converted graphql middleware blog post

Author: rsill-neo4j

modules/ROOT/pages/security/securing-a-graphql-api.adoc

@@ -0,0 +1,353 @@
+= Using GraphQL middleware with Neo4j GraphQL
+
+You can wrap your auto-generated Neo4j GraphQL resolver with custom logic that intercepts specific GraphQL operations.
+This approach allows you to retain all the benefits of auto-generation while adding the custom behavior you need.
+
+
+== GraphQL middleware
+
+This page makes use of https://github.com/dimatill/graphql-middleware[`graphql-middleware`].
+GraphQL middleware is a library that provides a way to wrap and extend the behavior of your GraphQL resolvers.
+It acts as a layer that allows you to apply reusable logic, such as logging, validation, or authentication, across multiple resolvers in a consistent and modular
+way.
+
+
+=== Logging every request
+
+Consider this Neo4j GraphQL setup:
+
+[source,typescript]
+----
+import { ApolloServer } from "@apollo/server";
+import { startStandaloneServer } from "@apollo/server/standalone";
+import { applyMiddleware } from "graphql-middleware";
+import * as neo4j from "neo4j-driver";
+import { Neo4jGraphQL } from "@neo4j/graphql";
+
+const typeDefs = /* GraphQL */ `
+    type User @node {
+        id: ID! @id
+        name: String!
+        email: String!
+        posts: [Post!]! @relationship(type: "AUTHORED", direction: OUT)
+    }
+
+    type Post @node {
+        id: ID!
+        title: String!
+        content: String!
+        author: [User!]! @relationship(type: "AUTHORED", direction: IN)
+    }
+
+    type Query {
+        me: User @cypher(statement: "MATCH (u:User {id: $userId}) RETURN u", columnName: "u")
+    }
+`;
+
+const driver = neo4j.driver("bolt://localhost:7687", neo4j.auth.basic("neo4j", "password"));
+
+const neoSchema = new Neo4jGraphQL({
+    typeDefs,
+    driver,
+});
+
+const server = new ApolloServer({
+    schema: await neoSchema.getSchema(),
+});
+
+const { url } = await startStandaloneServer(server, {
+    listen: { port: 4000 },
+});
+console.log(`🚀 Server ready at ${url}`);
+----
+
+Add logging to every single operation without touching the generated schema:
+
+[source,typescript]
+----
+import { applyMiddleware } from "graphql-middleware";
+
+/* ...existing code... */
+
+const logMiddleware = async (resolve, root, args, context, info) => {
+    const start = Date.now();
+    console.log(`🚀 ${info.fieldName} started`);
+
+    try {
+        const result = await resolve(root, args, context, info);
+        console.log(`✅ ${info.fieldName} completed in ${Date.now() - start}ms`);
+        return result;
+    } catch (error) {
+        console.log(`💥 ${info.fieldName} failed`);
+        throw error;
+    }
+};
+
+// Wrap your executable schema
+const schemaWithLogging = applyMiddleware(await neoSchema.getSchema(), {
+    Query: logMiddleware,
+    Mutation: logMiddleware,
+});
+
+const server = new ApolloServer({ schema: schemaWithLogging });
+----
+
+*That’s it. Every query and mutation is now logged.* Your auto-generated
+resolver is unchanged, but you’ve added custom behavior.
+
+Query the users:
+
+[source,graphql]
+----
+{
+    users {
+        name
+    }
+}
+----
+
+You should see in your server:
+
+....
+🚀 users started
+✅ users completed in 23ms
+....
+
+
+=== Email validation before database writes
+
+You can use middleware to enforce specific business rules before data is written to the database.
+For example, you can ensure that email addresses provided during user creation are valid.
+By using middleware, you can intercept and validate the input before it reaches the Neo4j GraphQL resolver.
+
+Add a middleware that validates the email input in the `createUsers` operation.
+A validation error will be thrown before it reaches the Neo4j GraphQL resolver, and the GraphQL client will receive the error message "Invalid email addresses detected".
+
+[source,typescript]
+----
+/* ...existing code... */
+
+const validateEmails = async (resolve, root, args, context, info) => {
+    // Only check createUsers mutations
+    if (info.fieldName === "createUsers") {
+        // Note: This is a simplistic and intentionally flawed email validation example, but good for demonstration purposes.
+        const invalidEmails = args.input.filter((user) => !user.email.includes("@"));
+        if (invalidEmails.length > 0) {
+            throw new Error("Invalid email addresses detected");
+        }
+    }
+
+    return resolve(root, args, context, info);
+};
+
+const schema = applyMiddleware(
+    await neoSchema.getSchema(),
+    {
+        Query: logMiddleware,
+        Mutation: logMiddleware,
+    },
+    {
+        Mutation: validateEmails,
+    }
+);
+----
+
+Try to create a user with the email "not-an-email":
+
+[source,graphql]
+----
+mutation createUsers {
+    createUsers(input: [{ email: "not-an-email.com", name: "firstname" }]) {
+        users {
+            email
+        }
+    }
+}
+----
+
+
+== Working with Neo4j GraphQL
+
+Most of the above is applicable even outside Neo4j GraphQL, but there is an important concept when writing middleware for Neo4j GraphQL resolvers.
+
+Here's the key difference from how traditional GraphQL resolvers are
