A boilerplate for TypeScript + Node Express + Apollo GraphQL APIs.
- Features
- Setup
- Prerequisites
- Getting Started
- Project Structure
- Sample Environment File
- Recommended Workflow
- Database Migrations and Seeding
- Session Management with Supertokens
- Naming Convention
- Deployment
- Pro Tips
- Contributing
- Known Dependency Issues
- License
- ESM (ECMAScript Modules) support
- Login with Email and Password
- Session Management (Access Token + Rotating Refresh Tokens) with SuperTokens.
- SuperTokens is only used for Session Management with this setup but it can do more than that. If you need a different kind of Authentication (e.g. Social Login, Passwordless), read more on the SuperTokens docs on how to setup one easily.
- Node Express REST endpoint examples
- Apollo GraphQL as middleware for Node Express
- Centralized error handling
- See
src/errors/base.error.ts
andsrc/errors/error-handler/index.ts
. - For GraphQL, see
src/graphql/index.ts
and look for theformatError
of Apollo. - For REST, see
src/middlewares/error.middleware.ts
.
- See
- Automatic type code generation for GraphQL resolvers with GraphQL Code Generator
- Facebook Dataloader for caching and batching
- PostgreSQL Database
- pgAdmin for managing the PostgreSQL Database
- Redis for Caching and GraphQL Subscription
- RedisCommander for managing the Redis Database
- Pre-commit hook for auto formatting files with Husky, Lint-Staged and Prettier
- Zod for schema validations
- GraphQL Subscription example
- Dockerized containers for both development and production
- Multi-stage build for production
- Kysely for query building
- Prisma for database migrations and seeding
- If your project has authorization needs and your frontend also uses JavaScript/TypeScript, I recommend using CASL. If you want to spin your own but not sure how, here's a good article to get the ideas down. If you want a language-agnostic solution, there is Permify.
If the nature of your app isn't CRUDy (for a lack of a better word), you probably need domain models and a good amount of layering + mappers. Here's a good article on the topic. Otherwise, proceed.
- Docker
- NodeJS
- TypeScript
- supertokens-website on your frontend app.
- For the client-side session management.
- Read more here for more info about the specific cookies being used by
SuperTokens
. - If you don't use JS for your frontend app, you'll have to make the HTTP requests that
supertokens-website
heavy lifts yourself, such as refreshing the session when it expires.
# Install dependencies for the host
pnpm install
# Generate GraphQL Types
pnpm generate:gql-types
# Build the project for the first time or when you add dependencies
docker-compose build
# Start the application (or to restart after making changes to the source code)
docker-compose up
# Or run it in detached mode then listen only to this application's logs
docker-compose up -d
docker-compose logs -f api
Then refer to Database Migrations and Seeding to setup the database.
Note: You might be prompted to share your drive with Docker if you haven't done so previously. The drive letter you need to share in this case would be the drive letter of where this repository resides.
If docker compose have bootstrapped all services successfully, you should be able to access:
Login to the pgAdmin page using the credentials below:
Field | Value |
---|---|
[email protected] | |
password | password |
If first time setting up:
- Right click on the
Servers
then create a new server. - In the
General
tab, give this server connection a name. - Switch over to the
Connection
tab. - Fill in the fields below.
Field | Value | Notes |
---|---|---|
Host name/address | db | Service name in docker-compose.yml file for our Database service is named db |
Username | postgres | Default username is postgres |
Password | password | As defined in the docker-compose.yml config |
http://localhost:8080/api/v1/maintenance/health-check
Note: If you prefer a different port, container name, or anything docker environment related. Just modify the docker-compose.yml
file and adjust to your preferred setup.
Name | Description |
---|---|
src/config/* | Any app level environment configs should go here. |
src/db/schema/index.ts | Zod Database schemas goes here. |
src/db/types.d.ts | Generated types by prisma-kysely . |
src/errors/<error-name> .error.ts |
Custom Errors. |
src/generated/**/*.ts | Generated files. |
src/graphql/scalars/<scalar-name> .scalar.ts |
Custom Scalars and their resolvers. |
src/graphql/enums/index.ts | GraphQL Enum resolvers and internal values. |
src/graphql/index.ts | Apollo GraphQL setup. |
src/graphql/schema.ts | GraphQL Schema/Resolver builder script. |
src/graphql/init-loaders.ts | Collect all dataloaders here. |
src/graphql/pubsub.ts | Initialize pubsub engine here. |
src/middlewares/<middleware-name> .middleware.ts |
Node Express Middleware files. |
src/modules/<module-name> /* |
Your feature modules. |
src/modules/_/* | Reserved module. Contains root schema for graphql to work along with some sample resolvers. |
src/redis/index.ts | Default Redis client is initialized here along with some helpers. |
src/shared/ | Anything that's shared (generic) throughout the app should be placed here. |
src/utils/<utility-name> .util.ts |
Utility files. |
src/app.ts | Main application file. |
.dockerignore | Folder and files ignored by Docker. |
.eslintignore | Folder and files ignored by ESLint. |
.eslintrc.js | Linter rules are defined here. |
.gitignore | Folder and files that should be ignored by git. |
.huskyrc | Husky config. Git hooks made easy. |
.lintstagedrc | Lint-Staged config. Run commands against staged git files. |
.prettierrc | Prettier config. An opinionated code formatter. |
codegen.yml | GraphQL Code Generator (file watcher) config file. |
docker-compose.yml | Docker compose config file. |
Dockerfile | Production Docker config file. |
Dockerfile.dev | Development Docker config file used by docker-compose . |
gulpfile.ts | Gulp task runner config file. |
tsconfig.json | Contains typescript config for this project. |
Note: This project structure makes use of barrel files, those index.ts you see on most of the folders. Make sure not to forget to export your newly created files to their respective barrel files (index.ts) if applicable!
