GraphQL Schema & Execution | Communication & Data Transfer

Core concepts

Schema — strongly typed contract (Query, Mutation, Subscription, object types)
Single endpoint — typically POST /graphql with JSON body { "query": "...", "variables": {} }
Selection set — client requests exactly the fields needed
Resolvers — functions backing each field; form a resolver tree executed depth-first

query {
  user(id: "1") {
    name
    orders(first: 5) { total }
  }
}

Queries vs mutations vs subscriptions

Operation	Semantics	HTTP analogy
Query	Read-only	GET (but POST transport)
Mutation	Side effects	POST
Subscription	Real-time stream	WebSocket (usually)

GraphQL mutations are not guaranteed sequential — document ordering if business logic requires it (e.g. debit before credit in one mutation field).

The N+1 problem

Naive resolver per field:

User.orders → 1 query for user + N queries for each user's orders

DataLoader batches: collect load(orderId) calls in one tick → single WHERE id IN (...). Also per-request caching.

Staff signal: mention DataLoader unprompted in GraphQL interviews.

Error model

Partial results: data + errors[] array. HTTP may still be 200 with errors in body — clients must check errors. Some APIs use 4xx for auth at HTTP layer.

BFF pattern

Backend-for-Frontend — GraphQL gateway aggregates microservices for mobile/web. Keeps domain services REST/gRPC; BFF owns graph shape and auth context. LLD helps you design a clean OrderService facade in-process; this topic places that facade behind HTTP as a GraphQL layer.

Avoids exposing entire microservice graph to clients.

vs REST (when to pick GraphQL)

GraphQL wins: multiple clients with different field needs, reducing over-fetch, rapid product iteration on UI.

REST wins: simple CRUD, heavy CDN caching, file upload standards, public API simplicity, strong HTTP semantics.

Neither replaces internal high-throughput RPC — see gRPC topic.

Hands-On Tasks (Optional)

API design drills and whiteboard exercises — protocol selection, contract design, and bulk-transfer architecture. Assumes Networking and sibling tracks on the hub page (Distributed Systems, Databases, Concurrency, LLD).

Design a schema for e-commerce
Define types for Product, Variant, Inventory, and queries: productById, searchProducts. Identify which fields need separate resolvers vs can be resolved from parent object.
20m