I want to make Jackson work with enums not by name and not by ordinal, but with a custom property I added called "stringId".
I wanted to support this with all Enums in the system so I made an interface called StringIdEnum which the FooEnum will implement.
I'm using Kotlin so I created a property in the interface called stringId which I override in each enum value.
Now I want to make Jackson serialize and deserialize using this stringId field, from what I seen I have several options:
Use #JsonProperty annotation on each enum value and make sure it is aligned with the stringId property.
I see two issues with this approach. one it's a lot of annotation to add (we have many enum classes across the system). two I need to make sure the annotation value and the property value should be always the same which can cause issues in the future.
I tried to use the READ_ENUMS_USING_TO_STRING feature, but because I'm using an interface I can't override the toString in the interface class (I can override it in every enum class but that again seems like a lot of redundant code)
Implement a custom serializer/deserializer.
The serializer is pretty straightforward, however, I had trouble with the deserializer.
I wanted to register the deserializer on the StringIdEnum interface, but I had an issue getting all the runtime enum values for the actual FooType enum.
StringIdEnum:
interface StringIdEnum {
val stringId: String
}
enum class FooType(override val stringId: String) : StringIdEnum {
FOO("FOO"),
GOO("GOO");
}
Managed to get it working:
#JsonSerialize(using = StringIdEnumSerializer::class)
#JsonDeserialize(using = StringIdEnumDeserializer::class)
interface StringIdEnum: DbEnum {
val stringId: String
}
class StringIdEnumSerializer: StdSerializer<StringIdEnum>(StringIdEnum::class.java) {
override fun serialize(value: StringIdEnum, gen: JsonGenerator, provider: SerializerProvider) {
gen.writeString(value.stringId)
}
}
class StringIdEnumDeserializer : JsonDeserializer<Enum<*>>(), ContextualDeserializer {
private lateinit var type: JavaType
override fun deserialize(p: JsonParser, ctxt: DeserializationContext): Enum<*> {
val t = p.text
val enumConstants = (type.rawClass as Class<Enum<*>>).enumConstants
return enumConstants.single { (it as StringIdEnum).stringId == t }
}
override fun createContextual(ctxt: DeserializationContext?, property: BeanProperty?): JsonDeserializer<*> {
val wrapperType: JavaType = property!!.type
val stringIdEnumDeserializer = StringIdEnumDeserializer()
stringIdEnumDeserializer.type = wrapperType
return stringIdEnumDeserializer
}
}
Related
I'm building a schema generator and I'm trying to get the JsonProperty of enum values, for example, for this class:
enum class Size {
#JsonProperty("really-tall") TALL,
#JsonProperty("really-grande") GRANDE;
}
I'd like to get the list "really-tall", "really-grande".
How do I access the annotation of an enum?
Thanks!
UPDATE:
Solution based on this reply for a generic KType:
return (jvmErasure.java as Class<Enum<*>>)
.enumConstants
.map {
it.javaClass.getField(it.name).getAnnotation(JsonProperty::class.java)?.value // Get the JsonProperty string first if exists
?: it.name
}
Update: Additional question from OP
How do I make the first approach work for a generic KType
inline fun <reified T : Enum<T>> getJsonPropertyAnnotations() = enumValues<T>().map {
it.declaringClass
.getField(it.name)
.getAnnotation(JsonProperty::class.java)
.value
}
class SomeTest : StringSpec({
"getJsonPropertyAnnotations" {
getJsonPropertyAnnotations<Size>()
shouldBe listOf("really-tall", "really-grande")
}
})
Please note that with Kotlin 1.7, IntelliJ may show a deprecation warning with wrong replacement for declaringClass in getJsonPropertyAnnotations. I guess this will be sorted out in later versions. Link to related source
The following code should do what you want.
class SomeTest : StringSpec({
"getting annotation values" {
val result = enumValues<Size>().map {
it.declaringClass.getField(it.name).getAnnotation(JsonProperty::class.java).value
}
result shouldBe listOf("really-tall", "really-grande")
}
})
An alternative (less code): Add a String property to your enum class (I called it someFieldName in the below code), annotate with #get:JsonValue, and construct each enum entry with the string value you want. #get:JsonValue will use someFieldName instead of the enum value during serialization.
enum class Size(#get:JsonValue val someFieldName: String) {
TALL("really-tall"),
GRANDE("really-grande");
}
Same test again
class SomeTest : StringSpec({
"getting prop values" {
val result = enumValues<Size>().map {
it.someFieldName
}
result shouldBe listOf("really-tall", "really-grande")
}
})
We're using the latter approach in an ongoing project.