# Make sure to set this to "production" in production environments
NODE_ENV=
# If you want to change the app port. Defaults to 8080.
PORT=
# DB Connection URLs
POSTGRES_CONNECTION_URL=
REDIS_CONNECTION_URL=
# SuperTokens
SUPERTOKENS_CONNECTION_URL=
SUPERTOKENS_API_KEY=
SUPERTOKENS_APP_NAME=
SUPERTOKENS_API_DOMAIN=
SUPERTOKENS_WEBSITE_DOMAIN=
See files inside src/config/*
that uses process.env
. Those are the environment variables that you can configure.
If your feature requires modifications to the database, update the Prisma Schema accordingly then run pnpm exec prisma migrate dev
.
See docs if you're unfamiliar with this command.
Since we have prisma-kysely
configured, this should also generate the corresponding typescript types for your database models and should be usable with Kysely right away for your database queries.
If you want to seed data, update the seed file at prisma/seed.ts
accordingly.
Based on your database model changes, make sure to reflect your schema changes as well in the src/db/schema/index.ts
file.
- Create one under the
src/modules/<module_name>
directory.
- Create a use case TS file under your module.
- Should be under
src/modules/<module_name>/use-cases/<use_case_name>.use-case.ts
- Should be under
- Create a Zod schema for your DTO.
- Infer the DTO type based on the Zod schema for your use case.
- Check for auth requirements if applicable.
- Validate DTO.
- Write your business logic for the use case.
- If some operations warrants creating a service, don't hesitate to create one.
- Make sure to return an object-like data on your use case functions so that you can easily extend the results if needed.
- Unless it's a write operation and is idempotent in nature, then you might not need to return anything.
Important:
- Your Interface Layer shouldn't do any DB Operations directly.
- Use cases should never call another use case. There will be tight coupling if you do that. In the case you decide to change a parameter or behavior on the child use case, the dependent use case would've to adjust accordingly. Instead, create a service that both use cases can use.
- If you foresee an operation that you think will be reused whether by the same or other modules, put it in a service and let this service be called by the module that wants to use it.
- Create a GraphQL file (
.graphql
) under a graphql folder of your module:src/modules/<module_name>/graphql/
.- The folder structure convention is important. This particular path is being used by GraphQL Code Generator and
graphql-tools/load-files
to locate your type definitions.
- The folder structure convention is important. This particular path is being used by GraphQL Code Generator and
- Define your GraphQL SDL.
- Think in graphs.
- Your schema doesn't necessarily have to be a reflection of your database schema.
-
If not yet present, create a new folder named
factories
undersrc/modules/<module_name>/
. -
Create a factory file under that folder.
- Recommended file name should be in the format:
<entity-name>.factory.ts
- Recommended file name should be in the format:
-
Create the main factory function inside the file:
-
Recommended function name should be in the format:
createGQL<graphql-object-type-name>
-
Make sure to specify the return type of this function to the equivalent GraphQL Object Type generated by GraphQL Code Generator.
-
Example:
function createGQLUser(user: Selectable<User>): GQL_User { // ... }
-
- If not yet present, create a new folder named
resolvers
undersrc/modules/<module_name>/graphql/
. - Create an
index.ts
file under the folder.- Note: This is important as
src/graphql/schema.ts
uses theindex
barrels (TS files) to load the resolvers.
- Note: This is important as
- Create the
query/mutation/subscription
resolver file under the folder. - Implement. Refer to the resolvers under
user
for examples.
Tip: Keep your resolvers thin by making the business logic layer (use cases) do the actual work and calling those in the resolvers.
If you have enums defined in your GraphQL Schema, most likely you have your own internal values which means you'll have to resolve these enums to match your internal values which, most of the time, are internal enums you use in the app. Simply define the enums under the src/graphql/enums
directory and make sure to export it on the index
barrel.
