面试题 17.05. 字母与数字

给定一个放有字符和数字的数组，找到最长的子数组，且包含的字符和数字的个数相同。

返回该子数组，若存在多个最长子数组，返回左端点最小的。若不存在这样的数组，返回一个空数组。

示例 1:

输入: ["A","1","B","C","D","2","3","4","E","5","F","G","6","7","H","I","J","K","L","M"]

输出: ["A","1","B","C","D","2","3","4","E","5","F","G","6","7"]
示例 2:

输入: ["A","A"]

输出: []
提示：

array.length <= 100000

链接：https://leetcode-cn.com/problems/find-longest-subarray-lcci
 

思路：

令数字为1,字母为-1,求字母和数字个数相同的子数组，即是求累加和为0的字数组。

遍历数组array，使用diff存储累加和，之所以命名为diff，是因为这个累加和实际上是字母和数字个数的差值，使用hash diffPos存储diff值最早出现的位置，即diffPos[diff] = pos意味着array[0...pos]的累加和是diff。

比如array = [1, 1, -1]，遍历array，生成的diffPos = { {1, 0}, {2, 1} } 

遍历array的过程也是生成diffPos的过程，在此过程中，有两种和为0的情况，假设当前位置是cur_pos：

（1）diff为0，即是array[0...cur_pos]累加和0

（2）diff不为0, 但查询diffPos，得知diff此前已在位置pos出现过，意味着array[pos + 1, cur_pos]的累加和是0

对于以上两种情况，记录maxLength和子数组起始pos，即可求解问题。

算法总是倾向于归纳，因为归纳所以简约，我们可以在diffPos中插入一个diffPos[0] = -1，这样（1）可用（2）来表述。

代码：

class Solution {
public:
    vector<string> findLongestSubarray(vector<string>& array) {
        unordered_map<int, size_t> diffPos;
        diffPos.insert({0, -1});

        int diff = 0;
        int maxLength = 0;
        int begin = 0;

        for (size_t i = 0; i < array.size(); ++i) {
            diff += (isdigit(array[i][0]) ? 1 : -1);
            auto iter = diffPos.find(diff);
            if (iter != diffPos.end()) {
                if (maxLength < i - iter->second) {
                    maxLength = i - iter->second;
                    begin = iter->second + 1;
                }
            }
            else {
                diffPos.insert({diff, i});
            }
        }

        return vector(array.begin() + begin, array.begin() + begin + maxLength);
    }
};