ABC-273解题报告
D. LRUD Instructions
题意:一个左上角为 \((1,1)\)、右下角为 \((H,W)\) 的矩阵,矩阵中有 \(n\) 个障碍。你初始在 \((r,c)\),给你一个操作序列,每个操作为向上/下/左/右走若干格,如果遇到障碍/走到边界则停止。每次操作后输出当前位置。
用数据结构存下每行中的障碍和每列中的障碍,每次从中找到最近的障碍,如果被挡住则走到障碍的上一格,否则正常走。
best code:
By tourist
#include <bits/stdc++.h>
using namespace std;
#ifdef LOCAL
#include "algo/debug.h"
#else
#define debug(...) 42
#endif
int main() {
ios::sync_with_stdio(false);
cin.tie(0);
int h, w, r, c;
cin >> h >> w >> r >> c;
int n;
cin >> n;
map<int, set<int>> mr;
map<int, set<int>> mc;
for (int i = 0; i < n; i++) {
int rr, cc;
cin >> rr >> cc;
mr[rr].insert(cc);
mc[cc].insert(rr);
}
int q;
cin >> q;
while (q--) {
string op;
int cnt;
cin >> op >> cnt;
if (op[0] == 'R') {
auto it = mr[r].lower_bound(c);
if (it != mr[r].end()) {
c = min(*it - 1, c + cnt);
} else {
c += cnt;
}
}
if (op[0] == 'D') {
auto it = mc[c].lower_bound(r);
if (it != mc[c].end()) {
r = min(*it - 1, r + cnt);
} else {
r += cnt;
}
}
if (op[0] == 'L') {
auto it = mr[r].lower_bound(c);
if (it != mr[r].begin()) {
c = max(*prev(it) + 1, c - cnt);
} else {
c -= cnt;
}
}
if (op[0] == 'U') {
auto it = mc[c].lower_bound(r);
if (it != mc[c].begin()) {
r = max(*prev(it) + 1, r - cnt);
} else {
r -= cnt;
}
}
r = max(r, 1);
r = min(r, h);
c = max(c, 1);
c = min(c, w);
cout << r << " " << c << '\n';
}
return 0;
}
这份代码有两个好的地方:
- 使用
map
套set
来维护障碍,避免了复杂的离散化处理。 - 将越界情况统一处理,每次操作最后使用四行 \(\min/\max\) 来处理越界。
By WA_TLE
#include "bits/stdc++.h"
using namespace std;
using namespace std::chrono;
typedef long long int llint;
typedef long double lldo;
#define mp make_pair
#define mt make_tuple
#define pub push_back
#define puf push_front
#define pob pop_back
#define pof pop_front
#define fir first
#define sec second
#define res resize
#define ins insert
#define era erase
#define REP(i, n) for(int i = 0;i < (n);i++)
/*cout<<fixed<<setprecision(20);cin.tie(0);ios::sync_with_stdio(false);*/
const llint mod=1000000007;
template<class T> void SO(T& ve){sort(ve.begin(),ve.end());}
int main(void){
int H,W,rs,cs,i,j,n;cin>>H>>W>>rs>>cs>>n;
vector<pair<int,int>>rc(n+2);
vector<pair<int,int>>cr(n+2);
for(i=0;i<n;i++){
int r,c;cin>>r>>c;
rc[i]=mp(r,c);
cr[i]=mp(c,r);
}
rc[n]=mp(mod,mod);
cr[n]=mp(mod,mod);
rc[n+1]=mp(-mod,-mod);
cr[n+1]=mp(-mod,-mod);
SO(rc);
SO(cr);
int Q;cin>>Q;
while(Q--){
char d;int l;
cin>>d>>l;
if(d=='U'){//r--
auto kabe=*prev(lower_bound(cr.begin(),cr.end(),mp(cs,rs)));
rs-=l;
chmax(rs,1);
if(kabe.fir==cs){chmax(rs,kabe.sec+1);}
}
if(d=='D'){//r++
auto kabe=*lower_bound(cr.begin(),cr.end(),mp(cs,rs));
rs+=l;
chmin(rs,H);
if(kabe.fir==cs){chmin(rs,kabe.sec-1);}
