Scalar and Object Types

Every field in a GraphQL schema ultimately resolves to a scalar type at its leaves — a concrete, indivisible value like a string or number — even when the field itself returns a complex object type. GraphQL ships with five built-in scalars: String for UTF-8 text, Int for signed 32-bit integers, Float for signed double-precision floating point numbers, Boolean for true/false, and ID for a unique identifier that is serialized as a string but signals semantic intent, such as being used as a cache key or lookup argument rather than displayed text.

The ID scalar deserves special attention because it looks identical to String on the wire — both serialize as JSON strings — but it carries semantic meaning that tooling relies on: normalized caches in Apollo Client and Relay use a type's id field combined with its type name to uniquely identify and deduplicate objects across different queries, and GraphQL's Relay-style global object identification convention is built entirely around a well-defined ID field. Choosing Int versus Float also matters for precision-sensitive domains: a warehouse inventory count should be Int, while a geographic coordinate is typically modeled as Float, though many schemas use a custom Decimal or Money scalar instead of Float for currency to avoid floating-point rounding errors.

Beyond the five built-ins, GraphQL lets servers define custom scalars for domain-specific values that need their own serialization logic, the most common example being a DateTime scalar for ISO-8601 timestamps; implementing one requires providing serialize (converting the internal value to a JSON-safe form for the response), parseValue (converting an incoming JSON value from a variable), and parseLiteral (converting an inline value written directly in the query text) functions. Enums are a related but distinct concept: an enum type restricts a field to one of a fixed set of named values, such as enum OrderStatus { PENDING, SHIPPED, DELIVERED, CANCELLED }, giving both the schema and generated client types a closed, self-documenting set of valid options instead of an unconstrained string.

Object types become genuinely powerful when their fields reference other object types rather than only scalars: a Book type can have an author field of type Author, and Author can have a books field of type [Book!]!, creating a graph with a real cycle that a client can traverse in either direction within a single query — fetching a book's author and that author's other books in one round trip. This is the structural feature GraphQL is named for, but it also means a resolver for a nested object field runs independently for each parent item, so naively resolving each Book's Author with a separate database call inside a loop produces the classic N+1 query problem, which is why production servers typically batch and cache these lookups using a tool like DataLoader.