I've created 2 kotlin methods: one to check a type and another to cast an object. They look like:
fun Any?.isOfType(type: Class<*>): Boolean{
return type.isInstance(this)
// return `this is T` does NOT work.
}
and
fun <T> Any?.castToType(): T {
return this as T
// Works, albeit with a warning.
}
I've read some posts on generics and erasures, but I can't get over what seems to be a discrepancy.
Why is it that checking for the type of an object cannot be done with generics, but casting to a generic can?
The question is why:
fun <T> Any?.castToType() = this as T // compiles with warning
"hello".castToType<Int>()
"works" but this won't even compile:
fun <T> Any?.isOfType() = this is T // won't compile
"hello".isOfType<Int>()
Actually both don't really work. In both cases the type is erased at runtime. So why does one compile and the other doesn't?
this is T cannot work at runtime since the type of T is unknown and thus the compiler has to reject it.
this as T on the other hand might work:
"hello".castToType<Int>() // no runtime error but NOP
"hello".castToType<Int>().minus(1) // throws ClassCastException
2.0.castToType<Int>().minus(1) // no runtime error, returns 1
In some cases it works, in others it throws an exception. Now every unchecked cast can either succeed or lead to runtime exceptions (with or without generic types) so it makes sense to show a warning instead of a compile error.
Summary
unchecked casts with generic types are no different from unchecked casts without generic types, they are dangerous but a warning is sufficient
type checks with generic types on the other hand are impossible at runtime
Addendum
The official documentation explains type erasure and why is-checks with type arguments can't succeed at runtime:
At runtime, the instances of generic types do not hold any information about their actual type arguments. The type information is said to be erased. For example, the instances of Foo and Foo<Baz?> are erased to just Foo<*>.
Due to the type erasure, there is no general way to check whether an instance of a generic type was created with certain type arguments at runtime, and the compiler prohibits such is-checks such as ints is List or list is T (type parameter)
(https://kotlinlang.org/docs/generics.html#type-erasure)
In my own words: I can't check whether A is B if I don't know what B is. If B is a class I can check against an instance of that class (that's why type.isInstance(this) works) but if B is a generic type, the runtime has no information on it (it was erased by the compiler).
This isn't about casting vs checking; it's about using generics vs class objects.
The second example is generic; it uses T as a type parameter. Unfortunately, because generics are implemented using type erasure, this means that the type isn't available at runtime (because it has been erased, and replaced by the relevant upper bound — Any? in this case). This is why operations such as type checking or casting to a type parameter can be unsafe and give compilation warnings.
The first example, though, doesn't use a type parameter; instead, it uses a parameter which is called type, but is a Class object, representing a particular class. This is a value which is provided at runtime, just like any other method parameter, and so you can call methods such as cast() and isInstance() to handle some type issues at runtime. However, they're closely related to reflection, and have some of the same disadvantages, such as fragility, ugly code, and limited compile-time checks.
(Kotlin code often uses KClass objects instead of Java Class objects, but the principle is the same.)
It may be worth highlighting the difference between class and type, which are related but subtly different. For example, String is both a class and a type, while String? is another type derived from the same class. LinkedList is a class, but not a type (because it needs a type parameter); LinkedList<Int> is a type.
Types can of course be derived from interfaces as well as from classes, e.g. Runnable, or MutableList<Int>.
This is relevant to the question, because generics use type parameters, while Class objects represent classes.
Related
To get the class definition to be used for example for json deserialization the following can be used in Kotlin:
Map::class.java
A example usage is the following:
val map = mapper.readValue(json, Map::class.java)
But now how to have the generic type definition?
Something like this does not compile:
val map = mapper.readValue(decodedString, Map<String, String>::class.java)
So my question is: What is the generic equivalent to *::class.java
Class<T> (in Java) or KClass<T> (in Kotlin) can only represent classes, not all types. If the API you're using only uses Class<T> or KClass<T>, it simply doesn't support generic types (at least in those functions).
Instead, KType (or Type in Java) is the proper type to use to represent the complete type information including generics. You could use it this way:
val myMapType: KType = typeOf<Map<String,String>>()
Unfortunately, KType doesn't have a type parameter (it's not KType<T>), and that makes it impossible to use for compile-time type checking: you can't have the equivalent of fun deserialize(Input, KClass<T>): T using KType instead of KClass, because you can't define the T for the return type by using only a KType argument.
There are several tricks to work around this:
In both Java and Kotlin, one of the ways is to get this information through inheritance by providing a generic superclass and inheriting from it.
In general, serialization APIs (especially the deserializing part) provide workarounds using this, such as Jackson's TypeReference or Gson's TypeToken. It's basically their version of Type but with a type parameter to have some compile-time type safety.
In Kotlin, there is sometimes another way depending on the situation: making use of reified type parameters. Using inline functions, the compiler can know more information at compile time about the type parameters by replacing them with the actual inferred type at the call site when inlining the function's body. This allows things like T::class in the inline function's body. This is how you can get functions like typeOf to get a KType.
