Apache Johnzon is a project providing an implementation of JsonProcessing (aka JSR-353) and a set of useful extension for this specification like an Object mapper, some JAX-RS providers and a WebSocket module provides a basic integration with Java WebSocket API (JSR-356).
Apache Johnzon is a Top Level Project at the Apache Software Foundation (ASF). It fully implements the JSON-P 2.1 and JSON-B 3.0 specifications.
Johnzon comes with four main modules.
<dependency> <groupId>org.apache.johnzon</groupId> <artifactId>johnzon-core</artifactId> <version>${johnzon.version}</version> </dependency>
This is the implementation of the JSON-P 2.1 specification. You'll surely want to add the API as dependency too:
<dependency> <groupId>jakarta.json</groupId> <artifactId>jakarta.json-api</artifactId> <version>2.1.2</version> <scope>provided</scope> <!-- or compile if your environment doesn't provide it --> </dependency>
Please note: The jakarta JSON-P API jar has hardcoded parsson as the default JSON-P implementation. This might cause unintended behaviour in cases where standard Java service loading is not possible.
The generator factory supports the standard properties (pretty one for example) but also:
org.apache.johnzon.encoding
: encoding to use for the generator when converting an OutputStream to a Writer.org.apache.johnzon.buffer-strategy
: how to get buffers (char buffer), default strategy is a queue/pool based one but you can switch it to a THREAD_LOCAL
one. BY_INSTANCE
(per call/prototype) and SINGLETON
(single instance) are also supported but first one is generally slower and last one does not enable overflows.org.apache.johnzon.default-char-buffer-generator
(int): buffer size of the generator, it enables to work in memory to flush less often (for performances).org.apache.johnzon.boundedoutputstreamwriter
(int): when converting an OuputStream
to a Writer
it defines the buffer size (if > 0) +- 2 charaters (for the encoding logic). It enables a faster flushing to the actual underlying output stream combined with org.apache.johnzon.default-char-buffer-generator
.This has been removed with Johnzon 2.0.x, johnzon-core is now JSON-P compliant by default.
<dependency> <groupId>org.apache.johnzon</groupId> <artifactId>johnzon-jsonp-strict</artifactId> <version>${johnzon.version}</version> </dependency>
This module enables to enforce a strict compliance of JsonPointer behavior on /-
usage.
Johnzon default implementation enables an extended usage of it for replace/remove and get operations.
In that case, it will point to the last element of the array so it's easy to replace/remove or get the last element of the array.
For add operation, it remains the same, aka points to the element right after the last element of the array.
This module enforces Johnzon to be JSONP compliant and fail if /-
is used for anything but add.
Note that you can even customize this behavior implementing your own JsonPointerFactory
and changing the ordinal value to take a highest priority.
<dependency> <groupId>org.apache.johnzon</groupId> <artifactId>johnzon-mapper</artifactId> <version>${johnzon.version}</version> </dependency>
The mapper module allows you to use the implementation you want of Json Processing specification to map Json to Object and the opposite.
final MySuperObject object = createObject(); final Mapper mapper = new MapperBuilder().build(); mapper.writeObject(object, outputStream); final MySuperObject otherObject = mapper.readObject(inputStream, MySuperObject.class);
The mapper uses a direct java to json representation.
For instance this java bean:
public class MyModel { private int id; private String name; // getters/setters }
will be mapped to:
{ "id": 1234, "name": "Johnzon doc" }
Note that Johnzon supports several customization either directly on the MapperBuilder of through annotations.
@JohnzonIgnore is used to ignore a field. You can optionally say you ignore the field until some version if the mapper has a version:
public class MyModel { @JohnzonIgnore private String name; // getters/setters }
Or to support name for version 3, 4, … but ignore it for 1 and 2:
public class MyModel { @JohnzonIgnore(minVersion = 3) private String name; // getters/setters }
Converters are used for advanced mapping between java and json.
There are several converter types:
The most common is to customize date format but they all take. For that simple case we often use a Converter:
public class LocalDateConverter implements Converter<LocalDate> { @Override public String toString(final LocalDate instance) { return instance.toString(); } @Override public LocalDate fromString(final String text) { return LocalDate.parse(text); } }
If you need a more advanced use case and modify the structure of the json (wrapping the value for instance) you will likely need Reader/Writer or a Codec.
