「1153」Decode Registration Card of PAT

A registration card number of PAT consists of 4 parts:

• the 1st letter represents the test level, namely, T for the top level, A for advance and B for basic;
• the 2nd - 4th digits are the test site number, ranged from 101 to 999;
• the 5th - 10th digits give the test date, in the form of yymmdd;
• finally the 11th - 13th digits are the testee’s number, ranged from 000 to 999.

Now given a set of registration card numbers and the scores of the card owners, you are supposed to output the various statistics according to the given queries.

Input Specification:

Each input file contains one test case. For each case, the first line gives two positive integers N (≤104) and M (≤100), the numbers of cards and the queries, respectively.

Then N lines follow, each gives a card number and the owner’s score (integer in [0,100]), separated by a space.

After the info of testees, there are M lines, each gives a query in the format Type Term, where

• Type being 1 means to output all the testees on a given level, in non-increasing order of their scores. The corresponding Term will be the letter which specifies the level;
• Type being 2 means to output the total number of testees together with their total scores in a given site. The corresponding Term will then be the site number;
• Type being 3 means to output the total number of testees of every site for a given test date. The corresponding Term will then be the date, given in the same format as in the registration card.

Output Specification:

For each query, first print in a line Case #: input, where # is the index of the query case, starting from 1; and input is a copy of the corresponding input query. Then output as requested:

• for a type 1 query, the output format is the same as in input, that is, CardNumber Score. If there is a tie of the scores, output in increasing alphabetical order of their card numbers (uniqueness of the card numbers is guaranteed);
• for a type 2 query, output in the format Nt Ns where Nt is the total number of testees and Ns is their total score;
• for a type 3 query, output in the format Site Nt where Site is the site number and Nt is the total number of testees at Site. The output must be in non-increasing order of Nt‘s, or in increasing order of site numbers if there is a tie of Nt.

If the result of a query is empty, simply print NA.

Sample Input:

8 4
B123180908127 99
B102180908003 86
A112180318002 98
T107150310127 62
A107180908108 100
T123180908010 78
B112160918035 88
A107180908021 98
1 A
2 107
3 180908
2 999

Sample Output:

Case 1: 1 A
A107180908108 100
A107180908021 98
A112180318002 98
Case 2: 2 107
3 260
Case 3: 3 180908
107 2
123 2
102 1
Case 4: 2 999
NA

---

Ω

做这种题目的时候总是心惊胆战，生怕哪个地方优化没做好，数据结构没搞好就TLE。

给出若干个PAT考试登记码： ，接下来会有多个查询，有三种查询模式：

• Type 1：输入level(B、A、T)，按照成绩降序输出所有参加这个等级考的学生登记码和成绩，成绩一样按登记码升序排列

• Type 2：输入考场号，输出在该考场参加考试的总人数以及总成绩

• Type 3：输入日期，按照当日考试人数降序输出所有有人考试的考场号，以及相应的考试人数，如果考试人数一致则按考场号升序排列

查询结果为空就输出NA。

前两种查询方式都挺好处理。读取完所有登记码和成绩，对三个level的信息分别进行排序；注意到考场号只有三位数，那么直接开一个vector<pair<int,int>> site(1000)，考场号即为索引，来存储人数和总成绩。第三个查询模式就比较棘手了，我的思路是直接开一个map<int,map<int,int>>来存储单个日期下单个考场的人数，在查询时把子map拷贝到一个vector中，然后再进行排序输出。

第一次提交，后三个测试点全是TLE，把我看傻了，心态直接崩。不知道还有什么优化手段，于是去网上看了些题解，发现暴力循环遍历都能过，百思不得其解。我试着把三个level的排序部分给注释掉，竟然发现时间骤减。仔细研究一番，发现问题出在lambda表达式中的传参。由于我把登记码和成绩单独存储在一个vector<pair<int,int>> info中，而三个level都只存info的索引，因此在sort的lambda函数中需要将info传入：

[info](int a, int b) {return info[a].second == info[b].second ? info[a].first < info[b].first : info[a].second > info[b].second;}

那么就意味着每次调用lambda函数时都会拷贝一遍info，从而造成了大量的时间开销。其实lambda函数是可以传入引用的，只要在变量名前加上&即可：

[&info](int a, int b)

传入所有参数是[=]，⚠️这也是拷贝传参。如果需要传入多个参数就用【,】分隔。

---

🐎

#include <iostream>
#include <vector>
#include <unordered_map>
#include <algorithm>

#define idx(x) (x=='B'?0:x=='A'?1:2)

using namespace std;
typedef const pair<int, int> cpii;

int main()
{
    int n, m, tp, s, d;
    scanf("%d %d", &n, &m);
    vector<pair<string, int>> info(n, {string(13, 0), 0});
    vector<pair<int, int>> site(1000);
    vector<int> level[3];
    unordered_map<int, unordered_map<int, int>> date;
    for (int i = 0; i < n; ++i)
    {
        scanf("%s %d", &info[i].first[0], &info[i].second);
        s = stoi(info[i].first.substr(1, 3)), d = stoi(info[i].first.substr(4, 6));
        level[idx(info[i].first[0])].push_back(i);
        ++site[s].first; site[s].second += info[i].second;
        ++date[d][s];
    }
    for (auto &k: level)
        sort(k.begin(), k.end(), [&info](int a, int b) 
        {return info[a].second == info[b].second ? info[a].first < info[b].first : info[a].second > info[b].second;});
    string st;
    for (int i = 0; i < m; ++i)
    {
        cin >> tp >> st;
        printf("Case %d: %d %s\n", i + 1, tp, st.c_str());
        switch (tp)
        {
            case 1:
                if (level[idx(st[0])].empty()) printf("NA\n");
                else
                    for (auto &k: level[idx(st[0])])
                        printf("%s %d\n", info[k].first.c_str(), info[k].second);
                break;
            case 2:
                if (!site[stoi(st)].first) printf("NA\n");
                else printf("%d %d\n", site[stoi(st)].first, site[stoi(st)].second);
                break;
            case 3:
                if (date[stoi(st)].empty()) printf("NA\n");
                else
                {
                    vector<pair<int, int>> tmp(date[stoi(st)].begin(), date[stoi(st)].end());
                    sort(tmp.begin(), tmp.end(), [](cpii &a, cpii &b) 
                    {return a.second == b.second ? a.first < b.first : a.second > b.second;});
                    for (auto &k: tmp)
                        printf("%03d %d\n", k.first, k.second);
                }
                break;
        }
    }
}