An alternative query approach.
WHAT IT IS
A single endpoint where the caller specifies exactly what they want.
WHAT IT PROVIDES
Fetching several related things in one request No over-fetching of unused fields A strongly typed schema
WHAT PROBLEM IT ADDRESSES
Interfaces requiring many round trips, or returning far more than needed.
WHAT IT COSTS
Caching is harder, since every query differs Rate limiting by request count is meaningless A caller can construct a very expensive query Errors are returned with a success status
WHY THAT LAST POINT MATTERS
Monitoring based on status codes sees nothing wrong.
WHAT TO IMPLEMENT ON THE SERVER
Query depth limits Complexity scoring, with a budget Timeouts
WHY COMPLEXITY SCORING SPECIFICALLY
It is the only meaningful way to limit cost.
WHAT THE REPEATED QUERY PROBLEM IS
Resolving a list then querying once per item.
WHAT SOLVES IT
Batching loaders.
WHY THAT MATTERS
Without them, one request becomes hundreds of database queries.
WHEN GRAPHQL SUITS
Many clients with differing needs Deeply related data Mobile clients where round trips are expensive
WHEN IT DOES NOT
Simple interfaces Public interfaces where you cannot trust callers