Then once your converter developed you can either register globally on the MapperBuilder or simply decorate the field you want to convert with @JohnzonConverter:
public class MyModel { @JohnzonConverter(LocalDateConverter.class) private LocalDate date; // getters/setters }
Sometimes the json name is not java friendly (_foo or foo-bar or even 200 for instance). For that cases @JohnzonProperty allows to customize the name used:
public class MyModel { @JohnzonProperty("__date") private LocalDate date; // getters/setters }
If you don't fully know your model but want to handle all keys you can use @JohnzonAny to capture/serialize them all:
public class AnyMe { private String name; // Regular serialization for the known 'name' field /* This example uses a TreeMap to store and retrieve the other unknown fields for the @JohnzonAny annotated methods, but you can choose anything you want. Use @JohnzonIgnore to avoid exposing this as an actual 'unknownFields' property in JSON. */ @JohnzonIgnore private Map<String, Object> unknownFields = new TreeMap<String, Object>(); public String getName() { return name; } public void setName(final String name) { this.name = name; } @JohnzonAny public Map<String, Object> getAny() { return unknownFields; } @JohnzonAny public void handle(final String key, final Object val) { this.unknownFields.put(key, val); } }
On MapperBuilder you have several AccessMode available by default but you can also create your own one.
The default available names are:
You can use these names with setAccessModeName().
<dependency> <groupId>org.apache.johnzon</groupId> <artifactId>johnzon-jaxrs</artifactId> <version>${johnzon.version}</version> </dependency>
JAX-RS module provides two providers (and underlying MessageBodyReaders and MessageBodyWriters):
Note: Wildcard providers are basically the same as other provider but instead of application/json they support */json, /+json, */x-json, */javascript, */x-javascript. This split makes it easier to mix json and other MediaType in the same resource (like text/plain, xml etc since JAX-RS API always matches as true wildcard type in some version whatever the subtype is).
Tip: ConfigurableJohnzonProvider maps most of MapperBuilder configuration letting you configure it through any IoC including not programming language based formats.
IMPORTANT: when used with johnzon-core
, NoContentException
is not thrown in case of an empty incoming input stream by these providers except JsrProvider
to limit the breaking changes.
TomEE uses by default Johnzon as JAX-RS provider for versions 7.x. If you want however to customize it you need to follow this procedure:
<?xml version="1.0" encoding="UTF-8"?> <openejb-jar> <pojo-deployment class-name="jaxrs-application"> <properties> # optional but requires to skip scanned providers if set to true cxf.jaxrs.skip-provider-scanning = true # list of providers we want cxf.jaxrs.providers = johnzon,org.apache.openejb.server.cxf.rs.EJBAccessExceptionMapper </properties> </pojo-deployment> </openejb-jar>
<?xml version="1.0" encoding="UTF-8"?> <resources> <Service id="johnzon" class-name="org.apache.johnzon.jaxrs.ConfigurableJohnzonProvider"> # 1M maxSize = 1048576 bufferSize = 1048576 # ordered attributes attributeOrder = $order # Additional types to ignore ignores = org.apache.cxf.jaxrs.ext.multipart.MultipartBody </Service> <Service id="order" class-name="com.company.MyAttributeSorter" /> </resources>
Note: as you can see you mainly just need to define a service with the id johnzon (same as in openejb-jar.xml) and you can reference other instances using $id for services and @id for resources.
Johnzon provides a module johnzon-jsonb implementing JSON-B standard based on Johnzon Mapper.
It fully reuses the JSON-B as API.
However it supports some specific properties to wire to the native johnzon configuration - see JohnzonBuilder
for details.
One example is johnzon.interfaceImplementationMapping
which will support a Map<Class,Class>
to map interfaces to implementations
to use for deserialization.
JsonbConfig specific properties:
@JsonbCreator
be used too.jsonb.fail-on-unknown-properties
.@JsonbCreator
misses some values.BigInteger
is mapped as a string. true
by default, set to false
to ensure strict JSON-B 3 complianceBigDecimal
is mapped as a string. true
by default, set to false
to ensure strict JSON-B 3 complianceTIP: more in JohnzonBuilder class.
A JAX-RS provider based on JSON-B is provided in the module as well. It is org.apache.johnzon.jaxrs.jsonb.jaxrs.JsonbJaxrsProvider
.
IMPORTANT: in JAX-RS 1.0 the provider can throw any exception he wants for an empty incoming stream on reader side. This had been broken in JAX-RS 2.x where it must throw a jakarta.ws.rs.core.NoContentException
.
To ensure you can pick the implementation you can and limit the breaking changes, you can set ̀throwNoContentExceptionOnEmptyStreamson the provider to switch between both behaviors. Default will be picked from the current available API. Finally, this behavior only works with
johnzon-core`.
