检查字符串是否为JSON格式

Question

How to check if a given string is in form of multiple json string separated by spaces/newline?

For example,
given: "test" 123 {"Name": "mike"} (3 json concatenated with space)
return: true, since each of item ("test" 123 and {"Name": "mike"}) is a valid json.

In Go, I can write a O(N^2) function like:

// check given string is json or multiple json concatenated with space/newline
func validateJSON(str string) error {
    // only one json string
    if isJSON(str) {
        return nil
    }
    // multiple json string concatenate with spaces
    str = strings.TrimSpace(str)
    arr := []rune(str)
    start := 0
    end := 0
    for start < len(str) {
        for end < len(str) && !unicode.IsSpace(arr[end]) {
            end++
        }
        substr := str[start:end]
        if isJSON(substr) {
            for end < len(str) && unicode.IsSpace(arr[end]) {
                end++
            }
            start = end
        } else {
            if end == len(str) {
                return errors.New("error when parsing input: " + substr)
            }
            for end < len(str) && unicode.IsSpace(arr[end]) {
                end++
            }
        }
    }
    return nil
}

func isJSON(str string) bool {
    var js json.RawMessage
    return json.Unmarshal([]byte(str), &js) == nil
}

But this won't work for large input.

Answer 1

There are two options. The simplest, from a coding standpoint, is going to be just to decode the JSON string normally. You can make this most efficient by decoding to an empty struct:

package main

import "encoding/json"

func main() {
    input := []byte(`{"a":"b", "c": 123}`)
    var x struct{}
    if err := json.Unmarshal(input, &x); err != nil {
        panic(err)
    }

    input = []byte(`{"a":"b", "c": 123}xxx`) // This one fails
    if err := json.Unmarshal(input, &x); err != nil {
        panic(err)
    }
}

(playground link)

This method has a few potential drawbacks:

• It only works with a single JSON object. That is, a list of objects (as requested in the question) will fail, without additional logic.
• As pointed out by @icza in comments, it only works with JSON objects, so bare arrays, numbers, or strings will fail. To accomodate these types, interface{} must be used, which introduces the potential for some serious performance penalties.
• The throw-away x value must still be allocated, and at least one reflection call is likely under the sheets, which may introduce a noticeable performance penalty for some workloads.

Given these limitations, my recommendation is to use the second option: loop through the entire JSON input, ignoring the actual contents. This is made simple with the standard library json.Decoder:

package main

import (
    "bytes"
    "encoding/json"
    "io"
)

func main() {
        input := []byte(`{"a":"b", "c": 123}`)
        dec := json.NewDecoder(bytes.NewReader(input))
        for {
            _, err := dec.Token()
            if err == io.EOF {
                break // End of input, valid JSON
            }
            if err != nil {
                panic(err) // Invalid input
            }
        }

        input = []byte(`{"a":"b", "c": 123}xxx`) // This input fails
        dec = json.NewDecoder(bytes.NewReader(input))
        for {
            _, err := dec.Token()
            if err == io.EOF {
                break // End of input, valid JSON
            }
            if err != nil {
                panic(err) // Invalid input
            }
        }
}

(playground link)