How can I solve the following case?
interface I
class A(i: I)
class C : I, A(this) // << --- 'this' is not defined in this context
In short, I want to pass the class instance to super class constructor.
Is it possible in Kotlin?
P.S.
All the answers are good and technically correct. But let's give a concrete example:
interface Pilot {
fun informAboutObstacle()
}
abstract class Car(private val pilot: Pilot) {
fun drive() {
while (true) {
// ....
if (haveObstacleDetected()) {
pilot.informAboutObstacle()
}
// ....
}
}
fun break() {
// stop the car
}
}
class AutopilotCar : Pilot, Car(this) { // For example, Tesla :)
override fun informAboutObstacle() {
break() // stop the car
}
}
This example don't look too contrived, and why can't I implement it with OOP-friendly language?
No, this is not possible on the JVM. this is only available after the super class has been initialized.
From
https://docs.oracle.com/javase/specs/jvms/se9/html/jvms-4.html#jvms-4.10.2.4
The instance initialization method (§2.9.1) for class myClass sees the new uninitialized object as its this argument in local variable 0. Before that method invokes another instance initialization method of myClass or its direct superclass on this, the only operation the method can perform on this is assigning fields declared within myClass.
So the bytecode instruction aload 0 to push this on the stack is forbidden before the super-class constructor is called. That's why it cannot be passed as an argument to the super-constructor.
Kotlin was born as a JVM language and aims for maximum interoperability with Java code and a minimum overhead of its language features. While Kotlin could have chosen to orchestrate object initialization in a different way, it would create problems in mixed Java-Kotlin class hierarchies and add significant overhead.
In the good tradition of OOP languages such as Java, C# or Swift, Kotlin doesn't allow you to leak the this reference before the call to superclass initialization has completed. In your special case you're just storing the reference, but in just a slightly different case the superclass code might try to use the received object, which at that point is still uninitialized.
As a specific example of why languages don't allow this, consider a case where A is a class from a library you use and this rule is not in effect. You pass this like you do and things work fine. Later you update the library to a newer version and it happens to add something as benign as i.toString() to its constructor. It has no idea it's actually calling an overridden method on itself. Your toString() implementation observes all its invariants broken, such as uninitialized vals.
This design suffers from other problems, not just the circular initialization dependency you are struggling with now. In a nutshell, the class A expects this:
But instead you create this:
The class A has a dependency on a collaborator object of type I. It doesn't expect itself as the collaborator. This may bring about all kinds of weird bugs. For example your C.toString() may delegate to super.toString() and A.toString() (A is the super of C) may call into I.toString(), resulting in a StackOverflowError.
I can't say from your question whether A is designed for extension, which would make the C : A part correct, but you should definitely disentangle A from I.
Related
I have the following base class and interface
open class Base {
open fun foo() { println("Base.foo") }
}
fun interface Delegate {
fun foo(): Unit
}
Now I want to override the foo method of a subclass Composed of Base using delegation, like this:
class Composed(delegate: Delegate) : Base(), Delegate by delegate
This works as I intend. Instances of Composed use the foo method of delegate, instead of the one from Base. But, I get a warning from IntelliJ:
Delegated member 'fun foo(): Unit' hides supertype override: public open fun foo(): Unit defined in Base. Please specify proper override explicitly.
Why do I get this error? What is "bad" about implicitly overriding the foo method? Are there any drawbacks I don't see? Specifying the override explicitly seems like a lot of unnecessary code, especially if I do this for multiple methods and multiple subclasses.
I think there's a warning for at least two reasons:
The behavior is ambiguous about which foo() implementation is used. There's no intuitive expectation about which foo() should win, since the superclass and the delegate are declared on the same line. I've been using Kotlin for a few years, and I couldn't have told you which behavior to expect without testing it, because this is kind of an unusual setup.
You will get possibly unexpected behavior if any of the functions in the Base superclass call foo(), because they will be calling the delegate's foo() instead of their own. It might be OK in this case, but in most cases, when you override a function, you want to call super in case there are necessary side-effects. That's not possible with the delegate.
It means that, by the time of the base class constructor execution, the properties declared or overridden in the derived class are not yet initialized. If any of those properties are used in the base class initialization logic (either directly or indirectly, through another overridden open member implementation), it may lead to incorrect behavior or a runtime failure. When designing a base class, you should therefore avoid using open members in the constructors, property initializers, and init blocks.
I was studying Inheritence from Kotlin docs, and I got stuck here. There was another post which asked a question about this, but the answers were just what the docs said in a different way.
To be clear, I understood the data flow between constructors and inheritence. What I couldn't understand was how we can use an overridden property in a base class initialization. It says
It could happen directly or indirectly
What does this actually mean? How can the base class can somehow access to the overridden property in the derived class?
Also, it said
You should therefore avoid using open members in the constructors,
property initializers and init blocks.