JsonValue
You can use some optimization to map a JsonObject
to a POJO using Johnzon JsonValueReader
- or any implementation of Reader
implementing Supplier<JsonStructure>
- and JsonValueWriter
- or any implementation of Writer
implementing Consumer<JsonValue>
-:
final JsonValueReader<Simple> reader = new JsonValueReader<>(Json.createObjectBuilder().add("value", "simple").build()); final Jsonb jsonb = getJohnzonJsonb(); final Simple simple = jsonb.fromJson(reader, SomeModel.class);
final JsonValueWriter writer = new JsonValueWriter(); final Jsonb jsonb = getJohnzonJsonb(); jsonb.toJson(object, writer); final JsonObject jsonObject = writer.getObject();
These two example will not use any IO and directly map the JsonValue
to/from a POJO.
Also note that, as an experimental extension and pre-available feature of the next specification version, org.apache.johnzon.jsonb.api.experimental.JsonbExtension
enables
to map POJO to JsonValue
and the opposite.
<dependency> <groupId>org.apache.johnzon</groupId> <artifactId>johnzon-websocket</artifactId> <version>${johnzon.version}</version> </dependency>
WebSocket module provides a basic integration with Java WebSocket API (JSR 356).
Integration is at codec level (encoder/decoder). There are two families of codecs:
Note that if you want to control the Mapper or JSON-B instance used by decoders you can set up the associated servlet listeners:
if you write in the servlet context an attribute named org.apache.johnzon.websocket.internal.mapper.MapperLocator.mapper
(it is a Supplier<Mapper>
) or org.apache.johnzon.websocket.jsonb.JsonbLocator.jsonb
(depending the implementation you use) it will be used instead of the default instance.
Encoders:
org.apache.johnzon.websocket.jsr.JsrObjectEncoder
org.apache.johnzon.websocket.jsr.JsrArrayEncoder
org.apache.johnzon.websocket.jsr.JsrStructureEncoder
Decoders:
org.apache.johnzon.websocket.jsr.JsrObjectDecoder
org.apache.johnzon.websocket.jsr.JsrArrayDecoder
org.apache.johnzon.websocket.jsr.JsrStructureDecoder
Encoder:
org.apache.johnzon.websocket.mapper.JohnzonTextEncoder
Decoder:
org.apache.johnzon.websocket.mapper.JohnzonTextDecoder
Encoder:
org.apache.johnzon.websocket.jsonb.JsonbTextEncoder
Decoder:
org.apache.johnzon.websocket.jsonb.JsonbTextDecoder
On server and client side configuration is easy: just provide the encoders
and decoders
parameters to @[Server|Client]Endpoint
(or EndpointConfig
if you use programmatic API)):
@ClientEndpoint(encoders = JsrObjectEncoder.class, decoders = JsrObjectDecoder.class)
public class JsrClientEndpointImpl {
@OnMessage
public void on(final JsonObject message) {
// ...
}
}
@ServerEndpoint(value = "/my-server", encoders = JsrObjectEncoder.class, decoders = JsrObjectDecoder.class)
public class JsrClientEndpointImpl {
@OnMessage
public void on(final JsonObject message) {
// ...
}
}
Server configuration is as simple as providing encoders
and decoders
parameters to @ServerEndpoint
:
@ServerEndpoint(value = "/server", encoders = JohnzonTextEncoder.class, decoders = JohnzonTextDecoder.class)
public class ServerEndpointImpl {
@OnMessage
public void on(final Session session, final Message message) {
// ...
}
}
Client configuration is almost the same excepted in this case it is not possible for Johnzon
to guess the type you expect so you'll need to provide it. In next sample it is done just extending JohnzonTextDecoder
in MessageDecoder
.
@ClientEndpoint(encoders = JohnzonTextEncoder.class, decoders = ClientEndpointImpl.MessageDecoder.class)
public class ClientEndpointImpl {
@OnMessage
public void on(final Message message) {
// ...
}
public static class MessageDecoder extends JohnzonTextDecoder {
public MessageDecoder() {
super(Message.class);
}
}
}
<dependency> <groupId>org.apache.johnzon</groupId> <artifactId>johnzon-jsonb-extras</artifactId> <version>${johnzon.version}</version> </dependency>
This module provides some extension to JSON-B.
This extension shouldn't be used anymore if you don't absolutely rely on the JSON format it generates/parses. Use JSON-B 3 polymorphism instead. It provides a way to handle polymorphism:
For the deserialization side you have to list the potential children on the root class:
@Polymorphic.JsonChildren({
Child1.class,
Child2.class
})
public abstract class Root {
public String name;
}
Then on children you bind an “id” for each of them (note that if you don't give one, the simple name is used):
@Polymorphic.JsonId("first")
public class Child1 extends Root {
public String type;
}
Finally on the field using the root type (polymorphic type) you can bind the corresponding serializer and/or deserializer:
public class Wrapper {
@JsonbTypeSerializer(Polymorphic.Serializer.class)
@JsonbTypeDeserializer(Polymorphic.DeSerializer.class)
public Root root;
@JsonbTypeSerializer(Polymorphic.Serializer.class)
@JsonbTypeDeserializer(Polymorphic.DeSerializer.class)
public List<Root> roots;
}
Binding the polymophic serializer and/or deserializer must not be done using JsonbConfig.withSerializers
/ JsonbConfig.withDeserializers
, as it is designed to work only with binding performed using annotations.