Here's a Kotlin implementation of the technique suggested by user #aSemy.
It's an extension function on ObjectMapper for asking the mapper how it would serialize the values of an enum. This is more robust that just inspecting the #JsonProperty annotation, since it works with #JsonValue as well as any custom annotation introspectors registered with the mapper.
inline fun <reified T : Enum<T>> ObjectMapper.enumValues() : List<String> {
return convertValue(
kotlin.enumValues<T>(),
jacksonTypeRef<List<String>>()
)
}
Usage:
println(jsonMapper().enumValues<Size>())
I have a base class:
abstract class JSONDeserializationStrategy<T : Any>: DeserializationStrategy<T> {
protected abstract fun parse(json: JsonObject): T
protected abstract fun getSerializationException(): SerializationException
}
and then a derived class
class MyClassParserDeserializationStrategy : JSONDeserializationStrategy<MyClass>() {
override val descriptor: SerialDescriptor
= buildClassSerialDescriptor("MyClass")
override fun getSerializationException(): SerializationException
= throw SerializationException("Invalid JSON received for MyClass.")
How could I move the property descriptor and the method getSerializationException from the derived class into the base class, since they only "adapt" by providing their name as String? I was trying to do something in the direction of T::class.java.simpleName as String but it didnt work. What is the best way to do this?
As #Tenfour04 explained, T is erased, so it is not directly accessible. However, as long as the subclass of JSONDeserializationStrategy provides the T as a specific class/type, it can be acquired with a bit of reflection voodoo:
fun main() {
val strategy = MyClassParserDeserializationStrategy()
println(strategy.descriptor.serialName) // MyClass
}
abstract class JSONDeserializationStrategy<T : Any>: DeserializationStrategy<T> {
protected val type: KType = this::class.supertypes
.first { it.classifier == JSONDeserializationStrategy::class }
.arguments[0].type!!
#Suppress("UNCHECKED_CAST")
protected val typeClass = requireNotNull(type.classifier as? KClass<T>) {
"T is unknown"
}
override val descriptor: SerialDescriptor = buildClassSerialDescriptor(typeClass.simpleName!!)
fun getSerializationException(): SerializationException =
throw SerializationException("Invalid JSON received for ${typeClass.simpleName!!}.")
...
}
I'm not 100% sure, but I believe both type!! and simpleName!! are safe, they can't be null.
Also, it won't work if T is really fully erased and unknown, e.g.:
val strategy = GenericParserDeserializationStrategy<MyClass>() // exception
For this reason it makes sense to open type/typeClass properties for overriding, so generic non-abstract subclasses could provide their own means to acquire T. However, then we would probably need to move the initialization of most of properties outside of the constructor.
Because of type erasure, T's class is not accessible. Work-around could be to add it as a constructor property that returns the type, and then the subclasses must pass the type. The property needs to be in the constructor, rather than provided as an abstract property for subclasses to override because you need it to initialize descriptor at instantiation time. (It's highly discouraged to call an open property at class initialization time.)
abstract class JSONDeserializationStrategy<T : Any>(protected val typeClass: KClass<out T>): DeserializationStrategy<T> {
protected abstract fun parse(json: JsonObject): T
override val descriptor: SerialDescriptor = buildClassSerialDescriptor(typeClass.simpleName!!)
fun getSerializationException(): SerializationException =
throw SerializationException("Invalid JSON received for ${typeClass.simpleName}.")
}
class TodaySaleParserDeserializationStrategy : JSONDeserializationStrategy<TodaySale>(TodaySale::class) {
}
I'm trying to build a string with properties that are initialized in a subclass.
