Beego: Building Web API with Auto Generated API Document Support

Introduction

We released Beego 1.3 last week with a very cool feature: Auto Generated API Document. This article will show you how to use this new feature.

Why do we need Auto Generated API Document?

Now we need to build a API for the mobile app. We designed the API structures and wrote the API document prototype based on the requirements. We shared the document with the mobile developers to let them know how to use and test our API. However while developing the API structures will keep changing and we need to keep the API document updated. It's a pain for us. Many times after changing the API we forgot to update the API Document, it's also a pain for the user of the API. Another thing is how can we make the API be easy tested. It will be very good to have a interface where the user can test the API together with the document as well.

To walk through from this dilemma we found Swagger. You can see a live demo here.

Swagger™ is a specification and complete framework implementation for describing, producing, consuming, and visualizing RESTful web services.

With the auto generated document by integrating Swagger into beego we can have these benifits.

1. Standard declarative comments for API functions which makes
2. The API code becomes more maintainable with the comments
3. The Swagger style API document will be generated automacally based on the comments.

Creating the Beego API aplication

go get -u github.com/beego/bee
go get -u github.com/astaxie/beego

Go to GOPATH/src directory and execute command bee api bapi to create the API application. Then go to the application directory cd bapi and run bee run -downdoc=true -docgen=true.

This displays output similar to the following:

Loading up http://127.0.0.1:8080/swagger/swagger-1/ in the browser shows the following output:

You can only use 127.0.0.1 here other than loalhost to prevent the Ajax CORS issue.

Bang! Our live testable API document is ready to use. Now let's look into the details of the project we just created.

Directory structure

You'll see the following directory structure created by bee api

|-- bapi
|-- conf
|   `-- app.conf
|-- controllers
|   |-- object.go
|   `-- user.go
|-- docs
|   |-- doc.go
|   `-- docs.go
|-- lastupdate.tmp
|-- main.go
|-- models
|   |-- object.go
|   `-- user.go
|-- routers
|   |-- commentsRouter.go
|   `-- router.go
|-- swagger
`-- tests
    `-- default_test.go

• main.go Our bootstrap file.
• bapi The compiled binary file.
• conf The directory for configuration file.
• controllers Our controllers
• models Our models
• docs Our auto generated document
• lastupdate.tmp Cache file for Annotation Router
• routers Routers
• swagger A static directory for Swagger API related files.
• test The testcase for the application. You can run the testcases without run the application.

The bootstrap main file

The following is the main.go file:

package main

import (
    _ "bapi/docs"
    _ "bapi/routers"

    "github.com/astaxie/beego"
)

func main() {
    if beego.RunMode == "dev" {
        beego.DirectoryIndex = true
        beego.StaticDir["/swagger"] = "swagger"
    }
    beego.Run()
}

This is just a normal Beego bootstrap file with several lines of code to add swagger to static directory.

Next let's see the router structure of our API application.

Namespace Router

The router is using the namespace router. Two points two notice:

1. The API with auto generated document only support namespace router now.
2. It only support NSNamespace and NSInclude within two levels.

The following is our routers:

func init() {
    ns := beego.NewNamespace("/v1",
        beego.NSNamespace("/object",
            beego.NSInclude(
                &controllers.ObjectController{},
            ),
        ),
        beego.NSNamespace("/user",
            beego.NSInclude(
                &controllers.UserController{},
            ),
        ),
    )
    beego.AddNamespace(ns)
}

It created a ns variable with beego.NewNamespace. Actually ns variable is keeping a rouer tree and we can add this tree under any other router tree. This is a benifit of namespace that you can design the namespace in your module and add it to other modules with any prefix.

Let's take a look at NewNamespace function. Here is the definition:

NewNamespace(prefix string, params ...innnerNamespace) *Namespace

the first parameter is the url prefix, the second parameter is the variable argument list with type innnerNamespace.

