wretch

Wretch

A tiny (~1.8KB g-zipped) wrapper built around fetch with an intuitive syntax.

f[ETCH] [WR]apper

Wretch 3.0 is now live 🎉 ! Check out the Migration Guide for upgrading from v2, and please have a look at the releases and the changelog after each update for new features and breaking changes.

And if you like the library please consider becoming a sponsor ❤️.

Features

wretch is a small wrapper around fetch designed to simplify the way to perform network requests and handle responses.

• 🪶 Small - core is less than 1.8KB g-zipped
• 💡 Intuitive - lean API, handles errors, headers and (de)serialization
• 🧊 Immutable - every call creates a cloned instance that can then be reused safely
• 🔌 Modular - plug addons to add new features, and middlewares to intercept requests
• 🧩 Isomorphic - compatible with modern browsers, Node.js 22+, Deno and Bun
• 🦺 Type safe - strongly typed, written in TypeScript
• ✅ Proven - fully covered by unit tests and widely used
• 💓 Maintained - alive and well for many years

Table of Contents

• Quick Start
• Motivation
• Installation
• Compatibility
• Usage
• Recipes
• Api
• Addons
• Middlewares
• Migration Guides
• License

Quick Start

# 1️⃣ Install
npm i wretch

// 2️⃣ Import and create a reusable API client
import wretch from "wretch"

const api = wretch("https://jsonplaceholder.typicode.com")
  .options({ mode: "cors" })

// 3️⃣ Make requests with automatic JSON handling
const post = await api.get("/posts/1").json()
console.log(post.title)

// 4️⃣ POST with automatic serialization
const created = await api
  .post({ title: "New Post", body: "Content", userId: 1 }, "/posts")
  .json()

// 5️⃣ Handle errors elegantly
await api
  .get("/posts/999")
  .notFound(() => console.log("Post not found!"))
  .json()

// 6️⃣ Different response types
const text = await api.get("/posts/1").text()      // Raw text
const response = await api.get("/posts/1").res()   // Raw Response object
const blob = await api.get("/photos/1").blob()     // Binary data

Motivation

Because having to write a second callback to process a response body feels awkward.

Fetch needs a second callback to process the response body.

fetch("https://jsonplaceholder.typicode.com/posts/1")
  .then(response => response.json())
  .then(json => {
    //Do stuff with the parsed json
  });

Wretch does it for you.

// Use .res for the raw response, .text for raw text, .json for json, .blob for a blob …
wretch("https://jsonplaceholder.typicode.com/posts/1")
  .get()
  .json(json => {
    // Do stuff with the parsed json
  });

Because manually checking and throwing every request error code is tedious.

Fetch won’t reject on HTTP error status.

fetch("https://jsonplaceholder.typicode.com/posts/1")
  .then(response => {
    if(!response.ok) {
      if(response.status === 404) throw new Error("Not found")
      else if(response.status === 401) throw new Error("Unauthorized")
      else if(response.status === 418) throw new Error("I'm a teapot !")
      else throw new Error("Other error")
    }
    else {/* … */}
  })
  .then(data => {/* … */})
  .catch(error => { /* … */ })

Wretch throws when the response is not successful and contains helper methods to handle common codes.

wretch("https://jsonplaceholder.typicode.com/posts/1")
  .get()
  .notFound(error => { /* … */ })
  .unauthorized(error => { /* … */ })
  .error(418, error => { /* … */ })
  .res(response => {/* … */ })
  .catch(error => { /* uncaught errors */ })

Because sending a json object should be easy.

With fetch you have to set the header, the method and the body manually.

fetch("https://jsonplaceholder.typicode.com/posts", {
  method: "POST",
  headers: { "Content-Type": "application/json" },
  body: JSON.stringify({ "hello": "world" })
}).then(response => {/* … */})
// Omitting the data retrieval and error management parts…

With wretch, you have shorthands at your disposal.

wretch("https://jsonplaceholder.typicode.com/posts")
  .post({ "hello": "world" })
  .res(response => { /* … */ })

Because configuration should not rhyme with repetition.

