Featured image of post Use Go to unmarshal JSON null, set, and missing fields
Software Development

Use Go to unmarshal JSON null, set, and missing fields

How to detect null, set, or missing JSON fields when unmarshalling into a Go struct

JSON unmarshalling use cases

When passing a JSON payload to a Go application, you may encounter situations where you must tell the difference between set, missing, or null fields.

For example, consider the following JSON payload:

{
  "name": "Alice",
  "age": 30,
  "address": {
    "street": "123 Main St",
    "city": "Springfield"
  }
}

We can unmarshal this JSON payload using JSON tags and the following Go structs:

type Person struct {
    Name    string `json:"name"`
    Age     int    `json:"age"`
    Address Address `json:"address"`
}
type Address struct {
    Street string `json:"street"`
    City   string `json:"city"`
}

However, we will not be able to tell the difference between these two JSON payloads:

  • { "name": null }
  • { "name": "" }

Go’s zero values are not distinguishable from missing fields when unmarshalling JSON.

We can change the above struct to use pointers to identify null fields:

type Person struct {
    Name    *string `json:"name"`
    Age     *int    `json:"age"`
    Address *Address `json:"address"`
}

However, we will still not be able to tell the difference between these two JSON payloads:

  • { "name": null }
  • { }

Both of these payloads will unmarshal into a Person struct with all fields set to nil, and we cannot distinguish between a missing field and a field set to null.

One reason to distinguish between missing and null fields is to avoid overwriting existing values with null values. For example, when name is not specified in the JSON payload, we may want to keep the existing name value in the Person struct. But we may want to clear the name when name is defined as null.

Detecting null, set, and missing JSON fields with Go

We can use custom unmarshalling logic by implementing the Unmarshaler interface to detect the difference between null and missing fields. The UnmarshalJSON method allows us to inspect the JSON token stream and decide how to unmarshal the JSON payload. The critical insight is that UnmarshalJSON is only called when the field is present in the JSON payload. So, we can mark a Set flag as true when the field is present and false when it is not.

Here is an example implementation:

type Any[T any] struct {
    Set   bool
    Valid bool
    Value T
}

// MarshalJSON implements the json.Marshaler interface.
// Only Value is marshaled, and only if Valid is true.
func (s Any[T]) MarshalJSON() ([]byte, error) {
    if !s.Valid {
       return []byte("null"), nil
    }
    return json.Marshal(s.Value)
}

// UnmarshalJSON implements the json.Unmarshaler interface.
// Set is always set to true, even if the JSON data was set to null.
// Valid is set if the JSON data is not set to null.
func (s *Any[T]) UnmarshalJSON(data []byte) error {
    s.Set = true
    s.Valid = false

    if bytes.Equal(data, []byte("null")) {
       // The key was set to null, set value to zero/default value
       var zero T
       s.Value = zero
       return nil
    }

    // The key isn't set to null
    var v T
    if err := json.Unmarshal(data, &v); err != nil {
       return err
    }
    s.Value = v
    s.Valid = true
    return nil
}

We used a generic type T to allow Any to work with any type. The Valid flag distinguishes between nil and non-nil values. The Set flag is set to true only when the field is present in the JSON payload.

Here is how we can use the Any type in a Person struct:

type Person struct {
    Name    Any[string] `json:"name"`
    Age     Any[int]    `json:"age"`
    Address Any[Address] `json:"address"`
}

Testing the custom unmarshalling logic

The following example demonstrates how the Any type works:

type Form struct {
    Name string `json:"name"`
    Age  int    `json:"age"`
}

type Config struct {
    Form   Any[Form] `json:"form"`
    Member Any[bool] `json:"member"`
}

func main() {
    tests := []struct {
       description string
       JSON        string
    }{
       {
          "Nothing set",
          `{}`,
       },
       {
          "Set all fields",
          `{"form": {"name": "John", "age": 30}, "member": false}`,
       },
       {
          "Set only member field, and leave form fields unchanged",
          `{"member": true}`,
       },
       {
          "Set only the form field, and leave the member field unchanged",
          `{"form": {"name": "Jane", "age": 25}}`,
       },
       {
          "Leave all fields unchanged",
          `{}`,
       },
       {
          "Clear all fields",
          `{"form": null, "member": null}`,
       },
       {
          "Set only form field",
          `{"form": {"name": "Chris", "age": 35}}`,
       },
    }

    c := Config{}
    for _, test := range tests {
       fmt.Printf("\nTest: %s\n", test.description)
       _ = json.Unmarshal([]byte(test.JSON), &c)
       fmt.Printf("Input: %s\n", test.JSON)
       fmt.Printf("%+v\n", c)
       data, _ := json.Marshal(c)
       fmt.Printf("Output: %s\n", data)
    }

}

The test output will show how the Any type behaves when unmarshalling JSON payloads with different fields set, missing, or set to null.

Test: Nothing set
Input: {}
{Form:{Set:false Valid:false Value:{Name: Age:0}} Member:{Set:false Valid:false Value:false}}
Output: {"form":null,"member":null}

Test: Set all fields
Input: {"form": {"name": "John", "age": 30}, "member": false}
{Form:{Set:true Valid:true Value:{Name:John Age:30}} Member:{Set:true Valid:true Value:false}}
Output: {"form":{"name":"John","age":30},"member":false}

Test: Set only member field, and leave form fields unchanged
Input: {"member": true}
{Form:{Set:true Valid:true Value:{Name:John Age:30}} Member:{Set:true Valid:true Value:true}}
Output: {"form":{"name":"John","age":30},"member":true}

Test: Set only the form field, and leave the member field unchanged
Input: {"form": {"name": "Jane", "age": 25}}
{Form:{Set:true Valid:true Value:{Name:Jane Age:25}} Member:{Set:true Valid:true Value:true}}
Output: {"form":{"name":"Jane","age":25},"member":true}

Test: Leave all fields unchanged
Input: {}
{Form:{Set:true Valid:true Value:{Name:Jane Age:25}} Member:{Set:true Valid:true Value:true}}
Output: {"form":{"name":"Jane","age":25},"member":true}

Test: Clear all fields
Input: {"form": null, "member": null}
{Form:{Set:true Valid:false Value:{Name: Age:0}} Member:{Set:true Valid:false Value:false}}
Output: {"form":null,"member":null}

Test: Set only form field
Input: {"form": {"name": "Chris", "age": 35}}
{Form:{Set:true Valid:true Value:{Name:Chris Age:35}} Member:{Set:true Valid:false Value:false}}
Output: {"form":{"name":"Chris","age":35},"member":null}

Complete code on Go Playground

The complete Go code for unmarshalling JSON null, set, and missing fields is available on the Go Playground.

Further reading

Recently, we published an article on how to optimize the performance of a Go application. We benchmarked JSON decoding vs gob decoding in that article.

In addition, we wrote about how to read program arguments from STDIN with Go, which is more secure than using environment variables or command-line arguments.

Watch how to use Go to unmarshal JSON null, set, and missing fields accurately

Note: If you want to comment on this article, please do so on the YouTube video.

© 2024 Victor on Software