mbknor / mbknor-jackson-jsonschema

Generate JSON Schema draft 4 with Polymorphism using Jackson annotations

GitHub

Jackson jsonSchema Generator

Build Status Maven Central

This projects aims to do a better job than the original jackson-module-jsonSchema in generating jsonSchema from your POJOs using Jackson @Annotations.

Highlights

  • JSON Schema Draft v4
  • Supports polymorphism (@JsonTypeInfo, MixIn, and registerSubtypes()) using JsonSchema's oneOf-feature.
  • Supports schema customization using:
    • @JsonSchemaDescription/@JsonPropertyDescription
    • @JsonSchemaFormat
    • @JsonSchemaTitle
    • @JsonSchemaDefault
    • @JsonSchemaOptions
    • @JsonSchemaInject
    • @JsonSchemaExamples
  • Supports many Javax-validation @Annotations
  • Works well with Generated GUI's using https://github.com/jdorn/json-editor
    • (Must be configured to use this mode)
    • Special handling of Option-/Optional-properties using oneOf.
  • Supports custom Class-to-format-Mapping
  • Supports injecting custom json-schema-fragments using the @JsonSchemaInject-annotation.

Benefits

  • Simple implementation - Just one file (for now..)
  • Implemented in Scala (Built for 2.10, 2.11 and 2.12)
  • Easy to fix and add functionality

Flexible

If this generator does not generate exactly the schema you want, you can inject it by using the @JsonSchemaInject-annotation.

If you need to use patternProperties (which is not currently 'natively' supported by mbknor-jackson-jsonSchema), you can make it work by injecting the following json-schema-fragment:

{
  "patternProperties" : {
    "^[a-zA-Z0-9]+" : {
      "type" : "string"
    }
  }
}

.. like this in Scala:

@JsonSerialize(using = MySpecialSerializer.class)
JsonSchemaInject(
  json =
    """
      {
        "patternProperties" : {
          "^[a-zA-Z0-9]+" : {
            "type" : "string"
          }
        }
      }
    """
)
case class MyPojo(...)

.. or like this in Java

@JsonSerialize(using = MySpecialSerializer.class)
@JsonSchemaInject( json = "{\n" +
        "  \"patternProperties\" : {\n" +
        "    \"^[a-zA-Z0-9]+\" : {\n" +
        "      \"type\" : \"string\"\n" +
        "    }\n" +
        "  }\n" +
        "}" )
public class MyPojo {
    ...
    ...
    ...
}

Alternatively, the following one liner will achieve the same result:

@JsonSerialize(using = MySpecialSerializer.class)
@JsonSchemaInject(strings = {@JsonSchemaString(path = "patternProperties/^[a-zA-Z0-9]+/type", value = "string")})
public class MyPojo {
    ...
    ...
    ...
}

The annotation will nest the value at the level defined by the path. You can use the raw json along with the individual path/value pairs in the same @JsonSchemaInject annotation. Although keep in mind that the pairs are applied last. For boolean and number values use the JsonSchemaInject#bools and JsonSchemaInject#ints collections correspondingly.

public class MyPojo {
    @JsonSchemaInject(
      bools = {@JsonSchemaBool(path = "exclusiveMinimum", value = true)},
      ints = {@JsonSchemaInt(path = "multipleOf", value = 7)}
    )
    @Min(5)
    public int myIntValue;
    ...
    ...
    ...
}

If a part of the schema is not known at compile time, you can use a json supplier:

case class MyPojo {
  @JsonSchemaInject(jsonSupplier = classOf[UserNamesLoader])
  uns:Set[String]
  ...
  ...
  ...
}

class UserNamesLoader extends Supplier[JsonNode] {
  val _objectMapper = new ObjectMapper()

  override def get(): JsonNode = {
    val schema = _objectMapper.createObjectNode()
    val values = schema.putObject("items").putArray("enum")
    loadUsers().foreach(u => values.add(u.name))

    schema
  }
  ...
  ...
  ...
}

This will associate an enum of possible values for the set that you generate at runtime.

If you need even more control over the schema-generating runtime, you can use @JsonSchemaInject.jsonSupplierViaLookup like this:

case class MyPojo {
  @JsonSchemaInject(jsonSupplierViaLookup = "theKeyToUseWhenLookingUpASupplier")
  uns:Set[String]
  ...
  ...
  ...
}

Then you have to add the mapping between the key 'theKeyToUseWhenLookingUpASupplier' and the Supplier-instance in the config-object used when creating the JsonSchemaGenerator.

The default behaviour of @JsonSchemaInject is to merge the injected json into the generated JsonSchema. If you want to have full control over it, you can specify @JsonSchemaInject.merge = false to replace the generated jsonSchema with the injected json.

