go-jsonstruct

Generate Go structs from multiple JSON or YAML objects.

What does go-jsonstruct do and why should I use it?
How do I use go-jsonstruct?
YAML support
What are go-jsonstruct's key features?
How does go-jsonstruct work?
License

What does go-jsonstruct do and why should I use it?

go-jsonstruct generates Go structs from multiple JSON or YAML objects. Existing Go struct generators such as json-to-go and json2struct take only a single JSON object as input. go-jsonstruct takes multiple objects as input and generates the most specific Go struct possible into which all the input objects can be unmarshalled.

This is useful if you have a collection of JSON objects, where no single object has all properties present, and you want to unmarshal those JSON objects into a Go program. Example collections include:

JSON responses received from a REST API with no defined schema.
Multiple values from a JSON column in an SQL database.
All the JSON documents in a document database.
All the YAML configuration files in a directory.

How do I use go-jsonstruct?

Install go-jsonstruct:

go install github.com/twpayne/go-jsonstruct/v3/cmd/gojsonstruct@latest

Feed it some JSON objects. For example you can feed it with

{
  "age": 37,
  "user_height_m": 2
}

{
  "age": 38,
  "user_height_m": 1.7,
  "favoriteFoods": [
    "cake"
  ]
}

by running

echo '{"age":37,"user_height_m":2}' \
    '{"age":38,"user_height_m":1.7,"favoriteFoods":["cake"]}' \
    | gojsonstruct

This will output:

package main

type T struct {
    Age           int      `json:"age"`
    FavoriteFoods []string `json:"favoriteFoods,omitempty"`
    UserHeightM   float64  `json:"user_height_m"`
}

This example demonstrates:

age is always observed as an integer, and so is given type int. The lower-case JSON property age is converted into an exported Go field name Age for compatibility with encoding/json.
favoriteFoods is observed as a JSON array, but is not always present, but when it is present it only contains JSON strings, and so is given type []string. The camelCase name favoriteFoods is converted into the exported Go field name FavoriteFoods and is tagged with omitempty.
user_height_m is observed as JSON numbers 2 and 1.7, for which the most specific Go type is float64. The snake_case name user_height_m is converted to the exported Go field name UserHeightM.
Properties are sorted alphabetically.

go-jsonstruct recursively handles nested array elements and objects. For example, given the following three JSON objects input:

{
  "nested": {
    "bar": true,
    "foo": "baz"
  }
}

{
  "nested": {
    "bar": false,
    "foo": null
  }
}

{
  "nested": {
    "bar": true,
    "foo": ""
  }
}

which you can try by running

echo '{"nested":{"bar":true,"foo":"baz"}}' \
    '{"nested":{"bar":false,"foo":null}}' \
    '{"nested":{"bar":true,"foo":""}}' \
    | gojsonstruct --package-name mypackage --typename MyType

generates the output

package mypackage

type MyType struct {
    Nested struct {
        Bar bool    `json:"bar"`
        Foo *string `json:"foo"`
    } `json:"nested"`
}

This demonstrates:

The package name and type name can be set on the command line.
The arbitrarily-complex property nested is recursively converted to a nested struct that all values can be unmarshalled to. go-jsonstruct handles array elements in an identical fashion, resolving array elements to the most-specific type.
nested.bar is always observed as a JSON bool, and is converted to a field of type bool.
nested.foo is observed as a JSON string, a JSON null, and an empty JSON string and is converted to a field of type *string without omitempty. With omitempty, Go's encoding/json omits the field in the two cases of nil and a pointer to an empty string, so there is no way to distinguish between the observed values of null and "". go-jsonstruct falls back to the option of *string without omitempty which means that a value is always present, even if empty.

You can feed it your own data via the standard input, for example if you have a file with one JSON object per line in objects.json you can run:

gojsonstruct < objects.json

To learn about more about the available options, run:

gojsonstruct --help

YAML support

For YAML files, pass the --format=yaml flag, for example:

$ gojsonstruct --format=yaml *.yaml

gojsonstruct will analyze all passed YAML files and generate a Go struct with yaml:"..." struct tags.

What are go-jsonstruct's key features?

Finds the most specific Go type that can represent all input values.
Handles JSON and YAML.
Generates Go struct field names from camelCase, kebab-case, and snake_case object property names.
Capitalizes common abbreviations (e.g. HTTP, ID, and URL) when generating Go struct field names to follow Go conventions, with the option to add your own abbreviations.
Gives you control over the output, including the generated package name, type name, and godoc-compatible comments.
Generates deterministic output based only on the determined structure of the input, making it suitable for incorporation into build pipelines or detecting schema changes.
Generates ,omitempty tags.
Generates ,string tags.
Uses the standard library's time.Time when possible.
Gracefully handles properties with spaces that cannot be unmarshalled by encoding/json.

How does go-jsonstruct work?

go-jsonstruct consists of two phases: observation and code generation.

Firstly, in the observation phase, go-jsonstruct explores all the input objects and records statistics on what types are observed in each part. It recurses into objects and iterates over arrays.

Secondly, in the code generation phase, go-jsonstruct inspects the gathered statistics and determines the strictest possible Go type that can represent all the observed values. For example, the values 0 and 1 can be represented as an int, the values 0, 1, and 2.2 require a float64, and true, 3.3, and "name" require an any.

License

BSD

twpayne / go-jsonstruct