Golang GraphQL API generator using gqlgen and gorm
While following microservices design patterns we ended up with many "model services". gqlgen is perfect tool, but implementing resolvers in every service is getting more and more cumbersome. Using this tool we only have to update model.graphql
and all resolvers get generated automatically.
Before you start, please make sure you have goimports
installed:
go get golang.org/x/tools/cmd/goimports
NOTE: Make sure you have Go installed on your system.
- Create new project repository
- run
go mod init [MODULE]
to initialize your project with go modules - run
go run github.com/loopcontext/go-graphql-orm init
- follow initialization instruction (creating makefile is suggested)
- open created
model.graphql
and create your custom model schema - each time you change model, run
make generate
orgo run github.com/loopcontext/go-graphql-orm
to recreate generated source codes
NOTE: graphql-orm requires Go modules for installation. If you are running in $GOPATH, make sure you are running init command with GO111MODULE=on
For running locally you can use:
make run
or without makefile:
DATABASE_URL=sqlite3://test.db PORT=8080 go run *.go
If you need to debug:
make debug
PORT
- sets the default port for the APIDEBUG
- setting to"true"
enables debugging for GORMDATABASE_URL
- connection string for database in formatdb://user:password@host:port/tablename
(eg.mysql://root:pass@localhost:3306/test
; required)EXPOSE_MIGRATION_ENDPOINT
- expose/migration
endpoint which triggers database migration (migrates to latest database schema; default: false)EXPOSE_PLAYGROUND_ENDPOINT
- expose/graphql/playground
playground endpoint (allows playground; default: false)TABLE_NAME_PREFIX
- set global prefix for all table names (default: "")EVENT_TRANSPORT_URL
- destination url for sending mutation events (array supported in formatEVENT_TRANSPORT_URL_[INDEX]
) see Events transportEVENT_TRANSPORT_SOURCE
- custom value for CloudEvent source attribute (default:http://{hostname}/graphql
)
In case you want to connect with sqlite, you can use local file storage:
sqlite3://path/to/file.db
Or use in-memory storage:
sqlite3://:memory:
You can find example project at graphql-orm-example repo
GraphQL Voyager is very nice tool for previewing your GraphQL Schema, you can run it locally by:
make voyager
or without makefile:
docker run --rm -v `pwd`/gen/schema.graphql:/app/schema.graphql -p 8082:80 graphql/voyager
...after voyager starts up go to https://localhost:8082
All generated files is stored in ./gen/
folder.
There's a provided a docker-compose.yml
in the root of your generated API, use it to easily run a dev instance of
the server docker-compose up dev
and navigate to http://localhost:8081
If you want, use the generated Dockerfile initialization it's as easy as running:
docker build -f docker/Dockerfile -t {IMAGE_NAME} .
If you want to create your own docker image, you can check the example repository for generated Dockerfile: https://github.com/loopcontext/go-graphql-orm-example/blob/master/Dockerfile
While following microservices design patterns we ended up with many "model services". gqlgen is perfect tool, but implementing resolvers in every service is getting more and more cumbersome. Using this tool we only have to update model.graphql
and all resolvers get generated automatically.
For event driven architecture it's necessary that the service is able to send events about changes in state. Services built using this library automatically send event for every mutation using CloudEvents (entity created/updated/deleted and changed column and their values). Supported targets are:
- HTTP/HTTPS
- AWS Services using cloudevents-aws-transport (SNS/SQS/EventBridge)
For more information about event structure see: https://github.com/loopcontext/go-graphql-orm/blob/master/events/model.go
Since version 0.4.0
the migrations using gormigrate are introduced and it's possible to write custom migrations with rollbacks.
The automigration (with foreign keys) is still available, but gormigrate migrations are used by default. You use following commands:
make migrate
- runs gormigrate migrationsmake automigrate
- runs gorm basic automigration
The same applies for HTTP endpoints (when EXPOSE_MIGRATION_ENDPOINT=true
):
POST /migrate
- runs gormigrate migrationspost /automigrate
- runs gorm basic automigration
To add new migration, edit src/migrations
file and its GetMigrations method. For more information see gormigrate Readme
If you need to do a relationship with a table, and you don't need it to be queried on its related records table, let's say it's a typification table (i.e: categories) then you hve to do like this:
type Category @entity {
name String @column
description Srting @column
product Product @relationship(inverse: "category")
}
type Product @entity {
code String @column
description String @column
category Category @relationship(inverse:"product")
}
This way when querying for categories, even though Product and ProductId will show in the list of fields, they won't get filled. This comes with the caveat that the table WILL have a productId column in the db you can always drop it, but it will be nullable and never filled.