You should start by defining what a tree is (for the domain), this is best done by defining the interface first. Not all trees structures are modifyable, being able to add and remove nodes should be an optional feature, so we make an extra interface for that.
There's no need to create node objects which hold the values, in fact I see this as a major design flaw and overhead in most tree implementations. If you look at Swing, the TreeModel is free of node classes (only DefaultTreeModel makes use of TreeNode), as they are not really needed.
public interface Tree <N extends Serializable> extends Serializable {
List<N> getRoots ();
N getParent (N node);
List<N> getChildren (N node);
}
Mutable tree structure (allows to add and remove nodes):
public interface MutableTree <N extends Serializable> extends Tree<N> {
boolean add (N parent, N node);
boolean remove (N node, boolean cascade);
}
Given these interfaces, code that uses trees doesn't have to care much about how the tree is implemented. This allows you to use generic implementations as well as specialized ones, where you realize the tree by delegating functions to another API.
Example: file tree structure
public class FileTree implements Tree<File> {
@Override
public List<File> getRoots() {
return Arrays.stream(File.listRoots()).collect(Collectors.toList());
}
@Override
public File getParent(File node) {
return node.getParentFile();
}
@Override
public List<File> getChildren(File node) {
if (node.isDirectory()) {
File[] children = node.listFiles();
if (children != null) {
return Arrays.stream(children).collect(Collectors.toList());
}
}
return Collections.emptyList();
}
}
Example: generic tree structure (based on parent/child relations):
public class MappedTreeStructure<N extends Serializable> implements MutableTree<N> {
public static void main(String[] args) {
MutableTree<String> tree = new MappedTreeStructure<>();
tree.add("A", "B");
tree.add("A", "C");
tree.add("C", "D");
tree.add("E", "A");
System.out.println(tree);
}
private final Map<N, N> nodeParent = new HashMap<>();
private final LinkedHashSet<N> nodeList = new LinkedHashSet<>();
private void checkNotNull(N node, String parameterName) {
if (node == null)
throw new IllegalArgumentException(parameterName + " must not be null");
}
@Override
public boolean add(N parent, N node) {
checkNotNull(parent, "parent");
checkNotNull(node, "node");
// check for cycles
N current = parent;
do {
if (node.equals(current)) {
throw new IllegalArgumentException(" node must not be the same or an ancestor of the parent");
}
} while ((current = getParent(current)) != null);
boolean added = nodeList.add(node);
nodeList.add(parent);
nodeParent.put(node, parent);
return added;
}
@Override
public boolean remove(N node, boolean cascade) {
checkNotNull(node, "node");
if (!nodeList.contains(node)) {
return false;
}
if (cascade) {
for (N child : getChildren(node)) {
remove(child, true);
}
} else {
for (N child : getChildren(node)) {
nodeParent.remove(child);
}
}
nodeList.remove(node);
return true;
}
@Override
public List<N> getRoots() {
return getChildren(null);
}
@Override
public N getParent(N node) {
checkNotNull(node, "node");
return nodeParent.get(node);
}
@Override
public List<N> getChildren(N node) {
List<N> children = new LinkedList<>();
for (N n : nodeList) {
N parent = nodeParent.get(n);
if (node == null && parent == null) {
children.add(n);
} else if (node != null && parent != null && parent.equals(node)) {
children.add(n);
}
}
return children;
}
@Override
public String toString() {
StringBuilder builder = new StringBuilder();
dumpNodeStructure(builder, null, "- ");
return builder.toString();
}
private void dumpNodeStructure(StringBuilder builder, N node, String prefix) {
if (node != null) {
builder.append(prefix);
builder.append(node.toString());
builder.append('\n');
prefix = " " + prefix;
}
for (N child : getChildren(node)) {
dumpNodeStructure(builder, child, prefix);
}
}
}