This is also true to scalars at src/graphql/scalars
.
- If not yet present, create a new folder named
routes
undersrc/modules/<module_name>/
. - Create an
index.ts
file inside that folder.- Note: This file will contain all your route definitions (router) for that module.
- Important: Do not forget to import this router on
app.ts
.
- Define the router. See existing index files for examples.
- If not yet present, create a new folder named
handlers
undersrc/modules/<module_name>/routes/
. - Create a route handler file under that folder.
- Recommended file name should be in the format:
<graphql-object-type-name>.handler.ts
- Recommended file name should be in the format:
- Implement. Refer to the route handlers under
user
for examples. - Add this route handler to your router.
- Make sure to wrap this with the async handler util.
Migrations and Seeding is handled by Prisma. Refer to docs on how to work with this.
The environment variables (see src/config/environment.ts
file) are currently configured with consideration to a running Docker container so running the scripts below directly from the Host machine wouldn't work. As you can see from the file, the connection strings are pointing to addresses that based on the Docker containers name (defined in docker-compose.yml
) and the default Docker Network created from running docker-compose up
makes this work.
This means you should run the Prisma commands within the Docker container. You can do this in two simple ways, either use (1) the Docker extension from VS Code or (2) do it via CLI by:
First, identify the Container ID:
docker ps
Then run the following command:
docker exec -it <container_id> sh
Tip: You don't need to supply the entire Container ID, just the first few unique sequence. For example, your target Container ID is dc123f66554b
. You can just run docker-exec -it dc sh
and you'll be inside the container.
Once inside the container, you can run your prisma commands.
pnpm exec prisma migrate dev
This will run database migrations. If there are changes in the Prisma Schema before you run this command, it will create a migration file and you will be prompted for a name for the migration. Make sure to double check the output.
Important: If you're not using Docker to run this app, you need to configure the connection strings of the database servers (e.g. Redis, PostgreSQL) via the environment variables.
pnpm exec prisma db seed
Important: Make sure to write idempotent seed scripts!
Resetting the database also runs the seed automatically.
pnpm exec prisma migrate reset
This boilerplate makes use of SuperTokens to handle sessions (with cookies) because security is very hard. The folks from Supertokens know really well what they're doing and it'd be in your best interest to know how this topic should be handled properly. Here are some relevant resources:
- Stop using JWT for sessions
- The best way to securely manage user sessions
- Detecting session hijacking using rotating refresh tokens - OSW 2020
That said, see SuperTokens docs for specifics.
When trying this boilerplate and heading to the playground right away to test some GraphQL queries, you'll probably get an Unauthenticated Error. Something like the one below for example when querying users
:
{
"errors": [
{
"message": "Unauthenticated. Please try logging in again.",
"locations": [
{
"line": 2,
"column": 3
}
],
"path": ["users"],
"extensions": {
"code": "UNAUTHENTICATED",
"data": {
"$errorId": "801aecca-b19f-410c-9a7e-e9724b6f0e9f"
}
}
}
],
"data": null
}
This is because you're trying to request a resource that requires authentication (and in turn, you'll need a session). See this issue for implementation details of this boilerplate. Basically, the idea is you need the session cookies to be attached on your requests and this can be done after logging in successfully.
You can hit the /api/v1/auth/login/superadmin
endpoint to login as superadmin and get a session cookie to make requests on the playground. Your access token might expire after some time so you'll need to refresh it. Under normal circumstances, the frontend SDK of Supertokens will handle this for you but since we're working purely on the backend side for this boilerplate, you'll have to clear the cookies manually and then hit the endpoint again.
If sessions with cookies is not an option for you, Supertokens also supports header-based sessions (basically token-based). See this link for specifics. In the scenario that you're not using any of the Frontend SDK of Supertokens, then refer here.
Note: If you're not using Docker to run this app or prefer using Supertokens' Managed Service, make sure to configure the environment variables. See src/config/supertokens.ts
.
Generally, use snake-case
.
In most cases, we include the file functionality
in its file name in the format:
<file-name>.<functionality>.<extension>
For example:
- get-users
.use-case
.ts - sort-direction
.enum
.ts - user
.model
.ts - async-handler
.util
.ts
TypeScript Interface
and Type
file names should match their definition name.
For example:
Interface/Type name | File name |
---|---|
IAccessTokenPayload |
IAccessTokenPayload .ts |
IContext |
IContext .ts |
UniqueID |
UniqueID .ts |
Maybe |
Maybe .ts |
Read more on the SuperTokens docs based on your infrastructure.
Make sure to configure the database setup.
There are a few ways to go about this and can vary based on your setup.
Simply build the image from your local machine (or let your CI tool do it) with the command below.
# The dot on the end is important.
docker build -f <docker_file> -t <image_name>[:<image_tag>] .