}
if(d=='L'){//c--
auto kabe=*prev(lower_bound(rc.begin(),rc.end(),mp(rs,cs)));
cs-=l;
chmax(cs,1);
if(kabe.fir==rs){chmax(cs,kabe.sec+1);}
}
if(d=='R'){//c++
auto kabe=*lower_bound(rc.begin(),rc.end(),mp(rs,cs));
cs+=l;
chmin(cs,W);
if(kabe.fir==rs){chmin(cs,kabe.sec-1);}
}
cout<<rs<<" "<<cs<<endl;
}
return 0;
}
这份代码巧妙地利用了 pair
自带比较函数的特点,使用正反两个 pair
数组维护障碍,同样避免了离散化且实现巧妙。
By Factorio
#include <bits/stdc++.h>
using namespace std;
map<int, set<int> > r, c;
int n, m, sx, sy, o, x, y, q, l;
char d;
int main()
{
cin >> n >> m >> sx >> sy >> o;
for (int i = 0; i < o; i++)
{
cin >> x >> y;
r[x].insert(y);
c[y].insert(x);
}
cin >> q;
for (int i = 0; i < q; i++)
{
cin >> d >> l;
if (d == 'L')
{
r[sx].insert(0);
int p = *--r[sx].upper_bound(sy);
sy = max(sy - l, p + 1);
}
else if (d == 'R')
{
r[sx].insert(m + 1);
int p = *r[sx].upper_bound(sy);
sy = min(sy + l, p - 1);
}
else if (d == 'U')
{
c[sy].insert(0);
int p = *--c[sy].upper_bound(sx);
sx = max(sx - l, p + 1);
}
else
{
c[sy].insert(n + 1);
int p = *c[sy].upper_bound(sx);
sx = min(sx + l, p - 1);
}
printf("%d %d\n", sx, sy);
}
return 0;
}
这份代码的巧妙之处在于,每次询问先把朝向的边界加入 set
,使三种情况得到统一,更好地避免了分类讨论。
worse code:
By cxm1024
#include<bits/stdc++.h>
using namespace std;
int xx[200010],yy[200010],tmpx[200010],tmpy[200010],cntx=0,cnty=0;
vector<int> xt[200010],yt[200010];
signed main() {
int h,w,r,c,n,q;
cin>>h>>w>>r>>c>>n;
for(int i=1;i<=n;i++) {
cin>>xx[i]>>yy[i];
tmpx[++cntx]=xx[i],tmpy[++cnty]=yy[i];
}
sort(tmpx+1,tmpx+cntx+1);
cntx=unique(tmpx+1,tmpx+cntx+1)-tmpx-1;
sort(tmpy+1,tmpy+cnty+1);
cnty=unique(tmpy+1,tmpy+cnty+1)-tmpy-1;
for(int i=1;i<=n;i++) {
xt[lower_bound(tmpx+1,tmpx+cntx+1,xx[i])-tmpx].push_back(yy[i]);
yt[lower_bound(tmpy+1,tmpy+cnty+1,yy[i])-tmpy].push_back(xx[i]);
}
for(int i=1;i<=cntx;i++)
sort(xt[i].begin(),xt[i].end());
for(int i=1;i<=cnty;i++)
sort(yt[i].begin(),yt[i].end());
cin>>q;
while(q--) {
char op;
int len;
cin>>op>>len;
int nowx=lower_bound(tmpx+1,tmpx+cntx+1,r)-tmpx;
if(tmpx[nowx]!=r) nowx=0;
int nowy=lower_bound(tmpy+1,tmpy+cnty+1,c)-tmpy;
if(tmpy[nowy]!=c) nowy=0;
if(op=='U') {
auto tmp=lower_bound(yt[nowy].begin(),yt[nowy].end(),r);
if(tmp==yt[nowy].begin()) r=max(1,r-len);
else {
tmp--;
r=max(r-len,(*tmp)+1);
}
}
else if(op=='D') {
auto tmp=lower_bound(yt[nowy].begin(),yt[nowy].end(),r);
if(tmp==yt[nowy].end()) r=min(r+len,h);
else r=min(r+len,(*tmp)-1);
}
else if(op=='L') {
auto tmp=lower_bound(xt[nowx].begin(),xt[nowx].end(),c);
if(tmp==xt[nowx].begin()) c=max(1,c-len);
else {
tmp--;
c=max(c-len,(*tmp)+1);
}
}
else {
