In this post, you'll learn how to build––from scratch––an entirely type-safe, fullstack web app, using GraphQL with a database attached!
To follow along with the source code, clone this repo.
🗓 Update 19/11/2020: The tutorial has been updated to use Nexus Schema as the Nexus Framework has been discontinued
First, let's have a look at our tools of choice:
- TypeScript - the programming language for both backend and frontend
- React and Next.js - as the frontend framework and middle-end
- Urql GraphQL client - the GraphQL client on the frontend
- PostgreSQL - the database for the app
- Apollo Server - the server framework we'll use to serve the GraphQL API
- Nexus Schema - a library for building code-first GraphQL APIs
- Prisma Client and
prisma db push- a toolkit to change the database schema, access, and query the database (Note:prisma db pushis still in a preview state at the moment)
Let's get started! 🚀
Before we start, make sure that you have installed these VS Code extensions for syntax highlighting and auto-formatting - Prisma and GraphQL.
The first thing you'll need is a PostgreSQL database instance to interact with during development.
There are many options for this, but Heroku allows us to host PostgreSQL databases for free with minimal setup required. Check out this post by Nikolas Burk guiding you through the process!
If you have Docker installed and would rather keep your development database local, you can also check out this video I did on how to do this with Docker Compose.
You will be able to get a PostgreSQL URI in this format:
postgresql://<USER>:<PASSWORD>@<HOST_NAME>:<PORT>/<DB_NAME>
Note: if you are using Heroku, you'll get a URI with
postgresinstead ofpostgresqlas the protocol. Both formats should work but we'd prefer to usepostgresql.
When everything is set up properly, you're good to move on to the next step! 😃
Now, create a Next.js project with create-next-app and enter the directory:
npx create-next-app my-awesome-app --use-npm -e with-typescript
cd my-awesome-appGit should be automatically initialized by create-next-app, and your project structure should look like this:
create-next-app
The
create-next-appmight not have the latest TypeScript package pre-installed. If you want to use the latest TypeScript features, runnpm install -D typescript@latestafter the project is bootstrapped.
With the Next.js project ready, open a terminal window at the root of the application and install both Nexus Schema and Prisma.
For Prisma, we need @prisma/client, @nexus/schema and nexus-plugin-prisma as a regular dependencies and @prisma/cli as a dev dependency.
Regular dependencies:
npm i @prisma/client @nexus/schema nexus-plugin-prismaDev dependencies:
npm i @prisma/cliOnce the dependencies are installed, initialize Prisma in the project.
npx prisma initThis command will create a prisma directory. If you have a look inside, you'll see a .env file and a schema.prisma file. The schema.prisma file will hold the database model and the .env file will hold the database connection string.
Because the database connection string contains sensitive information, it's a good practice to never commit this .env file with Git, so make sure it's also added to the .gitignore file.
Adjust the schema.prisma file to include a User model:
datasource db {
provider = "postgresql"
url = env("DATABASE_URL")
}
generator client {
provider = "prisma-client-js"
}
model User {
id String @default(cuid()) @id
name String
}The schema file tells Prisma to use PostgreSQL as the database type, and the database connection URL is defined as an environment variable. It also defines a simple User data model with an id and a name field.
Your project should currently look like this:
prisma directory
To improve the development experience, also add the Nexus TypeScript Language Service plugin, and tweak a few compiler options in the tsconfig.json file like below:
{
"compilerOptions": {
// ...
"noEmit": true,
"rootDir": ".",
"typeRoots": ["node_modules/@types", "types"],
"plugins": [{ "name": "nexus/typescript-language-service" }]
},
// ...
"include": ["**/*.ts", "**/*.tsx", ".", "types.d.ts"]
}Nexus Schema is a library that allows us to build code-first GraphQL APIs. It's our responsibility to bring along a server to serve that API. For our purposes, we'll use Apollo Server.
There are several varieties of Apollo Server that are used for various purposes. For our project, we'll want apollo-server-mirco as it's well-suited to serverless deployments.
npm install apollo-server-microTo create a GraphQL endpoint, create a new file in your project at /pages/api/graphql.ts. Thanks to the powerful API routes in Next.js, the GraphQL server will be accessible at http://our-app-domain/api/graphql when the Next.js server is started.
In the /pages/api/graphql.ts file, write the following boilerplate code:
import { ApolloServer } from 'apollo-server-micro';
// we'll create these in a second!
import { schema } from '../../graphql/schema';
import { createContext } from './../../graphql/context';
const apolloServer = new ApolloServer({
context: createContext,
schema,
tracing: process.env.NODE_ENV === 'development'
});
export const config = {
api: {
bodyParser: false
}
};
export default apolloServer.createHandler({
path: '/api/graphql'
});Since everything inside the /pages/api/ directory is considered as an API route, it's a good idea to implement the actual schema and resolvers outside this directory.