So how can we properly use open properties?
EDIT FOR THE COMMENT:
fun main ()
{
val d = Derived("Test2")
}
open class Base()
{
open val something:String = "Test1"
init
{
println(something) //prints null
}
}
class Derived(override val something: String): Base()
What does this actually mean? How can the base class can somehow access to the overridden property in the derived class?
This is one direct way:
abstract class Base {
abstract val something: String
init {
println(something)
}
}
class Child(override val something: String): Base()
fun main() {
Child("Test") // prints null! because the property is not initialized yet
}
This prints null, which is pretty bad for a non-nullable String property.
You should therefore avoid using open members in the constructors, property initializers and init blocks.
So how can we properly use open properties?
You can use these properties in regular methods on the base class (or in custom property getters):
abstract class Base {
abstract val something: String
fun printSomething() {
println(something)
}
}
class Child(override val something: String): Base()
fun main() {
Child("Test").printSomething() // correctly prints "Test"
}
EDIT: Here are some clarifications regarding the follow-up questions in the comments.
I couldn't quite get why the code in the init block went for the parameter in the child class constructor
I think you might be confused by Kotlin's compact syntax for the primary constructors in general, which probably makes the debugger's flow hard to understand. In the Child declaration, we actually declare many things:
the argument something passed to the Child's primary constructor
the property something on the Child class, which overrides the parent's one
the call to the parent constructor (Base())
When Child() is called, it immediately calls the Base() no-arg constructor, which runs the init block.
We didn't even delegate the base constructor with a parameter or anything, but it still went for the parameter who did the overriding
You might be mixing declarations and runtime here. Although we declare things in the Base class and in the Child class, there is only 1 instance at runtime (an instance of Child) in this example code.
So, in fact, there is only 1 property called something here (only one place in memory). If the init block accesses this property, it can only be the property of the child instance. We don't need to pass anything to the Base constructor because the init block is effectively executed with the data/fields of the Child instance.
Maybe you would be less confused if you saw the Java equivalent of this. It's obvious if you think of the abstract something as a declaration of a getter getSomething(). The child class overrides this getSomething() method and declares a private something field, the getter returns the current value of the field something. But that field is only initialized after the constructor of the parent (and the init block) finished executing.
I've bumped into this code and I'm not sure why would anyone do this. Basically the author decided for making the class constructor private so that it cannot be instantiated outside the file, and added a public method to a companion object in the class that creates a new instance of this class. What is the benefit of this approach?
This is what I found:
class Foo private constructor(private val arg1: Any) {
//more code here..
companion object {
fun newFoo(arg1: Any) = Foo(arg1 = arg1)
}
}
Why is it better than this?
class Foo(private val arg1: Any) {
//more code here..
}
There are several benefits to providing a factory method instead of a public constructor, including:
It can do lots of processing before calling the construstor. (This can be important if the superclass constructor takes parameters that need to be calculated.)
It can return cached values instead of new instances where appropriate.
It can return a subclass. (This allows you to make the top class an interface, as noted in another answer.) The exact class can differ between calls, and can even be an anonymous type.
It can have a name (as noted in another answer). This is especially important if you need multiple methods taking the same parameters. (E.g. a Point object which could be constructed from rectangular or polar co-ordinates.) However, a factory method doesn't need a specific name; if you implement the invoke() method in the companion object, you can call it in exactly the same way as a constructor.
It makes it easier to change the implementation of the class without affecting its public interface.
It also has an important drawback:
It can't be used by subclass constructors.
Factory methods seem to be less used in Kotlin than Java, perhaps due to Kotlin's simpler syntax for primary constructors and properties. But they're still worth considering — especially as Kotlin companion objects can inherit.
For much deeper info, see this article, which looks at the recommendation in Effective Java and how it applies to Kotlin.
If you want to change Foo into an interface in the future the code based on the method will keep working, since you can return a concrete class which still implements Foo, unlike the constructor which no longer exists.
An example specific to android is, that Fragments should be constructed with an empty constructed, and any data you'd like to pass through to them should be put in a bundle.
We can create a static/companion function, which takes in the arguments we need for that fragment, and this method would construct the fragment using the empty constructor and pass in the data using a bundle.
There are many useful cases, for example what Kiskae described. Another good one would be to be able to "give your constructors names":
class Foo<S: Any, T: Any> private constructor(private val a: S, private val b: T) {
//more code here...
companion object {
fun <S: Any> createForPurposeX(a: S) = Foo(a = a, b = "Default value")
fun createForPurposeY() = Foo(a = 1, b = 2)
}
}
Call site:
Foo.createForPurposeX("Hey")
Foo.createForPurposeY()
Note: You should use generic types instead of Any.
