目录

Codeforces Round 731 (Div. 3) 题解 (DEFG)

免责说明:题解短是因为题目太裸(doge)。

又好久没打 CF 了,而且代码风格大变,需要一段时间适应 qwq。

# D. Co-growing Sequence

大意:自己看。

由于输出字典序最小的 y,因此先试着让 $y_1$ 最小。显然,$y_2$ 可以是任何数,也就意味着 $x_2\oplus y_2$ 可以是任何数,那么 $y_1$ 可以是 0(让 $y_2$ 来适应 $y_1$)。

又由于 $x_1\oplus y_1 \subseteq x_2\oplus y_2$,前者已知,可以直接计算 $y_2$ 最小值(最小值为 $z$$x_2$ 的差值,z - (z & x2),其中 $z=x_1\oplus y_1$)。$y_3,y_4,\ldots$ 也都可以这么求。

#include <bits/stdc++.h>
#define repeat(i, a, b) for (int i = (a), ib = (b); i < ib; i++)
#define repeat_back(i, a, b) for (int i = (b) - 1, ib = (a);i >= ib; i--)
#define mst(a, x) memset(a, x, sizeof(a))
#define fi first
#define se second
#define int ll
using namespace std;
namespace start {
    typedef long long ll; const int inf = INT_MAX >> 1; const ll INF = INT64_MAX >> 1;
        ll read() { ll x; if (scanf("%lld", &x) != 1) exit(0); return x; } // will detect EOF
        void print(ll x, bool e = 0) { printf("%lld%c", x, " \n"[e]); }
} using namespace start;
void Solve() {
    int n = read();
    int pre = 0;
    repeat (i, 0, n) {
        int x = read();
        print(pre - (x & pre), i == n - 1);
        pre |= x;
    }
}
signed main() {
    // freopen("data.txt", "r", stdin);
    int T = 1; T = read();
    repeat (ca, 1, T + 1) {
        Solve();
    }
    return 0;
}

# E. Air Conditioners

大意:一些地方有一些空调。一个位置的温度为对每一空调,空调温度加空调到这里的距离,的最小值。求每一位置温度。

直接 for 两倍处理。没搞懂为什么放 D 题后面。()

#include <bits/stdc++.h>
#define repeat(i, a, b) for (int i = (a), ib = (b); i < ib; i++)
#define repeat_back(i, a, b) for (int i = (b) - 1, ib = (a);i >= ib; i--)
#define mst(a, x) memset(a, x, sizeof(a))
#define fi first
#define se second
#define int ll
using namespace std;
namespace start {
    typedef long long ll; const int inf = INT_MAX >> 1; const ll INF = INT64_MAX >> 1;
    mt19937 rnd(chrono::high_resolution_clock::now().time_since_epoch().count());
        ll read() { ll x; if (scanf("%lld", &x) != 1) exit(0); return x; } // will detect EOF
        void print(ll x, bool e = 0) { printf("%lld%c", x, " \n"[e]); }
    const int N = 300010;
} using namespace start;
int a[N], x[N], t[N];
void Solve() {
    int n = read(), k = read();
    fill(a + 1, a + n + 1, inf);
    repeat (i, 0, k)
        x[i] = read();
    repeat (i, 0, k)
        t[i] = read();
    repeat (i, 0, k) {
        a[x[i]] = t[i];
    }
    repeat (i, 2, n + 1) a[i] = min(a[i], a[i - 1] + 1);
    repeat_back (i, 1, n) a[i] = min(a[i], a[i + 1] + 1);
    repeat (i, 1, n + 1) print(a[i], i == n);
}
signed main() {
    // freopen("data.txt", "r", stdin);
    int T = 1; T = read();
    repeat (ca, 1, T + 1) {
        Solve();
    }
    return 0;
}

# F. Array Stabilization (GCD version)

大意:一个循环序列 a,令 b[i] = gcd(a[i], a[i + 1]),然后 ba 替换,为一次操作。问至少多少次操作 a 只有一种元素。

想了好久,感觉朴素分解质因数复杂度不行,后来就换成线性筛优化了。

第一步,最后剩下的肯定是 ngcd(a[1..n])。把初始 $a[i]$ 都除以这个数,作为预处理。

然后将每一 $a[i]$ 分解质因数,对于每一质因数单独讨论(显然可以独立计算)。答案就是有同一质因数的最大区间长度(要考虑循环)。比如 [2, 0, 4, 2][4, 2, 2] 就是一个合法区间,答案为 3。

