Let's create a GraphQL server with Go.
gqlgen is a good choice to implement GraphQL server with Go.
code is here.
GitHub
This is an English version of article posted on Qiita Advent Calendar 2019.
What is GraphQL?
GraphQL is a query language for APIs and a runtime for fulfilling those queries with your existing data. GraphQL provides a complete and understandable description of the data in your API, gives clients the power to ask for exactly what they need and nothing more, makes it easier to evolve APIs over time, and enables powerful developer tools.
GraphQL is a query language and implementation for API. At first it is developed by Facebook, and now promoted by GraphQL Foundation.
We can take many advantages like type safe request, editor completion, auto-generated documents.
What is gqlgen?
gqlgen is a Go library for building GraphQL servers without any fuss.
gqlgen is GraphQL server library that have following features.
- Schema first
- Type safe
- Code generate
Schema first development is a development style. define schema
-> code generate
-> implement specific logic
. Please check [Feature comparison] if you are interested.
Features
- list tasks (with pagination)
- order tasks
- by
created_at
- by
due
- by
- create task
- update task
This is one example of GraphQL schema for implementing above features.
query {
tasks(input: TasksInput!, orderBy: TaskOrderFields!, page: PaginationInput!): TaskConnection!
}
mutation {
createTask(input: CreateTaskInput!): Task!
updateTask(input: UpdateTaskInput!): Task!
}
type Task implements Node {
id: ID!
title: String!
notes: String!
completed: Boolean!
due: Time!
}
type TaskEdge implements Edge {
cursor: String!
node: Task!
}
type TaskConnection implements Connection {
pageInfo: PageInfo!
edges: [TaskEdge]!
}
input TasksInput {
completed: Boolean
}
enum TaskOrderFields {
LATEST
DUE
}
input CreateTaskInput {
title: String!
notes: String
completed: Boolean
due: Time
}
input UpdateTaskInput {
taskID: ID!
title: String
notes: String
completed: Boolean
due: Time
}
type PageInfo {
endCursor: String!
hasNextPage: Boolean!
}
interface Connection {
pageInfo: PageInfo!
edges: [Edge]!
}
interface Edge {
cursor: String!
node: Node!
}
interface Node {
id: ID!
}
input PaginationInput {
first: Int
after: String
}
Details will be explained on each implementation.
Project settings
1. Create Go environment using Docker
and docker-compose
.
Create Go environment by go modules and install following softwares or libraries.
- Go 1.13.4
- Docker, docker-compose
- MySQL 8.0.13
- ORM
- HTTP server
- DB migration tool
- hot reload for server
- struct validation
- unique id generator
FROM golang:1.13.4-alpine3.10 as build
WORKDIR /app
RUN apk update --no-cache \
&& apk add --no-cache \
git \
gcc \
musl-dev
COPY go.mod .
COPY go.sum .
RUN go mod download
COPY . .
RUN GOOS=linux GOARCH=amd64 go build -o app main.go
RUN GO111MODULE=off go get github.com/oxequa/realize
RUN GO111MODULE=off go get -tags 'mysql' -u github.com/golang-migrate/migrate/cmd/migrate
---
version: '3.7'
services:
app:
container_name: graphql-app-backend
build:
context: ./app
target: build
volumes:
- ./app:/app
environment:
DB_HOST: db
DB_PORT: 3306
DB_USER: root
DB_PASSWORD: root
DB_NAME: graphql-app-development
ports:
- 3000:3000
depends_on:
- db
links:
- db
tty: true
db:
container_name: graphql-app-db
image: mysql:8.0.13
volumes:
- ./db/mysql/data:/var/lib/mysql
- ./db/mysql/my.cnf:/etc/mysql/conf.d/my.cnf
environment:
MYSQL_USER: root
MYSQL_ROOT_PASSWORD: root
MYSQL_DATABASE: graphql-app-development
ports:
- 3306:3306
tty: true
2. Setup Go modules and create simple HTTP server using echo
Create simple HTTP server with Go using echo. echo is High performance and minimalist Go web framework.
package main
import (
"net/http"
"github.com/labstack/echo"
"github.com/labstack/echo/middleware"
)
func main() {
e := echo.New()
e.Use(middleware.Recover())
e.Use(middleware.Logger())
e.Use(middleware.Gzip())
e.Use(middleware.CORSWithConfig(middleware.CORSConfig{
AllowOrigins: []string{os.Getenv("CORS_ALLOW_ORIGIN")},
AllowHeaders: []string{echo.HeaderOrigin, echo.HeaderContentType, echo.HeaderAccept},
}))
e.GET("/health", func(c echo.Context) error {
return c.NoContent(http.StatusOK)
})
e.HideBanner = true
e.Logger.Fatal(e.Start(":3000"))
}
3. Setup hot reload
Add realize.yml
for setting up hot reload using realize.