The following is the definition of innnerNamespace:

type innnerNamespace func(*Namespace)

It's a function that accept parameter *Namespace. In beego all the functions list below support return this function type innnerNamespace:

• NSCond(cond namespaceCond) innnerNamespace
• NSBefore(filiterList …FilterFunc) innnerNamespace
• NSAfter(filiterList …FilterFunc) innnerNamespace
• NSInclude(cList …ControllerInterface) innnerNamespace
• NSRouter(rootpath string, c ControllerInterface, mappingMethods …string) innnerNamespace
• NSGet(rootpath string, f FilterFunc) innnerNamespace
• NSPost(rootpath string, f FilterFunc) innnerNamespace
• NSDelete(rootpath string, f FilterFunc) innnerNamespace
• NSPut(rootpath string, f FilterFunc) innnerNamespace
• NSHead(rootpath string, f FilterFunc) innnerNamespace
• NSOptions(rootpath string, f FilterFunc) innnerNamespace
• NSPatch(rootpath string, f FilterFunc) innnerNamespace
• NSAny(rootpath string, f FilterFunc) innnerNamespace
• NSHandler(rootpath string, h http.Handler) innnerNamespace
• NSAutoRouter(c ControllerInterface) innnerNamespace
• NSAutoPrefix(prefix string, c ControllerInterface) innnerNamespace
• NSNamespace(prefix string, params …innnerNamespace) innnerNamespace

So we can use any function list above for the parameter list of NewNamespace.

Now let's take a look at our namespace router ns. It's a multi-level nested function. The first parameter is /v1 which is the prefix of the router tree. The second and the third parameters are beego.NSNamespace. In the second level of beego.NSNamespace, we use beego.NSInclude to inject the Annotation Router. beego.NSInclude is designed for Annotation Router, we don't have any router settings here but only set up the controller. So how will this work? Let's go deeper into the Annotation Router.

Annotation Router

Annotation Router are some comments set for Controller that Beego will parse them into the real routers automatically. Let's see what do we do to for it in ObjectController and UserController:

// Operations about object
type ObjectController struct {
    beego.Controller
}

// @Title create
// @Description create object
// @Param   body        body    models.Object   true        "The object content"
// @Success 200 {string} models.Object.Id
// @Failure 403 body is empty
// @router / [post]
func (this *ObjectController) Post() {
    var ob models.Object
    json.Unmarshal(this.Ctx.Input.RequestBody, &ob)
    objectid := models.AddOne(ob)
    this.Data["json"] = map[string]string{"ObjectId": objectid}
    this.ServeJson()
}

We can see there are bunch of comments above the functions. The last line is for the Annotation Router such as //@router / [post] which means routing post method for / to this function.

It has the exactlly same functionality as the router we were using befoer beego.Router("/", &ObjectController{},"post:Post") but in this case Beego will help you to register the router. But how does Beego do that? While the application starting, if the application is using NSInclude and the RunMode is dev, Beego will check if the router is already generated and if there is update, if true, Beego will use AST in Go to analyze the application's source code of the controller that NSInclude called and generate the routers and write to file routers/commentsRouter.go.

The router analysis will only happens in dev mode. So if you updated your code related to Annotation Router, you need to run dev mode at least once to generate the routers.

Annotation Router starts with // @router and accept two parameters: 1. The router URI 2. The supported methods. It need to be put above the router function. The order of the comments doesn't matter.

The first parameter router URI supports all the URI rules. Such as /object/:key, /object and /cms_:id([0-9]+).html.

The second parameter supported methods. Multiple methods separated by , such as [post,get]. But the auto generated API document only supports one method which means if there are multiple methods in the list, the auto generated API document will only take the first method. We don't recommend you to set multiple methods for one function.

By having these basic understanding of Annotation Router let's start the main topic today: Auto Generated API Document.

Auto Generated API Document

Auto Generated API Document is the elegant swagger document generated by Beego based on the annotation comments automatically. The benifit of this approach is you only write the comments for the functions once and you get the live testable API document. They will always keep synchronized. It makes the source code maintainable as well as makes the API document testable. Two birds in one shot.

So how can we generate the API Document?

Here is how can we implement this approach:

1. Swagger's schema is a bunch of JSON data.
2. We write some annotations for the router functions.
3. Beego generate Swagger's schema based on the annotations.

The first step is describing the API:

API Document

We mentioned it before in router.go there are bunch of annotaions. It's the description of this API application:

// @APIVersion 1.0.0
// @Title beego Test API
// @Description beego has a very cool tools to autogenerate documents for your API
// @Contact astaxie@gmail.com
// @TermsOfServiceUrl http://beego.me/
// @License Apache 2.0
// @LicenseUrl http://www.apache.org/licenses/LICENSE-2.0.html

Following are the flags we can use:

• @APIVersion
• @Title
• @Description
• @Contact
• @TermsOfServiceUrl
• @License
• @LicenseUrl

All of them are optional followed by a string. The following is the doc we generated: http://127.0.0.1:8080/docs:

{
  "apiVersion": "1.0.0",
  "swaggerVersion": "1.2",
  "apis": [
    {
      "path": "/object",
      "description": "Operations about object\n"
    },
    {
      "path": "/user",
      "description": "Operations about Users\n"
    }
  ],
  "info": {
    "title": "beego Test API",
    "description": "beego has a very cool tools to autogenerate documents for your API",
    "contact": "astaxie@gmail.com",
    "termsOfServiceUrl": "http://beego.me/",
    "license": "Url http://www.apache.org/licenses/LICENSE-2.0.html"
  }
}

This is the information fetched by the first request. How do we generate the apis object? Beego fetched it from your namespace source code. Right now it only supports two and only two level nested namespace. The first level is the baseUrl. The path comes from the second level namespace's prefix. So where does the description come from?

