up-fetch

Tiny & Composable fetch configuration tool with sensible defaults.

➡️ Highlights

• 🚀 Lightweight - 1kB gzipped, no dependency
• 🤩 Familiar - same API as fetch with additional options and sensible defaults
• 🎯 Intuitive - define the params and body as plain objects, the Response is parsed out of the box
• 🔥 Composable - bring your own serialization, parsing and throwing strategies when needed
• 👁️ observable - thanks to the built in interceptors
• 💫 Reusable - create instances with custom defaults
• 💪 Strongly typed - best in class type inferrence and autocomplete
• 🤯 Validation adapters - (opt-in) validate the data for maximum type safety with zod or valibot
• 📦 Tree Shakable - You only get what you use

➡️ QuickStart

npm i up-fetch # or bun i up-fetch

Create a new upfetch instance

import { up } from 'up-fetch'

const upfetch = up(fetch)

Make a fetch request

const todo = await upfetch('https://a.b.c', {
   method: 'POST',
   body: { hello: 'world' },
})

You can set some defaults for all requests. They are evaluated before each request to avoid sending stale values

const upfetch = up(fetch, () => ({
   baseUrl: 'https://a.b.c',
   headers: { Authorization: localStorage.getItem('bearer-token') },
}))

Since up extends the provided fetch API options, anything that can be done with fetch can also be done with upfetch.

// the baseUrl and Authorization header can be omitted
const todo = await upfetch('/todos', {
   method: 'POST',
   body: { title: 'Hello World' },
   params: { some: 'query params' },
   headers: { 'X-Header': 'Another header' },
   signal: AbortSignal.timeout(5000),
   keepalive: true,
   cache: 'no-store',
})

Any fetch API implementation can be used, like undici or node-fetch

import { fetch } from 'undici'

const upfetch = up(fetch)

Raw fetch vs upfetch

fetch that throws when response.ok is false:

You should first create a custom ResponseError class that extends the built in Error class in order to expose the response and the parsed response data.

A naive implementation might look like this

export class ResponseError extends Error {
   constructor(response, data) {
      super(`Request failed with status ${res.status}`)
      this.data = data
      this.name = 'ResponseError'
      this.response = response
      this.status = response.status
   }
}

Then proceed with the definition of the fetcher itself. The following is a simplified example

const fetchTodos = async ({ search, take, skip }) => {
   const response = await fetch(
      `https://a.b.c/?search=${search}&skip=${skip}&take=${take}`,
   )
   const data = await response.json()
   if (response.ok) {
      return data
   }
   throw new ResponseError(response, data)
}

Same example using up-fetch:

Granted that you've already created an up(fetch) instance the previous example can be written like this:

const fetchData = (params) => upfetch('https://a.b.c', { params })

➡️ Features

✔️ Set defaults for an upfetch instance

up-fetch default behaviour can be entirely customized

const upfetch = up(fetch, () => ({
   baseUrl: 'https://a.b.c',
   headers: { 'X-Header': 'hello world' },
}))

See the full options list for more details.

✔️ Set the url params as object

// before
fetch(`https://a.b.c/?search=${search}&skip=${skip}&take=${take}`)

// after
upfetch('https://a.b.c', {
   params: { search, skip, take },
})

✔️ baseUrl option

Set the baseUrl when you create the instance

export const upfetch = up(fetch, () => ({
   baseUrl: 'https://a.b.c',
}))

You can then omit it on all requests

const todos = await upfetch('/todos')

✔️ Automatic Response parsing

The response is automatically parsed to json with a fallback to text in case of invalid json.

The parsing method is customizable via the parseResponse option

// before
const response = await fetch('https://a.b.c')
const todos = await response.json()

// after
const todos = await upfetch('https://a.b.c')

✔️ throws by default

up-fetch throws a ResponseError when response.ok is false.
This behavior can be customized using the throwResponseErrorWhen option

The parsed error body is available with error.data.
The raw Response can be accessed with error.response.
The raw status can be accessed with error.status.
The options used make the api call are available with error.options.

import { isResponseError } from 'up-fetch'
import { upfetch } from '...'

try {
   await upfetch('https://a.b.c')
} catch (error) {
   if (isResponseError(error)) {
      console.log(error.data)
      console.log(error.status)
   } else {
      console.log('Request error')
   }
}

✔️ Set the body as object

The 'Content-Type': 'application/json' header is automatically set when the body is a Jsonifiable object or array. Plain objects, arrays and classes with a toJSON method are Jsonifiable.

// before
fetch('https://a.b.c', {
   method: 'POST',
   headers: { 'Content-Type': 'application/json' },
   body: JSON.stringify({ post: 'Hello World' }),
})

// after
upfetch('https://a.b.c', {
   method: 'POST',
   body: { post: 'Hello World' },
})

✔️ Data Validation

up-fetch has built-in validation adapters for zod and valibot, see the adapters list below.

Example: with zod

npm i zod

import { z } from 'zod'
import { withZod } from 'up-fetch/with-zod'
import { upfetch } from './abc'

const todo = await upfetch('/todo/1', {
   parseResponse: withZod(
      z.object({
         id: z.number(),
         title: z.string(),
         description: z.string(),
         createdOn: z.string(),
      }),
   ),
})
// todo is properly typed in case of validation success

Using an adapter ensures that data flows properly through the onParsingError and onSuccess interceptors

In case of error the validation adapters will throw.

✔️ Transform

The parsed response data can be transformed before being passed to .then

const todos = await upfetch('https://a.b.c', {
   // Transform the data as you like
   // the type of `data` is inferred from `parseResponse`
   transform: (todos) => todos.map((todo) => new Todo(todo)),
})

✔️ Interceptors

You can setup the interceptors for all requests

const upfetch = up(fetch, () => ({
   onBeforeFetch: (options) => console.log('Before fetch'),
   onSuccess: (data, options) => console.log(data),
   onResponseError: (error, options) => console.log(error),
   onRequestError: (error, options) => console.log(error),
   onParsingError: (error, options) => console.log(error),
   onTransformError: (error, options) => console.log(error),
}))

Or for single requests

upfetch('/todos', {
   onBeforeFetch: (options) => console.log('Before fetch'),
   onSuccess: (todos, options) => console.log(todos),
   onResponseError: (error, options) => console.log(error),
   onRequestError: (error, options) => console.log(error),
   onParsingError: (error, options) => console.log(error),
   onTransformError: (error, options) => console.log(error),
})

Learn more here.

✔️ Timeout

Worth mentionning that while up-fetch does not provide any timeout option since the AbortSignal.timeout static method is now supported everywhere, you can still leverage up-fetch to apply a default timeout.

Set a default timeout for all requests:

const upfetch = up(fetch, () => ({
   signal: AbortSignal.timeout(5000),
}))

Use a different timeout for a specific request:

upfetch('/todos', {
   signal: AbortSignal.timeout(3000),
})

➡️ How to

Since the defaults are evaluated at request time, the Authentication header can be defined in up

import { up } from 'up-fetch'

const upfetch = up(fetch, () => ({
   headers: { Authentication: localStorage.getItem('bearer-token') },
}))

localStorage.setItem('bearer-token', 'Bearer abcdef123456')
upfetch('/profile') // Authenticated request

localStorage.removeItem('bearer-token')
upfetch('/profile') // Non authenticated request

// ❌ Don't read the storage / cookies outside of `up`

// This value will never change
const bearerToken = localStorage.getItem('bearer-token')

const upfetch = up(fetch, () => ({
   headers: { Authentication: bearerToken },
}))

// ✅ Keep it inside the function call

// Checks the localStorage on each request
const upfetch = up(fetch, () => ({
   headers: { Authentication: localStorage.getItem('bearer-token') },
}))

The same approach can be used with cookies

up-fetch throws a ResponseError when response.ok is false.
You can decide when to throw using the throwResponseErrorWhen option.
You can decide what to throw using the parseResponseError option.