---
settings:
legacy:
force: false
interval: 0s
schema:
- name: app
path: .
commands:
install:
status: true
method: go build -o app main.go
run:
status: true
method: ./app
watcher:
extensions:
- go
paths:
- /
ignored_paths:
- .realize
4. Create Makefile
Create Makefile for running each commands easily.
DB_HOST=db
DB_PORT=3306
DB_USER=root
DB_PASSWORD=root
DB_NAME=graphql-app-development
DB_CONN=mysql://${DB_USER}:${DB_PASSWORD}@tcp\(${DB_HOST}:${DB_PORT}\)/${DB_NAME}
.PHONY: run
run:
docker-compose up --build -d
.PHONY: start
start:
docker-compose exec app realize start --run
# create migration file
.PHONY: migrate-create
migrate-create:
docker-compose exec app migrate create -ext sql -dir migrations ${FILENAME}
# run migration
.PHONY: migrate-up
migrate-up:
docker-compose exec app migrate --source file://migrations --database ${DB_CONN} up
# run migration(rollback)
.PHONY: migrate-down
migrate-down:
docker-compose exec app migrate --source file://migrations --database ${DB_CONN} down 1
5. Overview
File structure should be like this.
$ tree backend
.
├── .gitignore
├── Makefile
├── README.md
├── app
│ ├── .realize.yaml
│ ├── Dockerfile
│ ├── go.mod
│ ├── go.sum
│ └── main.go
├── db
│ └── mysql
│ └── my.cnf
└── docker-compose.yml
When you run make
at backend
directory, app
service and db
server will be up. And you can start server with make start
command.
$ make start
docker-compose exec app realize start --run
[14:28:00][APP] : Watching 9 file/s 6 folder/s
[14:28:00][APP] : Install started
[14:28:01][APP] : Install completed in 0.748 s
[14:28:01][APP] : Running..
[14:28:02][APP] : ⇨ http server started on [::]:3000
Create tasks table
1. create and run migration fil
Next, create tasks table.
Run make migrate-create
for creating migration file.
$ FILENAME=create_tasks make migrate-create
<timestamp>_create_tasks.up.sql
and <timestamp>_create_tasks.down.sql
will be created under migrations
directory. So write up/down SQL.
CREATE TABLE tasks (
id INT NOT NULL AUTO_INCREMENT,
identifier varchar(255) BINARY NOT NULL,
title varchar(255) NOT NULL,
notes text NOT NULL,
completed tinyint(1) NOT NULL DEFAULT 0,
due timestamp NULL DEFAULT NULL,
created_at timestamp NOT NULL,
updated_at timestamp NOT NULL,
deleted_at timestamp NULL DEFAULT NULL,
PRIMARY KEY (id),
UNIQUE KEY uix_tasks_identifier (identifier)
) ENGINE=InnoDB;
DROP TABLE IF EXISTS tasks;
Run migrate-up
for creating table.
$ make migrate-up
Table will be like this.
$ docker-compose exec db -it sh
sh> mysql -uroot -proot graphql-app-development
mysql> desc tasks;
+------------+--------------+------+-----+---------+----------------+
| Field | Type | Null | Key | Default | Extra |
+------------+--------------+------+-----+---------+----------------+
| id | int(11) | NO | PRI | NULL | auto_increment |
| identifier | varchar(255) | NO | UNI | NULL | |
| title | varchar(255) | NO | | NULL | |
| notes | text | NO | | NULL | |
| completed | tinyint(1) | NO | | 0 | |
| due | timestamp | YES | | NULL | |
| created_at | timestamp | NO | | NULL | |
| updated_at | timestamp | NO | | NULL | |
| deleted_at | timestamp | YES | | NULL | |
+------------+--------------+------+-----+---------+----------------+
9 rows in set (0.02 sec)
Set up gqlgen
You will set up gqlgen. This step is almost the same as official tutorial.
gqlgen getting started
Create project template
Run gqlgen init
for creating gqlgen project template.
