retryOn, retryDelay, enhancements, and bugfixes/docs updates

[alex-cory]

Notes

• This should suffice File Upload fails due to added Content-Type header #213, Retry request in interceptors response #170, response causing infinite loops #228
• Don't add header content-type: application/json if we're sending FormData
• retryOn
• retryDelay
• updated readme with pre-existing path
• 1 rerender instead of 2 when pulling data from cache
• memoize response object so it doesn't cause infinite loops
• fix suspense tests from unpredictably failing...

Proposed Solution

if retries > 0 then by default always retry if error or if response.status >= 300
if retryOn: [400] and retries > 0 then only retry on 400

Todos

• fix types
• fix lint errors
• fix existing tests
• clean up TODOs in code
• tests
  • create tests for if we set timeout and retryDelay
  • retryDelay: 1000
  • retryDelay()
  • retryOn: [300]
  • retryOn()
  • retrying with no network (aka instead of going into the try retry we go into the catch retry)
  • when retryDelay is not a function or number
    • 🚫 there's an issue with this^ and I can't test if the function returns a number currently. We can however most likely test for a console.error and verify that.
  • when retryOn is not an array of http status codes or a function
  • should not be content-type: application/json by default if using FormData for request body
  • console.warn that says something like To have a retryDelay set, you need to have retries > 0
  • invariant retries >= 0 and typeof retries === 'number'
  • double check if we have a test for fail 1st, retry and succeed test
• error handling
  • when retryDelay is not a function returning a number or number
  • when retryOn is not an array of http status codes or a function
• somehow reset attempts after completing the last retry
• re-order all options in docs alphabetically
• codesandbox examples with retryOn and retryDelay
• better documentation for retryOn and retryDelay. Take a look at fetch-retry docs
• make sure all the defaults are the correct values in the docs. Also update documentation in table for retryOn and retryDelay to explain how we default to 3 retries if there's network errors with 1000ms as the retryDelay
• difference between timeout and retryDelay. WIP
  • timeout will cancel the current pending http request if it's not completed before this time is up

    request --------- > abort ------------> retry request
    |------ timeout ------|
    

  • retryDelay the time after a failed http request and when to retry the new one

    request ----------> error ---------------> retry request
                          |---- retryDelay----|
    

  • Mental model if you're using both

    initial request ----- > error ------------> retry ---------> success or error
    | ----- timeout -------- | --- retryDelay --- | --- timeout --- | 
    

  • Questions:
    1. What happens if we abort before completing the 1st request out of 3 (2 retries). Do we retry 2 more times after aborting?

    const request = useFetch('url', { retries: 2 })
    const handleMakeRequest = async () => {
      // if we cancel/abort this immediately, should it retry?
      const todo = await request.post('/todos', { no: 'way' })
    }
    <button onClick={handleMakeRequest}>Make Request</button>
    <button onClick={request.abort}>Abort</button>      

    A: it would just cancel/abort the 1st request and proceed to retry the other 2 times. We could potentially have an option to cancel them all.

[lgtm-com]

This pull request introduces 1 alert when merging 3d0cfca into 1d42e23 - view on LGTM.com

new alerts:

• 1 for Unused variable, import, function or class

[lgtm-com]

This pull request introduces 1 alert when merging 1ed2be7 into 1d42e23 - view on LGTM.com

new alerts:

• 1 for Unused variable, import, function or class