+usually built:
+
+In *traditional GraphQL*, each field resolver executes independently, potentially causing multiple database calls.
+By contrast, in *Neo4j GraphQL* the root field resolver (like `users` or `createUsers`) analyzes the entire query tree and executes one optimized Cypher query.
+
+The N+1 problem is solved in Neo4j GraphQL by analyzing the entire GraphQL operation (via the `info` object) and generating optimized Cypher queries that fetch all requested data in a single database round-trip.
+
+Consider this query:
+
+[source,graphql]
+----
+{
+    users {
+        name
+        email
+        posts {
+            title
+            content
+        }
+    }
+}
+----
+
+Neo4j GraphQL doesn't execute separate resolvers for `name`, `email`, `posts`, `title`, and `content`.
+Instead, the `users` field resolver generates and executes a single Cypher query that returns all the data at once.
+The nested field resolvers simply return the already fetched data from memory.
+
+
+=== Timing matters
+
+Timing matters for middleware - by the time the individual field resolvers execute, the database has already been queried and the data is available in the resolver's result.
+
+Consider the `logMiddleware` from above:
+
+[source,typescript]
+----
+const logMiddleware = async (resolve, root, args, context, info) => {
+    const start = Date.now();
+    console.log(`🚀 ${info.fieldName} started`);
+
+    try {
+        const result = await resolve(root, args, context, info);
+        console.log(`✅ ${info.fieldName} completed in ${Date.now() - start}ms`);
+        return result;
+    } catch (error) {
+        console.log(`💥 ${info.fieldName} failed: ${error.message}`);
+        throw error;
+    }
+};
+----
+
+Apply the `logMiddleware` to queries and the user's name:
+
+[source,typescript]
+----
+const schema = applyMiddleware(
+    schema,
+    {
+        Query: logMiddleware, // wraps all the Queries and it's executed before the database round-trip
+    },
+    {
+        User: {
+            name: logMiddleware, // wraps only the User's name field resolver and it's executed after the database roundtrip
+        },
+    }
+);
+----
+
+Run this query:
+
+[source,graphql]
+----
+query {
+    users {
+        name
+    }
+}
+----
+
+You should see:
+
+....
+🚀 users started
+... Neo4j resolver generates and executes Cypher ...
+✅ users completed in 48ms
+🚀 name started
+✅ name completed in 0ms
+....
+
+Note how the name resolution happens after the round-trip to the database.
+
+Note the following difference:
+
+* Query and mutation level middleware runs before and after the Neo4j GraphQL autogenerated resolvers.
+* Type and field level middleware runs only after the Neo4j GraphQL autogenerated resolvers.
+
+
+== Stack multiple middleware
+
+It's possible to apply multiple pieces of middleware for the same field.
+For instance, you can apply diverse middleware to the same `users` resolver:
+
+[source,typescript]
+----
+const schema = applyMiddleware(
+    schema,
+    {
+        Query: {
+            users: async (resolve, root, args, context, info) => {
+                console.log("A started");
+                await resolve(root, args, context, info);
+                console.log("A completed");
+            },
+        },
+    },
+    {
+        Query: {
+            users: async (resolve, root, args, context, info) => {
+                console.log("B started");
+                await resolve(root, args, context, info);
+                console.log("B completed");
+            },
+        },
+    },
+    {
+        Query: {
+            users: async (resolve, root, args, context, info) => {
+                console.log("C started");
+                await resolve(root, args, context, info);
+                console.log("C completed");
+            },
+        },
+    }
+);
+----
+
+The order in which middleware is applied is important, as they execute one after the other.
+Each middleware wraps the next one, creating a chain of execution from outermost to innermost.
+
+Run this query:
+
+[source,graphql]
+----
+query {
+    users {
+        name
+    }
+}
+----
+
+Schematic output:
+
+[source,bash]
+----
+....
+A started
+B started
+C started
+... Neo4j GraphQL user resolver ...
+C completed
+B completed
+A completed
+....
+----
+
+The user's resolver is wrapped in three layers of middleware.
+
+
+== Conclusion
+
+GraphQL middleware with Neo4j GraphQL gives you the best of both worlds: the power of auto-generated schemas and the flexibility to inject custom logic exactly where you need it.
+
+When you need custom logic, graphql-middleware lets you keep the rapid development benefits of Neo4j GraphQL while adding the custom behavior you need.
+
+The GraphQL ecosystem evolves rapidly.
+https://the-guild.dev/[The Guild] has developed https://envelop.dev/[Envelop] with its own https://www.npmjs.com/package/@envelop/graphql-middleware[graphql-middleware
+plugin].
+
+This guide uses `graphql-middleware` because it's server-agnostic and delivers the clearest path to understanding middleware with Neo4j GraphQL.
+If you need a more comprehensive plugin ecosystem, we recommend exploring envelop.