Annotation for Controller

We add the annotation for controller and it's methods to describe the controller.

// Operations about object
type ObjectController struct {
    beego.Controller
}

The comment above describes the controller. The comment following for the methods will describe the router, parameters and the return value.

// @Title Get
// @Description find object by objectid
// @Param   objectId        path    string  true        "the objectid you want to get"
// @Success 200 {object} models.Object
// @Failure 403 :objectId is empty
// @router /:objectId [get]
func (this *ObjectController) Get() {
}

Let's take a look at these annotations:

• @Title

  Title of the interface. Unique and optional.

  Format: followed by a string

• @Description

  Description of the interface. Unique and optional.

  Format: followed by a string

• @Param

  The request parameters for the interface. Multiple and optional.

  Format: parameter name parameter sending type parameter data type required description comment

  parameter name is a string

  parameter sending type can be any one of the following values:

  • query means the parameter in url send by GET. such as ?aa=bb&cc=dd
  • form means the parameter send by POST.
  • path means the parameter in the url path, such as for /user/{uid} uid is a parameter with path type.
  • body means the raw dada send from request body.
  • header means the parameter in request header

  parameter data type can be any one of the following values:

  • string
  • int
  • int64
  • object path The object path is related to the project path. For example models.Object means Object object under models directory.

  description comment is a string

  required is true or false

• @Success

  The success message returned to client.

  Format: status code return type return message or return object path

  • status code is the standard HTTP status code, 200 201 etc
  • return type {object} is the object and will get the object by return object path. Other value will return return message.
  • return object path The object path is the relative path to the project path. For example models.Object means Object object under models directory.

• @Failure

  The failure message returned to client.

  Format: status code error message

• @router

  The Annotation Router we discussed before.

Every format above, parameters can be separated by on or more '\t', '\n', '\v', '\f', '\r', ' ', U+0085 (NEL), U+00A0 (NBSP)

Following code is the JSON data we generated from these annotations.

{
  "path": "/object/{objectId}",
  "description": "",
  "operations": [
    {
      "httpMethod": "GET",
      "nickname": "Get",
      "type": "",
      "summary": "find object by objectid",
      "parameters": [
        {
          "paramType": "path",
          "name": "objectId",
          "description": "\"the objectid you want to get\"",
          "dataType": "string",
          "type": "",
          "format": "",
          "allowMultiple": false,
          "required": true,
          "minimum": 0,
          "maximum": 0
        }
      ],
      "responseMessages": [
        {
          "code": 200,
          "message": "models.Object",
          "responseModel": "Object"
        },
        {
          "code": 403,
          "message": ":objectId is empty",
          "responseModel": ""
        }
      ]
    }
  ]
}

Customized Object Annotation

Following is the definition of our object:

type Object struct {
    ObjectId   string
    Score      int64
    PlayerName string
}

The auto generated object description for document is:

Object {
ObjectId (string, optional): ,
PlayerName (string, optional): ,
Score (int64, optional):
}

We can see all the fields are optional, we can change it by adding tag as showing below:

type Object struct {
    ObjectId   string   `required:"true" description:"object id"`
    Score      int64        `required:"true" description:"players's scores"`
    PlayerName string   `required:"true" description:"plaers name, used in system"`
}

To change the returned fields name, you can add tag with name json or thrife as showing below:

type Object struct {
    ObjectId   string   `json:"object_id"`
    Score      int64        `json:"player_score"`
    PlayerName string   `json:"player_name"`
}

The auto generated object description for document is:

Object {
object_id (string, optional): ,
player_score (string, optional): ,
player_name (int64, optional):
}

FAQ

1. Q: bee doesn't have the command we ran befoer.

   A: Please upgrade bee to latest version.

2. Q: The document is generated, but I can't test the API.

   A: It might because of the CORS issue. please make sure the url and port you request is same as the application.

3. Q: Some other exceptions.

   A: Please report the issue to us. We will try to fix it ASAP.