On the default ResponseError:

• The parsed response body is available with error.data. \
• The raw Response is available with error.response. \
• The response status is available with error.status. \
• The options used the make the request are available with error.options.

The type guard isResponseError can be used to check if an error is a ResponseError

import { upfetch } from '...'
import { isResponseError } from 'up-fetch'

// with try/catch
try {
   return await upfetch('https://a.b.c')
} catch (error) {
   if (isResponseError(error)) {
      console.log(error.name)
      console.log(error.message)
      console.log(error.data)
      console.log(error.status)
      console.log(error.options)
   } else {
      console.log(error.name)
      console.log(error.message)
   }
}

// with Promise.catch
upfetch('https://a.b.c').catch((error) => {
   if (isResponseError(error)) {
      console.log(error.name)
      console.log(error.message)
      console.log(error.data)
      console.log(error.status)
      console.log(error.options)
   } else {
      console.log(error.name)
      console.log(error.message)
   }
})

up-fetch also exports some listeners, useful for logging

import { up } from 'up-fetch'
import { log } from './my-logging-service'

const upfetch = up(fetch, () => ({
   onResponseError(error) {
      log.responseError(error)
   },
   onRequestError(error) {
      log.requestError(error)
   },
}))

upfetch('/fail-to-fetch')

Simply pass undefined

import { up } from 'up-fetch'

const upfetch = up(fetch, () => ({
   cache: 'no-store',
   params: { expand: true, count: 1 },
   headers: { Authorization: localStorage.getItem('bearer-token') },
}))

upfetch('https://a.b.c', {
   cache: undefined, // remove cache
   params: { expand: undefined }, // only remove `expand` from the params
   headers: undefined, // remove all headers
})

You may sometimes need to conditionally override the default options provided in up. Javascript makes it a bit tricky:

import { up } from 'up-fetch'

const upfetch = up(fetch, () => ({
   headers: { 'X-Header': 'value' }
}))

❌ Don't
// if `condition` is false, the header will be deleted
upfetch('https://a.b.c', {
   headers: { 'X-Header': condition ? 'newValue' : undefined }
})

In order to solve this problem, upfetch exposes the defaultOptions when the options (2nd arg) are defined as a function.
defaultOptions are stricly typed (const generic)

✅ Do
upfetch('https://a.b.c', (defaultOptions) => ({
   headers: { 'X-Header': condition ? 'newValue' : defaultOptions.headers['X-Header'] }
}))

Since up-fetch extends the fetch API, Next.js specific fetch options also work with up-fetch.

Choose a default caching strategy

import { up } from 'up-fetch'

const upfetch = up(fetch, () => ({
   next: { revalidate: false },
}))

Override it for a specific request

upfetch('/posts', {
   next: { revalidate: 60 },
})

➡️ Adapters & Recipies

You can use the zod validation adapter to guarantee the type safety of the data.

import { z } from 'zod'
import { withZod } from 'up-fetch/with-zod'
import { upfetch } from './abc'

const todo = await upfetch('/todo/1', {
   parseResponse: withZod(
      z.object({
         id: z.number(),
         title: z.string(),
         description: z.string(),
         createdOn: z.string(),
      }),
   ),
})
// todo is properly typed in case of validation success

Using an adapter ensures the data properly flows through the interceptors

import { z } from 'zod'
import { withZod } from 'up-fetch/with-zod'

const upfetch = up(fetch, () => ({
   onParsingError: (error) => console.log(error),
   onSuccess: (data) => console.log(data),
}))

const todo = await upfetch('/todo/1', {
   onParsingError: (error) => console.log(error),
   onSuccess: (data) => console.log(data),
   parseResponse: withZod(
      z.object({
         id: z.number(),
         title: z.string(),
         description: z.string(),
         createdOn: z.string(),
      }),
   ),
})

You can use the valibot validation adapter to guarantee the type safety of the data.

