@stonecrop/graphql-client

Transport layer for Stonecrop GraphQL APIs. Handles HTTP communication, response parsing, and metadata caching.

Why transport-only?

This client intentionally never constructs GraphQL queries. All query generation — including PostGraphile-specific field naming (inflection), nested link sub-selections, link display enrichment, and fetch strategy dispatch — lives in the server-side middleware (@stonecrop/graphql-middleware).

This boundary exists because PostGraphile's schema naming is configurable. An application might use ById for UUID primary keys, ByRowId for row_id columns, or entirely custom conventions. If the client hardcoded any of these conventions, it would silently produce wrong queries for non-default setups. The middleware owns the single StonecropInflectionConfig — the client only knows how to pass options.includeNested through to the stonecropRecord resolver and receive pre-merged flat data.

Responsibilities

Transport — The client sends requests and parses responses. It doesn't construct queries.

Caching — Metadata is cached in memory after first fetch.

Contract — The client expects the server to expose these operations:

| Operation | Arguments | Returns | |-----------|-----------|---------| | stonecropRecord | doctype, id, options? | { data, unknownLinks? } | | stonecropRecords | doctype, filters?, orderBy?, limit?, offset?, options? | { data[], count } | | stonecropMeta | doctype | DoctypeMeta | | stonecropAllMeta | — | DoctypeMeta[] | | stonecropAction | doctype, action, args? | { success, data, error, record } |

The client has no opinions about how the server implements these — naming conventions, query construction, nested data merging are all the server's concern.

Assumptions

• All record operations accept a doctype string argument
• stonecropRecord accepts options: { includeNested?, maxDepth? }
• The server handles query building and field naming

Usage

import { StonecropClient } from '@stonecrop/graphql-client'

const client = new StonecropClient({
  endpoint: 'http://localhost:4000/graphql',
  headers: { Authorization: `Bearer ${token}` }, // optional
})

// Fetch a record
const result = await client.getRecord({ name: 'SalesOrder' }, 'so-1')
result.record  // plain object with the record fields
result.unknownLinks  // links requested but not found in schema

// Fetch with nested links
const withNested = await client.getRecord({ name: 'SalesOrder' }, 'so-1', {
  includeNested: true,
})

// Custom queries
const custom = await client.query<{ myData: unknown[] }>(`query { myData { id } }`)

Data Shapes

• getRecord returns { record: Record<string, unknown> | null, unknownLinks?: string[] }. The record field contains the record's fields. Nested links are merged into the same object when includeNested is used. Inline link fields are enriched by the middleware as { id, displayText } objects where displayText comes from the target doctype's displayField.
• getRecords returns { data: Record<string, unknown>[], hasMore: boolean, count?: number } — flat objects with the same inline link enrichment.
• unknownLinks will contain link names you requested that don't exist in the doctype schema — useful for catching typos.

See API Reference for full method signatures.