A Wretch object is immutable which means that you can reuse previous instances safely.

const token = "MY_SECRET_TOKEN"

// Cross origin authenticated requests on an external API
const externalApi = wretch("https://jsonplaceholder.typicode.com") // Base url
  // Authorization header
  .auth(`Bearer ${token}`)
  // Cors fetch options
  .options({ credentials: "include", mode: "cors" })
  // Handle 403 errors
  .resolve((w) => w.forbidden(error => { /* Handle all 403 errors */ }));

// Fetch a resource
const resource = await externalApi
  // Add a custom header for this request
  .headers({ "If-Unmodified-Since": "Wed, 21 Oct 2015 07:28:00 GMT" })
  .get("/posts/1")
  .json(() => {/* do something with the resource */});

// Post a resource
externalApi
  .url("/posts")
  .post({ "Shiny new": "resource object" })
  .json(() => {/* do something with the created resource */});

Installation

Package Manager

npm i wretch # or yarn/pnpm add wretch

<script> tag

The package contains multiple bundles depending on the format and feature set located under the /dist/bundle folder.

Bundle variants

💡 If you pick the core bundle, then to plug addons you must import them separately from /dist/bundle/addons/[addonName].min.js

Feature set	File Name
Core features only	wretch.min.js
Core + all addons	wretch.all.min.js

Format	Extension
ESM	.min.mjs
CommonJS	.min.cjs
UMD	.min.js

<!--
  Pick your favourite CDN:
    - https://unpkg.com/wretch
    - https://cdn.jsdelivr.net/npm/wretch/
    - https://www.skypack.dev/view/wretch
    - https://cdnjs.com/libraries/wretch
    - …
-->

<!-- UMD import as window.wretch -->
<script src="https://unpkg.com/wretch"></script>

<!-- Modern import -->
<script type="module">
  import wretch from 'https://cdn.skypack.dev/wretch/dist/bundle/wretch.all.min.mjs'

  // … //
</script>

Compatibility

Browsers

wretch@^3 is compatible with modern browsers only. For older browsers please use wretch@^1.

Node.js

Wretch is compatible with and tested in Node.js >= 22.

For older Node.js versions, please use wretch@^2.

Node.js 22+ includes native fetch support and all required Web APIs (FormData, URLSearchParams, AbortController, etc.) out of the box, so no polyfills are needed.

Deno

Works with Deno out of the box.

deno add npm:wretch

import wretch from "wretch";

const text = await wretch("https://httpbingo.org").get("/status/200").text();
console.log(text); // -> { "code": 200, "description": "OK" }

Bun

Works with Bun out of the box.

bun add wretch

import wretch from "wretch";

const text = await wretch("https://httpbingo.org").get("/status/200").text();
console.log(text); // -> { "code": 200, "description": "OK" }

Usage

Import

// ECMAScript modules
import wretch from "wretch"
// CommonJS
const wretch = require("wretch")
// Global variable (script tag)
window.wretch

Common Use Cases

// 🌐 REST API Client
const api = wretch("https://jsonplaceholder.typicode.com")
  .auth("Bearer token")
  .resolve(r => r.json())

const users = await api.get("/users")
const user = await api.post({ name: "John" }, "/users")

// 📤 File Upload with Progress
import ProgressAddon from "wretch/addons/progress"
import FormDataAddon from "wretch/addons/formData"

await wretch("https://httpbingo.org/post")
  .addon([FormDataAddon, ProgressAddon()])
  .formData({ file: file })
  .post()
  .progress((loaded, total) => console.log(`${(loaded/total*100).toFixed()}%`))
  .json()

// 🔄 Auto-retry on Network Failure
import { retry } from "wretch/middlewares"

const resilientApi = wretch()
  .middlewares([retry({ maxAttempts: 3, retryOnNetworkError: true })])

// 🎯 Type-safe TypeScript
interface User { id: number; name: string; email: string }

const user = await wretch("https://jsonplaceholder.typicode.com")
  .get("/users/1")
  .json<User>() // Fully typed!

