@silkline/hasura-sim

An in-memory Hasura GraphQL engine for tests. The real Hasura GraphQL Engine v2 (Haskell) is compiled to WebAssembly and paired with pglite (WASM Postgres). Feed it static Hasura metadata and a migrated pglite instance, then run GraphQL operations and get back exactly what the real engine returns — no Docker, no Postgres server, no network.

How it works

GraphQL op ─▶ sim_execute (wasm Hasura engine)
                  │  parse → IR → SQL  (real engine: permissions,
                  │                     session vars, relationships, …)
                  ▼
            pg_exec bridge ──▶ pglite (in-process Postgres)
                  ▲                │
                  └──── rows ──────┘

The engine runs the real schema-cache builder and query/mutation translator. SQL is executed against pglite over the PostgreSQL wire protocol (src/wire.mjs), so binary params + OIDs and result formats round-trip faithfully.

Usage

import { PGlite } from "@electric-sql/pglite";
import { createSimulator } from "@silkline/hasura-sim";

const db = new PGlite();
await db.exec(/* your migrations */);

const sim = await createSimulator({
  db,
  metadataJson: {
    version: 3,
    sources: [
      {
        name: "default",
        kind: "postgres",
        configuration: {
          connection_info: { database_url: "postgres://localhost/db" },
        },
        tables: [{ table: { schema: "public", name: "widget" } }],
      },
    ],
  },
});

const { data, errors } = await sim.execute({
  query: "query ($id: Int!) { widget(where: { id: { _eq: $id } }) { id name } }",
  variables: { id: 2 },
  role: "user", // default "admin"
  sessionVariables: { "x-hasura-user-id": "42" },
});

metadataJson is consolidated Hasura metadata (hasura metadata export -o json). The database_url is ignored — all SQL is routed to the provided pglite db.

Status

Validated end-to-end: queries (param-free + parameterized) and mutations (insert, committing). Remote schemas and actions are out of scope (stubbed).