org.apache.commons.configuration2.tree.NodeTreeWalker Maven / Gradle / Ivy
Show all versions of commons-configuration2 Show documentation
/*
* Licensed to the Apache Software Foundation (ASF) under one or more
* contributor license agreements. See the NOTICE file distributed with
* this work for additional information regarding copyright ownership.
* The ASF licenses this file to You under the Apache License, Version 2.0
* (the "License"); you may not use this file except in compliance with
* the License. You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package org.apache.commons.configuration2.tree;
import java.util.LinkedList;
import java.util.List;
/**
*
* A class providing different algorithms for traversing a hierarchy of configuration nodes.
*
*
* The methods provided by this class accept a {@link ConfigurationNodeVisitor} and visit all nodes in a hierarchy
* starting from a given root node. Because a {@link NodeHandler} has to be passed in, too, arbitrary types of nodes can
* be processed. The {@code walk()} methods differ in the order in which nodes are visited. Details can be found in the
* method documentation.
*
*
* An instance of this class does not define any state; therefore, it can be shared and used concurrently. The
* {@code INSTANCE} member field can be used for accessing a default instance. If desired (e.g. for testing purposes),
* new instances can be created.
*
*
* @since 2.0
*/
public class NodeTreeWalker {
/** The default instance of this class. */
public static final NodeTreeWalker INSTANCE = new NodeTreeWalker();
/**
* Helper method for performing a BFS traversal. Implementation node: This method organizes the nodes to be visited in
* structures on the heap. Therefore, it can deal with larger structures than would be the case in a recursive approach
* (where the stack size limits the size of the structures which can be traversed).
*
* @param root the root node to be navigated
* @param visitor the visitor
* @param handler the handler
* @param the type of the nodes involved
*/
private static void bfs(final T root, final ConfigurationNodeVisitor visitor, final NodeHandler handler) {
final List pendingNodes = new LinkedList<>();
pendingNodes.add(root);
boolean cancel = false;
while (!pendingNodes.isEmpty() && !cancel) {
final T node = pendingNodes.remove(0);
visitor.visitBeforeChildren(node, handler);
cancel = visitor.terminate();
pendingNodes.addAll(handler.getChildren(node));
}
}
/**
* Helper method for checking the parameters for the walk() methods. If mandatory parameters are missing, an exception
* is thrown. The return value indicates whether an operation can be performed.
*
* @param root the root node
* @param visitor the visitor
* @param handler the handler
* @param the type of the nodes involved
* @return true if a walk operation can be performed, false otherwise
* @throws IllegalArgumentException if a required parameter is missing
*/
private static boolean checkParameters(final T root, final ConfigurationNodeVisitor visitor, final NodeHandler handler) {
if (visitor == null) {
throw new IllegalArgumentException("Visitor must not be null!");
}
if (handler == null) {
throw new IllegalArgumentException("NodeHandler must not be null!");
}
return root != null;
}
/**
* Recursive helper method for performing a DFS traversal.
*
* @param node the current node
* @param visitor the visitor
* @param handler the handler
* @param the type of the nodes involved
*/
private static void dfs(final T node, final ConfigurationNodeVisitor visitor, final NodeHandler handler) {
if (!visitor.terminate()) {
visitor.visitBeforeChildren(node, handler);
handler.getChildren(node).forEach(c -> dfs(c, visitor, handler));
if (!visitor.terminate()) {
visitor.visitAfterChildren(node, handler);
}
}
}
/**
* Visits all nodes in the hierarchy represented by the given root node in breadth first search manner. This
* means that the nodes are visited in an order corresponding to the distance from the root node: first the root node is
* visited, then all direct children of the root node, then all direct children of the first child of the root node,
* etc. In this mode of traversal, there is no direct connection between the encounter of a node and its children.
* Therefore, on the visitor object only the {@code visitBeforeChildren()} method gets called!.
*
* @param root the root node of the hierarchy to be processed (may be null, then this call has no effect)
* @param visitor the {@code ConfigurationNodeVisitor} (must not be null)
* @param handler the {@code NodeHandler} (must not be null)
* @param the type of the nodes involved
* @throws IllegalArgumentException if a required parameter is null
*/
public void walkBFS(final T root, final ConfigurationNodeVisitor visitor, final NodeHandler handler) {
if (checkParameters(root, visitor, handler)) {
bfs(root, visitor, handler);
}
}
/**
* Visits all nodes in the hierarchy represented by the given root node in depth first search manner. This
* means that first {@link ConfigurationNodeVisitor#visitBeforeChildren(Object, NodeHandler)} is called on a node, then
* recursively all of its children are processed, and eventually
* {@link ConfigurationNodeVisitor#visitAfterChildren(Object, NodeHandler)} gets invoked.
*
* @param root the root node of the hierarchy to be processed (may be null, then this call has no effect)
* @param visitor the {@code ConfigurationNodeVisitor} (must not be null)
* @param handler the {@code NodeHandler} (must not be null)
* @param the type of the nodes involved
* @throws IllegalArgumentException if a required parameter is null
*/
public void walkDFS(final T root, final ConfigurationNodeVisitor visitor, final NodeHandler handler) {
if (checkParameters(root, visitor, handler)) {
dfs(root, visitor, handler);
}
}
}