为什么可以在 Kotlin 的类之外编写函数？

Question

I don't understand why is possible to write a function outside a class in Kotlin? Is that a good practice?

For example, it's possible in Kotlin to write a function outside my MainActivity class:

fun hello(){}

class MainActivity : AppCompatActivity() {

    override fun onCreate(savedInstanceState: Bundle?) {
        super.onCreate(savedInstanceState)
        setContentView(R.layout.activity_main)

        hello()
    }
}

In Java, this is impossible, That's not how an object-oriented language works normally? right?

In the documentation, they talk of Local Functions for the classic function and Member Functions for the function defined inside a class or object but they don't explain when it's better to use one or the other.

Answer 1

In Java, this is impossible, That's not how an object-oriented language works normally? right?

Just stop for a second and reconsider the nature of java's static method. A class is supposed to be a blueprint for objects, describe their behavior and state. But you can call a static method without creating any instances.

How does that fit into the object-oriented picture? How does a static method "belong" to the class it's declared in?

Actually, static methods are top-level functions and the class only provides a namespace for them. Speaking strictly from the OOP perspective you bring up, they are a hack. But you got used to them over the years so you don't feel that anymore.

In contrast to that, Kotlin allows you to declare top-level functions without forcing you to couple their name to a class. Sometimes you expressly want to couple a top-level function to a class, and that's what companion objects are for in Kotlin.

Answer 2

Yes, this is good practice. Kotlin is not a purely object-oriented language, so it's not obligated to follow how "an object-oriented language works normally" (even though other object-oriented languages, such as C++, Ruby and Python, also allow top-level functions).

It's better to use a top-level function when the logic of this function does not clearly belong to any class.

Answer 3

Yes, it is a good practice to create package-level functions if the function logic is independent of properties and lifecycle of a class. Example:

• a function to convert miles per gallon to kilometers per litre is independent of any object and fits well as package-level.
• otoh, a function to cancel reservation would naturally be associated with a specifc reservation object and fits well inside such a class.

The main benefit of a package-level function is simplicity (ergo better maintainability): callers of your function don't need to declare and create an object to call the function. (If your package-level function needs to be called from Java code, this benefit is lost because the Java calling code has to use a class name that is generated by Kotlin.)

IMPORTANT: Although you don't have a class lexical scope for your function, the Single-Responsibility Principle (SRP) still applies . Do not create a Kotlin source file, say Util.kt, and bloat it up with functions that lack cohesion, that is, functions that do unrelated things.