I do not fully understand how variance in Generics work. In the code below the classes are as follows Any -> Mammals -> Cats. Any is the supertype, there is a parameter called from in the copy function
From what I understand about the out and in keywords, out allows reference to any of it's subtype, can only be produced not consumed.
in allows reference to any of it's supertype, can only be consumed not produced.
However in the copytest function we are instantiating the function copy. I gave it a catlist1 argument in the from parameter. Since the parameter has an out keyword wouldn't it mean that we can only input parameters that are a subtype of catlist2?
To top of my confusion I have seen many conflicting definitions, for instance , In Kotlin, we can use the out keyword on the generic type which means we can assign this reference to any of its supertypes.
Now I am really confused could anybody guide me on how all of these works? Preferably from scratch, thanks!
class list2<ITEM>{
val data = mutableListOf<ITEM>()
fun get(n:Int):ITEM = data[n]
fun add(Item:ITEM){data.add(Item)}
}
fun <T> Copy(from: list2<out T>, to:list2<T>){
}
fun copytest(){
val catlist1 = list2<Cat>()
val catlist2 = list2<Cat>()
val mammallist = list2<Mammal>()
Copy(catlist1,mammallist)
}
I think maybe you're mixing up class-declaration-site generics and use-site generics.
Class-declaration-site generics
Defined at the class declaration site with covariant out, it is true you cannot use the generic type as the type of a function parameter for any functions in the class.
class MyList<out T>(
private val items: Array<T>
) {
fun pullRandomItem(): T { // allowed
return items.random()
}
fun addItem(item: T) { // Not allowed by compiler!
// ...
}
}
// Reason:
val cowList = MyList<Cow>(arrayOf(Cow()))
// The declaration site out covariance allows us to up-cast to a more general type.
// It makes logical sense, any cow you pull out of the original list qualifies as an animal.
val animalList: MyList<Animal> = cowList
// If it let us put an item in, though:
animalList.addItem(Horse())
// Now there's a horse in the cow list. That doesn't make logical sense
cowList.pullRandomItem() // Might return a Horse, impossible!
It is not logical to say, "I'm going to put a horse in a list that may have the requirement that all items retrieved from it must be cows."
Use-site generics
This has nothing to do with the class level restriction. It's only describing what kind of input the function gets. It is perfectly logical to say, "my function does something with a container that I'm going to pull something out of".
// Given a class with no declaration-site covariance of contravariance:
class Bag<T: Any>(var contents: T?)
// This function will take any bag of food as a parameter. Inside the function, it will
// only get things out of the bag. It won't put things in it. This makes it possible
// to pass a Bag of Chips or a Bag of Pretzels
fun eatBagContents(bagOfAnything: Bag<out Food>) {
eat(bagOfAnything.contents) // we know the contents are food so this is OK
bagOfAnything.contents = myChips // Not allowed! we don't know what kind of stuff
// this bag is permitted to contain
}
// If we didn't define the function with "out"
fun eatBagContentsAndPutInSomething(bagOfAnything: Bag<Food>) {
eat(bagOfAnything.contents) // this is fine, we know it's food
bagOfAnything.contents = myChips // this is fine, the bag can hold any kind of Food
}
// but now you cannot do this
val myBagOfPretzels: Bag<Pretzels> = Bag(somePretzels)
eatBagContentsAndPutInSomething(myBagOfPretzels) // Not allowed! This function would
// try to put chips in this pretzels-only bag.
Combining both
What could be confusing to you is if you saw an example that combines both of the above. You can have a class where T is a declaration site type, but the class has functions where there are input parameters where T is part of the definition of what parameters the function can take. For example:
abstract class ComplicatedCopier<T> {
abstract fun createCopy(item: T): T
fun createCopiesFromBagToAnother(copyFrom: Bag<out T>, copyTo: Bag<in T>) {
val originalItem = copyFrom.contents
val copiedItem = createCopy(originalItem)
copyTo.contents = copiedItem
}
}
This logically makes sense since the class generic type has no variance restriction at the declaration site. This function has one bag that it's allowed to take items out of, and one bag that it's allowed to put items into. These in and out keywords make it more permissive of what types of bags you can pass to it, but it limits what you're allowed to do with each of those bags inside the function.
Related
In Kotlin, I want to add a "namespace" to a class that has a set of related functions. Clients of my class will use that namespace to help classify what type of operation they want to do. (I know you're thinking the functions should be in different classes, problem solved. But for other reasons, it's convenient to house all the functions in a single class).
