Given two words (start and end), and a dictionary, find all shortest transformation sequence(s) from start to end, such that:
Only one letter can be changed at a time
Each intermediate word must exist in the dictionary
For example,
Given:
start = "hit"
end = "cog"
dict = ["hot","dot","dog","lot","log"]
Return
[
["hit","hot","dot","dog","cog"],
["hit","hot","lot","log","cog"]
]
Note:
All words have the same length.
All words contain only lowercase alphabetic characters.
[分析] 这是一道呼唤耐心的题目,基本思路和word ladder一致,利用BFS搜索出最短路径,所不同的是这题需要找出所有最短路径,因此在搜索时需要记录搜索路径以便于最后进行回溯。有几个关键的细节需要注意:
1)搜索到一个可供跳转的单词时不能像word ladder题中那样立即删除,需要遍历完当前层后再统一从dict中删除下一层的单词,考虑这种情况:input= {"bet", "hat", ["bet", "bit", "bot", "hit", "hot", "hat"]},遍历第三层hit时,若直接删除hat,则最终output会少一条从hot到hat的路径。遍历每层时使用visited集合记录所有下一层单词,该层遍历结束后从dict中删除该层对应的visited集合。
2)反方向搜索,即从end往start搜,可提高回溯时的效率,因为构造路径时可顺序添加而无需插入在数组头部(会有隐藏的数组移位操作)。
3)为了方便回溯,反向搜索过程中要记录当前元素的前驱(output路径方向看),可使用HashMap,key为当前元素,value是key的前驱列表。正向搜索记录后继也是可行的,但实现起来会麻烦些,因为在遍历过程中找到的后继不一定最终能导向end,在回溯时需要进行额外判断,不然可能会空指针。
先参照
http://blog.csdn.net/linhuanmars/article/details/23071455 稍微修改了如Method1所示,然后再此基础上修改了自己初始的想法得到Method 2实现
[ref]
http://www.cnblogs.com/TenosDoIt/p/3443512.html
http://blog.csdn.net/linhuanmars/article/details/23071455
http://blog.csdn.net/u013325815/article/details/42043051
// Method 2: 去除Method1 的辅助数据结构
public class Solution1 {
public List<List<String>> findLadders(String start, String end, Set<String> dict) {
List<List<String>> result = new ArrayList<List<String>>();
LinkedList<String> queue = new LinkedList<String>();
queue.add(end);
HashMap<String, ArrayList<String>> nextMap = new HashMap<String, ArrayList<String>>();
HashSet<String> visited = new HashSet<String>();
dict.add(start);
dict.remove(end);
int currLevel = 1, nextLevel = 0;
int L = start.length();
boolean found = false;
while (!queue.isEmpty()) {
String currStr = queue.poll();
if (currLevel == 0) {
if (found) break;
currLevel = nextLevel;
nextLevel = 0;
dict.removeAll(visited);
visited.clear();
}
char[] currCharArray = currStr.toCharArray();
for (int i = 0; i < L; i++) {
char oldChar = currCharArray[i];
boolean foundCurCycle = false;
for (char c = 'a'; c <= 'z'; c++) {
if (c == oldChar) continue;
currCharArray[i] = c;
String newStr = new String(currCharArray);
if (dict.contains(newStr)) {
if (!nextMap.containsKey(newStr)) {
nextMap.put(newStr, new ArrayList<String>());
}
nextMap.get(newStr).add(currStr);
// 务必先更新nextMap再判断搜索是否结束,否则nextMap可能最终为空
if (newStr.equals(start)) {
found = true;
foundCurCycle = true;
break;
}
if (visited.add(newStr)) {
queue.offer(newStr);
nextLevel++;
}
}
}
currCharArray[i] = oldChar;
if (foundCurCycle) break;
}
currLevel--;
}
if (found) {
List<String> item = new ArrayList<String>();
item.add(start);
getPath(start, end, item, nextMap, result);
}
return result;
}
public void getPath(String start, String end, List<String> item,
HashMap<String, ArrayList<String>> nextMap, List<List<String>> result) {
if (start.equals(end)) {
result.add(new ArrayList<String>(item));
} else {
ArrayList<String> nextList = nextMap.get(start);
for (String next : nextList) {
item.add(next);
getPath(next, end, item, nextMap, result);
item.remove(item.size() - 1);
}
}
}
}
// Method 1, 参考下面链接修改
// http://blog.csdn.net/linhuanmars/article/details/23071455
public class Solution {
class StringWithLevel {
String str;
int level;
public StringWithLevel(String str, int level) {
this.str = str;
this.level = level;
}
}
public List<List<String>> findLadders(String start, String end, Set<String> dict) {
List<List<String>> result = new ArrayList<List<String>>();
LinkedList<StringWithLevel> queue = new LinkedList<StringWithLevel>();
queue.add(new StringWithLevel(end, 0));
HashMap<String, ArrayList<String>> nextMap = new HashMap<String, ArrayList<String>>();
HashSet<String> visited = new HashSet<String>();
dict.add(start);
dict.remove(end);
int currLevel = 0, preLevel = 0;
int L = start.length();
boolean found = false;
while (!queue.isEmpty()) {
StringWithLevel curr = queue.poll();
String currStr = curr.str;
currLevel = curr.level;
if (currLevel > preLevel) {
if (found) break;
preLevel = currLevel;
dict.removeAll(visited);
visited.clear();
}
char[] currCharArray = currStr.toCharArray();
for (int i = 0; i < L; i++) {
char oldChar = currCharArray[i];
boolean foundCurCycle = false;
for (char c = 'a'; c <= 'z'; c++) {
if (c == oldChar) continue;
currCharArray[i] = c;
String newStr = new String(currCharArray);
if (dict.contains(newStr)) {
ArrayList<String> nextList = null;
if (nextMap.containsKey(newStr)) {
nextList = nextMap.get(newStr);
} else {
nextList = new ArrayList<String>();
}
nextList.add(currStr);
nextMap.put(newStr, nextList);
if (newStr.equals(start)) {
found = true;
foundCurCycle = true;
break;
}
if (visited.add(newStr))
queue.offer(new StringWithLevel(newStr, currLevel + 1)); // enqueue a new item of next level
}
}
currCharArray[i] = oldChar;
if (foundCurCycle) break; // can not use found here, or will miss some paths
}
}
if (found) {
List<String> item = new ArrayList<String>();
item.add(start);
getPath(start, end, item, nextMap, result);
}
return result;
}
public void getPath(String start, String end, List<String> item,
HashMap<String, ArrayList<String>> nextMap, List<List<String>> result) {
if (start.equals(end)) {
result.add(new ArrayList<String>(item));
} else {
ArrayList<String> nextList = nextMap.get(start);
for (String next : nextList) {
item.add(next);
getPath(next, end, item, nextMap, result);
item.remove(item.size() - 1);
}
}
}
}