Consider this simple example:
package main
import (
"encoding/json"
"fmt"
"log"
"reflect"
)
var args = `[1, 2.5, "aaa", true, false]`
func main() {
var x []interface{}
err := json.Unmarshal([]byte(args), &x)
if err != nil {
log.Fatalf("%s", err.Error())
panic(fmt.Sprintf("%s", err.Error()))
}
for _, arg := range x {
t := reflect.TypeOf(arg).Kind().String()
v := reflect.ValueOf(arg)
if t == "int64" {
fmt.Printf("int64 %v
", v.Int())
}
if t == "float64" {
fmt.Printf("float64 %v
", v.Float())
}
if t == "string" {
fmt.Printf("string %v
", v.String())
}
if t == "bool" {
fmt.Printf("bool %v
", v.Bool())
}
}
}
The program outputs:
float64 1
float64 2.5
string aaa
bool true
bool false
As you can see, my input is a valid JSON which represents an array with five items:
- integer
- floating point number
- string
- boolean
- boolean
When I unmarshal the valid JSON string into []interface{} and try to check the types with reflection, the integer value from JSON has a type of float64. Any idea why? Is this expected behaviour?
This is documented behavior of Unmarshal
. All numbers are unmarshaled into float64.
To unmarshal JSON into an interface value, Unmarshal stores one of these in the interface value:
- bool, for JSON booleans
- float64, for JSON numbers
- string, for JSON strings
- []interface{}, for JSON arrays
- map[string]interface{}, for JSON objects
- nil for JSON null
This is because JSON does not have integers, every number in JSON is defined to be a 64 bit floating point.
This is the default behaviour of the JSON decoder. You can change it to output json.Number
instead by using the UseNumber
method.