The Point

Marshal converts an in-memory data structure into a transmittable format (like JSON). Unmarshal does the reverse – it turns JSON back into a data structure. Most APIs use JSON because it is human-readable, natively supported by JavaScript, and universally available across languages.

Explanation

Marshal / Unmarshal

 1type ShippingClass struct {
 2    Name string `json:"name"`
 3    Lang string `json:"lang"`
 4}
 5
 6// Unmarshal: JSON string → struct
 7jsonStr := `{"name": "標準配送", "lang": "zh"}`
 8var obj ShippingClass
 9json.Unmarshal([]byte(jsonStr), &obj)
10// obj.Name == "標準配送"
11
12// Marshal: struct → JSON string
13data, _ := json.Marshal(obj)
14// data == {"name":"標準配送","lang":"zh"}

This is not limited to JSON – serialization for XML, YAML, and protobuf is also called marshal/unmarshal. Same concept, different formats.

Why do most APIs use JSON?

• Human-readable: you can read it directly when debugging; binary formats like protobuf cannot do this
• Native JS support: browsers parse JSON at zero cost, no extra handling on the frontend
• Universal: almost every language has a mature JSON library
• Less verbose than XML: XML’s opening and closing tags add a lot of overhead

Knowledge Sugar

What are Go struct tags?

The json:"name" tag tells Go’s JSON library which key name this field maps to in JSON. Without a tag, it defaults to the field name (case-sensitive):

1type Example struct {
2    DisplayName string `json:"display_name"`  // JSON key is display_name
3    Age         int    `json:"age,omitempty"` // omitempty: omit this key when value is zero
4    Internal    string `json:"-"`             // never output to JSON
5}

Downsides of JSON

JSON is not without problems. In high-performance scenarios (internal microservice communication), protobuf is the usual choice: