Go界面方法的用法

Question

I have written a very simple Go app using the Martini library. When putting together a simple REST API example I wanted to render JSON data back to the user when an error occurred like:

{
    error: "Document Not Found",
    code: 404
}

I'm using the following code to return a map that Martini can render into JSON. package ErrorResponces

import "net/http"

type ErrJson interface {

    RenderErr(v int)

}

func RenderErr(v int) map [string]interface{} {

    var returnMap = map[string]interface{} {
        "error": http.StatusText(v),
        "code": v,
    }

    return returnMap

}

Later in my controller code I try and test this method using

fmt.Println(ErrJson.RenderErr(400))

However I get the following error:

controllers\FoodController.go:25: cannot use 400 (type int) as type ErrorResponces.ErrJson in function argument: int does not implement ErrorResponces.ErrJson (missing RenderErr method)

controllers\FoodController.go:25: not enough arguments in call to ErrorResponces.ErrJson.RenderErr controllers\FoodController.go:25: ErrorResponces.ErrJson.RenderErr(400) used as value

I'm having a hard time figuring out exactly what this error is talking about.

Answer 1

It seems that you're trying to called a function directly on an interface type instead of an object which implements that interface.

Here's a simple example that returns JSON data:

package main

import (
  "encoding/json"
  "github.com/codegangsta/martini"
  "net/http"
)

func Encode(v ...interface{}) (string, error) {
  var data interface{} = v
  if v == nil {
    // So that empty results produces `[]` and not `null`
    data = []interface{}{}
  } else if len(v) == 1 {
    data = v[0]
  }
  b, err := json.Marshal(data)
  return string(b), err
}

func RenderErr(v int) map[string]interface{} {
  var returnMap = map[string]interface{}{
    "error": http.StatusText(v),
    "code":  v,
  }

  return returnMap
}

func main() {
  m := martini.Classic()
  m.Get("/", func(c martini.Context, w http.ResponseWriter, r *http.Request) string {
    w.Header().Set("Content-Type", "application/json")

    str, err := Encode(RenderErr(400))
    if err != nil {
      panic(err)
    }

    return str
  })
  m.Run()
}

If you wanted to use your interface idea, you could do something like this (I tried to copy what you were basically doing):

package main

import (
  "encoding/json"
  "github.com/codegangsta/martini"
  "net/http"
)

func Encode(v ...interface{}) (string, error) {
  var data interface{} = v
  if v == nil {
    // So that empty results produces `[]` and not `null`
    data = []interface{}{}
  } else if len(v) == 1 {
    data = v[0]
  }
  b, err := json.Marshal(data)
  return string(b), err
}

type ErrJson interface {
  RenderErr() string
}

type ErrJsonCode int

func (e ErrJsonCode) RenderErr() string {
  var returnMap = map[string]interface{}{
    "error": http.StatusText(int(e)),
    "code":  int(e),
  }

  str, err := Encode(returnMap)
  if err != nil {
    panic(err)
  }

  return str
}

func main() {
  m := martini.Classic()
  m.Get("/", func(c martini.Context, w http.ResponseWriter, r *http.Request) string {
    w.Header().Set("Content-Type", "application/json")
    return ErrJsonCode(400).RenderErr()
  })
  m.Run()
}

I'm not sure I would design it like this though. I would probably make it more generic and have it support multiple content types, and not have an error tied to its content type. Here is a decent article about building a restful API with Martini: http://0value.com/build-a-restful-API-with-Martini (it uses some advanced concepts).