auto tmp=lower_bound(xt[nowx].begin(),xt[nowx].end(),c);
if(tmp==xt[nowx].end()) c=min(w,c+len);
else c=min(c+len,(*tmp)-1);
}
cout<<r<<" "<<c<<endl;
}
return 0;
}
这份代码直接使用离散化后的 vector
数组来存储障碍,且用了过多分类讨论,实现繁琐。
改进版本:
By cxm1024
#include<bits/stdc++.h>
using namespace std;
int x[200010],y[200010],cntx=0,cnty=0;
map<int,set<int> > xt,yt;
signed main() {
int h,w,r,c,n,q;
cin>>h>>w>>r>>c>>n;
for(int i=1;i<=n;i++)
cin>>x[i]>>y[i];
for(int i=1;i<=n;i++) {
xt[x[i]].insert(y[i]);
yt[y[i]].insert(x[i]);
}
cin>>q;
while(q--) {
char op;
int len;
cin>>op>>len;
if(op=='U') {
auto tmp=yt[c].lower_bound(r);
if(tmp==yt[c].begin()) r=max(1,r-len);
else {
tmp--;
r=max(r-len,(*tmp)+1);
}
}
else if(op=='D') {
auto tmp=yt[c].lower_bound(r);
if(tmp==yt[c].end()) r=min(r+len,h);
else r=min(r+len,(*tmp)-1);
}
else if(op=='L') {
auto tmp=xt[r].lower_bound(c);
if(tmp==xt[r].begin()) c=max(1,c-len);
else {
tmp--;
c=max(c-len,(*tmp)+1);
}
}
else {
auto tmp=xt[r].lower_bound(c);
if(tmp==xt[r].end()) c=min(w,c+len);
else c=min(c+len,(*tmp)-1);
}
cout<<r<<" "<<c<<endl;
}
return 0;
}
wrong code:
By MMNMM(TLE)
#include <bits/extc++.h>
template<typename T, typename S>
std::ostream& operator<<(std::ostream& os, const std::pair<T, S>& p){
os << p.first << " " << p.second;
return os;
}
int main() {
using namespace std;
long H, W;
cin >> H >> W;
pair<long, long> now{};
cin >> now.first >> now.second;
unsigned long N;
cin >> N;
unordered_map<long, set<long>> vertical, horizontal;
for(unsigned long i{}; i < N; ++i){
long x, y;
cin >> x >> y;
vertical[x].insert(y);
horizontal[y].insert(x);
}
unsigned long Q;
cin >> Q;
for(unsigned long i{}; i < Q; ++i){
char d;
long l;
cin >> d >> l;
switch(d){
case 'L':
if(vertical.count(now.first)){
const auto walls{vertical[now.first]};
const auto next_wall{walls.upper_bound(now.second)};
if(next_wall == begin(walls))now.second -= l;
else now.second = max(now.second - l, *prev(next_wall) + 1);
}else now.second -= l;
break;
case 'R':
if(vertical.count(now.first)){
const auto walls{vertical[now.first]};
const auto next_wall{walls.lower_bound(now.second)};
if(next_wall == end(walls))now.second += l;
else now.second = min(now.second + l, *next_wall - 1);
}else now.second += l;
break;
case 'U':
if(horizontal.count(now.second)){
const auto walls{horizontal[now.second]};
const auto next_wall{walls.upper_bound(now.first)};
if(next_wall == begin(walls))now.first -= l;
else now.first = max(now.first - l, *prev(next_wall) + 1);
}else now.first -= l;
break;
case 'D':
if(horizontal.count(now.second)){
const auto walls{horizontal[now.second]};
