diff --git a/README.md b/README.md index f787ec8..2f77fd0 100644 --- a/README.md +++ b/README.md @@ -322,6 +322,7 @@ My LeetCode solutions with Chinese explanation. 我的LeetCode中文题解。 | 581 |[Shortest Unsorted Continuous Subarray](https://leetcode.com/problems/shortest-unsorted-continuous-subarray)|[C++](https://github.com/ShusenTang/LeetCode/blob/master/solutions/581.%20Shortest%20Unsorted%20Continuous%20Subarray.md)|Easy| | | 605 |[Can Place Flowers](https://leetcode.com/problems/can-place-flowers)|[C++](https://github.com/ShusenTang/LeetCode/blob/master/solutions/605.%20Can%20Place%20Flowers.md)|Easy| | | 617 |[Merge Two Binary Trees](https://leetcode.com/problems/merge-two-binary-trees/)|[C++](solutions/617.%20Merge%20Two%20Binary%20Trees.md)|Easy| | +| 621 |[Task Scheduler](https://leetcode.com/problems/task-scheduler/)|[C++](solutions/621.%20Task%20Scheduler.md)|Medium| | | 628 |[Maximum Product of Three Numbers](https://leetcode.com/problems/maximum-product-of-three-numbers)|[C++](https://github.com/ShusenTang/LeetCode/blob/master/solutions/628.%20Maximum%20Product%20of%20Three%20Numbers.md)|Easy| | | 643 |[Maximum Average Subarray I](https://leetcode.com/problems/maximum-average-subarray-i)|[C++](https://github.com/ShusenTang/LeetCode/blob/master/solutions/643.%20Maximum%20Average%20Subarray%20I.md)|Easy| | | 661 |[Image Smoother](https://leetcode.com/problems/image-smoother)|[C++](https://github.com/ShusenTang/LeetCode/blob/master/solutions/661.%20Image%20Smoother.md)|Easy| | diff --git a/solutions/621. Task Scheduler.md b/solutions/621. Task Scheduler.md new file mode 100644 index 0000000..a3507cf --- /dev/null +++ b/solutions/621. Task Scheduler.md @@ -0,0 +1,84 @@ +# [621. Task Scheduler](https://leetcode.com/problems/task-scheduler/) + +# 思路 + +## 思路一、暴力模拟 + +首先应该明确我们始终**应该优先处理出现次数最多的任务**。为此我们可以模拟这个过程,先统计每个字母的次数,然后将这些次数送入到一个优先队列(最大堆)里,然后每一轮都从优先队列里面取`n+1`个元素出来将其次数减一,不断重复这个过程直到队列空。 + +设任务总数为N,不同的任务数为M(M<=26),那么时间复杂度为O(NlogM),空间复杂度为O(M)。亲测此方法500ms左右,效率较低。 + +## 思路二 + +还是按照优先处理出现次数最多的任务这个原则。我们假设 A 为出现次数最多的任务,假设其出现了 p 次,考虑到冷却时间,那么执行完所有任务的时间至少为 `(p - 1) * (n + 1) + 1`,如下左图所示,其中浅色代表空闲时间。 + +
+ +
+ +然后我们应当考虑把剩余的任务安排到这些空闲时间里,先将这些任务的出现次序,从大到小进行安排,会有下面两种情况: +* 某个任务和 A 出现的次数相同,例如图 2 中的任务 B。此时我们只能让 B 占据 p - 1 个空闲时间,而在非空闲时间里额外安排一个时间给 B 执行; +* 某个任务比 A 出现的次数少,例如图 2 中的任务 C 和 D。此时我们可以按照列优先的顺序,将其填入空闲时间中。 + +如果在安排某一个任务时,遇到了剩余的空闲时间不够的情况,那么答案一定就等于任务的总数。这是因为我们可以将空闲时间增加虚拟的一列,继续安排任务。 + +所以我们只需要求出出现次数最大(设为`mx`)的任务A,以及出现次数和A同样多的任务(即B这种)数`mx_cnt`,这样最终结果就为`max(N, (n + 1) * (mx - 1) + mx_cnt)`,其中N为任务总数。 + +空间复杂度O(1),时间复杂度O(N) + +[参考来源](https://leetcode-cn.com/problems/task-scheduler/solution/ren-wu-diao-du-qi-by-leetcode/) + + + + + +# C++ +## 思路一 +``` C++ +class Solution { +public: + int leastInterval(vector& tasks, int n) { + if(n == 0) return tasks.size(); + vectorcnt(26, 0); + for(char t: tasks) cnt[t - 'A']++; + + priority_queuemaxheap; + for(int i: cnt) if(i > 0) maxheap.push(i); + + int res = 0; + while(!maxheap.empty()){ + vectortmp; // 存放从maxheap取出的元素 + for(int i = 0; i <= n; i++){ + if(maxheap.empty()) { + if(!tmp.empty()) res += n + 1 - i; // idle + break; + } + res += 1; + auto num = maxheap.top(); maxheap.pop(); + if(--num > 0) tmp.push_back(num); + } + + for(auto a: tmp) maxheap.push(a); // 放回队列中 + } + + return res; + } +}; +``` + +## 思路二 +``` C++ +class Solution { +public: + int leastInterval(vector& tasks, int n) { + vectorcnt(26, 0); + for(char t: tasks) cnt[t - 'A']++; + + int mx = cnt[0], mx_cnt = 0; + for(int i: cnt) mx = max(mx, i); + for(int i: cnt) if(i == mx) mx_cnt++; + + return max((int)tasks.size(), (n + 1) * (mx - 1) + mx_cnt); + } +}; +``` \ No newline at end of file diff --git a/solutions/img/621.png b/solutions/img/621.png new file mode 100644 index 0000000..2068a1a Binary files /dev/null and b/solutions/img/621.png differ