// 🔐 Automatic Token Refresh
const api = wretch("https://httpbingo.org/basic-auth/user/pass")
  .addon(BasicAuthAddon)
  .resolve(w => w.unauthorized(async (error, req) => {
    const newToken = await refreshToken()
    return req
    .basicAuth("user", "pass")
    .unauthorized(e => {
      console.log("Still unauthorized after token refresh");
      throw e
    })
    .fetch()
    .json()
  }))

Custom Fetch Implementation

You can provide a custom fetch implementation using the .fetchPolyfill() method. This is useful for for a variety of use cases including mocking, adding logging, timing, or other custom behavior to all requests made through a wretch instance.

import wretch from "wretch"

// Per-instance custom fetch
const api = wretch("https://jsonplaceholder.typicode.com")
  .fetchPolyfill((url, opts) => {
    console.log('Fetching:', url)
    console.time(url)
    return fetch(url, opts).finally(() => {
      console.timeEnd(url)
    })
  })

await api.get("/posts").json()

toFetch() - Fetch Adapter 🔗

Converts a wretch instance into a fetch-like function, preserving all configuration (middlewares, catchers, headers, etc.). Useful for integrating wretch with libraries that expect a fetch signature.

const myFetch = wretch("https://jsonplaceholder.typicode.com")
  .auth("Bearer token")
  .catcher(401, (err) => console.log("Unauthorized"))
  .toFetch()

// Use like regular fetch
const response = await myFetch("/users", { method: "GET" })

// Pass to libraries expecting fetch
import { createClient } from "@apollo/client"

const client = createClient({
  fetch: wretch().auth("Bearer token").toFetch()
})

Chaining

A high level overview of the successive steps that can be chained to perform a request and parse the result.

┌─────────────────────────────────────────────────────────────────┐
│                        Request Chain                            │
│                                                                 │
│  wretch(url) ──> .helper() ──> .body() ──> .httpMethod()        │
│                    ↓              ↓            ↓                │
│                 .headers()     .json()      .get()              │
│                 .auth()        .body()      .post()             │
│                 .options()                  .put()              │
│                                             .delete()           │
└─────────────────────────────────────────────────────────────────┘
                              ↓
                    [ fetch() is called ]
                              ↓
┌─────────────────────────────────────────────────────────────────┐
│                       Response Chain                            │
│                                                                 │
│  .catcher() ──> .responseType() ──> Promise ──> .then()/.catch()│
│      ↓              ↓                                           │
│  .notFound()     .json()                                        │
│  .unauthorized() .text()                                        │
│  .error()        .blob()                                        │
│                  .res()                                         │
└─────────────────────────────────────────────────────────────────┘

Step-by-step breakdown:

// First, instantiate wretch
wretch(baseUrl, baseOptions)

The "request" chain starts here.

  // Optional - A set of helper methods to set the default options, set accept header, change the current url…
  .<helper method(s)>()
  // Optional - Serialize an object to json or FormData formats and sets the body & header field if needed
  .<body type>()
    // Required - Sends the get/put/post/delete/patch request.
  .<http method>()

The "response" chain starts here.

Fetch is called after the request chain ends and before the response chain starts.
The request is on the fly and now it is time to chain catchers and finally call a response type handler.

  // Optional - You can chain error handlers here
  .<catcher(s)>()
  // Required - Specify the data type you need, which will be parsed and handed to you
  .<response type>()
  // >> Ends the response chain.

From this point on, wretch returns a standard Promise.

  .then(…)
  .catch(…)

Concrete Example

Here's how the chaining works in practice:

await wretch("https://api.example.com")        // Base URL
  .headers({ "X-Api-Key": "secret" })          // Helper method
  .query({ limit: 10 })                        // Helper method
  .json({ name: "Alice", role: "admin" })      // Body type
  .post("/users")                              // HTTP method (starts request)
  .badRequest(err => console.log("Invalid"))   // Catcher
  .unauthorized(err => console.log("No auth")) // Catcher
  .json(user => console.log(user))             // Response type

Recipes

Looking for common patterns and solutions? Check out the Recipes Guide for practical examples covering:

• Error Handling - Parsing error response bodies, custom error types, global handlers
• TypeScript Patterns - Typing precomposed instances, reusable catchers
• File Uploads - Progress tracking, FormData handling
• Query Strings - Filtering undefined values
• Request Control - Combining timeouts with AbortControllers, aborting on errors
• Advanced Patterns - Token refresh & replay, schema validation, async polling

API 🔗

💡 The API documentation is now autogenerated and hosted separately, click the links access it.