import { object, number, string } from 'valibot'
import { withValibot } from 'up-fetch/with-valibot'
import { upfetch } from './abc'

const todo = await upfetch('/todo/1', {
   parseResponse: withValibot(
      object({
         id: number(),
         title: string(),
         description: string(),
         createdOn: string(),
      }),
   ),
})
// todo is properly typed in case of validation success

Using an adapter ensures the data properly flows through the interceptors

import { object, number, string } from 'valibot'
import { withValibot } from 'up-fetch/with-valibot'

const upfetch = up(fetch, () => ({
   onParsingError: (error) => console.log(error),
   onSuccess: (data) => console.log(data),
}))

const todo = await upfetch('/todo/1', {
   onParsingError: (error) => console.log(error),
   onSuccess: (data) => console.log(data),
   parseResponse: withValibot(
      object({
         id: number(),
         title: string(),
         description: string(),
         createdOn: string(),
      }),
   ),
})

If you grab the FormData from a form, you dont need any adapter.

const form = document.querySelector('#my-form')

upfetch('/todos', {
   method: 'POST',
   body: new FormData(form),
})

However if you need to transform an object to FormData you might use object-to-formdata (<1kb)

Note: when sending FormData the fetch API automatically adds the correct header. See MDN docs

import { serialize } from 'object-to-formdata'

const upfetch = up(fetch, () => ({
   serializeBody: (body) => serialize(body),
}))

upfetch('https://a.b.c', {
   method: 'POST',
   body: { file: new File(['foo'], 'foo.txt') },
})

Coming soon

April 2024

Node, bun and browsers implementation of the fetch API do not support HTTP agents.

In order to use http agents you'll have to use undici (node only)

Add an HTTP Agent on a single request

import { fetch, Agent } from 'undici'

const upfetch = up(fetch)

const data = await upfetch('https://a.b.c', {
   dispatcher: new Agent({
      keepAliveTimeout: 10,
      keepAliveMaxTimeout: 10,
   }),
})

Dynamically add an HTTP Agent on each request request

import { fetch, Agent } from 'undici'

const upfetch = up(fetch, () => ({
   dispatcher: new Agent({
      keepAliveTimeout: 10,
      keepAliveMaxTimeout: 10,
   }),
}))

const data = await upfetch('https://a.b.c')

➡️ Types

See the type definitions file for more details

➡️ Options

All options can be set either on up or on an upfetch instance except for the body

// set defaults for the instance
const upfetch = up(fetch, () => ({
   baseUrl: 'https://a.b.c',
   cache: 'no-store',
   headers: { Authorization: `Bearer ${token}` },
}))

// override the defaults for a specific call
upfetch('/todos', {
   baseUrl: 'https://x.y.z',
   cache: 'force-cache',
})

upfetch adds the following options to the fetch API.

<baseUrl>

Type: string

Sets the base url for the requests

Example:

const upfetch = up(fetch, () => ({
   baseUrl: 'https://a.b.c',
}))

// make a GET request to 'https://a.b.c/id'
upfetch('/id')

// change the baseUrl for a single request
upfetch('/id', { baseUrl: 'https://x.y.z' })

<params>

Type: { [key: string]: any }

The url search params.
The params defined in up and the params defined in upfetch are shallowly merged.
Only non-nested objects are supported by default. See the serializeParams option for nested objects.

Example:

const upfetch = up(fetch, () => ({
   params: { expand: true },
}))

// `expand` can be omitted
// ?expand=true&page=2&limit=10
upfetch('https://a.b.c', {
   params: { page: 2, limit: 10 },
})

// override the `expand` param
// ?expand=false&page=2&limit=10
upfetch('https://a.b.c', {
   params: { page: 2, limit: 10, expand: false },
})

// delete `expand` param
// ?expand=false&page=2&limit=10
upfetch('https://a.b.c', {
   params: { expand: undefined },
})

// conditionally override the expand param `expand` param
// ?expand=false&page=2&limit=10
upfetch('https://a.b.c', (defaultOptions) => ({
   params: { expand: isTruthy ? true : defaultOptions.params.expand },
}))

