protoc-gen-jsonschema

protoc-gen-jsonschema is a plugin that converts protocol buffer files into JSON Schema. While it primarily focuses on the protobuf standard, it also supports certain non-standard specifications, such as those used by Kubernetes, Node.js, and ArgoCD. Instead of aiming to cover the entire JSON Schema specification, the plugin is specifically designed to translate ProtoJSON specifications. It supports JSON Schema versions draft-04, draft-06, draft-07, draft-2019-09, and draft-2020-12. as well as support protocol buffer syntax proto2, syntax proto3, and edition 2023.

If you’d like to support other specification, contributions are always welcome! Feel free to submit a PR.

Installation

If you have go runtime, you can go install it.

go install github.com/pubg/protoc-gen-jsonschema

Alternatively, you can download pre-built binary from Github Release.

Usage

Refer to the Plugin Options section below for various options available for this plugin.

I'm not sure which options to use

This plugin provides default options that are ready to use. For testing or generating a basic json-schema file, the following command is sufficient without extra options.

protoc --jsonschema_out=. *.proto

Generate with yaml format

protoc --jsonschema_out=. --jsonschema_opt=output_file_suffix=.yaml *.proto

Shrink bytes for transfer over network

protoc --jsonschema_out=. --jsonschema_opt=pretty_json_output=false *.proto

I'd like to comply with the protobuf JSON mapping standard

By default, this plugin does not comply to the Protobuf standard because most plugins and other JSON libraries do not address integers larger than a 53-bit value. To ensure greater compatibility with other libraries, this plugin converts int64 values to integers instead of strings. However, to comply with the Protobuf standard, int64 values should be converted to strings. Below options will assist you.

protoc --jsonschema_out=. --jsonschema_opt=int64_as_string=true *.proto

I'm not satisfied with the plugin's options. I want to customize every fields

This plugin offers options for fields, messages, and enums. You can utilize these options in the jsonschema.proto file within your proto.

cp jsonschema.proto examples/jsonschema.proto
protoc --jsonschema_out=. --proto_path=examples examples/jsonschema.proto

Options

Plugin Options

entrypoint_message

entrypoint_message is used which message should be entrypoint object of schema.

default: null or empty
example:
    - --jsonschema_opt=entrypoint_message=MyMessage

output_file_suffix

output_file_suffix is used to determine output file name suffix.
Values should end with '.json' or '.yaml' or '.yml'.

default: .schema.json
example:
    - --jsonschema_opt=output_file_suffix=.schema.json
    - --jsonschema_opt=output_file_suffix=.schema.yaml

pretty_json_output

pretty_json_output is used to determine output json should be pretty printed.
This option is only used when output_file_suffix is '.json'.

default: true
example:
    - --jsonschema_opt=pretty_json_output=true
    - --jsonschema_opt=pretty_json_output=false

draft

draft is used to determine which draft version should be used.
The value should be one of Draft04, Draft05, Draft06, Draft07, Draft201909, Draft202012.

default: Draft202012
example:
    - --jsonschema_opt=draft=Draft202012

mandatory_nullable

mandatory_nullable determines whether this plugin should treat optional field as nullable.
Many programming languages do not differentiate between undefined and null.
However, scripting languages like JavaScript and TypeScript can distinguish between them.
By default, optional field is treated as nullable and undefined.

default: false
example:
    - --jsonschema_opt=mandatory_nullable=true
    - --jsonschema_opt=mandatory_nullable=false

int64_as_string

int64_as_string determines whether int64 field treat as string.
Depends on Javascript specification, The JS stores integer to only 53bits.
So, if you want to use int64 field in JS, you should use string type.
References:

default: false
example:
    - --jsonschema_opt=int64_as_string=true
    - --jsonschema_opt=int64_as_string=false

preserve_proto_field_names

preserve_proto_field_names is used to determine if output json field names
should be identical to the proto field names.
Otherwise field names either use the value of the `json_name` field option
or they are automatically converted to lowerCamelCase.
This default behaviour mirrors the behaviour of Protobuf's canonical JSON format (ProtoJSON).

default: false
example:
    - --jsonschema_opt=preserve_proto_field_names=true
    - --jsonschema_opt=preserve_proto_field_names=false

Protobuf Options

Below tables are not auto-generated features. Check out the Options file to see auto-generated options.

protobuf label	jsonschema
required	$.required.append(field)
optional	oneof {type: null, $original_schema}
repeated	type: array, items: $original_schema

WellKnown Types
k8s.io.apimachinery.pkg.util.intstr.IntOrString
k8s.io.api.core.v1.Volume
k8s.io.api.core.v1.SecretProjection
k8s.io.api.core.v1.ConfigMapVolumeSource
k8s.io.api.core.v1.ConfigMapProjection
k8s.io.api.core.v1.ConfigMapKeySelector
k8s.io.api.core.v1.SecretKeySelector
k8s.io.api.core.v1.ConfigMapEnvSource
k8s.io.api.core.v1.SecretEnvSource
k8s.io.api.core.v1.Probe
k8s.io.api.core.v1.EphemeralContainer
google.protobuf.Timestamp
google.protobuf.Duration
google.protobuf.Any
google.protobuf.NullValue

If you'd like contribute well-known types, Please check generator file.

Output Examples

You can find basic example cases in the ./examples directory and more complex cases in ./testdata/cases.