@JsonSchemaInject can also be used on properties.

Another way of altering the generated schema is to use the config-param classTypeReMapping.

This can be used to remap the Class found by Jackson into another Class before generating the schema for it.

It might be the case that you have a complex Class-structure using Polymorphism, but for some reason you know upfront that the user needs to enter/specify a specific subType. To enforce this into the generated schema, you can map the SuperType into the specific-type.

Project status

We're currently using this codebase in an ongoing (not yet released) project at work, and we're improving the jsonSchema-generating code when we finds issues and/or features we need that not yet is supported.

I would really appreciate it if other developers wanted to start using and contributing improvements and features.

Dependency

This project publishes artifacts to central maven repo.

The project is also compiled using Java 8. This means that you also need to use Java 8.

Artifacts for both Scala 2.10, 2.11 and 2.12 is now available (Thanks to @bbyk for adding crossBuild functionality).

Using Maven

Add this to you pom.xml:

<dependency>
    <groupId>com.kjetland</groupId>
    <artifactId>mbknor-jackson-jsonschema_2.12</artifactId>
    <version>[---LATEST VERSION---]</version>
</dependency>    

Using sbt

Add this to you sbt build-config:

"com.kjetland" %% "mbknor-jackson-jsonschema" % "[---LATEST VERSION---]"

Code - Using Scala

This is how to generate jsonSchema in code using Scala:

    val objectMapper = new ObjectMapper
    val jsonSchemaGenerator = new JsonSchemaGenerator(objectMapper)
    val jsonSchema:JsonNode = jsonSchemaGenerator.generateJsonSchema(classOf[YourPOJO])

    val jsonSchemaAsString:String = objectMapper.writeValueAsString(jsonSchema)

This is how to generate jsonSchema used for generating HTML5 GUI using json-editor:

    val objectMapper = new ObjectMapper
    val jsonSchemaGenerator = new JsonSchemaGenerator(objectMapper, config = JsonSchemaConfig.html5EnabledSchema)
    val jsonSchema:JsonNode = jsonSchemaGenerator.generateJsonSchema(classOf[YourPOJO])

    val jsonSchemaAsString:String = objectMapper.writeValueAsString(jsonSchema)

This is how to generate jsonSchema using custom type-to-format-mapping using Scala:

    val objectMapper = new ObjectMapper
    val config:JsonSchemaConfig = JsonSchemaConfig.vanillaJsonSchemaDraft4.copy(
      customType2FormatMapping = Map( "java.time.OffsetDateTime" -> "date-time-ABC-Special" )
    )
    val jsonSchemaGenerator = new JsonSchemaGenerator(objectMapper, config = config)
    val jsonSchema:JsonNode = jsonSchemaGenerator.generateJsonSchema(classOf[YourPOJO])

    val jsonSchemaAsString:String = objectMapper.writeValueAsString(jsonSchema)

Note about Scala and Option[Int]:

Due to Java's Type Erasure it impossible to resolve the type T behind Option[T] when T is Int, Boolean, Double. As a workaround, you have to use the @JsonDeserialize-annotation in such cases. See https://github.com/FasterXML/jackson-module-scala/wiki/FAQ#deserializing-optionint-and-other-primitive-challenges for more info.

Example:

    case class PojoUsingOptionScala(
                                     _string:Option[String], // @JsonDeserialize not needed here
                                     @JsonDeserialize(contentAs = classOf[Int])     _integer:Option[Int],
                                     @JsonDeserialize(contentAs = classOf[Boolean]) _boolean:Option[Boolean],
                                     @JsonDeserialize(contentAs = classOf[Double])  _double:Option[Double],
                                     child1:Option[SomeOtherPojo] // @JsonDeserialize not needed here
                                   )

PS: Scala Option combined with Polymorphism does not work in jackson-scala-module and therefore not this project either.

Code - Using Java

    ObjectMapper objectMapper = new ObjectMapper();
    JsonSchemaGenerator jsonSchemaGenerator = new JsonSchemaGenerator(objectMapper);

    // If using JsonSchema to generate HTML5 GUI:
    // JsonSchemaGenerator html5 = new JsonSchemaGenerator(objectMapper, JsonSchemaConfig.html5EnabledSchema() );

    // If you want to configure it manually:
    // JsonSchemaConfig config = JsonSchemaConfig.create(...);
    // JsonSchemaGenerator generator = new JsonSchemaGenerator(objectMapper, config);


    JsonNode jsonSchema = jsonSchemaGenerator.generateJsonSchema(YourPOJO.class);

    String jsonSchemaAsString = objectMapper.writeValueAsString(jsonSchema);

Nullable types

Out of the box, the generator does not support nullable types. There is a preconfigured JsonSchemaGenerator configuration shortcut that can be used to enable them:

JsonSchemaConfig config = JsonSchemaConfig.nullableJsonSchemaDraft4();
JsonSchemaGenerator generator = new JsonSchemaGenerator(objectMapper, config);

Under the hood nullableJsonSchemaDraft4 toggles the useOneOfForOption and useOneOfForNullables properties on JsonSchemaConfig.