<headers>

Type: HeadersInit | Record<string, string | number | null | undefined>

Same as the fetch API headers with widened types.
The headers defined in up and the headers defined in upfetch are shallowly merged. \

Example:

const upfetch = up(fetch, () => ({
   headers: { Authorization: 'Bearer ...' },
}))

// the request will have both the `Authorization` and the `Test-Header` headers
upfetch('https://a.b.c', {
   headers: { 'Test-Header': 'test value' },
})

// override the `Authorization` header
upfetch('https://a.b.c', {
   headers: { Authorization: 'Bearer ...2' },
})

// delete the `Authorization` header
upfetch('https://a.b.c', {
   headers: { Authorization: null }, // undefined also works
})

// conditionally override the `Authorization` header
upfetch('https://a.b.c', (defaultOptions) => ({
   headers: {
      Authorization: isTruthy ? 'Bearer ...3' : defaultOptions.headers.val,
   },
}))

<body>

Type: BodyInit | JsonifiableObject | JsonifiableArray | null

Note that this option is not available on up

The body of the request.
Can be pretty much anything.
See the serializeBody for more details.

Example:

upfetch('/todos', {
   method: 'POST',
   body: { hello: 'world' },
})

<serializeParams>

Type: (params: { [key: string]: any } ) => string

Customize the params serialization into a query string.
The default implementation only supports non-nested objects.

Example:

import qs from 'qs'

// add support for nested objects using the 'qs' library
const upfetch = up(fetch, () => ({
   serializeParams: (params) => qs.stringify(params),
}))

// ?a[b]=c
upfetch('https://a.b.c', {
   params: { a: { b: 'c' } },
})

<serializeBody>

Type: (body: JsonifiableObject | JsonifiableArray) => BodyInit | null | undefined

Default: JSON.stringify

Customize the body serialization into a valid BodyInit, a string in most cases
The body is passed to serializeBody when it is a plain object, an array or a class instance with a toJSON method. The other body types remain untouched

Example: serialize objects to FormData

This example uses object-to-formdata (<1kb)

Note: when sending FormData the fetch API automatically adds the correct header. See MDN docs

import { serialize } from 'object-to-formdata'

const upfetch = up(fetch, () => ({
   serializeBody: (body) => serialize(body),
}))

upfetch('https://a.b.c', {
   method: 'POST',
   body: { file: new File(['foo'], 'foo.txt') },
})

<parseResponse>

Type: <TData> (response: Response, options: ComputedOptions) => Promise<TData>

Customize the fetch response parsing.
By default json and text responses are parsed

This option is best used with a validation adapter

Example:

// create a fetcher for blobs
const fetchBlob = up(fetch, () => ({
   parseResponse: (res) => res.blob(),
}))

// disable the default parsing
const upfetch = up(fetch, () => ({
   parseResponse: (res) => res,
}))

Example: with the zod adapter

import { z } from 'zod'
import { withZod } from 'up-fetch/with-zod'

// ...create or import your upfetch instance

const todo = await upfetch('/todo/1', {
   parseResponse: withZod(
      z.object({
         id: z.number(),
         title: z.string(),
         description: z.string(),
         createdOn: z.string(),
      }),
   ),
})

<parseResponseError>

Type: <TError> (response: Response, options: ComputedOptions) => Promise<TError>

Customize the parsing of a thrown fetch response.
By default the response is thrown when response.ok is false, it is customizable with throwResponseErrorWhen.
By default a ResponseError is thrown

Example:

// throw a `CustomResponseError` when `response.ok` is `false`
const upfetch = up(fetch, () => ({
   parseResponseError: (res) => new CustomResponseError(res),
}))

parseResponse can also be used with a validation adapter

<transform>

Type: <TData, TParsedData> = (parsedData: TParsedData, options: ComputedOptions) => MaybePromise<TData>

Transform the data after parseResponse is done, errors will trigger the onTransformError interceptors.

