Current Situation

Java has supported immutable record types since Java 14, and Kotlin's data classes can easily implement immutable classes, even supporting copy functions. However, they are designed for simple immutable objects, and have problems dealing with complex deep data structure.

The reason why deep, complex immutable objects are difficult to handle is not how to create a brand new data structure from scratch, which is very simple for any programming language. The real difficulty is wanting to make some local adjustments to an existing data structure and build a new data structure. This is very difficult for current Java and Kotlin, please see

First define an immutable tree node

Java

TreeNode.java

record TreeNode(
    String name,
    List<TreeNode> childNodes
) {}

Kotlin

TreeNode.java

data class TreeNode(
    val name: String,
    val childNodes: List<TreeNode>  
)

Prepare an old object

Java

var oldTreeNode = ...blabla... 

Kotlin

val oldTreeNode = ...blabla...

Let's implement three data change operations in order from simple to complex to create a brand new object

1. Change the name property of the root node

   Java

   TreeNode newTreeNode = new TreeNode(
       "Hello", // Set root node name  
       oldTreeNode.childNodes()
   );

   Kotlin

   val newTreeNode = oldTreeNode.copy(
       name = "Hello" // Set root node name
   );

   This case is very simple, and the problems of Java/Kotlin are not obvious yet.

2. Change the name property of the first level child node

   Breadcrumb conditions:

   • Position of first level child node: pos1
   Java

   TreeNode newTreeNode = new TreeNode(
       oldTreeNode.name(),
       IntStream
           .range(0, oldTreeNode.childNodes().size())
           .mapToObj(index1 -> {
               TreeNode oldChild1 = oldTreeNode.childNodes().get(index1);
               if (index1 != pos1) {
                   return oldChild1;
               }
               return new TreeNode(
                   "Hello", // Set name of first level child node  
                   oldChild1.childNodes()
               );
           })
           .toList()
   );

   Kotlin

   val newTreeNode = oldTreeNode.copy(
       childNodes = oldTreeNode
           .childNodes
           .mapIndexed { index1, child1 ->
               if (index1 == pos1) {
                   child1.copy(
                       name = "Hello" // Set name of first level child node   
                   )
               } else {
                   child1
               }
           }
   )

3. Change the name property of the second level child node

   Breadcrumb conditions:

   • Position of first level child node: pos1
   • Position of second level child node: pos2
   Java

   TreeNode newTreeNode = new TreeNode(
       oldTreeNode.name(),
       IntStream
           .range(0, oldTreeNode.childNodes().size())
           .mapToObj(index1 -> {
               TreeNode oldChild1 = oldTreeNode.childNodes().get(index1);
               if (index1 != pos1) {
                   return oldChild1;
               }
               return new TreeNode(
                   oldChild1.name(),
                   IntStream
                       .range(0, oldChild1.childNodes().size())
                       .mapToObj(index2 -> {
                           TreeNode oldChild2 = oldChild1.childNodes().get(index2);
                           if (index2 != pos2) {
                               return oldChild2;
                           } else {
                               return new TreeNode(
                                   "Hello", // Set name of second level child node
                                   oldChild2.childNodes()
                               );
                           }
                       })
                       .toList()
               );
           })
           .toList()
   );

   Kotlin

   val newTreeNode = oldTreeNode.copy(
       childNodes = oldTreeNode
           .childNodes
           .mapIndexed { index1, child1 ->
               if (index1 == pos1) {
                   child1.copy(
                       childNodes = child1
                           .childNodes
                           .mapIndexed { index2, child2 ->
                               if (index2 == pos2) {
                                   child2.copy(
                                       name = "Hello" // Set name of second level child node
                                   )
                               } else {
                                   child2
                               }
                           }
                   )
               } else {
                   child1
               }
           }
   )

info

It can be seen that as long as the object tree has some depth, creating a new immutable object based on another immutable object (i.e. secondary "modification") will be a nightmare.