At the root of Kotlin’s Collection API lies the
Iterator interface, similar to Java’s.
But the similitude stops after that.
java.util.ListIterator features are broken down into different contracts:
ListIteratorto move the iterator index forward and backward
MutableIteratorto remove content from the iterator
MutableListIteratorinherits from the 2 interfaces above to mimic the entire contract of
Collection, List and Set
The hierarchy of collections in Kotlin are very similar as in Java:
(I won’t detail maps, but they follow the same design).
The only, but huge, difference is that it’s divided between mutable and immutable types.
Mutable types have methods to change their contents (_e.g.
add() and `set()), while immutable types don’t.
Of course, the hierarchy is a bit more detailed compared to Java, but that’s expected from a language that benefits from its parent’s experience.
IMHO, the important bit about Kotlin collections is not their hierarchy - though it’s important to understand about the difference between mutable and immutable.
As Java developers know, there’s no such things as out-of-the-box immutable collection types in Java.
When an immutable collection is required, the mutable collection must be wrapped into an unmodifiable type via a call to the relevant
But unmodifiable types are not
public, they are
types returned are generic ones (
It means they implement all methods of the standard collections.
Immutability comes from the mutable-related methods throwing exceptions:
at compile time, there’s no way to differentiate between a mutable and an immutable collection.
On the opposite, Kotlin offers a clean separation between mutable and immutable types. It also provides dedicated functions to create objects of the relevant type:
As opposed to Scala, Kotlin doesn’t implement its own collection types, it reuses those from Java. That means that even when the compile-time type is immutable, the runtime type is always mutable. The downside is that it’s possible to change the collection elements by casting it to the correct runtime type. IMHO, this is no more severe than what allows standard reflection. There are several advantages, though:
- Faster time-to-market
- Java collection types benefited from years of improvement
- The underlying implementation can be changed in the future with full backward compatibility