Some Kotlin-specific APIs of deserialization libraries use inline functions to remove the hassle from the user, and get type information directly. This is what jackson-module-kotlin does by providing an inline readValue extension without a Class argument, which reifies the type parameter to get the target type information
Why does Kotlin in one case infer type returned from Java to be nullable and in another case it is can be either, nullable or non-nullable?
I've checked both HashMap.get and JsonNode.get and I could not identify any #NotNull-like annotations neither in calsses nor anywhere in inheritance chain. What makes Kotlin treating those 2 calls differently?
I have read documentation https://kotlinlang.org/docs/java-interop.html#null-safety-and-platform-types but it explanation use "Platform Types" without explaining what those are and it does not explain differences in behavior anyway.
import com.fasterxml.jackson.databind.JsonNode
private fun docType(node: JsonNode, map: java.util.HashMap<String,String>) {
val x: JsonNode = node.get("doc_type") // DOES compile and can throw NPE at runtime
val y: JsonNode? = node.get("doc_type") // DOES compile and Kotlin's type system will force you to check for null
val z: String = map.get("a") // ERROR: Type mismatch: inferred type is String? but String was expected
}
Kotlin provides seamless interoperability with Java, without compromising its own null-safety... almost. One exception is that Kotlin assumes that all types that are defined in Java are not-null.
To understand, let's look at JsonNode.get()
Platform types
public JsonNode get(String fieldName) { return null; }
Note that JsonNode is defined in Java, and is a therefore 'platform type' - and Kotlin does not 'translate' it to JsonNode?, even though that would be technically correct (because in Java all types are nullable).
When calling Java from Kotlin, for convenience it's assumed that the platform type is non-nullable. If this wasn't the case, you would always have to check that any instance of any platform type is not null.
So, to answer your question about what a 'platform type' is, it's a term that means
some type that is defined in an external target language,
you can't mention it explicitly in Kotlin code (but there's probably a synonymous Kotlin equivalent),
and we're going to assume that it's non-nullable for convenience.
Also the notation is <type>!, for example String! - which we can take to mean String or String?
Nullability annotations
The closest Java equivalent of Kotlin's nullable ? symbol are nullability annotations, which the Kotlin compiler can parse and take into account. However, none are used on JsonNode methods. And so Kotlin will quite happily assume that node.get("") will return JsonNode, not JsonNode?.
As you noted, there are none defined for HashMap.get(...).
So how does Kotlin know that map.get("a") returns a nullable type?
Type inference
Type inference can't help. The (Java) method signature
public V get(Object key) {
//...
}
indicates that a HashMap<String, String> should return String, not String?. Something else must be going on...
Mapped types
For most Java types, Kotlin will just use the definition as provided. But for some, Kotlin decides to treat them specially, and completely replace the Java definition with its own version.
You can see the list of mapped types in the docs. And while HashMap isn't in there, Map is. And so, when we're writing Kotlin code, HashMap doesn't inherit from java.util.Map - because it's mapped to kotlin.collections.Map
Aside: in fact if you try and use java.util.Map you'll get a warning
So if we look at the code for the get function that kotlin.collections.Map defines, we can see that it returns a nullable value type
/**
* Returns the value corresponding to the given [key], or `null` if such a key is not present in the map.
*/
public operator fun get(key: K): V?
And so the Kotlin compiler can look at HashMap.get(...) and deduce that, because it's implementing kotlin.collections.Map.get(...), the returned value must be a nullable value, which in our case is String?.
Workaround: External annotations
For whatever reason, Jackson doesn't use the nullability annotations that would solve this problem. Fortunately IntelliJ provides a workaround that, while not as strict, will provide helpful warnings: external annotations.
Once I follow the instructions...
Alt+Enter → 'Annotate method...'
Select 'Nullable' annotation
Save annotations.xml
Now node.get("") will show an warning.
This annotation isn't visible to the Kotlin compiler, so it can only be a warning - not a compilation error.
java.util.HashMap.get implements the interface method java.util.Map.get. Kotlin maps some Java types to its own types internally. The full table of these mappings is available on the website. In our particular case, we see that java.util.Map gets mapped internally to kotlin.collections.Map, whose get function looks like
abstract operator fun get(key: K): V?
So as far as Kotlin is concerned, java.util.Map is just a funny name for kotlin.collections.Map, and all of the methods on java.util.Map actually have the signatures of the corresponding ones from kotlin.collections.Map (which are basically the same except with correct null annotations).
So while the first two node.get calls are Java calls and return platform types, the third one (as far as Kotlin is concerned) is actually calling a method Kotlin understands: namely, get from its own Map type. And that type has an explicit nullability annotation already available, so Kotlin can confidently say that that value can be null and needs to be checked.
I've got a C++/CLI layer that I've been using successfully for a long time. But I just discovered something that makes me think I need to relearn some stuff.
