Tiny (~550 bytes) wrapper to simplify native fetch calls using any HTTP method (existing or imagined).

Features

• Fully typed/TypeScript support
• Automatically parses responses (optional)
• Automatically serializes object payloads
• Accepts any HTTP method (including user-defined)
• 404, 400, 500, errors actually throw to allow easier catching
• Still allows any native fetch options (including headers, etc) to be sent

Simple Usage

import { fetcher } from 'itty-fetcher'

// create a basic fetcher with default options
const basics = fetcher()

// skips the body parsing for normal GET requests
await basics.get('https://api.kittens.com/v1/names/?max=2') // ['Fluffy', 'Mittens']

// set a base for simplifying repeated calls
const api = fetcher({ base: 'https://api.kittens.com/v1' })

// then use it... base will be prepended to urls
await api.get('/names/?max=2') // ['Fluffy', 'Mittens']

// automatic handle sending payloads (no need to stringify and set headers)
await api.post('/create-a-cat', { name: 'Halsey', age: 3 }) // { id: 'Q4AW', name: 'Halsey', age: 3 }

// use any conceivable HTTP method
api.put('/kitten/13', { name: 'Different Cat' }) // sends using PUT method
api.foo('/kitten/13', { name: 'Different Cat' }) // sends using FOO method

// supports GET query params
await api.get('/names', { max: 2, foo: ['bar', 'baz'] })
// GET https://api.kittens.com/v1/names?max=2&foo=bar&foo=baz

// ERROR HANDLING: 400, 404, 500, etc will actually throw, allowing an easy catch
api.get('/not-a-valid-path').catch(({ status, message }) => {
  console.log('received a status', status, 'error with message:', message)
})

Why yet another fetching library?

We've all done this countless times in our apps...

We want to make a nice, lightweight app that (of-course) talks to some API. We could import a full-featured fetch library like axios, but we want to keep our bundle size down, right?

So we just write some basic native fetch statements. That's not hard... we've tread this ground before! Of course as the project grows a bit, we start to become bothered by the repeated boilerplate of setting headers, checking for errors, translating response bodies, etc.

So what do we do?

Why, we write a little abstraction layer of course! Just like this one, but probably a bit bigger.

So who is this for?

This is not a kitchen-sink sort of library. It will intentionally not cover every edge case. By only handling a variety of the most common use-cases, I can keep the bundle size down to [likely] smaller than the code you would have written yourself, making it a no-brainer for easy inclusion into your projects.

Need more advanced fetch handling? Perhaps try a different library (or stick to native fetch and handle the edge case manually)!

ADVANCED USAGE

// skipping autoParse returns full Response control
const unparsed = fetcher({ autoParse: false })

unparsed.get('https://api.kittens.com/v1/names/?max=2').then((response) => {
  if (response.ok) return response.json()
})

// can send all native fetch options through in 3rd param
fetcher().post(
  'https://api.kittens.com/v1/names/?max=2',
  { payload: 'is second param' },
  {
    credentials: 'same-origin',
    cache: 'no-cache',
    headers: {
      ['my-fancy-header']: 'will be sent',
    },
  },
)

API

fetcher(options?: FetcherOptions): FetcherType

Returns a fetcher object, with method calls (like .get, .post, etc) mapped to fetch commands.

RequestLike matches the following signature:

type RequestLike = RequestInit & {
  method: string // method is required
  headers: HeadersInit // headers are populated with { 'Content-Type': 'application/json' } if not set
  url: string // url is required and is the fully qualified URL
}

transformRequest examples

// Add a header to every request
fetcher({
  transformRequest(req) {
    req.headers['Authorization'] = token
    return req
  },
})

// Add a query param to every request
fetcher({
  transformRequest(req) {
    const url = new URL(req.url)
    url.searchParams.set('message', 'hello world')
    req.url = url.toString()
    return req
  },
})

// Only add a header on a "/admin" route
fetcher({
  transformRequest(req) {
    const url = new URL(req.url)
    if (url.pathname.startsWith('/admin')) {
      req.headers['Authorization'] = token
    }
    return req
  },
})

// Change the origin of a URL only in dev mode
fetcher({
  transformRequest(req) {
    if (dev) {
      req.url = req.url.replace('prod.example.com', 'dev.example.com')
    }
    return req
  },
})

// Only apply a token on a specific domain
fetcher({
  transformRequest(req) {
    if (req.url.startsWith("https://api.example.com") {
      req.headers["Authorization"] = `Bearer ${token}`
    }
    return req
  }
})

// Util function to inject a user auth token into a request
function api_client(token: string) {
  return fetcher({
    base: 'https://api.example.com',
    transformRequest(req) {
      req.headers['Authorization'] = token
      return req
    },
  })
}

// Somewhere else...
const token = read_jwt(jwt_cookie)
const api = api_client(token)
api.get('/foo')

fetcher().{method}(url, payload, options)

Each method call maps to the corresponding HTTP method (in uppercase) with the following signature:

Example

fetcher().patch('https://foo.bar', { value: 2 })
// sends PATCH to https://foo.bar with payload of { value: 2 }

Special Thanks

I have to thank my friends and colleagues that helped me through the idea itself, implementation details, and importantly... made it possible for me to muck through making this a TS-first library. Huge thanks for that!!

Contributors

As always, these are the real heroes!

@danawoodman - For a truly amazing amount of work with me on this project, including (but not limited to) build improvements, testing improvements, query param (GET) support, and the incredibly powerful transformRequest feature, which unlocks future plugin support potential! Huge thanks, Dana!! :)