MilosSimic/go-rest-api

An idiomatic Go REST API starter kit (boilerplate) following the SOLID principles and Clean Architecture

Go RESTful API Starter Kit (Boilerplate)

This starter kit is designed to get you up and running with a project structure optimized for developing RESTful API services in Go. It promotes the best practices that follow the SOLID principles and clean architecture. It encourages writing clean and idiomatic Go code.

The kit provides the following features right out of the box:

• RESTful endpoints in the widely accepted format
• Standard CRUD operations of a database table
• JWT-based authentication
• Environment dependent application configuration management
• Structured logging with contextual information
• Error handling with proper error response generation
• Database migration
• Data validation
• Full test coverage
• Live reloading during development

The kit uses the following Go packages which can be easily replaced with your own favorite ones since their usages are mostly localized and abstracted.

• Routing: ozzo-routing
• Database access: ozzo-dbx
• Database migration: golang-migrate
• Data validation: ozzo-validation
• Logging: zap
• JWT: jwt-go

Getting Started

If this is your first time encountering Go, please follow the instructions to install Go on your computer. The kit requires Go 1.13 or above.

Docker is also needed if you want to try the kit without setting up your own database server. The kit requires Docker 17.05 or higher for the multi-stage build support.

After installing Go and Docker, run the following commands to start experiencing this starter kit:

# download the starter kit
git clone https://github.com/qiangxue/go-rest-api.git

cd go-rest-api

# start a PostgreSQL database server in a Docker container
make db-start

# seed the database with some test data
make testdata

# run the RESTful API server
make run

# or run the API server with live reloading, which is useful during development
# requires fswatch (https://github.com/emcrisostomo/fswatch)
make run-live

At this time, you have a RESTful API server running at http://127.0.0.1:8080. It provides the following endpoints:

• GET /healthcheck: a healthcheck service provided for health checking purpose (needed when implementing a server cluster)
• POST /v1/login: authenticates a user and generates a JWT
• GET /v1/albums: returns a paginated list of the albums
• GET /v1/albums/:id: returns the detailed information of an album
• POST /v1/albums: creates a new album
• PUT /v1/albums/:id: updates an existing album
• DELETE /v1/albums/:id: deletes an album

Try the URL http://localhost:8080/healthcheck in a browser, and you should see something like "OK v1.0.0" displayed.

If you have cURL or some API client tools (e.g. Postman), you may try the following more complex scenarios:

# authenticate the user via: POST /v1/login
curl -X POST -H "Content-Type: application/json" -d '{"username": "demo", "password": "pass"}' http://localhost:8080/v1/login
# should return a JWT token like: {"token":"...JWT token here..."}

# with the above JWT token, access the album resources, such as: GET /v1/albums
curl -X GET -H "Authorization: Bearer ...JWT token here..." http://localhost:8080/v1/albums
# should return a list of album records in the JSON format

To use the starter kit as a starting point of a real project whose package name is github.com/abc/xyz, do a global replacement of the string github.com/qiangxue/go-rest-api in all of project files with the string github.com/abc/xyz.

Project Layout

The starter kit uses the following project layout:

.
├── cmd                  main applications of the project
│   └── server           the API server application
├── config               configuration files for different environments
├── internal             private application and library code
│   ├── album            album-related features
│   ├── auth             authentication feature
│   ├── config           configuration library
│   ├── entity           entity definitions and domain logic
│   ├── errors           error types and handling
│   ├── healthcheck      healthcheck feature
│   └── test             helpers for testing purpose
├── migrations           database migrations
├── pkg                  public library code
│   ├── accesslog        access log middleware
│   ├── graceful         graceful shutdown of HTTP server
│   ├── log              structured and context-aware logger
│   └── pagination       paginated list
└── testdata             test data scripts

The top level directories cmd, internal, pkg are commonly found in other popular Go projects, as explained in Standard Go Project Layout.

Within internal and pkg, packages are structured by features in order to achieve the so-called screaming architecture. For example, the album directory contains the application logic related with the album feature.

Within each feature package, code are organized in layers (API, service, repository), following the dependency guidelines as described in the clean architecture.

Common Development Tasks

This section describes some common development tasks using this starter kit.

Implementing a New Feature

Implementing a new feature typically involves the following steps:

1. Develop the service that implements the business logic supporting the feature. Please refer to internal/album/service.go as an example.
2. Develop the RESTful API exposing the service about the feature. Please refer to internal/album/api.go as an example.
3. Develop the repository that persists the data entities needed by the service. Please refer to internal/album/repsitory.go as an example.
4. Wire up the above components together by injecting their dependencies in the main function. Please refer to the album.RegisterHandlers() call in cmd/server/main.go.

Working with DB Transactions

It is the responsibility of the service layer to determine whether DB operations should be enclosed in a transaction. The DB operations implemented by the repository layer should work both with and without a transaction.

You can use dbcontext.DB.Transactional() in a service method to enclose multiple repository method calls in a transaction. For example,

func serviceMethod(ctx context.Context, repo Repository, transactional dbcontext.TransactionFunc) error {
    return transactional(ctx, func(ctx context.Context) error {
        repo.method1(...)
        repo.method2(...)
        return nil
    })
}

If needed, you can also enclose method calls of different repositories in a single transaction. The return value of the function in transactional above determines if the transaction should be committed or rolled back.

You can also use dbcontext.DB.TransactionHandler() as a middleware to enclose a whole API handler in a transaction. This is especially useful if an API handler needs to put method calls of multiple services in a transaction.

Updating Database Schema

The starter kit uses database migration to manage the changes of the database schema over the whole project development phase. The following commands are commonly used with regard to database schema changes:

# Execute new migrations made by you or other team members.
# Usually you should run this command each time after you pull new code from the code repo. 
make migrate

# Create a new database migration.
# In the generated `migrations/*.up.sql` file, write the SQL statements that implement the schema changes.
# In the `*.down.sql` file, write the SQL statements that revert the schema changes.
make migrate-new

# Revert the last database migration.
# This is often used when a migration has some issues and needs to be reverted.
make migrate-down

# Clean up the database and rerun the migrations from the very beginning.
# Note that this command will first erase all data and tables in the database, and then
# run all migrations. 
make migrate-reset

Managing Configurations

The application configuration is represented in internal/config/config.go. When the application starts, it loads the configuration from a configuration file as well as environment variables. The path to the configuration file is specified via the -config command line argument which defaults to ./config/local.yml. Configurations specified in environment variables should be named with the APP_ prefix and in upper case. When a configuration is specified in both a configuration file and an environment variable, the latter takes precedence.

The config directory contains the configuration files named after different environments. For example, config/local.yml corresponds to the local development environment and is used when running the application via make run.

Do not keep secrets in the configuration files. Provide them via environment variables instead. For example, you should provide Config.DSN using the APP_DSN environment variable. Secrets can be populated from a secret storage (e.g. HashiCorp Vault) into environment variables in a bootstrap script (e.g. cmd/server/entryscript.sh).

Deployment

The application can be run as a docker container. You can use make build-docker to build the application into a docker image. The docker container starts with the cmd/server/entryscript.sh script which reads the APP_ENV environment variable to determine which configuration file to use. For example, if APP_ENV is qa, the application will be started with the config/qa.yml configuration file.

You can also run make build to build an executable binary named server. Then start the API server using the following command,

./server -config=./config/prod.yml