I am trying to be able to define the following code:
public class MyObject {
private String name;
... // Other attributes
}
@Path(...)
@Stateless
public class MyRestResource {
@POST
@Consumes(MediaType.APPLICATION_JSON)
public Response create(List<MyObject> myObjects) {
// Do some stuff there
}
}
I know that I need to use:
DeserializationConfig.Feature.ACCEPT_SINGLE_VALUE_AS_ARRAY, true
to setup correctly my object mapper to be able to accept single value as array on my rest resources. I succeed to setup that part.
My problem with this approach is that the following content is not differentiable:
{
"name": "a name",
... // other attributes
}
and
[{
"name": "a name",
... // other attributes
}]
will result into a list (List) of size one. Then, in the method create(List myObjects), I will not be able to do the difference between the List and the Single Object sent to the Rest Resource.
Then, my question is how to do something like that. The idea is to have only one @POST that accepts both Arrays and Single values?
Ideally, I will get rid of the configuration of the ObjectMapper to avoid letting the possibility to set Single Object into the other level of the JSON document. For example, I do not want to allow that:
{
...
"attributes": {
...
}
}
where normally this format should be mandatory:
{
...
"attributes": [{
...
}]
}
Based on that, I tried to put in place an object wrapper of my List to set if I am able to the difference between the list and the object. With something like that:
public class ObjectWrapper<T> {
private List<T> list;
private T object;
public boolean isObject() {
return list == null;
}
}
with the resource that becomes:
@Path(...)
@Stateless
public class MyRestResource {
@POST
@Consumes(MediaType.APPLICATION_JSON)
public Response create(ObjectWrapper myObjects) {
// Do some stuff there
}
}
and trying to put in place the deserialization of my content through the JAX-RS/Jersey/Jackson mechanisms. If I let the solution as it is now, the deserialization fails due to the fact that the JSON format expected is the following:
{
"list": [{
"name": "a name",
... // other attributes
}]
}
Then I tried to write a custom deserializer but I am a bit lost in this task. I have something like that:
public class ObjectWrapperDeserializer<T> extends JsonDeserializer<T> {
@Override
public T deserialize(JsonParser jp, DeserializationContext ctxt) throws IOException, JsonProcessingException {
... // What to put there to deserialize Array or Object
}
}
I just want to deserialize the root level to set the content deserialized into the object wrapper. I also want to keep the feature configured in a class annotated with @ApplicationPath when the configuraiton of the different @Provider are done.
I hope that all the info will give a sufficient picture of what I want to do and what I already tested.
Waiting for suggestion on how to do a resource that accept Arrays or Objects on the same path.
Thanks a lot in advance.
Ok, finally I succeed to put in place a mechanism that do exactly what I am looking for. But, I am not sure if there are negative consequences such the performance or such things.
First, I defined a class that can accept both List or Single Object:
public class RootWrapper<T> {
private List<T> list;
private T object;
}
Then, I need a custom deserializer that is able to know which kind of T type to deserialize and to handle the collection or the single object.
public class RootWrapperDeserializer extends JsonDeserializer<CollectionWrapper<?>> {
private Class contentType;
public RootWrapperDeserializer(Class contentType) {
this.contentType = contentType;
}
@Override
public RootWrapper deserialize(JsonParser jp, DeserializationContext ctxt)
throws IOException, JsonProcessingException {
// Retrieve the object mapper and read the tree.
ObjectMapper mapper = (ObjectMapper) jp.getCodec();
JsonNode root = mapper.readTree(jp);
RootWrapper wrapper = new RootWrapper();
// Check if the root received is an array.
if (root.isArray()) {
List list = new LinkedList();
// Deserialize each node of the array using the type expected.
Iterator<JsonNode> rootIterator = root.getElements();
while (rootIterator.hasNext()) {
list.add(mapper.readValue(rootIterator.next(), contentType));
}
wrapper.setList(list);
}
// Deserialize the single object.
else {
wrapper.setObject(mapper.readValue(root, contentType));
}
return wrapper;
}
}
As far as I know, I try to only deserialize the root level manually and then let Jackson take the next operations in charge. I only have to know which real type I expect to be present in the Wrapper.
At this stage, I need a way to tell Jersey/Jackson which deserializer to use. One way I found for that is to create a sort of deserializer registry where are stored the type to deserialize with the right deserializer. I extended the Deserializers.Base class for that.
public class CustomDeserializers extends Deserializers.Base {
// Deserializers caching
private Map<Class, RootWrapperDeserializer> deserializers = new HashMap<>();
@Override
public JsonDeserializer<?> findBeanDeserializer(JavaType type,
DeserializationConfig config, DeserializerProvider provider,
BeanDescription beanDesc, BeanProperty property) throws JsonMappingException {
// Check if we have to provide a deserializer
if (type.getRawClass() == RootWrapper.class) {
// Check the deserializer cache
if (deserializers.containsKey(type.getRawClass())) {
return deserializers.get(type.getRawClass());
}
else {
// Create the new deserializer and cache it.
RootWrapperDeserializer deserializer =
new RootWrapperDeserializer(type.containedType(0).getRawClass());
deserializers.put(type.getRawClass(), deserializer);
return deserializer;
}
}
return null;
}
}
Ok, then I have my deserializers registry that create new deserializer only on demand and keep them once created. What I am not sure about that approach is if there is any concurrency issue. I know that Jackson do a lot of caching and do not call every time the findBeanDeserializer once it was called a first time on a specific deserialization context.
Now I have created my different classes, I need to do some plumbing to combine everything together. In a provider where I create the ObjectMapper, I can setup the deserializers registry to the created object mapper like below:
@Provider
@Produces(MediaType.APPLICATION_JSON)
public class JsonObjectMapper implements ContextResolver<ObjectMapper> {
private ObjectMapper jacksonObjectMapper;
public JsonObjectMapper() {
jacksonObjectMapper = new ObjectMapper();
// Do some custom configuration...
// Configure a new deserializer registry
jacksonObjectMapper.setDeserializerProvider(
jacksonObjectMapper.getDeserializerProvider().withAdditionalDeserializers(
new RootArrayObjectDeserializers()
)
);
}
@Override
public ObjectMapper getContext(Class<?> arg0) {
return jacksonObjectMapper;
}
}
Then, I can also define my @ApplicationPath that is my REST application like following:
public abstract class AbstractRestApplication extends Application {
private Set<Class<?>> classes = new HashSet<>();
public AbstractRestApplication() {
classes.add(JacksonFeature.class);
classes.add(JsonObjectMapper.class);
addResources(classes);
}
@Override
public Set<Class<?>> getClasses() {
return classes;
}
@Override
public Set<Object> getSingletons() {
final Set<Object> singletons = new HashSet<>(1);
singletons.add(new JacksonJsonProvider());
return singletons;
}
private void addResources(Set<Class<?>> classes) {
classes.add(SomeRestResource.class);
// ...
}
}
Now, everything is in place and I can write a REST resource method like that:
@POST
@Path("somePath")
@Consumes(MediaType.APPLICATION_JSON)
@Produces(MediaType.APPLICATION_JSON)
public Response create(RootWrapper<SpecificClass> wrapper) {
if (wrapper.isObject()) {
// Do something for one single object
SpecificClass sc = wrapper.getObject();
// ...
return Response.ok(resultSingleObject).build();
}
else {
// Do something for list of objects
for (SpecificClass sc = wrapper.getList()) {
// ...
}
return Response.ok(resultList).build();
}
}
That's all. Do not hesitate to comment the solution. Feedbacks are really welcome especially around the way of deserialization process where I am really not sure that it is safe for performance and concurrency.