Factory Method 🔗

The default export is a factory function used to instantiate wretch.

import wretch from "wretch"

const api = wretch("http://domain.com/", { cache: "default" })

Helper Methods 🔗

Helper Methods are used to configure the request and program actions.

Available methods: .url() · .options() · .headers() · .auth() · .accept() · .content() · .signal() · .toFetch() · .fetchPolyfill() · .catcher() · .catcherFallback() · .customError() · .defer() · .resolve() · .middlewares() · .addon() · .polyfills()

let api = wretch("http://domain.com/")

api = api
  .url("/posts/1")
  .headers({ "Cache-Control": "no-cache" })
  .content("text/html")

Body Types 🔗

Specify a body type if uploading data. Can also be added through the HTTP Method argument.

Available methods: .body() · .json()

let api = wretch("http://domain.com/")

api = api.body("<html><body><div/></body></html>")

HTTP Methods 🔗

Sets the HTTP method and sends the request.

Calling an HTTP method ends the request chain and returns a response chain. You can pass optional url and body arguments to these methods.

Available methods: .get() · .post() · .put() · .patch() · .delete() · .head() · .opts()

const api = wretch("http://jsonplaceholder.typicode.com")

// These shorthands:
api.get("/posts");
api.post({ json: "body" }, "/posts");
// Are equivalent to:
api.url("/posts").get();
api.json({ json: "body" }).url("/posts").post();

NOTE: if the body argument is an Object it is assumed that it is a JSON payload and it will have the same behaviour as calling .json(body) unless the Content-Type header has been set to something else beforehand.

Catchers 🔗

Catchers are optional, but if none are provided an error will still be thrown for http error codes and it will be up to you to catch it.

Available methods: .badRequest() · .unauthorized() · .forbidden() · .notFound() · .timeout() · .internalError() · .error() · .fetchError()

wretch("http://domain.com/resource")
  .get()
  .badRequest((err) => console.log(err.status))
  .unauthorized((err) => console.log(err.status))
  .forbidden((err) => console.log(err.status))
  .notFound((err) => console.log(err.status))
  .timeout((err) => console.log(err.status))
  .internalError((err) => console.log(err.status))
  .error(418, (err) => console.log(err.status))
  .fetchError((err) => console.log(err))
  .res();

The error passed to catchers is enhanced with additional properties.

type WretchError = Error & {
  status: number;
  response: WretchResponse;
  url: string;
};

By default, error.message is set to the response body text (or statusText if body parsing fails).

Request Replay

The original request is passed along the error and can be used in order to perform an additional request.

await wretch("https://httpbingo.org/basic-auth/user/pass")
  .addon(BasicAuthAddon)
  .basicAuth("user", "wrongpass")
  .get()
  .unauthorized(async (error, req) => {
    // Renew credentials
    const password = await wretch("https://httpbingo.org/base64/decode/cGFzcw==").get().text();
    // Replay the original request with new credentials
    return req
      .basicAuth("user", password)
      .fetch()
      .unauthorized((err) => { throw err })
      .json();
  })
  .json()
  // The promise chain is preserved as expected
  // ".then" will be performed on the result of the original request
  // or the replayed one (if a 401 error was thrown)
  .then(() => { /* … */ });

Response Types 🔗

Setting the final response body type ends the chain and returns a regular promise.

All these methods accept an optional callback, and will return a Promise resolved with either the return value of the provided callback or the expected type.

Available methods: .res() · .json() · .text() · .blob() · .arrayBuffer() · .formData()

const ENDPOINT = "https://jsonplaceholder.typicode.com/posts/1"

// Without a callback
wretch(ENDPOINT)
  .get()
  .json()
  .then(json => {
    /* the json argument is the parsed json of the response body */
  })
// Without a callback using await
const json = await wretch(ENDPOINT).get().json()
// With a callback the value returned is passed to the Promise
wretch(ENDPOINT).get().json(json => "Hello world!").then(console.log) // => Hello world!