I want to override toString() in an interface and have objects that implement that interface to default to using that method (eg: not shadowed)
interface SingletonObjectTrait {
fun toString(): String = this.javaClass.simpleName
}
Is there a straightforward way to define such an interface, preferably with minimal configuration at implementation
object MyEvent: SomeEventLogic(), SomeEventType, SingletonObjectTrait
class SomeEventLogic {}
interface SomeEventType {}
That's not possible, I'm afraid.
Method implementations in interfaces work much like default methods in Java: they're used only if the implementing class doesn't have an implementation already. But every class already inherits toString() from Any, so the default would never be used.
In fact, the compiler has a specific error for this — if you try to implement toString() in an interface, it says:
An interface may not implement a method of 'Any'
I can't see a good way around this.
As Erik says, one option is to change the interface to an abstract class; but of course that's not viable if any implementations already extend another class.
Another option might be to implement a different method in the interface, and in the comments instruct implementing classes to override toString() and call that method. Not automatic, but less work for implementers, and less repetition.
There isn't a great way to do this other than using maybe an annotation processor to add the missing override at compile time (by adding an annotation to the interface that you detect and you generate the overrides in the implementation class). This would work, but may be biting off more than you want to, and is likely out of scope of an answer here on Stack Overflow (seek "how to write an annotation processor for Java or Kotlin" and "KAPT" for the one engine that supports Kotlin processors).
Back to your code and why it will not work as-is:
An interface cannot have a method with a signature that matches any of the methods in Any class. And if you try to override one of them you will get a compilation error.
An interface may not implement a method of 'Any'
The minimal code to do something like you want is:
interface SingletonObjectTrait {
fun asString(): String = this.javaClass.simpleName
}
open class SomeEventLogic {}
interface SomeEventType {}
object MyEvent: SomeEventLogic(), SomeEventType, SingletonObjectTrait {
override fun toString(): String = asString() // hope they don't forget to call this!
}
There is no guarantee the implementer will call the trait asString() function but at least you can share the functionality with a one-liner in the implementing class.
I'm attempting to create a class in Swift 3 to implement a Cordova plugin. I have this building and running, but the application crashes whenever any properties of the class are accessed. I've tried two ways of initializing the class:
#objc(DSFMediaCentre)
class DSFMediaCentre : CDVPlugin
{
var players = [UUID:DSFPlayerHandler] ();
...
}
and
#objc(DSFMediaCentre)
class DSFMediaCentre : CDVPlugin
{
var players :[UUID:DSFPlayerHandler];
override init () {
players = [:];
}
...
}
However, when my players property is used, the result is a EXC_BAD_ACCESS exception, with an address that looks like a null pointer dereference.
The object is being created by Objective C code, which is a language I have no familiarity with at all, but I think this is the line that creates it:
obj = [[NSClassFromString(className)alloc] initWithWebViewEngine:_webViewEngine];
The CDVPlugin class contains a comment stating that initWithWebViewEngine should not be overridden (and indeed I do not seem to be able to override this method, because while it is declared in the CDVPlugin.m file, it isn't mentioned in CDVPlugin.h, so the Swift compiler doesn't seem to know about it), but rather initialization code should be placed in a method called pluginInitialize instead. However, if I do that I get a compiler error ("Class DSFMediaCentre has no initializers").
Furthermore, if I put my init() method back in and set it to call pluginInitialize(), like this:
override init () {
super.init(); // necessary otherwise next line is an error
pluginInitialize();
}
override func pluginInitialize() {
players = [:];
}
the error then changes to "Property 'self.players' not initialized at super.init call".
How do I make this class initialize correctly?
You have a mismatch between the strict initialization system required by the language and the procedure used by the framework you're working with.
Swift demands that a) properties be initialized as part of object construction, and b) that construction be chained to the type's supertype. But the CDVPlugin type is doing the construction on your behalf; you don't have the ability to customize it. (This makes more sense in ObjC, because it doesn't have the same compile-time restrictions as Swift.)
The situation is similar to unpacking an object from a nib file. In that case too, because it's the nib loading system that's constructing your object, you don't have the ability to customize the initializer. Your type will always be constructed by init(coder:). In a certain sense, your initialization point moves further down, to awakeFromNib(), and among other things, that forces outlets to other objects in the archive to be declared as optional, usually implicitly unwrapped.
The same solution should avail you here. You should consider pluginInitialize() to be your initialization point. The language then requires that properties be optional, since they are not filled at its initialization point. Therefore, make the property an IUO:
#objc(DSFMediaCentre)
class DSFMediaCentre : CDVPlugin
{
var players :[UUID:DSFPlayerHandler]!
override func pluginInitialize() {
players = [:];
}
}
and all should be well.
The other solution is to use lazy keyword
lazy var players :[UUID:DSFPlayerHandler] = [:]
So, you don't need to initialize players in initializer but still make sure players always non-nulable