So, I might have a class Uber that contains fooInsert fooOpen fooDispose along with barInsert barTerminate and barHop. As you can see there's no common interface. Just a bunch of functions that for some reason belong in the same class. Some have an affinity with others (i.e. the fooXXX functions "belong" together, as do the "barYYY" functions).
What I've come up with is:
class Uber {
inner class FooNamespace {
fun insert(): Unit {}
fun open(): Unit {}
fun dispose(): Unit {}
}
val foo = FooNamespace()
inner class BarNamespace {
fun insert(): Unit {}
fun terminate(): Unit {}
fun hop(): Unit {}
}
val bar = BarNamespace()
}
Users of the class can do something like this:
val uber = Uber()
uber.foo.insert()
uber.bar.hop()
What I'd like is something that combines the inner class ... and val xxx = XxxNamespace() into one expression. Something like:
// This doesn't actually compile
val foo = object: inner class {
fun insert(): Unit {}
fun open(): Unit {}
fun dispose(): Unit {}
}
The problem here is that you need a properly defined type if you to want to access these members publicly.
For private properties, the syntax val foo = object { ... } is sufficient, but for publicly exposed properties these are inferred as Any and it makes them unusable.
One option is obviously to define an interface for these types, but it's even more boilerplate than what you came up with already, so I am pretty sure this won't suit your needs:
interface FooNamespace {
fun insert()
fun open()
fun dispose()
}
class Uber {
val foo = object : FooNamespace {
override fun insert(): Unit {}
override fun open(): Unit {}
override fun dispose(): Unit {}
}
}
I know you're thinking the functions should be in different classes, problem solved. But for other reasons, it's convenient to house all of the functions in a single class
I'm indeed really thinking that, and would love to hear more about what makes it so convenient to put everything in the same class :) Since the classes are inner classes, I'm assuming this has to do with accessing private state from Uber, but that could also be done by wrapping this private state into another class that's passed to foo and bar.
I believe this is not possible, at least for now.
The main technical problem here is that uber.foo.insert() is really interpreted as chaining uber.foo and then .insert(). So for this to work, uber.foo needs to have an explicitly defined type. It can't be anonymous class/object, because then there is no way to describe what is the type of uber.foo.
That being said, I've always wondered why Kotlin does not support this syntax:
val foo = object Foo {}
This is consistent with the object declaration where the name of the singleton is at the same time the name of the class. And the compiler even understands this above syntax, because it throws the error: "An object expression cannot bind a name". So Kotlin authors seem to intentionally disallow such use.
I found an issue in the YouTrack, so we can at least upvote it: https://youtrack.jetbrains.com/issue/KT-21329
I'm exploring the Substitution principal and from what I've understood about the principal is that a sub type of any super type should be passable into a function/class. Using this idea in a new section of code that I'm writing, I wanted to implement a abstract interface for a Filter like so
interface Filter {
fun filter(): Boolean
}
I would then imagine that this creates the contract for all classes that inherit this interface that they must implement the function filter and return a boolean output. Now my interpretation of this is that the input doesn't need to be specified. I would like it that way as I want a filter interface that guarantee the implementation of a filter method with a guarantee of a return type boolean. Does this concept even exists in Kotlin? I would then expect to implement this interface like so
class LocationFilter {
companion object : Filter {
override fun filter(coord1: Coordinate, coord2: Coordinate): Boolean {
TODO("Some business logic here")
}
}
}
But in reality this doesn't work. I could remove remove the filter method from the interface but that just defeats the point of the whole exercise. I have tried using varargs but again that's not resolving the issue as each override must implement varargs which is just not helpful. I know this may seem redundant, but is there a possibility to have the type of abstraction that I'm asking for? Or am I missing a point of an Interface?
Let's think about it a little. The main point of abstraction is that we can use Filter no matter what is the implementation. We don't need to know implementations, we only need to know interfaces. But how could we use Filter if we don't know what data has to be provided to filter? We would need to use LocationFilter directly which also defeats the point of creating an interface.