When support is enabled, the following types may be made nullable:

  • Use Optional<T> (or Scala's Option)
  • Use a non-optional, non-primitive type (IE: String, Boolean, Integer etc)

If you've otherwise enabled support for nullable types, but need to suppress this at a per-property level, you can do this like so:

// A standard validation @NotNull annotation.
@NotNull
public String foo;

// Using the Jackson @JsonProperty annotation, specifying the attribute as required.
@JsonProperty(required = true)
public String bar;

Using JSON Views

Using JSON Views is most helpful for an API for various clients that will receive different output fields out of the same class, by calling different service endpoints. While support for these is not built in jsonSchema, it is handy to know how to use them with it since it is not an obvious process unless you are very familiar with the Jackson API.

Hence, let's suppose that you want to filter YourPojo using properties marked with the view Views.MyView.

Here is how to do it in Scala:

    val objectMapper = new ObjectMapper

    objectMapper.disable(MapperFeature.DEFAULT_VIEW_INCLUSION)
    objectMapper.setConfig(objectMapper.getSerializationConfig().withView(Views.MyView.class))

    val jsonSchemaGenerator = new JsonSchemaGenerator(objectMapper)
    val jsonSchema:JsonNode = jsonSchemaGenerator.generateJsonSchema(classOf[YourPOJO])

    val jsonSchemaAsString:String = objectMapper.writeValueAsString(jsonSchema)

And here is the equivalent for Java:

    ObjectMapper objectMapper = new ObjectMapper();
		
    // Disabling default View so only the properties that matter are output
    objectMapper.disable(MapperFeature.DEFAULT_VIEW_INCLUSION);
	    
    // And the trick: grab the serializationConfig and define the desired view
    objectMapper.setConfig(objectMapper.getSerializationConfig().withView(Views.MyView.class));
	    
    // Then, proceed as usual. Only fields and classes annotated with MyView will appear in the schema
    JsonSchemaGenerator generator = new JsonSchemaGenerator(objectMapper);
    JsonNode jsonSchema = generator.generateJsonSchema(SearchResult.class);
    String jsonSchemaAsString = objectMapper.writeValueAsString(jsonSchema);

Subclass-resolving using reflection

In some cases it is needed to find extra info about classes not found in jackson data. https://github.com/classgraph/classgraph is used to solve this problem.

By default we scan the entire classpath. This can be slow, so it is better to customize what to scan.

This is how you can configure what mbknor-jackson-jsonSchema should scan

in Scala:

    // Scan only some packages (this is faster)
    
    val resolver = SubclassesResolverImpl()
                    .withPackagesToScan(List("this.is.myPackage"))
                    .withClassesToScan(List("this.is.myPackage.MyClass")) // and/or this one
                    //.withClassGraph() - or use this one to get full control..       
    
    config = config.withSubclassesResolver( resolver )

.. or in Java:

    // Scan only some packages (this is faster)
    
    final SubclassesResolver resolver = new SubclassesResolverImpl()
                                            .withPackagesToScan(Arrays.asList(
                                               "this.is.myPackage"
                                            ))
                                            .withClassesToScan(Arrays.asList( // and/or this one
                                               "this.is.myPackage.MyClass"
                                            ))
                                            //.withClassGraph() - or use this one to get full control..       
    
    config = config.withSubclassesResolver( resolver )

Backstory

At work we've been using the original jackson-module-jsonSchema to generate schemas used when rendering dynamic GUI using https://github.com/jdorn/json-editor.

Recently we needed to support POJO's using polymorphism like this:

    @JsonTypeInfo(
            use = JsonTypeInfo.Id.NAME,
            include = JsonTypeInfo.As.PROPERTY,
            property = "type")
    @JsonSubTypes({
            @JsonSubTypes.Type(value = Child1.class, name = "child1"),
            @JsonSubTypes.Type(value = Child2.class, name = "child2") })
    public abstract class Parent {

        public String parentString;

    }

This is not supported by the original jackson-module-jsonSchema. I have spent many hours trying to figure out how to modify/improve it without any luck, and since it is implemented in such a complicated way, I decided to instead write my own jsonSchema generator from scratch.