$ docker-compose exec app go run github.com/99designs/gqlgen init
Then following files will be created.
- gqlgen.yml
- settings for gqlgen
- generated.go
- GraphQL runtime (updated by
go generate
)
- GraphQL runtime (updated by
- models_gen.go
- missing model definitions like GraphQL type, input, enum. (updated by
go generate
)
- missing model definitions like GraphQL type, input, enum. (updated by
- resolver.go
- GraphQL resolver (you will implement business logic for query/mutation)
- schema.graphql
- schema definitions
- server/server.go
- run server
Then edit following files.
generated.go
- move to
resolver/generated.go
and rename package
models_gen.go
- move to
model/models_gen.go
and rename package
Update gqlgen settings for applying new file structures.
---
schema:
- "schema/*.graphql"
exec:
filename: resolver/generated.go
package: resolver
model:
filename: model/models_gen.go
package: model
resolver:
filename: resolver/resolver.go
type: Resolver
Move to resolver/resolver.go
and change content as following.
//go:generate go run github.com/99designs/gqlgen
package resolver
type Resolver struct{}
type queryResolver struct{ *Resolver }
type mutationResolver struct{ *Resolver }
func New() *Resolver {
return &Resolver{}
}
func (r *Resolver) Mutation() MutationResolver {
return &mutationResolver{r}
}
func (r *Resolver) Query() QueryResolver {
return &queryResolver{r}
}
Remove unnecessary types, queries and mutations.
type Query {}
type Mutation {}
Delete server/server.go
and add /graphql
endpoint at main.go
.
package main
import (
"app/resolver"
"net/http"
"github.com/99designs/gqlgen/handler"
"github.com/labstack/echo"
"github.com/labstack/echo/middleware"
)
func main() {
e := echo.New()
e.Use(middleware.Recover())
e.Use(middleware.Logger())
e.Use(middleware.Gzip())
e.GET("/health", func(c echo.Context) error {
return c.NoContent(http.StatusOK)
})
e.POST("/graphql", func(c echo.Context) error {
config := resolver.Config{
Resolvers: resolver.New(),
}
h := handler.GraphQL(resolver.NewExecutableSchema(config))
h.ServeHTTP(c.Response(), c.Request())
return nil
})
e.HideBanner = true
e.Logger.Fatal(e.Start(":3000"))
}
Finally backend/app
directory will be like this.
$ tree app
.
├── .realize.yaml
├── Dockerfile
├── go.mod
├── go.sum
├── gqlgen.yml
├── main.go
├── migrations
│ ├── <timestamp>_create_tasks.down.sql
│ └── <timestamp>_create_tasks.up.sql
├── model
│ └── models_gen.go
├── resolver
│ ├── generated.go
│ └── resolver.go
└── schema
└── schema.graphql
Add generate
command to Makefile
# add
.PHONY: generate
generate:
docker-compose exec app go generate ./...
gqlgen settings are all done. You will implement queries and mutations by these steps.
- Add/edit schema at
backend/app/schema/*.graphql
- Generate code by
make generate
- Implement resolver for fill interface which is created/updated by step 2.
Create task model
You will setup db connection and create task model.
1. setup DB
Add db settings to config/db.go
package config
import (
"fmt"
"os"
_ "github.com/go-sql-driver/mysql"
"github.com/jinzhu/gorm"
_ "github.com/jinzhu/gorm/dialects/mysql"
)
var db *gorm.DB
func InitDB() error {
conn, err := gorm.Open("mysql", dbsn())
if err != nil {
return err
}
db = conn.Set("gorm:auto_update", false)
return nil
}
func dbsn() string {
return fmt.Sprintf(
"%s:%s@tcp(%s:%s)/%s?charset=utf8mb4&parseTime=True&loc=Local",
os.Getenv("DB_USER"),
os.Getenv("DB_PASSWORD"),
os.Getenv("DB_HOST"),
os.Getenv("DB_PORT"),
os.Getenv("DB_NAME"),
)
}
func DB() *gorm.DB {
return db
}
...
func main() {
e := echo.New()
if err := config.InitDB(); err != nil {
panic(err.Error())
}
...
