Remove Substrings

update Sep 14,2017 21:40

LeetCode

Given a string s and a set of n substrings. You are supposed to remove every instance of those n substrings from s so that s is of the minimum length and output this minimum length.

Example

Given s = ccdaabcdbb, substrs = ["ab", "cd"]
Return 2

Explanation:

ccdaabcdbb -> ccdacdbb -> cabb -> cb (length = 2)

Basic Idea:

因为移除 substring 的顺序会对结果造成影响，我们需要使用 bfs 搜索所有可能的删除顺序，输出最短的最终string长度。

1. 使用 s.indexOf(pattern, start)(java) or s.find(pattern, start)(python) 获得要删除的 substring 的index;

2. 对于每个 pattern, 每层bfs删除一个，但是需要考虑到不同位置，所以操作流程是：先得到第一个出现的 pos，将删掉pattern之后的string加入queue，把 start 调整到 pos + 1，对原string继续，直到输出 -1；

3. 当所有 pattern 都不在结果中出现时，也就是 queue 为空时，结束；

Java Code:

    public class Solution {
        /**
         * @param s a string
         * @param dict a set of n substrings
         * @return the minimum length
         */
        public int minLength(String s, Set<String> dict) {
            if (s == null || s.length() == 0) return 0;
            Deque<String> queue = new LinkedList<>();
            Set<String> visited = new HashSet<>();
            queue.addFirst(s);
            visited.add(s);
            int minLen = s.length();

            while (! queue.isEmpty()) {
                String curr = queue.removeLast();
                for (String pattern : dict) {
                    int pos = -1;
                    for (;;) {
                    // 考虑每个pattern的所有出现位置，分别把删除相应位置之后的string加入queue
                        pos = curr.indexOf(pattern, pos + 1);
                        if (pos == -1) break;
                        String out = curr.substring(0, pos)
                                     + curr.substring(pos + pattern.length(), curr.length());
                        if (visited.contains(out)) continue;
                        visited.add(out);
                        queue.addFirst(out);
                    }
                }
                minLen = Math.min(minLen, curr.length());
            }
            return minLen;
        }
    }

Python Code

    class Solution:
        # @param {string} s a string
        # @param {set} dict a set of n substrings
        # @return {int} the minimum length
        def minLength(self, s, dict):
            if not s:
                return 0
            queue = collections.deque()
            visited = set()
            minLen = len(s)

            queue.appendleft(s)
            visited.add(s)
            while queue:
                curr = queue.pop()
                for pattern in dict:
                    pos = -1
                    while 1:
                        pos = curr.find(pattern, pos + 1)
                        if pos == -1:
                            break
                        out = curr[: pos] + curr[pos + len(pattern) :]
                        if out in visited:
                            continue
                        visited.add(out)
                        queue.appendleft(out)
                minLen = min(minLen, len(curr))

            return minLen

Thoughts：

起初拿到题目只想到需要考虑不同顺序的删除，但是没想出实现的方法。现在想来，应该是对 s.indexOf(pattern, start) 函数不够熟悉。

最终的思路可以这么理解：对于input s，BFS 的第一层考虑了所有只删除一个 pattern 的情况，第二层再深入一层，考虑所有删除 两个 pattern 的情况，以此类推。

例如，对于 s=‘ababcc’，pattern={‘a’，‘b’}； 第一层为删除一个 ‘a’ 或者删除一个 ‘b’ 的结果，即 ‘babcc’，‘abbcc’，‘aabcc’，’abacc‘；

update 2018-06-27 20:41:45

Update C++ Solution

时隔数月，没有第一时间回忆起这道题目，思路还是很直接的 brute force 加去重，但是一定要注意去重。

class Solution {
public:
    /*
     * @param s: a string
     * @param dict: a set of n substrings
     * @return: the minimum length
     */
    int minLength(string &s, unordered_set<string> &dict) {
        deque<string> q;
        q.push_back(s);
        unordered_set<string> visited;
        visited.insert(s);
        int ret = 0x7ffffff;
        while (! q.empty()) {
            string curr = q.front(); q.pop_front();
            cout << curr << endl;
            ret = min(ret, (int)curr.size());
            if (ret == 0) return 0;
            for (string pattern : dict) {
                int startPos = 0;
                while (true) {
                    int pos = curr.find(pattern, startPos);
                    if (pos == -1) break;
                    string next = curr.substr(0, pos) + curr.substr(pos + pattern.size());
                    startPos = pos + 1;
                    if (visited.insert(next).second) {
                        q.push_back(next);
                    }
                }
            }
        }
        return ret;
    }
};

update 2018-06-28 22:30:30

Update C++ DFS Solution

public class Solution {
    /*
     * @param s: a string
     * @param dict: a set of n substrings
     * @return: the minimum length
     */
    public int minLength(String s, Set<String> dict) {
        return dfs(s, dict, new HashSet<String>());
    }

    private int dfs(String curr, Set<String> dict, Set<String> visited) {
        if (! visited.add(curr)) return Integer.MAX_VALUE;
        int ret = curr.length();
        for (String pattern : dict) {
            int startPos = 0;
            for (;;) {
                int pos = curr.indexOf(pattern, startPos);
                if (pos != -1) {
                    String next = curr.substring(0, pos) + curr.substring(pos + pattern.length());
                    int len = dfs(next, dict, visited);
                    ret = Math.min(ret, len);
                } else {
                    break;
                }
                startPos = pos + 1;
            }
        }
        return ret;
    }
}