IMHO, the simplest way to do that would be to use introspection to dynamically generate a hash containing selected fields and then serialize that hash using Json. You simply have to decide what is the list of usable fields (see below).
Here are two example functions able to do that, first gets all public fields and public getters, the second gets all declared fields (including private ones) in current class and all its parent classes :
public Map<String, Object> getPublicMap(Object obj, List<String> names)
throws IllegalAccessException, IllegalArgumentException, InvocationTargetException {
List<String> gettedFields = new ArrayList<String>();
Map<String, Object> values = new HashMap<String, Object>();
for (Method getter: obj.getClass().getMethods()) {
if (getter.getName().startsWith("get") && (getter.getName().length > 3)) {
String name0 = getter.getName().substring(3);
String name = name0.substring(0, 1).toLowerCase().concat(name0.substring(1));
gettedFields.add(name);
if ((names == null) || names.isEmpty() || names.contains(name)) {
values.put(name, getter.invoke(obj));
}
}
}
for (Field field: obj.getClass().getFields()) {
String name = field.getName();
if ((! gettedFields.contains(name)) && ((names == null) || names.isEmpty() || names.contains(name))) {
values.put(name, field.get(obj));
}
}
return values;
}
public Map<String, Object> getFieldMap(Object obj, List<String> names)
throws IllegalArgumentException, IllegalAccessException {
Map<String, Object> values = new HashMap<String, Object>();
for (Class<?> clazz = obj.getClass(); clazz != Object.class; clazz = clazz.getSuperclass()) {
for (Field field : clazz.getDeclaredFields()) {
String name = field.getName();
if ((names == null) || names.isEmpty() || names.contains(name)) {
field.setAccessible(true);
values.put(name, field.get(obj));
}
}
}
return values;
}
Then you only have to get the result of one of this function (or of one you could adapt to your requirements) and serialize it with Jackson.
If you have custom encoding of you domain objects, you would have to maintain the serialization rules in two different places : hash generation and Jackson serialization. In that case, you could simply generate the full class serialization with Jackson and filter the generated string afterwards. Here is an example of such a filter function :
public String jsonSub(String json, List<String> names) throws IOException {
if ((names == null) || names.isEmpty()) {
return json;
}
ObjectMapper mapper = new ObjectMapper();
Map<String, Object> map = mapper.readValue(json, HashMap.class);
for (String name: map.keySet()) {
if (! names.contains(name)) {
map.remove(name);
}
}
return mapper.writeValueAsString(map);
}
Edit : integration in Spring MVC
As you are speaking of a web service and of Jackson, I assume that you use Spring RestController or ResponseBody annotations and (under the hood) a MappingJackson2HttpMessageConverter. If you use Jackson 1 instead, it should be a MappingJacksonHttpMessageConverter.
What I propose is simply to add a new HttpMessageConverter that could make use of one of the above filtering functions, and delegate actual work (and also ancilliary methods) to a true MappingJackson2HttpMessageConverter. In the write method of that new converter, it is possible to have access to the eventual fields request parameter with no need for an explicit ThreadLocal variable thanks to Spring RequestContextHolder. That way :
- you keep a clear separation of roles with no modification on existing controllers
- you have no modification in Jackson2 configuration
- you need no new ThreadLocal variable and simply use a Spring class in a class already tied to Spring since it implements
HttpMessageConverter
Here is an example of such a message converter :
public class JsonConverter implements HttpMessageConverter<Object> {
private static final Logger logger = LoggerFactory.getLogger(JsonConverter.class);
// a real message converter that will respond to ancilliary methods and do the actual work
private HttpMessageConverter<Object> delegate =
new MappingJackson2HttpMessageConverter();
// allow configuration of the fields name
private String fieldsParam = "fields";
public void setFieldsParam(String fieldsParam) {
this.fieldsParam = fieldsParam;
}
@Override
public boolean canRead(Class<?> clazz, MediaType mediaType) {
return delegate.canRead(clazz, mediaType);
}
@Override
public boolean canWrite(Class<?> clazz, MediaType mediaType) {
return delegate.canWrite(clazz, mediaType);
}
@Override
public List<MediaType> getSupportedMediaTypes() {
return delegate.getSupportedMediaTypes();
}
@Override
public Object read(Class<? extends Object> clazz, HttpInputMessage inputMessage) throws IOException, HttpMessageNotReadableException {
return delegate.read(clazz, inputMessage);
}
@Override
public void write(Object t, MediaType contentType, HttpOutputMessage outputMessage)
throws IOException, HttpMessageNotWritableException {
// is there a fields parameter in request
String[] fields = ((ServletRequestAttributes) RequestContextHolder.getRequestAttributes())
.getRequest().getParameterValues(fieldsParam);
if (fields != null && fields.length != 0) {
// get required field names
List<String> names = new ArrayList<String>();
for (String field : fields) {
String[] f_names = field.split("\\s*,\\s*");
names.addAll(Arrays.asList(f_names));
}
// special management for Map ...
if (t instanceof Map) {
Map<?, ?> tmap = (Map<?, ?>) t;
Map<String, Object> map = new LinkedHashMap<String, Object>();
for (Entry entry : tmap.entrySet()) {
String name = entry.getKey().toString();
if (names.contains(name)) {
map.put(name, entry.getValue());
}
}
t = map;
} else {
try {
Map<String, Object> map = getMap(t, names);
t = map;
} catch (Exception ex) {
throw new HttpMessageNotWritableException("Error in field extraction", ex);
}
}
}
delegate.write(t, contentType, outputMessage);
}
/**
* Create a Map by keeping only some fields of an object
* @param obj the Object
* @param names names of the fields to keep in result Map
* @return a map containing only requires fields and their value
* @throws IllegalArgumentException
* @throws IllegalAccessException
*/
public static Map<String, Object> getMap(Object obj, List<String> names)
throws IllegalArgumentException, IllegalAccessException {
Map<String, Object> values = new HashMap<String, Object>();
for (Class<?> clazz = obj.getClass(); clazz != Object.class; clazz = clazz.getSuperclass()) {
for (Field field : clazz.getDeclaredFields()) {
String name = field.getName();
if (names.contains(name)) {
field.setAccessible(true);
values.put(name, field.get(obj));
}
}
}
return values;
}
}
If you want the converter to be more versatile, you could define an interface
public interface FieldsFilter {
Map<String, Object> getMap(Object obj, List<String> names)
throws IllegalAccessException, IllegalArgumentException, InvocationTargetException;
}
and inject it with an implementation of that.
Now you must ask Spring MVC to use that custom message controller.
If you use XML config, you simply declare it in the <mvc:annotation-driven> element :
<mvc:annotation-driven >
<mvc:message-converters>
<bean id="jsonConverter" class="org.example.JsonConverter"/>
</mvc:message-converters>
</mvc:annotation-driven>
And if you use Java configuration, it is almost as simple :
@EnableWebMvc
@Configuration
public class WebConfig extends WebMvcConfigurerAdapter {
@Autowired JsonConverter jsonConv;
@Override
public void configureMessageConverters(List<HttpMessageConverter<?>> converters) {
converters.add(jsonConv);
StringHttpMessageConverter stringConverter = new StringHttpMessageConverter();
stringConverter.setWriteAcceptCharset(false);
converters.add(new ByteArrayHttpMessageConverter());
converters.add(stringConverter);
converters.add(new ResourceHttpMessageConverter());
converters.add(new SourceHttpMessageConverter<Source>());
converters.add(new AllEncompassingFormHttpMessageConverter());
converters.add(new MappingJackson2HttpMessageConverter());
}
}
but here you have to explicitely add all the default message converters that you need.