If an error is caught by catchers, the response type handler will not be called.

Addons

Addons are separate pieces of code that you can import and plug into wretch to add new features.

import FormDataAddon from "wretch/addons/formData"
import QueryStringAddon from "wretch/addons/queryString"

// Add both addons
const w = wretch().addon([FormDataAddon, QueryStringAddon])

// Additional features are now available
w.formData({ hello: "world" }).query({ check: true })

Typescript should also be fully supported and will provide completions.

https://user-images.githubusercontent.com/3428394/182319457-504a0856-abdd-4c1d-bd04-df5a061e515d.mov

QueryString 🔗

Used to construct and append the query string part of the URL from an object.

import QueryStringAddon from "wretch/addons/queryString"

let w = wretch("http://example.com").addon(QueryStringAddon);
// url is http://example.com
w = w.query({ a: 1, b: 2 });
// url is now http://example.com?a=1&b=2
w = w.query({ c: 3, d: [4, 5] });
// url is now http://example.com?a=1&b=2c=3&d=4&d=5
w = w.query("five&six&seven=eight");
// url is now http://example.com?a=1&b=2c=3&d=4&d=5&five&six&seven=eight
w = w.query({ reset: true }, { replace: true });
// url is now  http://example.com?reset=true

FormData 🔗

Adds a helper method to serialize a multipart/form-data body from an object.

import FormDataAddon from "wretch/addons/formData"

const form = {
  duck: "Muscovy",
  duckProperties: {
    beak: {
      color: "yellow",
    },
    legs: 2,
  },
  ignored: {
    key: 0,
  },
};

// Will append the following keys to the FormData payload:
// "duck", "duckProperties[beak][color]", "duckProperties[legs]"
wretch("https://httpbingo.org/post").addon(FormDataAddon).formData(form, { recursive: ["ignored"] }).post();

FormUrl 🔗

Adds a method to serialize a application/x-www-form-urlencoded body from an object.

import FormUrlAddon from "wretch/addons/formUrl"

const form = { a: 1, b: { c: 2 } };
const alreadyEncodedForm = "a=1&b=%7B%22c%22%3A2%7D";

// Automatically sets the content-type header to "application/x-www-form-urlencoded"
wretch("https://httpbingo.org/post").addon(FormUrlAddon).formUrl(form).post();
wretch("https://httpbingo.org/post").addon(FormUrlAddon).formUrl(alreadyEncodedForm).post();

Abort 🔗

Adds the ability to abort requests and set timeouts using AbortController and signals under the hood.

import AbortAddon from "wretch/addons/abort"

Use cases :

const [c, w] = wretch("https://httpbingo.org/get")
  .addon(AbortAddon())
  .get()
  .onAbort((_) => console.log("Aborted !"))
  .controller();

w.text((_) => console.log("should never be called"));
c.abort();

// Or :

const controller = new AbortController();

wretch("https://httpbingo.org/get")
  .addon(AbortAddon())
  .signal(controller)
  .get()
  .onAbort((_) => console.log("Aborted !"))
  .text((_) => console.log("should never be called"));

controller.abort();

wretch("https://httpbingo.org/delay/2")
  .addon(AbortAddon())
  .get()
   // 1 second timeout
  .setTimeout(1000)
  .onAbort(_ => {
    console.log("Request timed out")
  })
  .json(_ => {
    console.log("Response received in time")
  })

BasicAuth 🔗

Adds the ability to set the Authorization header for the basic authentication scheme without the need to manually encode the username/password.

Also, allows using URLs with wretch that contain credentials, which would otherwise throw an error.

import BasicAuthAddon from "wretch/addons/basicAuth"

const user = "user"
const pass = "pass"

// Automatically sets the Authorization header to "Basic " + <base64 encoded credentials>
wretch("https://httpbingo.org/get")
  .addon(BasicAuthAddon)
  .basicAuth(user, pass)
  .get()

// Allows using URLs with credentials in them
wretch(`https://${user}:${pass}@httpbingo.org/basic-auth/${user}/${pass}`)
  .addon(BasicAuthAddon)
  .get()

Progress 🔗

Adds the ability to monitor progress when downloading or uploading.