<dependency> <groupId>org.apache.johnzon</groupId> <artifactId>johnzon-jsonschema</artifactId> <version>${johnzon.version}</version> </dependency>
This module provides a way to validate an instance against a JSON Schema.
// long live instances (@ApplicationScoped/@Singleton) JsonObject schema = getJsonSchema(); JsonSchemaValidatorFactory factory = new JsonSchemaValidatorFactory(); JsonSchemaValidator validator = factory.newInstance(schema); // runtime starts here JsonObject objectToValidateAgainstSchema = getObject(); ValidatinResult result = validator.apply(objectToValidateAgainstSchema); // if result.isSuccess, result.getErrors etc... // end of runtime validator.close(); factory.close();
Known limitations are (feel free to do a PR on github to add these missing features):
<dependency> <groupId>org.apache.johnzon</groupId> <artifactId>johnzon-jsonlogic</artifactId> <version>${johnzon.version}</version> </dependency> <dependency> <!-- requires an implementation of JSON-P --> <groupId>org.apache.johnzon</groupId> <artifactId>johnzon-core</artifactId> <version>${johnzon.version}</version> </dependency>
This module provides a way to execute any JSON Logic expression.
final JohnzonJsonLogic jsonLogic = new JohnzonJsonLogic(); final JsonValue result = jsonLogic.apply( builderFactory.createObjectBuilder() .add("merge", builderFactory.createArrayBuilder() .add(builderFactory.createArrayBuilder() .add(1) .add(2)) .add(3) .add("4")) .build(), JsonValue.EMPTY_JSON_ARRAY);
Default operators are supported - except “log” one to let you pick the logger (impl + name) you want.
To register a custom operator just do it on your json logic instance:
final JohnzonJsonLogic jsonLogic = new JohnzonJsonLogic(); jsonLogic.registerOperator( "log", (jsonLogic, config, args) -> log.info(String.valueOf(jsonLogic.apply(config, args)));
Note that by default the set of standard JSON Logic operators is enriched with JSON-P jsonpatch, json merge diff and json merge patch operators.
Though Johnzon artifacts are OSGi bundles to begin with, this module provides further integration with the OSGi JAX-RS Whiteboard and OSGi CDI Integration specifications.
This module provides MessageBodyWriter
and MessageBodyReader
extensions for the media type application/json
(by default) to whiteboard JAX-RS Applications.
Configuration of this extension is managed via Configuration Admin using the pid org.apache.johnzon.jaxrs.jsonb
and defines a Metatype schema with the following properties:
| Property | Synopsis | Type | Default |
| —- | ————- | – | – |
| ignores
| List of fully qualified class names to ignore | String[] | empty |
| osgi.jaxrs.application.select
| Filter expression used to match the extension to JAX-RS Whiteboard Applications | String | (!(johnzon.jsonb=false))
(which is a convention allowing the extension to bind to all applications unless the application is configured with johnzon.jsonb=false
) |
| osgi.jaxrs.media.type
| List of media types handled by the extension | String[] | application/json
|
| throw.no.content.exception.on.empty.streams
| | boolean | false
|
| fail.on.unknown.properties
| | boolean | false
|
| use.js.range
| | boolean | false
|
| other.properties
| | String | empty |
| ijson
| | boolean | false
|
| encoding
| | String | empty |
| binary.datastrategy
| | String | empty |
| property.naming.strategy
| | String | empty |
| property.order.strategy
| | String | empty |
| null.values
| | boolean | false
|
| pretty
| | boolean | false
|
| fail.on.missing.creator.values
| | boolean | false
|
| polymorphic.serialization.predicate
| | String | empty |
| polymorphic.deserialization.predicate
| | String | empty |
| polymorphic.discriminator
| | String | empty |
Since JSON-B specification provides an integration with the CDI specification to handle caching, this module also provides such integration for OSGi CDI Integration specification by providing an jakarta.enterprise.inject.spi.Extension
service with the required service property osgi.cdi.extension
with the value JavaJSONB
.
In order to reduce the burden of configuration Apache Aries CDI (the OSGi CDI Integration RI) provides a feature of implicit extensions. These are extensions which the developer doesn't have to configure a requirement for in their CDI bundle. The Johnzon JSON-B CDI extension is such an extension and as such when running in Aries CDI does not need to be required.
This is achieve using the service property aries.cdi.extension.mode=implicit
.