Your problem isn't really related to Kotlin, but to OOP in general. In most languages it is solved by generics/templates/parameterized types. It means that an interface/class is parameterized by another type. You use it in Kotlin like this:
interface Filter<in T> {
fun filter(value: T): Boolean
}
object LocationFilter : Filter<Coordinate> {
override fun filter(value: Coordinate): Boolean {
TODO()
}
}
fun acquireCoordinateFilter(): Filter<Coordinate> = LocationFilter
fun main() {
val coord: Coordinate = TODO()
val filter: Filter<Coordinate> = acquireCoordinateFilter()
val result = filter.filter(coord)
}
Filter is parameterized, meaning that we can have a filter for filtering strings (type is: Filter<String>), for filtering integers (Filter<Int>) or for filtering coordinates (Filter<Coordinate>). Then we can't use e.g. Filter<String> to filter integers.
Note that the code in main() does not use LocationFilter directly, it only knows how to acquire Filter<Coordinate>, but the specific implementation is abstracted from it.
Also note there is already a very similar interface in Java stdlib. It is called Predicate.
my interpretation of this is that the input doesn't need to be specified.
Where did you get that interpretation from?
You can see that it can't be correct, by looking at how the method would be called. You should be able to write code that works for any instance of Filter — and that can only happen if the number and type of argument(s) is specified in the interface. To use your example:
val f: Filter = someMethodReturningAFilterInstance()
val result = f.filter(coord1, coord2)
could only work if all implementations used two Coordinate parameters. If some used one String param, and others used nothing at all, then how would you call it safely?
There are a few workarounds you could use.
If every implementation takes the same number of parameters, then you could make the interface generic, with type parameter(s), e.g.:
interface Filter<T1, T2> {
fun filter(t1: T1, t2: T2): Boolean
}
Then it's up to the implementation to specify which types are needed. However, the calling code either needs to know the types of the particular implementation, or needs to be generic itself, or the interface needs to provide type bounds with in variance.
Or if you need a variable number of parameters, you could bundle them up into a single object and pass that. However, you'd probably need an interface for that type, in order to handle the different numbers and types of parameters, and/or make that type a type parameter on Filter — all of which smells pretty bad.
Ultimately, I suspect you need to think about how your interface is going to be used, and in particular how its method is going to be called. If you're only ever going to call it when the caller knows the implementation type, then there's probably no point trying to specify that method in the interface (and maybe no point having the interface at all). Or if you'll want to handle Filter instances without knowing their concrete type, then look at how you'll want to make those calls.
The whole this is wrong!
First, OOP is a declarative concept, but in your example the type Filter is just a procedure wrapped in an object. And this is completely wrong.
Why do you need this type Filter? I assume you need to get a collection filtered, so why not create a new object that accepts an existing collection and represents it filtered.
class Filtered<T>(private val origin: Iterable<T>) : Iterable<T> {
override fun iterator(): Iterator<T> {
TODO("Filter the original iterable and return it")
}
}
Then in your code, anywhere you can pass an Iterable and you want it to be filtered, you simply wrap this original iterable (any List, Array or Collection) with the class Filtered like so
acceptCollection(Filtered(listOf(1, 2, 3, 4)))
You can also pass a second argument into the Filtered and call it, for example, predicate, which is a lambda that accepts an element of the iterable and returns Boolean.
class Filtered<T>(private val origin: Iterable<T>, private val predicate: (T) -> Boolean) : Iterable<T> {
override fun iterator(): Iterator<T> {
TODO("Filter the original iterable and return it")
}
}
Then use it like:
val oddOnly = Filtered(
listOf(1, 2, 3, 4),
{ it % 2 == 1 }
)
Type Hierarchy
open class Fruit()
open class CitrusFruit : Fruit()
class Orange : CitrusFruit()
Declaration-site Variance
The Crate is used as a producer or consumer of Fruits.
Invariant class
class Crate<T>(private val elements: MutableList<T>) {
fun add(t: T) = elements.add(t) // Consumer allowed
fun last(): T = elements.last() // Producer allowed
}
Covariant classout
class Crate<out T>(private val elements: MutableList<T>) {
fun add(t: T) = elements.add(t) // Consumer not allowed: Error
fun last(): T = elements.last() // Producer allowed
}
Contravariant classin
class Crate<in T>(private val elements: MutableList<T>) {
fun add(t: T) = elements.add(t) // Consumer allowed
fun last(): T = elements.last() // Producer not allowed: Error
}
Use-site Variance
All these use-site projections are for the invariant class Crate<T> defined above.
