = {
  username: string;
};

const requestFunctions = requestFunctionsBuilder(fetchRequestAdapter);

// requestFunctions.updateUser will have this signature:
declare function updateUser({
  params,
  body,
}: {
  params: UpdateUserRequestParameters;
  body: schemas.User;
}): TaskEither<ApiError, ApiResponse<schemas.User>>;

You can use it passing your defined HttpRequestAdapter:

const requestFunctions = requestFunctionsBuilder(fetchRequestAdapter);

requestFunctions
  .updateUser({
    params: { username: "johndoe" },
    body: { email: "[email protected]" },
  })()
  .then(/*...*/);

If this operation has a tag called user you can also use the generated userServiceBuilder

const userService = userServiceBuilder(requestFunctions);

userService
  .updateUser({
    params: { username: "johndoe" },
    body: { email: "[email protected]" },
  })()
  .then(/*...*/);

You can see more examples in the examples/react-query-petstore/src/api folder.

As described above, the type returned by the generated request functions is a TaskEither, which represents an asynchronous computation that can fail with an ApiError or succeeds with an ApiResponse.

If you want to run the asynchronous computation you can call the () method of the TaskEither and then distinguish between successful and unsuccessful responses using the fold method of the Either data type.

import * as E from "fp-ts/Either";

requestFunctions
  .updateUser({
    params: { username: "johndoe" },
    body: { email: "[email protected]" },
  })()
  .then(
    E.fold(
      (e) => {
        /* e is an ApiError */
      },
      (res) => {
        /* res.data is schemas.User, res.response is the Response object */
      }
    )
  );

Contributors

Made with contributors-img.

License

MIT