Are Kotlin data types built off primitive or non-p

问题:

I am new to Kotlin and was playing around with the data types. I took an Int type and then tried to cast it as a Double by saying num as Double, a call that is valid in java (non syntactically but you get the point). However, this failed, saying that Int cannot be cast to Double. I am assuming this is because it is built off the Integer class rather than the raw int data type. Am I correct, and what is the most efficient way to cast values? There is a .toDouble() function, but this seems inefficient and unwieldy.

回答1:

I took an Int type and then tried to cast it as a Double by saying num as Double <...> However, this failed, saying that Int cannot be cast to Double. I am assuming this is because it is built off the Integer class rather than the raw int data type.

No, and there are two important points to note:

• Kotlin positions its numeric types (Int, Long, Double etc.) as not being nested into each other, there is no subtyping relationship between these types. That's why the cast intNum as Double does not succeed in Kotlin. That's also why there's no implicit conversions between these types. Instead, the numeric conversion is done with the corresponding functions (e.g. .toDouble())

• The numeric type usages in Kotlin are compiled into JVM primitives where possible. Some usages require boxed types (e.g. a nullable Int? requires boxing, and so does a generic type implementation with an Int as a type argument), but the compiler decides whether they are necessary for each case.

<...> What is the most efficient way to cast values? There is a .toDouble() function, but this seems inefficient and unwieldy.

The most efficient way is to use the numeric conversion functions like .toDouble(). In fact, these functions are intrinsified, and there is no function call overhead when you use them. They are compiled closely to what javac would produce for a Java numeric cast or an implicit conversion. You can inspect the bytecode that the Kotlin compiler produces to find out what it's under the hood and whether a specific conversion introduces any overhead.

See also: an answer to a similar question, (link)

回答2:

This is because Kotlin does not work like Java in widening numbers.

There are no implicit widening conversions for numbers in Kotlin. for example, you can write something in Java as below:

int a = 1;
double b = a;

However, you can't write something in Kotlin. for example:

val a:Int = 1  
//             v--- can't be widening
val b:Double = a

This is because everything in Kotlin is object, there is no primitive types, so you should convert an Int to a Double explicitly, for example:

val a:Int = 1  
//               v--- convert it explicitly by `toDouble()` method
val b:Double = a.toDouble()