Now, create a new directory in the project root called /graphql/ and two files within: /graphql/schema.ts and /graphql/context.ts.
Inside /graphql/schema.ts, start by using the makeSchema function to construct a GraphQL schema with Nexus. We'll also want to use nexus-plugin-prisma with the CRUD feature enabled:
// graphql/schema.ts
import { objectType, queryType, mutationType, makeSchema } from '@nexus/schema';
import { nexusPrisma } from 'nexus-plugin-prisma';
import path from 'path';
const Query = queryType({
definition(t) {
t.string('hello', { resolve: () => 'hello world' });
}
});
export const schema = makeSchema({
types: [Query],
plugins: [nexusPrisma({ experimentalCRUD: true })],
outputs: {
typegen: path.join(process.cwd(), 'generated', 'nexus-typegen.ts'),
schema: path.join(process.cwd(), 'generated', 'schema.graphql')
},
typegenAutoConfig: {
contextType: 'Context.Context',
sources: [
{
source: '@prisma/client',
alias: 'prisma'
},
{
source: path.join(process.cwd(), 'graphql', 'context.ts'),
alias: 'Context'
}
]
}
});The call to makeSchema includes a property called plugins. This is an array of any plugins we want to use with Nexus Schema and, in our case, we want to use nexus-plugin-prisma. The configuration we have here tells the plugin to use the CRUD feature which is what allows us to have automatically-generated CRUD resolvers for our API. You can read more on the CRUD feature provided by Nexus Schema.
Next, initialize the PrismaClient within /graphql/context.ts and export a function to create the context in Apollo Server.
// graphql/context.ts
import { PrismaClient } from '@prisma/client';
const prisma = new PrismaClient();
export interface Context {
prisma: PrismaClient;
}
export function createContext(): Context {
return { prisma };
}The file structure should now look something like this:
my-awesome-app/
├─ components/
├─ graphql/
│ ├─ context.ts
│ ├─ schema.ts
├─ interfaces/
├─ pages/
├─ prisma/
│ ├─ .env
│ ├─ schema.prisma
├─ utils/
├─ next-env.d.ts
├─ package-lock.json
├─ package.json
├─ tsconfig.jsonWith these files in place, run the application:
npx next devIf you go to http://localhost:3000/api/graphql, you'll see the GraphQL Playground up and running (with our "hello world" schema)! 😃
With the GraphQL server running in the background and the GraphQL Playground ready at http://localhost:3000/api/graphql, it's time to start implementing the API!
Start by defining a User object type to reflect the database schema. Once defined, add it to the types array in makeSchema.
// graphql/schema.ts
import { objectType, queryType, makeSchema } from '@nexus/schema';
const User = objectType({
name: 'User',
definition(t) {
t.model.id();
t.model.name();
}
});
// ...
export const schema = makeSchema({
types: [User, Query]
// ...
});If you are typing the code above instead of copy n' pasting, you will notice that VS Code will autocomplete the fields (id, name) that are available on the User data model defined earlier in /prisma/schema.prisma.
Tip: You can also always invoke intellisense with CTRL + SPACE, in case it doesn't automatically show up sometimes, super helpful!
Now, go back to the GraphQL Playground and toggle the Schema side panel - you will see a GraphQL object type User is generated from the code you just wrote in the /graphql/schema.ts file.
type User {
id: String!
name: String!
}For the root Query type, Nexus provides a queryType function.
To query a list of existing users in the database, you can write a resolver for allUsers field as follows:
const Query = queryType({
definition(t) {
t.list.field('allUsers', {
type: 'User',
resolve(_parent, _args, ctx) {
return ctx.prisma.user.findMany({});
}
});
}
});You can do whatever you want in the resolve function. The Prisma client for your database can be directly accessed as the db property on the ctx object. You can read more about the API of Prisma Client in its official documentation.
Tip: You can always use the intellisense feature in VS Code to explore the APIs of Nexus and Prisma Client!
In addition to manually writing resolvers, the Nexus-Prisma plugin conveniently exposes basic "read" operations on the database on t.crud. The following code will let you find a User (or a list of Users) from the database directly.
const Query = queryType({
definition(t) {
t.list.field('allUsers', {
type: 'User',
resolve(_parent, _args, ctx) {
return ctx.prisma.user.findMany({});
}
});
t.crud.user();
t.crud.users();
}
});The code above will generate a GraphQL root Query type:
type Query {
allUsers: [User!]
user(where: UserWhereUniqueInput!): User
users(
skip: Int
after: UserWhereUniqueInput
before: UserWhereUniqueInput
first: Int
last: Int
): [User!]!
