Data from an API is not application state.
WHY IT IS DIFFERENT
It is owned elsewhere It can become stale It needs caching, refetching and invalidation It has loading and error states
WHAT HAPPENS WHEN IT IS TREATED AS ORDINARY STATE
Manual loading flags everywhere Duplicate requests Stale data with no refresh strategy Substantial repeated code
WHAT DATA-FETCHING LIBRARIES PROVIDE
Caching by key Deduplication of identical concurrent requests Background refetching Loading and error states Invalidation after mutations Retry behaviour
WHY THAT MATTERS
It removes the majority of the code people write around fetching.
WHAT TO DEFINE
A key for each piece of data How long it is considered fresh What invalidates it after a change
WHAT TO DO AFTER A MUTATION
Invalidate what it affected, so it refetches.
WHAT OPTIMISTIC UPDATES ARE
Updating the interface before the server confirms.
WHAT THEY REQUIRE
A way to revert on failure.
WHAT TO ALWAYS HANDLE
Failure, and slow responses.