Compatible with all platforms implementing the TransformStream WebAPI.

Download progress:

import ProgressAddon from "wretch/addons/progress"

await wretch("https://httpbingo.org/bytes/5000")
  .addon(ProgressAddon())
  .get()
  // Called with the number of bytes loaded and the total number of bytes to load
  .progress((loaded, total) => {
    console.log(`Download: ${(loaded / total * 100).toFixed(0)}%`)
  })
  .blob()

Upload progress:

import ProgressAddon from "wretch/addons/progress"
import FormDataAddon from "wretch/addons/formData"

const formData = new FormData()
formData.append('file', file)

wretch("https://httpbingo.org/post")
  .addon([ProgressAddon(), FormDataAddon])
  .onUpload((loaded, total) => {
    console.log(`Upload: ${(loaded / total * 100).toFixed(0)}%`)
  })
  .post(formData)
  .res()

Note for browsers: Upload progress requires HTTPS (HTTP/2) in Chrome/Chromium and doesn't work in Firefox due to streaming limitations. Works fully in Node.js.

Performance 🔗

Adds the ability to measure requests using the Performance Timings API.

Uses the Performance API (browsers & Node.js) to expose timings related to the underlying request.

💡 Make sure to follow the additional instructions in the documentation to setup Node.js if necessary.

Middlewares

Middlewares are functions that can intercept requests before being processed by Fetch. Wretch includes a helper to help replicate the middleware style.

import wretch from "wretch"
import { retry, dedupe } from "wretch/middlewares"

const w = wretch().middlewares([retry(), dedupe()])

💡 The following middlewares were previously provided by the wretch-middlewares package.

Retry 🔗

Retries a request multiple times in case of an error (or until a custom condition is true).

💡 By default, the request will be retried only for server errors (5xx) and other non-successful responses, but not for client errors (4xx).

// To retry on all non-2xx responses (including 4xx):
until: (response, error) => !!response && response.ok

import wretch from 'wretch'
import { retry } from 'wretch/middlewares'

wretch().middlewares([
  retry({
    /* Options - defaults below */
    delayTimer: 500,
    delayRamp: (delay, nbOfAttempts) => delay * nbOfAttempts,
    maxAttempts: 10,
    until: (response, error) => !!response && (response.ok || (response.status >= 400 && response.status < 500)),
    onRetry: undefined,
    retryOnNetworkError: false,
    resolveWithLatestResponse: false
  })
])

// You can also return a Promise, which is useful if you want to inspect the body:
wretch().middlewares([
  retry({
    until: response =>
      response?.clone().json().then(body =>
        body.field === 'something'
      ) || false
  })
])

Dedupe 🔗

Prevents having multiple identical requests on the fly at the same time.

import wretch from 'wretch'
import { dedupe } from 'wretch/middlewares'

wretch().middlewares([
  dedupe({
    /* Options - defaults below */
    skip: (url, opts) => opts.skipDedupe || opts.method !== 'GET',
    key: (url, opts) => opts.method + '@' + url,
    resolver: response => response.clone()
  })
])

Throttling Cache 🔗

A throttling cache which stores and serves server responses for a certain amount of time.

import wretch from 'wretch'
import { throttlingCache } from 'wretch/middlewares'

wretch().middlewares([
  throttlingCache({
    /* Options - defaults below */
    throttle: 1000,
    skip: (url, opts) => opts.skipCache || opts.method !== 'GET',
    key: (url, opts) => opts.method + '@' + url,
    clear: (url, opts) => false,
    invalidate: (url, opts) => null,
    condition: response => response.ok,
    flagResponseOnCacheHit: '__cached'
  })
])

Delay 🔗

Delays the request by a specific amount of time.

import wretch from 'wretch'
import { delay } from 'wretch/middlewares'

wretch().middlewares([
  delay(1000)
])

Writing a Middleware

Basically a Middleware is a function having the following signature :

// A middleware accepts options and returns a configured version
type Middleware = (options?: { [key: string]: any }) => ConfiguredMiddleware;
// A configured middleware (with options curried)
type ConfiguredMiddleware = (next: FetchLike) => FetchLike;
// A "fetch like" function, accepting an url and fetch options and returning a response promise
type FetchLike = (
  url: string,
  opts: WretchOptions,
) => Promise<WretchResponse>;

Context

If you need to manipulate data within your middleware and expose it for later consumption, a solution could be to pass a named property to the wretch options (suggested name: context).