#include <bits/stdc++.h>
#define repeat(i, a, b) for (int i = (a), ib = (b); i < ib; i++)
#define repeat_back(i, a, b) for (int i = (b) - 1, ib = (a);i >= ib; i--)
#define mst(a, x) memset(a, x, sizeof(a))
#define fi first
#define se second
#define int ll
using namespace std;
namespace start {
    typedef long long ll; const int inf = INT_MAX >> 1; const ll INF = INT64_MAX >> 1;
        ll read() { ll x; if (scanf("%lld", &x) != 1) exit(0); return x; } // will detect EOF
        void print(ll x, bool e = 0) { printf("%lld%c", x, " \n"[e]); }
    const int N = 1000010;
} using namespace start;
struct Sieve {
    static const int N = 1000010;
    bool vis[N]; int lpf[N]; vector<int> prime;
    Sieve() {
        vis[1] = 1;
        repeat (i, 2, N) {
            if (!vis[i]) prime.push_back(i), lpf[i] = i;
            for (auto j : prime) {
                if (i * j >= N) break;
                vis[i * j] = 1; lpf[i * j] = j;
                if (i % j == 0) break;
            }
        }
    }
} sieve;
int a[N];
bool f[N];
int ans, n;
void calc(vector<int> &v) {
    repeat (i, 0, v.size()) v.push_back(v[i] + n);
    int cnt = 1;
    repeat (i, 1, v.size()) {
        if (v[i] == v[i - 1] + 1) cnt++; else cnt = 1;
        ans = max(ans, cnt);
    }
}
vector<int> rec[N], appear;
void Solve() {
    n = read();
    int d = 1;
    repeat (i, 0, n) {
        a[i] = read();
        d = (i == 0 ? a[i] : __gcd(d, a[i]));
    }
    repeat (i, 0, n) a[i] /= d;
    repeat (i, 0, n) {
        while (a[i] != 1) {
            int t = sieve.lpf[a[i]];
            rec[t].push_back(i);
            appear.push_back(t);
            while (sieve.lpf[a[i]] == t) a[i] /= t;
        }
    }
    ans = 0;
    for (auto i : appear) {
        calc(rec[i]);
        rec[i].clear();
    }
    appear.clear();
    print(ans, 1);
}
signed main() {
    // freopen("data.txt", "r", stdin);
    int T = 1; T = read();
    repeat (ca, 1, T + 1) {
        Solve();
    }
    return 0;
}

# G. How Many Paths?

大意:给一张有向图,可以有自环,问 1 到 i 的路径数是哪种情况(没有路径,有唯一路径,有多个且有限路径,有无数个路径)。

很裸的 SCC 缩点后 DAG 上 DP。板子硬套即可。

对于 DAG 的 DP,取反图比较好写。如果 u 的下一个点是 v,且 v 到 1 路径数情况已知(可以往下 DFS,也可以拓扑排序),u 的情况也很好计算。注意如果 u 缩点前是多个顶点的 SCC 或者有自环,那 u 只有两种情况,没有路径和有无数个路径。

#include <bits/stdc++.h>
#define repeat(i, a, b) for (int i = (a), ib = (b); i < ib; i++)
#define repeat_back(i, a, b) for (int i = (b) - 1, ib = (a);i >= ib; i--)
#define mst(a, x) memset(a, x, sizeof(a))
#define fi first
#define se second
#define int ll
using namespace std;
namespace start {
    typedef long long ll; const int inf = INT_MAX >> 1; const ll INF = INT64_MAX >> 1;
    typedef pair<int, int> pii;
        ll read() { ll x; if (scanf("%lld", &x) != 1) exit(0); return x; } // will detect EOF
        void print(ll x, bool e = 0) { printf("%lld%c", x, " \n"[e]); }
    const int N = 1000010;
} using namespace start;
vector<int> a[N];
stack<int> stk;
bool vis[N], instk[N];
int dfn[N], low[N], co[N], w[N];
vector<int> sz;
int n, dcnt;
void dfs(int x) { // Tarjan
    vis[x] = instk[x] = 1; stk.push(x);
    dfn[x] = low[x] = ++dcnt;
    for(auto p : a[x]) {
        if (!vis[p]) dfs(p);
        if (instk[p]) low[x] = min(low[x], low[p]);
    }
    if (low[x] == dfn[x]) {
        int t; sz.push_back(0);
        do {
            t = stk.top();
            stk.pop();
            instk[t] = 0;
            sz.back() += w[x];
            co[t] = sz.size() - 1;
        } while (t != x);
    }
}
void getscc() {
    fill(vis, vis + n, 0);
    sz.clear();
    repeat (i, 0, n) if (!vis[i]) dfs(i);
}
void shrink() { // result: a, n (inplace)
    static set<pii> eset;
    eset.clear();
    getscc();
    repeat (i, 0, n)
    for (auto p : a[i])
    if (co[i] != co[p])
        eset.insert({co[i], co[p]});
    n = sz.size();
    repeat (i, 0, n){
        a[i].clear();
        w[i] = sz[i];
    }
    for(auto i : eset){
        a[i.fi].push_back(i.se);
        // a[i.se].push_back(i.fi);
    }
}
int ans[N];
void ddfs(int x) {
    vis[x] = 1; ans[x] = 0;
    if (x == co[0]) { ans[x] = (w[x] >= 2 ? -1 : 1); return; }
    int infty = 0, cnt = 0;
    for (auto p : a[x]) {
        if (!vis[p]) ddfs(p);
        if (ans[p] == -1) infty = 1;
        if (ans[p] == 2) cnt = 2;
        if (ans[p] == 1) cnt++;
    }
    if (cnt && w[x] >= 2) infty = 1;
    if (infty) ans[x] = -1; else ans[x] = min(cnt, 2ll);
}
void Solve() {
    int n0 = n = read(); int m = read();
    fill (w, w + n, 1);
    repeat (i, 0, n) a[i].clear();
    repeat (i, 0, m) {
        int x = read() - 1, y = read() - 1;
        if (x == y) w[x] = 2;
        else a[y].push_back(x);
    }
    shrink();
    fill (vis, vis + n, 0);
    repeat (i, 0, n) if (!vis[i]) ddfs(i);
    repeat (i, 0, n0) print(ans[co[i]], i == n0 - 1);
}
signed main() {
    // freopen("data.txt", "r", stdin);
    int T = 1; T = read();
    repeat (ca, 1, T + 1) {
        Solve();
    }
    return 0;
}