來源於英文“retrieval”. Trie樹就是字符樹,其核心思想就是空間換時間。
舉個簡單的例子。
給你100000個長度不超過10的單詞。對於每一個單詞,我們要判斷他出沒出現過,如果出現了,第一次出現第幾個位置。
這題當然可以用hash來,但是我要介紹的是trie樹。在某些方面它的用途更大。比如說對於某一個單詞,我要詢問它的前綴是否出現過。這樣hash就不好搞了,而用trie還是很簡單。
現在回到例子中,如果我們用最傻的方法,對於每一個單詞,我們都要去查找它前面的單詞中是否有它。那么這個算法的復雜度就是O(n^2)。顯然對於100000的范圍難以接受。現在我們換個思路想。假設我要查詢的單詞是abcd,那么在他前面的單詞中,以b,c,d,f之類開頭的我顯然不必考慮。而只要找以a開頭的中是否存在abcd就可以了。同樣的,在以a開頭中的單詞中,我們只要考慮以b作為第二個字母的……這樣一個樹的模型就漸漸清晰了……
我們可以看到,trie樹每一層的節點數是26^i級別的。所以為了節省空間。我們用動態鏈表,或者用數組來模擬動態。空間的花費,不會超過單詞數×單詞長度。(轉自一大牛)
Trie樹的java代碼 實現如下: import java.util.ArrayList; import java.util.Iterator; import java.util.List; /** *//** * A word trie which can only deal with 26 alphabeta letters. * @author Leeclipse * @since 2007-11-21 */ public class Trie{ private Vertex root;//一個Trie樹有一個根節點 //內部類 protected class Vertex{//節點類 protected int words; protected int prefixes; protected Vertex[] edges;//每個節點包含26個子節點(類型為自身) Vertex() { words = 0; prefixes = 0; edges = new Vertex[26]; for (int i = 0; i < edges.length; i++) { edges[i] = null; } } } public Trie () { root = new Vertex(); } /** *//** * List all words in the Trie. * * @return */ public List< String> listAllWords() { List< String> words = new ArrayList< String>(); Vertex[] edges = root.edges; for (int i = 0; i < edges.length; i++) { if (edges[i] != null) { String word = "" + (char)('a' + i); depthFirstSearchWords(words, edges[i], word); } } return words; } /** *//** * Depth First Search words in the Trie and add them to the List. * * @param words * @param vertex * @param wordSegment */ private void depthFirstSearchWords(List words, Vertex vertex, String wordSegment) { Vertex[] edges = vertex.edges; boolean hasChildren = false; for (int i = 0; i < edges.length; i++) { if (edges[i] != null) { hasChildren = true; String newWord = wordSegment + (char)('a' + i); depthFirstSearchWords(words, edges[i], newWord); } } if (!hasChildren) { words.add(wordSegment); } } public int countPrefixes(String prefix) { return countPrefixes(root, prefix); } private int countPrefixes(Vertex vertex, String prefixSegment) { if (prefixSegment.length() == 0) { //reach the last character of the word return vertex.prefixes; } char c = prefixSegment.charAt(0); int index = c - 'a'; if (vertex.edges[index] == null) { // the word does NOT exist return 0; } else { return countPrefixes(vertex.edges[index], prefixSegment.substring(1)); } } public int countWords(String word) { return countWords(root, word); } private int countWords(Vertex vertex, String wordSegment) { if (wordSegment.length() == 0) { //reach the last character of the word return vertex.words; } char c = wordSegment.charAt(0); int index = c - 'a'; if (vertex.edges[index] == null) { // the word does NOT exist return 0; } else { return countWords(vertex.edges[index], wordSegment.substring(1)); } } /** *//** * Add a word to the Trie. * * @param word The word to be added. */ public void addWord(String word) { addWord(root, word); } /** *//** * Add the word from the specified vertex. * @param vertex The specified vertex. * @param word The word to be added. */ private void addWord(Vertex vertex, String word) { if (word.length() == 0) { //if all characters of the word has been added vertex.words ++; } else { vertex.prefixes ++; char c = word.charAt(0); c = Character.toLowerCase(c); int index = c - 'a'; if (vertex.edges[index] == null) { //if the edge does NOT exist vertex.edges[index] = new Vertex(); } addWord(vertex.edges[index], word.substring(1)); //go the the next character } } public static void main(String args[]) //Just used for test { Trie trie = new Trie(); trie.addWord("China"); trie.addWord("China"); trie.addWord("China"); trie.addWord("crawl"); trie.addWord("crime"); trie.addWord("ban"); trie.addWord("China"); trie.addWord("english"); trie.addWord("establish"); trie.addWord("eat"); System.out.println(trie.root.prefixes); System.out.println(trie.root.words); List< String> list = trie.listAllWords(); Iterator listiterator = list.listIterator(); while(listiterator.hasNext()) { String s = (String)listiterator.next(); System.out.println(s); } int count = trie.countPrefixes("ch"); int count1=trie.countWords("china"); System.out.println("the count of c prefixes:"+count); System.out.println("the count of china countWords:"+count1); } } 運行: C:\test>java Trie 10 0 ban china crawl crime eat english establish the count of c prefixes:4 the count of china countWords:4