# Example:
docker build -f Dockerfile -t graphql-starter:latest .
Then push the image you just built from your local machine (or from your CI tool) to your docker image repository (e.g. Docker Hub, AWS ECR, GitHub Container Registry).
docker push <typically_a_url_to_your_docker_image_repository>
Let your host server pull this image and run it from there.
Note: Typically, the tag
for the image you built should match the url
to your docker image repository. Refer to your image repository provider for specific details.
Extra: If you want to test the image you just built with your local setup, then configure the docker-compose file such that it points to your image instead of building from a Dockerfile. For example:
From:
services:
api:
build:
context: .
dockerfile: Dockerfile.dev
command: npm start
volumes:
- .:/usr/src/app
- /usr/src/app/node_modules
restart: always
To:
services:
api:
# Set it to the image you just built
image: graphql-starter:latest
restart: always
If you're not using Docker, you can still get your hands on the build files. Normally, you'd want this approach if you're deploying to a VPS and have to move the files via FTP manually to your server or you have a cloud provider that accepts node apps where you'll have to provide it your project files with a package.json
file in the project folder (typically zipped).
Build the project:
pnpm build:prod
This will create a build
folder in the project directory which you can deploy.
Note: You might need to manually install the dependencies yourself if you're using a VPS. Otherwise, your cloud provider's NodeJS container will typically just need a package.json
from the root folder and they'll do the installation on every deploy.
-
When resolver types are generated by GraphQL Code Generator, the type of the 1st parameter of a field resolver is the parent type by default. This is not always true because at runtime, what the parent resolver returns is the actual type/object that will arrive in the field resolver 1st (parent) parameter. In cases like this, you'd need to type assert the parent. See
full-name.query.ts
.- Another example for this is let's say we have a
pets
table and a pet has anownerId
but in your GraphQL Schema, what you expose isowner
and notownerId
. You won't have access to theownerId
in your resolver because GraphQL Code Generator generates what you defined in the schema. You'll have then to type assert in your resolver the type you know you returned.
- Another example for this is let's say we have a
-
If you want to rename a field in your GraphQL Schema:
- DO NOT rename first on the GraphQL schema file (
*.graphql
).- Because if you do, the files that references the to-be-renamed field will break and TypeScript will fail to compile.
- GraphQL Code Generator only generates what's defined in the schema and overwrites the generated file so the old name that was previously being referenced is now missing.
- What you SHOULD DO instead is:
- Rename first the field in the generated graphql type definition file by GraphQL Code Generator at
src/generated/graphql/index.ts
then apply the new name in the GraphQL schema file.- Saves your time and sanity.
- Rename first the field in the generated graphql type definition file by GraphQL Code Generator at
- DO NOT rename first on the GraphQL schema file (
-
When trying to auto import a GraphQL resolver to its respective index barrel (
index.ts
file), you might notice you're not getting code completion when typing the resolver name in theexport default
object. This is normal because your IDE thinks you're typing the key/property name (remember key-value pair). -
When trying to debug async functions in VSCode and the breakpoints on the inner lines won't hit, try adding
trace: true
tolaunch.json
file. -
Generated custom scalars by GraphQL Code Generator are given a type of
any
. As with TypeScript, if you can help it, give it a type/interface then map it to GraphQL Code Generator. You can read more here. But basically:- You define the type/interface in the corresponding scalar file and export it.
- Then map it to GraphQL Code Generator by adding scalars config in
codegen.yml
such that it points to its corresponding scalar file.- Format is:
<Scalar_Name_in_GQL_Schema>: <path_to_custom_type>#<name_of_type>
- Format is:
-
If you're using this repo template with Docker Windows, you'll notice that when you save any file, the watchers aren't reacting. This has something to do with Linux not being able to detect file changes from Windows because the file system events are not being propagated down to Linux. You'll have to resort to polling mechanisms in that case for any tools that watches file changes. I recommend using WSL and developing inside WSL instead.
-
If you want to boost your knowledge about GraphQL, I highly recommend reading the e-book: Production Ready GraphQL by Marc-Andre Giroux.
-
If you have a use case for File Uploads, I recommend reading Apollo's blog post for options. There used to be a simple example in this boilerplate but has been removed. See Issue #42 and the associated pull request to see relevant code.
If something is unclear, confusing, or needs to be refactored, please let me know. Pull requests are always welcome but do consider the opinionated nature of this project. Please open an issue before submitting a pull request.
It might not look good to list it here but I think it's important for you to be aware of the issues currently being worked on (and hopefully get a fix from their respective authors/maintainers soon) as you'd still end up with these issues if you were to implement these yourself and use the same dependencies.
- Wrong TypeScript types from Apollo's
graphql-subscriptions
vsgraphql-codegen
. More info.
MIT License
Copyright (c) 2019-Present Cerino O. Ligutom III
Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:
The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.