2. Create task model
Create task model at model/task.go
. And setup validation using GORM's BeforeSave
hook and initialize validator at model/model.go
.
package model
import "time"
type Task struct {
ID int
Identifier string `validate:"required,max=255"`
Title string `validate:"required,max=255"`
Notes string `validate:"max=65535"`
Completed bool
Due *time.Time
CreatedAt time.Time
UpdatedAt time.Time
DeletedAt *time.Time
}
func (t *Task) BeforeSave() error {
return validator.Struct(t)
}
package model
import (
v "gopkg.in/go-playground/validator.v9"
)
var validator *v.Validate
func init() {
validator = v.New()
}
3. Add Task type to GraphQL schema
Create Task type to schema/task.graphql
.
type Task {
id: ID!
title: String!
notes: String!
completed: Boolean!
due: Time
}
Time
type is gqlgen's build-in scalar type. So you should define it.
https://gqlgen.com/reference/scalars/#time
# 追記
scalar Time
4. Connect GraphQL type and Go struct at gqlgen.yml
You can connect GraphQL type and Go struct at gqlgen.yml
. GraphQL's ID type should be globally unique. So you should use identifier
over id
.
...
models:
Task:
model: app/model.Task
fields:
id:
resolver: true
5. Update resolver
Apply updates on gqlgen.yml
by make generate
.
$ make generate
Then TaskResolver
interface will be created at resolver/generated.go
. So create taskResolver
at resolver/task.go
.
ID(ctx context.Context, obj *model.Task)
method will be used for resolving id
of GraphQL Task
type.
package resolver
import (
"app/model"
"context"
)
type taskResolver struct{ *Resolver }
func (r *Resolver) Task() TaskResolver {
return &taskResolver{r}
}
func (r *taskResolver) ID(ctx context.Context, obj *model.Task) (string, error) {
if obj == nil {
return "", nil
}
return obj.Identifier, nil
}
Implement CreateTask mutation
You will implement task creation feature. (CreateTask
mutation)
1. Define schema
...
type Mutation {
createTask(input: CreateTaskInput!): Task! # 追記
}
# 追記
input CreateTaskInput {
title: String!
notes: String
completed: Boolean
due: Time
}
You can define without input like createTask(title: String!, ...): Task!
. But it is easy to read by using input. And you can easily handle request arguments because input struct will be created.
2. Implement resolver
Run make generate
.
$ make generate
Then createTask
will be added at MutationResolver
interface. So you should create CreateTask
at resolver/resolver.go
.
func (r *mutationResolver) CreateTask(ctx context.Context, input model.CreateTaskInput) (*model.Task, error) {
db := config.DB()
id, err := config.ShortID().Generate()
if err != nil {
return &model.Task{}, err
}
task := model.Task{
Identifier: id,
Title: input.Title,
Due: input.Due,
}
if input.Notes != nil {
task.Notes = *input.Notes
}
if input.Completed != nil {
task.Completed = *input.Completed
}
if err := db.Create(&task).Error; err != nil {
return &model.Task{}, err
}
return &task, nil
}
4. Check behaviors
createTask
implementation is done. Check behaviors by using clients like GraphiQL or graphql-playground.
Implement UpdateTask mutation
You will implement UpdateTask
mutation.
1. Define schema
Define schema same as createTask
.
...
type Mutation {
...
updateTask(input: UpdateTaskInput!): Task! # 追記
}
# 追記
input UpdateTaskInput {
taskID: ID!
title: String
notes: String
completed: Boolean
due: Time
}
2. implement UpdateTask resolver
Generate code.
$ make generate
Then updateTask
will be added to MutationResolver
interface.
func (r *mutationResolver) UpdateTask(ctx context.Context, input model.UpdateTaskInput) (*model.Task, error) {
db := config.DB()
var task model.Task
if err := db.Where("identifier = ?", input.TaskID).First(&task).Error; err != nil {
return &model.Task{}, err
}
params := map[string]interface{}{}
if input.Title != nil {
params["title"] = *input.Title
}
if input.Notes != nil {
params["notes"] = *input.Notes
}
if input.Completed != nil {
params["completed"] = *input.Completed
}
if input.Due == nil {
params["due"] = nil
} else {
params["due"] = *input.Due
}
if err := db.Model(&task).Updates(params).Error; err != nil {
return &model.Task{}, err
}
return &task, nil
}
3. Check behaviors
updateTask
implementation is done. Check behaviors by using clients like GraphiQL or graphql-playground.