}
input UserWhereUniqueInput {
id: String
}Notice that all the related Input types are also generated for us for free! 💯
Query type is generated by Nexus
Similar to the Query type, a Mutation type can be defined with the mutationType function.
😈 Let's have some fun and create a bigRedButton mutation to destroy all user data in the database.
We also have access to the t.crud helper here, which exposes the basic "create", "update" and "delete" operations on the database. We then must add Mutation to the types array in makeSchema.
import { objectType, queryType, mutationType, makeSchema } from '@nexus/schema';
// ...
const Mutation = mutationType({
definition(t) {
t.field('bigRedButton', {
type: 'String',
async resolve(_parent, _args, ctx) {
const { count } = await ctx.prisma.user.deleteMany({});
return `${count} user(s) destroyed. Thanos will be proud.`;
}
});
t.crud.createOneUser();
t.crud.deleteOneUser();
t.crud.deleteManyUser();
t.crud.updateOneUser();
t.crud.updateManyUser();
}
});
// ...
export const schema = makeSchema({
types: [User, Query, Mutation]
// ...
});This will generate a GraphQL schema like below:
type Mutation {
bigRedButton: String
createOneUser(data: UserCreateInput!): User!
deleteOneUser(where: UserWhereUniqueInput!): User
deleteManyUser(where: UserWhereInput): BatchPayload!
updateOneUser(data: UserUpdateInput!, where: UserWhereUniqueInput!): User
updateManyUser(
data: UserUpdateManyMutationInput!
where: UserWhereInput
): BatchPayload!
}
Mutation type is generated by Nexus
Now, our simple but fully-featured GraphQL API is ready! 🥳
Before you can do anything with your GraphQL API, you'll need to create tables in the database corresponding to the Prisma schema file.
This can be done by manually connecting to the database and running SQL commands, but I'll show you how to do it with the prisma db push commnand - the database tool that's part of Prisma 2.
First, save the initial changes of our schema.prisma file with the command below. At the moment, the prisma db push command is still in a preview state, so the extra flag --preview-feature flag is needed.
npx prisma db push --preview-featureAwesome! With the database prepared, it's time to go back to http://localhost:3000/api/graphql, and have some fun with the your first GraphQL API with Nexus. Let me give you an example to play with!
mutation {
createOneUser(data: { name: "Alice" }) {
id
}
}
We'll use Urql as the GraphQL client on the frontend, but you can use any library you like.
First, install the dependencies:
npm install graphql-tag next-urql react-is urql isomorphic-unfetchThen, create a new file at /pages/_app.tsx. This is a special Next.js component that will be used to initialize all pages.
import React from 'react';
import { withUrqlClient, NextUrqlAppContext } from 'next-urql';
import NextApp, { AppProps } from 'next/app';
import fetch from 'isomorphic-unfetch';
// the URL to /api/graphql
const GRAPHQL_ENDPOINT = `http://localhost:3000/api/graphql`;
const App = ({ Component, pageProps }: AppProps) => {
return <Component {...pageProps} />;
};
App.getInitialProps = async (ctx: NextUrqlAppContext) => {
const appProps = await NextApp.getInitialProps(ctx);
return { ...appProps };
};
export default withUrqlClient((_ssrExchange, _ctx) => ({
url: GRAPHQL_ENDPOINT,
fetch
}))(
// @ts-ignore
App
);And that's it! Now you can use the GraphQL client in any page in your Next.js app.
First, create a TSX file at /components/AllUsers.tsx. This file will have a component that performs an allUsers GraphQL query and renders the result as a list. This way, we can use the component to fetch all the user info from our PostgreSQL database.
You can create the query first, for example, with the following code. By using gql, the GraphQL VS Code extension will be able to identify the template string as a GraphQL query and apply nice syntax highlighting to it.
import React from 'react';
import gql from 'graphql-tag';
import { useQuery } from 'urql';
const AllUsersQuery = gql`
query {
allUsers {
id
name
}
}
`;Since it's known that the data you are going to get is an array of User objects (thank you, GraphQL schema!), you can also define a new type:
type AllUsersData = {
allUsers: {
id: string;
name: string;
}[];
};Next, create the React component that will be using the query.