const auto next_wall{walls.lower_bound(now.first)};
if(next_wall == end(walls))now.first += l;
else now.first = min(now.first + l, *next_wall - 1);
}else now.first += l;
break;
}
now.first = clamp(now.first, 1L, H);
now.second = clamp(now.second, 1L, W);
cout << now << endl;
}
return 0;
}
这份代码 TLE 的原因在于第 35、43、51、59 行为了简化代码将 set 复制了一份,而 set 的直接复制是 \(O(n)\) 的,所以导致超时。如果要复制的话,可以选择按引用赋值,即将 const auto walls{vertical[now.first]};
改为 const auto &walls=vertical[now.first];
便不会超时。该语句相当于指向了原 set
而不会重新复制一份。
选自《Effective C++》:
The meaning of passing an object by value is defined by the copy constructor of that object's class. This can make pass-by-value an extremely expensive operation.
E. Notebook
题意:你需要维护一个序列,支持:
push_back
、pop_back
、记录一个新版本、回退版本。
首先看到这题,第一反应想到用可持久化的数据结构来维护,但对于这道题而言显然过于麻烦了。于是想到可以通过依赖树,将可持久化问题转为非可持久化来做,建图跑 DFS 即可。
By cxm1024
#include<bits/stdc++.h>
using namespace std;
map<int,int> m;
int a[500010];
pair<string,int> ask[500010];
vector<int> now;
vector<pair<int,pair<string,int> > > e[500010];
int ans[500010];
void dfs(int x) {
if(now.empty()) ans[x]=-1;
else ans[x]=now.back();
for(auto v:e[x]) {
int flag=-1;
if(v.second.first=="ADD")
now.push_back(v.second.second);
else if(v.second.first=="DELETE")
if(!now.empty())
flag=now.back(),now.pop_back();
dfs(v.first);
if(v.second.first=="ADD")
now.pop_back();
else if(v.second.first=="DELETE")
if(flag!=-1)
now.push_back(flag);
}
}
signed main() {
ios::sync_with_stdio(false);
int n,tot=0;
cin>>n;
for(int i=1;i<=n;i++) {
a[i]=++tot;
cin>>ask[i].first;
if(ask[i].first!="DELETE")
cin>>ask[i].second;
if(ask[i].first=="LOAD")
e[m[ask[i].second]].push_back({a[i],{"",0}});
else e[a[i-1]].push_back({a[i],ask[i]});
if(ask[i].first=="SAVE")
m[ask[i].second]=a[i];
}
dfs(0);
for(int i=1;i<=n;i++)
cout<<ans[i]<<" ";
cout<<endl;
return 0;
}
然而这个做法仍然不够简洁。注意到由于只需要支持 push_back
和 pop_back
这两个较为基础的操作,我们可以使用链表维护。这里使用反向链表,并维护当前的末尾指针,则每次生成新版本只需要记录当前的末尾即可拉出来一整条链。push_back
只需要增加一个元素作为新末尾并指向原末尾;pop_back
则不需要真的删除元素,单纯把当前末尾指针回退一格即可。显然时空复杂度是可以接受的。
By tourist
#include <bits/stdc++.h>
using namespace std;
#ifdef LOCAL
#include "algo/debug.h"
#else
#define debug(...) 42
#endif
int main() {
ios::sync_with_stdio(false);
cin.tie(0);
int tt;
cin >> tt;
map<int, int> note;
vector<pair<int, int>> all(1, make_pair(0, -1));
int at = 0;
for (int qq = 0; qq < tt; qq++) {
string op;
cin >> op;
if (op == "ADD") {
int num;
cin >> num;
all.emplace_back(at, num);
at = (int) all.size() - 1;
}
if (op == "DELETE") {
at = all[at].first;
}
if (op == "SAVE") {
int num;
cin >> num;
note[num] = at;
}
if (op == "LOAD") {
int num;
cin >> num;