I read about lazy initialization but somehow this doesn't work as I expected.
abstract class SubProcessFullNameBuilder(technicalDomain: TechnicalDomainEnumeration) {
protected val moduleName = "td.${technicalDomain.value().toLowerCase()}.shared"
private val packageName by lazy { packageName() }
private val processName by lazy { processName() }
val processFullName: String = "$moduleName/$packageName.$processName"
protected abstract fun packageName(): String
protected abstract fun processName(): String
}
class WorkerFullNameBuilder(
private val jmsDirection: JmsDirectionEnumeration,
technicalDomain: TechnicalDomainEnumeration,
private val cdmCode: String) : SubProcessFullNameBuilder(technicalDomain) {
override fun packageName() = "$moduleName.workers.${jmsDirection.value().toLowerCase()}.${cdmCode.toLowerCase()}"
override fun processName() = "Worker"
}
Since I have overridden the packageName() and processName() properties, I would expect that on calling the packageName property it would use the implementation from the subclass.
But when I call the processFullName property, it throws a java.lang.NullPointerException.
val builder = WorkerFullNameBuilder(JmsDirectionEnumeration.ESB_IN, TechnicalDomainEnumeration.INFOR, "ccmd")
val name = builder.processFullName
How can I initialize the packageName and processName properties in a proper way?
This is a case of calling a non-final method in a constructor and thus accessing uninitialized variables.
This line is still evaluated eagerly, at the time when the base class is constructed:
val processFullName: String = "$moduleName/$packageName.$processName"
To get the values of the two lazy properties, this will make calls to the abstract methods, of which packageName() refers to jmsDirection and cdmCode to return its value - these properties are not initialized yet, because their values are set after the superclass constructor runs. Here's a simplified version of the subclass' constructor, decompiled back to Java:
public WorkerFullNameBuilder(#NotNull JmsDirectionEnumeration jmsDirection, #NotNull TechnicalDomainEnumeration technicalDomain, #NotNull String cdmCode) {
super(technicalDomain);
this.jmsDirection = jmsDirection;
this.cdmCode = cdmCode;
}
As a demonstration, if you don't refer to these, for example, if you return constants in both of the subclass methods, your code will actually run fine:
override fun packageName() = "foo"
override fun processName() = "Worker"
However, the solution you need here is most likely to make the processFullName property itself lazy instead of the two values it uses (which you're evaluating at constructor time right now anyway, so you're not making use of them being lazy). This means you don't even need those two as separate properties:
abstract class SubProcessFullNameBuilder(technicalDomain: TechnicalDomainEnumeration) {
protected val moduleName = "td.${technicalDomain.value().toLowerCase()}.shared"
val processFullName by lazy { "$moduleName/${packageName()}.${processName()}" }
protected abstract fun packageName(): String
protected abstract fun processName(): String
}
I have a framework written in Java that, using reflection, get the fields on an annotation and make some decisions based on them. At some point I am also able to create an ad-hoc instance of the annotation and set the fields myself. This part looks something like this:
public #interface ThirdPartyAnnotation{
String foo();
}
class MyApp{
ThirdPartyAnnotation getInstanceOfAnnotation(final String foo)
{
ThirdPartyAnnotation annotation = new ThirdPartyAnnotation()
{
#Override
public String foo()
{
return foo;
}
};
return annotation;
}
}
Now I am trying to do the exact thing in Kotlin. Bear in mind that the annotation is in a third party jar.
Anyway, here is how I tried it in Kotlin:
class MyApp{
fun getAnnotationInstance(fooString:String):ThirdPartyAnnotation{
return ThirdPartyAnnotation(){
override fun foo=fooString
}
}
But the compiler complains about: Annotation class cannot be instantiated
So the question is: how should I do this in Kotlin?
You can do this with Kotlin reflection:
val annotation = ThirdPartyAnnotation::class.constructors.first().call("fooValue")
In the case of annotation having no-arg constructor (e.g. each annotation field has a default value), you can use following approach:
annotation class SomeAnnotation(
val someField: Boolean = false,
)
val annotation = SomeAnnotation::class.createInstance()
This is the solution I might have found but feels like a hack to me and I would prefer to be able to solve it within the language.
Anyway, for what is worth,it goes like this:
class MyApp {
fun getInstanceOfAnnotation(foo: String): ThirdPartyAnnotation {
val annotationListener = object : InvocationHandler {
override fun invoke(proxy: Any?, method: Method?, args: Array<out Any>?): Any? {
return when (method?.name) {
"foo" -> foo
else -> FindBy::class.java
}
}
}
return Proxy.newProxyInstance(ThirdPartyAnnotation::class.java.classLoader, arrayOf(ThirdPartyAnnotation::class.java), annotationListener) as ThirdPartyAnnotation
}
}
With the code below, I am getting the following error in IntelliJ IDEA 13.1.6 and Kotlin plugin 0.11.91.AndroidStudio.3:
Platform declaration clash: The following declarations have the same JVM signature (getName()Ljava/lang/String;):
• public open fun getName(): kotlin.String?