Your middleware can then take advantage of that by mutating the object reference.

const contextMiddleware = (next) =>
  (url, opts) => {
    if (opts.context) {
      // Mutate "context"
      opts.context.property = "anything";
    }
    return next(url, opts);
  };

// Provide the reference to a "context" object
const context = {};
const res = await wretch("https://httpbingo.org/get")
  // Pass "context" by reference as an option
  .options({ context })
  .middlewares([contextMiddleware])
  .get()
  .res();

console.log(context.property); // prints "anything"

Advanced examples

 👀 Show me the code

/* A simple delay middleware. */
const delayMiddleware = delay => next => (url, opts) => {
  return new Promise(res => setTimeout(() => res(next(url, opts)), delay))
}

/* Returns the url and method without performing an actual request. */
const shortCircuitMiddleware = () => next => (url, opts) => {
  // We create a new Response object to comply because wretch expects that from fetch.
  const response = new Response(url)
  // Instead of calling next(), returning a Response Promise bypasses the rest of the chain.
  return Promise.resolve(response)
}

/* Logs all requests passing through. */
const logMiddleware = () => next => (url, opts) => {
  console.log(opts.method + "@" + url)
  return next(url, opts)
}

/* A throttling cache. */
const cacheMiddleware = (throttle = 0) => {

  const cache = new Map()
  const inflight = new Map()
  const throttling = new Set()

  return next => (url, opts) => {
    const key = opts.method + "@" + url

    if(!opts.noCache && throttling.has(key)) {
      // If the cache contains a previous response and we are throttling, serve it and bypass the chain.
      if(cache.has(key))
        return Promise.resolve(cache.get(key).clone())
      // If the request in already in-flight, wait until it is resolved
      else if(inflight.has(key)) {
        return new Promise((resolve, reject) => {
          inflight.get(key).push([resolve, reject])
        })
      }
    }

    // Init. the pending promises Map
    if(!inflight.has(key))
      inflight.set(key, [])

    // If we are not throttling, activate the throttle for X milliseconds
    if(throttle && !throttling.has(key)) {
      throttling.add(key)
      setTimeout(() => { throttling.delete(key) }, throttle)
    }

    // We call the next middleware in the chain.
    return next(url, opts)
      .then(_ => {
        // Add a cloned response to the cache
        cache.set(key, _.clone())
        // Resolve pending promises
        inflight.get(key)?.forEach((([resolve, reject]) => resolve(_.clone())))
        // Remove the inflight pending promises
        inflight.delete(key)
        // Return the original response
        return _
      })
      .catch(_ => {
        // Reject pending promises on error
        inflight.get(key)?.forEach(([resolve, reject]) => reject(_))
        inflight.delete(key)
        throw _
      })
  }
}

// To call a single middleware
const cache = cacheMiddleware(1000)
wretch("https://httpbingo.org/get").middlewares([cache]).get()

// To chain middlewares
wretch("https://httpbingo.org/get").middlewares([
  logMiddleware(),
  delayMiddleware(1000),
  shortCircuitMiddleware()
]).get().text(text => console.log(text))

// To test the cache middleware more thoroughly
const wretchCache = wretch("https://httpbingo.org").middlewares([cacheMiddleware(500)])
const printResource = (url, timeout = 0) => {
  return new Promise(resolve => setTimeout(async () => {
    wretchCache.url(url).get().notFound(console.error).text(resource => {
      console.log(resource)
      resolve(resource)
    })
  }, timeout))
}
// The resource url, change it to an invalid route to check the error handling
const resourceUrl = "/base64/decode/YWVhY2YyYWYtODhlNi00ZjgxLWEwYjAtNzdhMTIxNTA0Y2E4"
// Only two actual requests are made here even though there are 30 calls
await Promise.all(Array.from({ length: 10 }).flatMap(() =>
  [
    printResource(resourceUrl),
    printResource(resourceUrl, 200),
    printResource(resourceUrl, 700)
  ]
))