Implement Tasks query
1. Overview
You will implement tasks
query with pagination. When implement pagination on GraphQL, relay-style pagination is recommended. So you will implement relay-style pagination.
relay-style pagination specification
2. Define schema
Define schema.
type Query {
tasks(input: TasksInput!, orderBy: TaskOrderFields!, page: PaginationInput!): TaskConnection!
}
...
type Task implements Node { # add `implements Node`
...
}
# add
type TaskEdge implements Edge {
cursor: String!
node: Task!
}
type TaskConnection implements Connection {
pageInfo: PageInfo!
edges: [TaskEdge]!
}
input TasksInput {
completed: Boolean
}
enum TaskOrderFields {
LATEST
DUE
}
...
type PageInfo {
endCursor: String!
hasNextPage: Boolean!
}
interface Connection {
pageInfo: PageInfo!
edges: [Edge]!
}
interface Edge {
cursor: String!
node: Node!
}
interface Node {
id: ID!
}
input PaginationInput {
first: Int
after: String
}
2. Generate code
Run make generate
.
$ make generate
3. Update task struct
Add IsNode method to task struct for implementing Node.
func (Task) IsNode() {} // add
4. Implement pagination
You will implement pagination like following steps.
- Decode cursor from request and get key.
- Create SQL.
- Execute SQL.
- Create connection from result.
You will implement pagination for ordering by created at
or due
. But you can not order by just targeted column.
You can order tasks by created_at
by using ORDER BY id
. But you can't do for due
because it can be duplicated.
These are example for implementing pagination by non-unique key.
id(unique)
- cursor
task:5
- SQL
SELECT * FROM tasks WHERE id > 5 ORDER BY id DESC;
due(non-unique)
- cursor
-
task:5:created_at:123456
(1234.. is unix timestamp)
-
- SQL
SELECT * FROM tasks WHERE (UNIX_TIMESTAMP(due) < 123456) OR (UNIX_TIMESTAMP(due) = 123456 AND id < 5) ORDER BY due IS NULL ASC, id ASC;
Now you can implement resolver.
func (r *queryResolver) Tasks(ctx context.Context, input model.TasksInput, orderBy model.TaskOrderFields, page model.PaginationInput) (*model.TaskConnection, error) {
db := config.DB()
if input.Completed != nil {
db = db.Where("completed = ?", *input.Completed)
}
var err error
switch orderBy {
case model.TaskOrderFieldsLatest:
db, err = pageDB(db, "id", desc, page)
if err != nil {
return &model.TaskConnection{PageInfo: &model.PageInfo{}}, err
}
var tasks []*model.Task
if err := db.Find(&tasks).Error; err != nil {
return &model.TaskConnection{PageInfo: &model.PageInfo{}}, err
}
return convertToConnection(tasks, orderBy, page), nil
case model.TaskOrderFieldsDue:
db, err = pageDB(db, "UNIX_TIMESTAMP(due)", asc, page)
if err != nil {
return &model.TaskConnection{PageInfo: &model.PageInfo{}}, err
}
var tasks []*model.Task
if err := db.Find(&tasks).Error; err != nil {
return &model.TaskConnection{PageInfo: &model.PageInfo{}}, err
}
return convertToConnection(tasks, orderBy, page), nil
default:
return &model.TaskConnection{PageInfo: &model.PageInfo{}}, errors.New("invalid order by")
}
}
package resolver
import (
"app/model"
"encoding/base64"
"errors"
"fmt"
"strconv"
"strings"
"github.com/jinzhu/gorm"
)
type direction string
var (
// unused: asc direction = "asc"
desc direction = "desc"
)
func pageDB(db *gorm.DB, col string, dir direction, page model.PaginationInput) (*gorm.DB, error) {
var limit int
if page.First == nil {
limit = 11
} else {
limit = *page.First + 1
}
if page.After != nil {
resource1, resource2, err := decodeCursor(*page.After)
if err != nil {
return db, err
}
if resource2 != nil {
switch dir {
case asc:
db = db.Where(
fmt.Sprintf("(%s > ?) OR (%s = ? AND id > ?)", col, col),