Example:

// Instanciate some class
upfetch('/todos', {
   transform: (todos) => todos.map((todo) => new Todo(todo)),
})

The data is inferred from parseResponse.

const upfetch = up(fetch, () => ({
   parseResponse: async (res) => ({
      json: await res.json(),
      status: res.status,
   }),
}))

// the data is properly typed
upfetch('/todos', {
   transform: (data) => {
      console.log(data.json)
      console.log(data.status)
      // do whatever you like
      return data
   },
})

<onSuccess>

Type: <TData>(data: TData, options: ComputedOptions) => void

Called when response.ok is true.
Receives the data from transform if defined, otherwise from parseResponse.

Example:

// listen to all requests
const upfetch = up(fetch, () => ({
   onSuccess: (data, options) => console.log('2nd'),
}))

// listen to  requests
upfetch('https://a.b.c', {
   onSuccess: (data, options) => console.log('1st'),
})

<onResponseError>

Type: <TError>(error: TError, options: ComputedOptions) => void

Called when a response error was thrown, by default when response.ok is false, customizable with the throwResponseErrorWhen option

Example:

// listen to all requests
const upfetch = up(fetch, () => ({
   onResponseError: (error, options) => console.log('Response error', error),
}))

// listen to one requests
upfetch('https://a.b.c', {
   onResponseError: (error, options) => console.log('Response error', error),
})

<onRequestError>

Type: (error: Error, options: ComputedOptions) => void

Called when the fetch request fails (no response from the server).

Example:

// listen to all requests
const upfetch = up(fetch, () => ({
   onRequestError: (error, options) => console.log('Request error', error),
}))

// listen to one requests
upfetch('https://a.b.c', {
   onRequestError: (error, options) => console.log('Request error', error),
})

<onParsingError>

Type: (error: any, options: ComputedOptions) => void

Called when either parseResponse or parseResponseError throw.
Usefull when using a validation adapter

Example:

import { z } from 'zod'
import { withZod } from 'up-fetch/with-zod'

// listen to all requests
const upfetch = up(fetch, () => ({
   onParsingError: (error, options) => console.log('Validation error', error),
}))

// listen to one requests
upfetch('https://a.b.c', {
   onParsingError: (error, options) => console.log('Validation error', error),
   parseResponse: withZod(
      z.object({
         id: z.number(),
         title: z.string(),
         description: z.string(),
         createdOn: z.string(),
      }),
   ),
})

<onTransformError>

Type: (error: Error, options: ComputedOptions) => void

Called when either transform throws. \

Example:

import { z } from 'zod'
import { withZod } from 'up-fetch/with-zod'

// listen to all requests
const upfetch = up(fetch, () => ({
   onTransformError: (error, options) => console.log('Transform error', error),
}))

// listen to one requests
upfetch('https://a.b.c', {
   onTransformError: (error, options) => console.log('Transform error', error),
   transform: (data) => /* throw some error */
})

<onBeforeFetch>

Type: (options: ComputedOptions) => void

Called before the request is sent.

Example:

// listen to all requests
const upfetch = up(fetch, () => ({
   onBeforeFetch: (options) => console.log('2nd'),
}))

// listen to one requests
upfetch('https://a.b.c', {
   onBeforeFetch: (options) => console.log('1st'),
})

<throwResponseErrorWhen>

Type: (response: Response) => MaybePromise<boolean>

Default: (response: Response) => !response.ok

Decide when to trigger parseResponseError and throw an error.
It can be an async function.

Example: never throw upon response

// for all requests
const upfetch = up(fetch, () => ({
   throwResponseErrorWhen: () => false,
}))

// for one requests
upfetch('https://a.b.c', {
   throwResponseErrorWhen: () => false,
})

Example: throw for specific statuses

// for all requests
const upfetch = up(fetch, () => ({
   throwResponseErrorWhen: (response) => [ 400, 404, ... ].includes(response.status),
}))

➡️ Compatibility

• ✅ All modern browsers
• ✅ Bun
• ✅ Node 18+
• ✅ Cloudflare Workers

From the same author

• tw-colors: Tailwind plugin to easily add multiple color themes to your projects.