Null Safety Best Practices

Prefer Non-Null by Default

Design functions and classes so most properties and parameters are non-nullable, reserving ? only for cases where absence is truly meaningful, which keeps the rest of your code free of constant null checks.

Example: Prefer Non-Null by Default

data class Product(val name: String, val price: Double, val discountCode: String? = null)

fun main() {
    val product = Product("Book", 15.0)
    println("${product.name}: $${product.price}, discount=${product.discountCode ?: "none"}")
}

Validate at the Boundary

When data comes from outside your program, such as parsed text or a network response, convert it into a proper non-null domain type as early as possible instead of passing raw nullable values deep into your logic.

Example: Validate at the Boundary

fun parseAge(input: String?): Int {
    return input?.toIntOrNull() ?: error("Invalid or missing age: $input")
}

fun main() {
    println("Age: ${parseAge("42")}")
}

Avoid Nullable Collections

Returning an empty list instead of a nullable list means callers can safely iterate or check size without an extra null check, simplifying calling code considerably.

Note: An empty list is far easier to work with than null for representing 'no results'.

Example: Avoid Nullable Collections

fun findMatches(query: String, items: List<String>): List<String> {
    return items.filter { it.contains(query) }
}

fun main() {
    val results = findMatches("kt", listOf("main.kt", "readme.md"))
    println("Found ${results.size} matches: $results")
}

Reserve !! for Truly Impossible Nulls

Treat !! as a signal that something is architecturally wrong if it ever actually throws; use it only in the rare case where the surrounding logic guarantees non-null, and prefer safer operators everywhere else.

Example: Reserve !! for Truly Impossible Nulls

fun main() {
    val cache = HashMap<String, String>()
    cache["greeting"] = "Hello"
    if ("greeting" in cache) {
        println(cache["greeting"]!!)
    }
}