Limitations

Cloudflare Workers

It seems like using wretch in a Cloudflare Worker environment is not possible out of the box, as the Cloudflare Response implementation does not implement the type property and throws an error when trying to access it.

Please check the issue #159 for more information.

Workaround

The following middleware should fix the issue (thanks @jimmed 🙇):

wretch().middlewares([
  (next) => async (url, opts) => {
    const response = await next(url, opts);
    try {
      Reflect.get(response, "type", response);
    } catch (error) {
      Object.defineProperty(response, "type", {
        get: () => "default",
      });
    }
    return response;
  },
])

Headers Case Sensitivity

The Request object from the Fetch API uses the Headers class to store headers under the hood. This class is case-insensitive, meaning that setting both will actually appends the value to the same key:

const headers = new Headers();
headers.append("Accept", "application/json");
headers.append("accept", "application/json");
headers.forEach((value, key) => console.log(key, value));
// prints: accept application/json, application/json

When using wretch, please be mindful of this limitation and avoid setting the same header multiple times with a different case:

wretch("https://httpbingo.org/post")
  .headers({ "content-type": "application/json" })
  // .json is a shortcut for .headers("Content-Type": "application/json").post().json()
  .json({ foo: "bar" })
  // Wretch stores the headers inside a plain javascript object and will not deduplicate them.
  // Later on when fetch builds the Headers object the content type header will be set twice
  // and its value will be "application/json, application/json".
  // Ultimately this is certainly not what you want.

Please check the issue #80 for more information.

Workaround

You can use the following middleware to deduplicate headers (thanks @jimmed 🙇):

export const manipulateHeaders =
  callback => next => (url, { headers, ...opts }) => {
    const nextHeaders = callback(new Headers(headers))
    return next(url, { ...opts, headers: nextHeaders })
  }

export const dedupeHeaders = (dedupeHeaderLogic = {}) => {
  const deduperMap = new Map(
    Object.entries(dedupeHeaderLogic).map(([k, v]) => [k.toLowerCase(), v]),
  )
  const dedupe = key =>
    deduperMap.get(key.toLowerCase()) ?? (values => new Set(values))

  return manipulateHeaders((headers) => {
    Object.entries(headers.raw()).forEach(([key, values]) => {
      const deduped = Array.from(dedupe(key)(values))
      headers.delete(key)
      deduped.forEach((value, index) =>
        headers[index ? 'append' : 'set'](key.toLowerCase(), value),
      )
    })
    return headers
  })
}

// By default, it will deduplicate identical values for a given header. This can be used as follows:
wretch().middlewares([dedupeHeaders()])
// If there is a specific header for which the defaults cause problems, then you can provide a callback to handle deduplication yourself:
wretch().middlewares([
  dedupeHeaders({
    Accept: (values) => values.filter(v => v !== '*/*')
  })
])

Migration Guides

Comprehensive migration guides are available for upgrading between major versions:

• Migration Guide: v2 to v3
• Migration Guide: v1 to v2

License

MIT

Similar packages

Other packages similar to wretch

axios

Axios is a promise-based HTTP client for the browser and Node.js. It provides a more comprehensive feature set compared to wretch, including request cancellation, automatic JSON transformation, and more. However, Axios is larger in size compared to wretch.

superagent

Superagent is a small progressive client-side HTTP request library. It offers a similar promise-based API and supports a wide range of HTTP methods. Superagent is more feature-rich but also larger in size compared to wretch.

ky

Ky is a tiny and elegant HTTP client based on the Fetch API. It offers a similar promise-based API and is known for its small size and simplicity, much like wretch. Ky focuses on modern browsers and environments, whereas wretch provides more flexibility in handling requests.