resource1.ID,
resource1.ID, resource2.ID,
)
case desc:
db = db.Where(
fmt.Sprintf("(%s < ?) OR (%s = ? AND id < ?)", col, col),
resource1.ID,
resource1.ID, resource2.ID,
)
}
} else {
switch dir {
case asc:
db = db.Where(fmt.Sprintf("%s > ?", col), resource1.ID)
case desc:
db = db.Where(fmt.Sprintf("%s < ?", col), resource1.ID)
}
}
}
switch dir {
case asc:
db = db.Order(fmt.Sprintf("%s IS NULL ASC, id ASC", col))
case desc:
db = db.Order(fmt.Sprintf("%s DESC, id DESC", col))
}
return db.Limit(limit), nil
}
type cursorResource struct {
Name string
ID int
}
func createCursor(first cursorResource, second *cursorResource) string {
var cursor []byte
if second != nil {
cursor = []byte(fmt.Sprintf("%s:%d:%s:%d", first.Name, first.ID, second.Name, second.ID))
} else {
cursor = []byte(fmt.Sprintf("%s:%d", first.Name, first.ID))
}
return base64.StdEncoding.EncodeToString(cursor)
}
func decodeCursor(cursor string) (cursorResource, *cursorResource, error) {
bytes, err := base64.StdEncoding.DecodeString(cursor)
if err != nil {
return cursorResource{}, nil, err
}
vals := strings.Split(string(bytes), ":")
switch len(vals) {
case 2:
id, err := strconv.Atoi(vals[1])
if err != nil {
return cursorResource{}, nil, errors.New("invalid_cursor")
}
return cursorResource{Name: vals[0], ID: id}, nil, nil
case 4:
id, err := strconv.Atoi(vals[1])
if err != nil {
return cursorResource{}, nil, errors.New("invalid_cursor")
}
id2, err := strconv.Atoi(vals[3])
if err != nil {
return cursorResource{}, nil, errors.New("invalid_cursor")
}
return cursorResource{
Name: vals[0],
ID: id,
}, &cursorResource{
Name: vals[2],
ID: id2,
}, nil
default:
return cursorResource{}, nil, errors.New("invalid_cursor")
}
}
func convertToConnection(tasks []*model.Task, orderBy model.TaskOrderFields, page model.PaginationInput) *model.TaskConnection {
if len(tasks) == 0 {
return &model.TaskConnection{PageInfo: &model.PageInfo{}}
}
pageInfo := model.PageInfo{}
if page.First != nil {
if len(tasks) >= *page.First+1 {
pageInfo.HasNextPage = true
tasks = tasks[:len(tasks)-1]
}
}
switch orderBy {
case model.TaskOrderFieldsLatest:
taskEdges := make([]*model.TaskEdge, len(tasks))
for i, task := range tasks {
cursor := createCursor(
cursorResource{Name: "task", ID: task.ID},
nil,
)
taskEdges[i] = &model.TaskEdge{
Cursor: cursor,
Node: task,
}
}
pageInfo.EndCursor = taskEdges[len(taskEdges)-1].Cursor
return &model.TaskConnection{PageInfo: &pageInfo, Edges: taskEdges}
case model.TaskOrderFieldsDue:
taskEdges := make([]*model.TaskEdge, 0, len(tasks))
for _, task := range tasks {
if task.Due == nil {
pageInfo.HasNextPage = false
return &model.TaskConnection{PageInfo: &pageInfo, Edges: taskEdges}
}
cursor := createCursor(
cursorResource{Name: "task", ID: int(task.Due.Unix())},
&cursorResource{Name: "due", ID: task.ID},
)
taskEdges = append(taskEdges, &model.TaskEdge{
Cursor: cursor,
Node: task,
})
}
pageInfo.EndCursor = taskEdges[len(taskEdges)-1].Cursor
return &model.TaskConnection{PageInfo: &pageInfo, Edges: taskEdges}
}
return &model.TaskConnection{PageInfo: &model.PageInfo{}}
}
5. Check behaviors
tasks
query implementation is done. Check behaviors by using clients like GraphiQL or graphql-playground.
order by latest
order by due
You finally finished to implement GraphQL server. I hope you enjoyed schema first development on GraphQL. Implementing pagination is a little bit difficult at the last part. Code generation is one of the good solution.
Top comments (2)
Very useful, thanks
This is exactly the post I've been looking for! Thank you for your thoroughness!