The component encapsulates the following logic:
- If the query is still in a fetching state, the text "Loading..." will be returned
- If an error is occurred during the process, we will display the error
- If the query is no longer fetching and there's no error, the data will be used to render a list of users
const AllUsers: React.FC = () => {
const [result] = useQuery<AllUsersData>({
query: AllUsersQuery
});
const { data, fetching, error } = result;
if (fetching) return <p>Loading...</p>;
if (error) return <p>Oh no... {error.message}</p>;
return (
<div>
<p>There are {data?.allUsers.length} user(s) in the database:</p>
<ul>
{data?.allUsers.map(user => (
<li key={user.id}>{user.name}</li>
))}
</ul>
</div>
);
};
export default AllUsers;Now, save the TSX file, and mount it onto the home page /pages/index.tsx:
import Link from 'next/link';
import Layout from '../components/Layout';
import AllUsers from '../components/AllUsers';
const IndexPage = () => (
<Layout title="Home | Next.js + TypeScript Example">
<h1>Hello Next.js 👋</h1>
<p>
<Link href="/about">
<a>About</a>
</Link>
</p>
{/* === Tada! === */}
<AllUsers />
</Layout>
);
export default IndexPage;Time to spin up the Next.js dev server!
npm run devVoilà! The user list is rendered! 🥳
Instead of manually defining all the types we expect to receive via GraphQL, we can also use a very cool package GraphQL Code Generator to generate types directly from the Nexus GraphQL endpoint. This way, you essentially only have to define the types once in the schema.prisma file as the single-source-of-truth, then all types you'll use in the application can be derived from that schema with little manual effort! 🎉
First, copy and refactor the GraphQL queries from the TSX files into the graphql directory. With the example from Step 8, create a new file at /graphql/queries.graphql.ts and copy the query from /components/AllUsers.tsx:
import gql from 'graphql-tag';
export const AllUsersQuery = gql`
query AllUsers {
allUsers {
id
name
}
}
`;Separating GraphQL operations from components makes it easier to navigate the codebase.
Next, install the packages needed by graphql-code-generator as dev dependencies:
npm install -D \
@graphql-codegen/cli \
@graphql-codegen/typescript \
@graphql-codegen/typescript-operations \
@graphql-codegen/typescript-urqlThen, create a codegen.yml file in the project root with the following content:
overwrite: true
schema: 'http://localhost:3000/api/graphql' # GraphQL endpoint via the nexus dev server
documents: 'graphql/**/*.graphql.ts' # parse graphql operations in matching files
generates:
generated/graphql.tsx: # location for generated types, hooks and components
plugins:
- 'typescript'
- 'typescript-operations'
- 'typescript-urql'
config:
withComponent: false # we'll use Urql client with hooks instead
withHooks: trueThe configs above will tell graphql-code-generator to pull the GraphQL schema from http://localhost:3000/api/graphql, then generate types, Urql useQuery hooks into a file located at /generated/graphql.tsx.
Cool, let the code generation begin (in watch mode)!
npx graphql-codegen --watchYou will see some nice, written-by-a-robot code in /generated/graphql.tsx. How neat!
useAllUsersQuery hook and all the types are generated by GraphQL Code Generator
Now, you can go back to components/AllUsers.tsx, and replace the manually written AllUsersData type, the GraphQL query, and the useQuery hook, with what's in the /generated/graphql.tsx file:
import React from 'react';
import { useAllUsersQuery } from '../generated/graphql';
const AllUsers: React.FC = () => {
const [result] = useAllUsersQuery();
const { data, fetching, error } = result;
if (fetching) return <p>Loading...</p>;
if (error) return <p>Oh no... {error.message}</p>;
return (
<div>
<p>There are {data?.allUsers?.length} user(s) in the database:</p>
<ul>
{data?.allUsers?.map(user => (
<li key={user.id}>{user.name}</li>
))}
</ul>
</div>
);
};
export default AllUsers;Revisit the index page of the app at http://localhost:3000, everything works like a charm! 🙌
To make the development experience even better, let's optimize the NPM scripts for the project.
First, install the Concurrently NPM module, which is a great tool for running multiple CLI watchers at the same time:
npm install -D concurrentlyThen, replace dev script in the package.json file with the following:
{
// ...
"scripts": {
// ...
"dev": "concurrently -r \"npx nexus dev\" \"npx next\" \"npx graphql-codegen --watch\""
// ...
}
// ...
}Now, we can use a single npm run dev command to launch Nexus, Next.js, and GraphQL Code Generator, all at the same time!
I hope you have enjoyed this tutorial and have learned something useful! You can find the source code in this GitHub repo.
Also, check out the Awesome Prisma list for more tutorials and starter projects in the Prisma ecosystem!