at = note[num];
}
cout << all[at].second << " \n"[qq == tt - 1];
}
return 0;
}
维护 notebook 也可以不用 map 而使用哈希:
时间复杂度上可以优化掉一个 \(\log\),思路,没有本质区别。
这道题本质上是维护一个可持久化栈,经典的做法就是用链表维护。我们也可以将其封装起来,作为一个专门的数据结构:
By sapphire15
#include <bits/stdc++.h>
using namespace std;
#define rep(i,n) for(int i = 0; i < int(n); i++)
#define per(i,n) for(int i = (n) - 1; 0 <= i; i--)
#define rep2(i,l,r) for(int i = (l); i < int(r); i++)
#define per2(i,l,r) for(int i = (r) - 1; int(l) <= i; i--)
#define MM << " " <<
#define pb push_back
#define eb emplace_back
#define all(x) begin(x), end(x)
#define rall(x) rbegin(x), rend(x)
#define sz(x) (int)x.size()
template <typename T>
void print(const vector<T> &v, T x = 0) {
int n = v.size();
for (int i = 0; i < n; i++) cout << v[i] + x << (i == n - 1 ? '\n' : ' ');
if (v.empty()) cout << '\n';
}
using ll = long long;
using pii = pair<int, int>;
using pll = pair<ll, ll>;
template <typename T>
bool chmax(T &x, const T &y) {
return (x < y) ? (x = y, true) : false;
}
template <typename T>
bool chmin(T &x, const T &y) {
return (x > y) ? (x = y, true) : false;
}
template<class T>
class PersistentStack {
private:
class PSNode;
using node_ptr = std::shared_ptr<PSNode>;
struct PSNode {
T _val;
node_ptr _next;
PSNode() : _next(nullptr) {}
PSNode(T x, node_ptr p)
: _val(x), _next(p) {}
};
node_ptr _target;
std::size_t _size;
PersistentStack(node_ptr p, std::size_t sz)
: _target(p), _size(sz) {}
PersistentStack reverse() {
PersistentStack now = *this;
PersistentStack ret;
while(!now.empty()) {
ret = ret.push(now.top());
now = now.pop();
}
return ret;
}
// friend PersistentQueue<T>;
public:
PersistentStack() : _size(0) {
_target = std::make_shared<PSNode>();
}
bool empty() const {
return _target->_next == nullptr;
}
T& top() const {
assert(!empty());
return _target->_val;
}
size_t size() const noexcept {
return _size;
}
PersistentStack pop() const {
assert(!empty());
return PersistentStack(_target->_next, _size - 1);
}
PersistentStack push(const T x) const {
node_ptr ret_p = std::make_shared<PSNode>(x, _target);
return PersistentStack(ret_p, _size + 1);
}
};
int main() {
ios::sync_with_stdio(false);
std::cin.tie(nullptr);
unordered_map<int, PersistentStack<int>> mp;
PersistentStack<int> a;
int q; cin >> q;
vector<int> ans(q);
rep(i,q) {
string s; cin >> s;
if(s == "ADD") {
int x; cin >> x;
a = a.push(x);
} else if(s == "DELETE") {
if(!a.empty()) a = a.pop();
} else if(s == "SAVE") {
int x; cin >> x;
mp[x] = a;
} else if(s == "LOAD") {
int x; cin >> x;
a = mp[x];
}
if(a.empty()) ans[i] = -1;
else ans[i] = a.top();
}
print(ans);
}