• internal final fun <get-name>(): kotlin.String?
Java class, JavaInterface.java:
public interface JavaInterface {
public String getName();
}
Kotlin class, KotlinClass.kt
public class KotlinClass(val name: String?) : JavaInterface
I've tried overriding the 'getter' method by
adding override fun getName(): String? = name, but that produces the same error.
I can see one workaround by doing this instead:
public class KotlinClass(val namePrivate: String?) : JavaInterface {
override fun getName(): String? = namePrivate
}
But in my real-world case I have a number of properties to implement and need setters too. Doing this for each property doesn't seem very Kotlin-ish. What am I missing?
Making that variable private solves the problem.
public class KotlinClass(private val name: String?) : JavaInterface
You could use #JvmField for instructs the compiler not generate getter/setter, and you can implement your setters and getters. With this your code work well in Java (as attribute getter/setter) and Kotlin as property
Example:
JAVA:
public interface Identifiable<ID extends Serializable>
{
ID getId();
}
KOTLIN:
class IdentifiableImpl(#JvmField var id: String) :Identifiable<String>
{
override fun getId(): String
{
TODO("not implemented")
}
}
The annotation feature of Kotlin named #JvmName will solve the duplication problem in Java and Kotlin when having the same signature.
fun function(p: String) {
// ...
}
// Signature: function(Ljava/lang/String)
With the use of JvmName will be:
#JvmName("functionOfKotlin")
fun function(p: String) {
// ...
}
// Signature: functionOfKotlin(Ljava/lang/String)
IMHO most readable combination is field + explicit interface implementation by the single-expression function (combination of #Renato Garcia's and #Steven Spungin's answers):
Java:
public inteface SomeInterface {
String getFoo();
}
Kotlin:
class Implementation(#JvmField val foo: String) : SomeInterface {
override fun getFoo() = foo
}
Another work-around is to declare the properties in an abstract Kotlin class, then write a small java class that extends KotlinClass and implements JavaInterface.
// JavaInterface.java
public interface JavaInterface {
int getFoo();
void setFoo(int value);
}
// KotlinClass.kt
abstract class KotlinClass(open var foo : Int = 0) {
}
// JavaAdapter.java
class JavaAdapter extends KotlinClass implements JavaInterface {
// all code in KotlinClass, but can't implement JavaInterface there
// because kotlin properties cannot override java methods.
}
We have found that to use the same names without clashing, the ctor args must be private AND you must still override the interfaces methods. You don't need any additional backing fields. Also, your expression body assignment will not recurse, so you can safely use that syntax.
Java Interface
interface IUser {
String getUserScope();
String getUserId();
}
Kotlin Class
class SampleUser(private val userScope: String, private val userId: String) : IUser {
override fun getUserId() = userId
override fun getUserScope() = userScope
}
If you have direct control over the interface then the best approach is to write the interface in Kotlin. You can then write your class
public class KotlinClass(override val name: String?) : KotlinInterface
and still reference it from any Java code using the same interface as before. This looks a lot neater than setting all the properties to private and overriding the get function. Obviously if you can't migrate the interface to Java because you don't own it then that seems to be the only solution.
public interface JavaInterface {
public String getName();
}
public class KotlinClass(val namePrivate: String?) : JavaInterface {
private var name = namePrivate
override fun getName(): String? {
return name
}
}
Rename the variable to something else, or make it private if u dont want it to be public.
convert function to property instead of initializing property from a function.
for ex:
fun getCountriesList(): List<Country> {
val countries = mutableListOf<Country>()
countries.add(Country("in", "+91", "India", R.drawable.indian_flag))
countries.add(Country("us", "+1", "United States",R.drawable.us_flag))
return countries
}
to
val countriesList: List<Country>
get() {
val countries = mutableListOf<Country>()
countries.add(Country("in", "+91", "India", R.drawable.indian_flag))
countries.add(Country("us", "+1", "United States", R.drawable.us_flag))
return countries
}