No Projection
No subtyping allowed: Only the Crate<Fruit> can be assigned to a Crate<Fruit>.
fun main() {
val invariantCrate: Crate<Fruit> = Crate<Fruit>(mutableListOf(Fruit(), Orange()))
invariantCrate.add(Orange()) // Consumer allowed
invariantCrate.last() // Producer allowed
}
Covariant Projectionout
Subtyping allowed: Crate<CitrusFruit> can be assigned to Crate<Fruit> when CitrusFruit is a subtype of Fruit.
fun main() {
val covariantCrate: Crate<out Fruit> = Crate<CitrusFruit>(mutableListOf(Orange()))
covariantCrate.add(Orange()) // Consumer not allowed: Error
covariantCrate.last() // Producer allowed
}
Contravariant Projectionin
Subtyping allowed: Crate<CitrusFruit> can be assigned to Crate<Orange> when the CitrusFruit is a supertype of Orange.
fun main() {
val contravariantCrate: Crate<in Orange> = Crate<CitrusFruit>(mutableListOf(Orange()))
contravariantCrate.add(Orange()) // Consumer allowed
contravariantCrate.last() // Producer allowed: No Error?
}
Questions
Is my understanding and the use of type projection correct in the given example?
For contravariance: why is the last()(producer) function not allowed at declaration-site but allowed at use-site? Shouldn't the compiler show an error like it shows in the declaration-site example? Maybe I'm missing something? If the producer is allowed for contravariance only at use-site, what could be the use case for it?
I prefer detailed answers with examples but any kind input will be much appreciated.
Let's start with the use-site.
When you write
val contravariantCrate: Crate<in Orange> = ...
the right side could be a Crate<Orange>, Crate<Fruit>, Crate<Any?>, etc. So the basic rule is that any use of contravariantCrate should work if it had any of these types.
In particular, for all of them
contravariantCrate.last()
is legal (with type Orange, Fruit, and Any? respectively). So it's legal for Crate<in Orange> and has type Any?.
Similarly for covariantCrate; calling the consumer method technically is allowed, just not with Orange. The problem is that a Crate<Nothing> is a Crate<out Fruit>, and you couldn't do
val covariantCrate: Crate<Nothing> = ...
covariantCrate.add(Orange())
Instead the parameter type is the greatest common subtype of Fruit, CitrusFruit, Nothing, etc. which is Nothing. And
covariantCrate.add(TODO())
is indeed legal because the return type of TODO() is Nothing (but will give warnings about unreachable code).
Declaration-site in or out effectively say that all uses are in/out. So for a contravariant class Crate<in T>, all calls to last() return Any?. So you should just declare it with that type.
My guess is that the difference between declaration-site and use-site contravariance is that delcaration-site can be statically checked by the compiler, but when using projections there is always the original, unprojected object in existence at run-time. Therefore, it is not possible to prevent the creation of the producer methods for in projections.
When you write:
class Crate<in T>(private val elements: MutableList<T>) {
fun add(t: T) = elements.add(t) // Consumer allowed
fun last(): T = elements.last() // Producer not allowed: Error
}
The compiler can know at compile-time that no method on Crate<T> should exist that produces a T, so the definition of fun last(): T is invalid.
But when you write:
val contravariantCrate: Crate<in Orange> = Crate<CitrusFruit>(mutableListOf(Orange()))
What has actually been created is a Crate<Any?>, because generics are erased by the compiler. Although you specified that you don't care about producing an item, the generic-erased Crate object still exists with the fun last(): Any? method.
One would expect the projected method to be fun last(): Nothing, in order to give you a compiler-time error if you try to call it. Perhaps that is not possible because of the need for the object to exist, and therefore be able to return something from the last() method.
coming across a sample with a class and a function and trying to understand the koltin syntax there,
what does this IMeta by dataItem do? looked at https://kotlinlang.org/docs/reference/classes.html#classes and dont see how to use by in the derived class
why the reified is required in the inline fun <reified T> getDataItem()? If someone could give a sample to explain the reified?
class DerivedStreamItem(private val dataItem: IMeta, private val dataType: String?) :
IMeta by dataItem {
override fun getType(): String = dataType ?: dataItem.getType()
fun getData(): DerivedData? = getDataItem()
private inline fun <reified T> getDataItem(): T? = if (dataItem is T) dataItem else null
}
for the reference, copied the related defines here:
interface IMeta {
fun getType() : String
fun getUUIDId() : String
fun getDataId(): String?