When my C++/CLI functions receive an instance of any managed class, they use the "hat" operator ('^') and when they receive an instance of a managed struct, they do not. I thought this was how I was supposed to write it.
To illustrate as blandly as I can
using Point = System::Windows::Point;
public ref class CppCliClass
{
String^ ReturnText(String^ text) { return text; } // Hat operator for class
Point ReturnStruct(Point pt) { return pt; } // No hat operator for struct
};
I thought this was required. It certainly works. But just today I discovered that CancellationToken is a struct, not a class. My code accepts it with a hat. I thought it was a class when I wrote it. And this code works just fine. My cancellations are honored in the C++/CLI layer.
void DoSomethingWithCancellation(CancellationToken^ token)
{
// Code that uses the token. It works just fine
}
So apparently I can choose either method.
But then what is the difference between passing in a struct by value (as I've done with every other struct type I use, like Point) and by reference (as I'm doing with CancellationToken?). Is there a difference?
^ for reference types and without for value types matches C#, but C++/CLI does give you more flexibility:
Reference type without ^ is called "stack semantics" and automatically tries to call IDisposable::Dispose on the object at the end of the variable's lifetime. It's like a C# using block, except more user-friendly. In particular:
The syntax can be used whether the type implements IDisposable or not. In C#, you can only write a using block if the type can be proved, at compile time, to implement IDisposable. C++/CLI scoped resource management works fine in generic and polymorphic cases, where some of the objects do and some do not implement IDisposable.
The syntax can be used for class members, and automatically implements IDisposable on the containing class. C# using blocks only work on local scopes.
Value types used with ^ are boxed, but with the exact type tracked statically. You'll get errors if a boxed value of a different type is passed in.
I have a generically typed class Builder<T> that takes a constructor argument Class<T> so I can keep the type around. This is a class that I use a lot in java code so I don't want to change the signature.
When I try to use the constructor like this:
Builder<List<Number>>(List<Number>::class)
I get an error: "Only classes are allowed on the left hand side of a class literal"
Any way to resolve this?
I can't change the constructor for Builder, too many java classes rely upon it.
I understand the whole type erasure issue, I really just want to make the compiler happy.
Due to generic type erasure List class has a single implementation for all its generic instantiations. You can only get a class corresponding to List<*> type, and thus create only Builder<List<*>>.
That builder instance is suitable for building a list of something. And again due to type erasure what that something is you can decide by yourself with the help of unchecked casts:
Builder(List::class.java) as Builder<List<Number>>
Builder(List::class.java as Class<List<Number>>)
Another approach is to create inline reified helper function:
inline fun <reified T : Any> Builder() = Builder(T::class.java)
and use it the following way:
Builder<List<Number>>()
The solution is to use reified generics in couple with super class tokens.
Please refer to this question for the method explained. Constructors in Kotlin don't support reified generics, but you can use TypeReference described there to write a builder factory function which will retain actual generic parameters at runtime:
inline <reified T: Any> fun builder(): Builder<T> {
val type = object : TypeReference<T>() {}.type
return Builder(type)
}
Then inside Builder you can check if type is ParameterizedType, and if it is, type.actualTypeArguments will contain the actual generic parameters.
For example, builder<List<Number>>() will retain the information about Number at runtime.
The limitation of this approach is that you cannot use non-reified generic as a reified type parameter because the type must be known at compile-time.
I know there are various capabilities in Java with reflection.
For example:
Class<?> clazz = Class.forName("java.util.Date");
Object ins = clazz.newInstance();
I wonder if I could pass class dynamicaly in some method declaration in <> tags (or there is other way to do it if it must be fixed). I would like to change that class declaration dynamicaly; because I would like to write generic method for all types of classes.
In there, I have this:
List<Country>
Can I write it something diffrent with reflection? For example can it be somehow be achieved to pass class as parameter (or how else should be this done):
List<ins>
? I would appreciate examples.
This cannot be done because generics are a compile time feature. Once code is compiled, the only place where generics are exists are at method signatures, and they are only used for compiling new code.
When working with reflection, you are basicly working with raw types, and need to code according to that, that means, you can cast the returned result of newInstance() to the list type your need, for example:
List<Country> ins = (List<Country>)clazz.newInstance();
This is a safe operation to do, because you know at that point its empty, and isn't passed to any outside code.
I don't think this is possible. Generics in Java are implemented in a way that prohibits runtime access.
Generics are there so that the compiler can verify correct typing, but are no longer present at runtime (this is called "type erasure"). Reflection deals with the runtime representation of types only. As far as I know the only case where reflection has to deal with generics is to find out "fixed" type parameters of sub-classes, e.g. when you have class Bar<T> and class Foo extends Bar<String>, you can find out that the T of Bar is fixed to String in Foo using reflection. However, this is information found in the class file, too. Except that, reflection can only see or create raw-types.