}
class DerivedData : IMeta {
override fun getType(): String {
return "" // stub
}
override fun getUUIDId(): String {
return "" // stub
}
override fun getDataId(): String? {
return "" // stub
}
}
why the reified is required in the inline fun <reified T> getDataItem()? If someone could give a sample to explain the reified?
There is some good documentation on reified type parameters, but I'll try to boil it down a bit.
The reified keyword in Kotlin is used to get around the fact that the JVM uses type erasure for generic. That means at runtime whenever you refer to a generic type, the JVM has no idea what the actual type is. It is a compile-time thing only. So that T in your example... the JVM has no idea what it means (without reification, which I'll explain).
You'll notice in your example that you are also using the inline keyword. That tells Kotlin that rather than call a function when you reference it, to just insert the body of the function inline. This can be more efficient in certain situations. So, if Kotlin is already going to be copying the body of our function at compile time, why not just copy the class that T represents as well? This is where reified is used. This tells Kotlin to refer to the actual concrete type of T, and only works with inline functions.
If you were to remove the reified keyword from your example, you would get an error: "Cannot check for instance of erased type: T". By reifying this, Kotlin knows what actual type T is, letting us do this comparison (and the resulting smart cast) safely.
(Since you are asking two questions, I'm going to answer them separately)
The by keyword in Kolin is used for delegation. There are two kinds of delegation:
1) Implementation by Delegation (sometimes called Class Delegation)
This allows you to implement an interface and delegate calls to that interface to a concrete object. This is helpful if you want to extend an interface but not implement every single part of it. For example, we can extend List by delegating to it, and allowing our caller to give us an implementation of List
class ExtendedList(someList: List) : List by someList {
// Override anything from List that you need
// All other calls that would resolve to the List interface are
// delegated to someList
}
2) Property Delegation
This allows you to do similar work, but with properties. My favorite example is lazy, which lets you lazily define a property. Nothing is created until you reference the property, and the result is cached for quicker access in the future.
From the Kotlin documentation:
val lazyValue: String by lazy {
println("computed!")
"Hello"
}
I'm playing with reflection and I came out with this problem. When using bound class reference via the ::class syntax, I get a covariant KClass type:
fun <T> foo(entry: T) {
with(entry::class) {
this // is instance of KClass<out T>
}
}
As I could learn from the docs, this will return the exact type of the object, in case it is instance of a subtype of T, hence the variance modifier.
However this prevents retrieving properties declared in the T class and getting their value (which is what I'm trying to do)
fun <T> foo(entry: T) {
with(entry::class) {
for (prop in memberProperties) {
val v = prop.get(entry) //compile error: I can't consume T
}
}
}
I found that a solution is using javaClass.kotlin extension function on the object reference, to get instead the invariant type:
fun <T> foo(entry: T) {
with(entry.javaClass.kotlin) {
this // is instance of KClass<T>
}
}
This way, I get both the exact type at runtime and the possibility to consume the type.
Interestingly, if I use a supertype instead of a generic, with the latter method I still get access to the correct type, without the need of variance:
class Derived: Base()
fun foo(entry: Base) {
with(entry.javaClass.kotlin) {
println(this == Derived::class)
}
}
fun main(args: Array<String>) {
val derived = Derived()
foo(derived) // prints 'true'
}
If I got it correct, ::class is equal to calling the java getClass, which returns a variant type with a wildcard, while javaClass is a getClass with a cast to the specific type.
Still, I don't get why would I ever need a covariant KClass, when it limits me to only produce the type, given that there are other ways to access the exact class at runtime and use it freely, and I wonder if the more immediate ::class should return an invariant type by design.
The reason for covariance in bound ::class references is, the actual runtime type of an object the expression is evaluated to might differ from the declared or inferred type of the expression.
Example:
open class Base
class Derived : Base()
fun someBase(): Base = Derived()
val kClass = someBase()::class
The expression someBase() is typed as Base, but at runtime it's a Derived object that it gets evaluated to.
Typing someBase()::class as invariant KClass<Base> is simply incorrect, in fact, the actuall result of evaluating this expression is KClass<Derived>.
To solve this possible inconsistency (that would lead to broken type-safety), all bound class references are covariant: someBase()::class is KClass<out Base>, meaning that at runtime someBase() might be a subtype of Base, and therefore this might be a class token of a subtype of Base.
This is, of course, not the case with unbound class references: when you take Base::class, you know for sure that it's the class token of Base and not of some of its subtypes, so it's invariant KClass<Base>.