ZS the Coder has drawn an undirected graph of n vertices numbered from 0 to n - 1 and m edges between them. Each edge of the graph is weighted, each weight is a positive integer.
The next day, ZS the Coder realized that some of the weights were erased! So he wants to reassign positive integer weight to each of the edges which weights were erased, so that the length of the shortest path between vertices s and t in the resulting graph is exactly L. Can you help him?
The first line contains five integers n, m, L, s, t (2 ≤ n ≤ 1000, 1 ≤ m ≤ 10 000, 1 ≤ L ≤ 109, 0 ≤ s, t ≤ n - 1, s ≠ t) — the number of vertices, number of edges, the desired length of shortest path, starting vertex and ending vertex respectively.
Then, m lines describing the edges of the graph follow. i-th of them contains three integers, ui, vi, wi (0 ≤ ui, vi ≤ n - 1, ui ≠ vi, 0 ≤ wi ≤ 109).ui and vi denote the endpoints of the edge and wi denotes its weight. If wi is equal to 0 then the weight of the corresponding edge was erased.
It is guaranteed that there is at most one edge between any pair of vertices.
Print "NO" (without quotes) in the only line if it's not possible to assign the weights in a required way.
Otherwise, print "YES" in the first line. Next m lines should contain the edges of the resulting graph, with weights assigned to edges which weights were erased. i-th of them should contain three integers ui, vi and wi, denoting an edge between vertices ui and vi of weight wi. The edges of the new graph must coincide with the ones in the graph from the input. The weights that were not erased must remain unchanged whereas the new weights can be any positive integer not exceeding 1018.
The order of the edges in the output doesn't matter. The length of the shortest path between s and t must be equal to L.
If there are multiple solutions, print any of them.
5 5 13 0 4 0 1 5 2 1 2 3 2 3 1 4 0 4 3 4
YES 0 1 5 2 1 2 3 2 3 1 4 8 4 3 4
2 1 123456789 0 1 0 1 0
YES 0 1 123456789
2 1 999999999 1 0 0 1 1000000000
NO
Here's how the graph in the first sample case looks like :
In the first sample case, there is only one missing edge weight. Placing the weight of 8 gives a shortest path from 0 to 4 of length 13.
In the second sample case, there is only a single edge. Clearly, the only way is to replace the missing weight with 123456789.
In the last sample case, there is no weights to assign but the length of the shortest path doesn't match the required value, so the answer is "NO".
Source
Codeforces Round #372 (Div. 2)
My Solution
图论、最短路、Dijkstra、路径、分配部分边权
首先去点2种不可能的情况:
1、不记录w == 0 的边的情况下,求出dis[e], 如果 dis[e] < L, 则 w == 0 的边无论怎样分配权值,src 到 e 的最短路 都是这个 dis[e], ans = “NO";
2、把w == 0 的边赋值为1,然后把这些 w == 0的边记录下来,跑一边 Dijkstra,如果 dis[e] > L, 则 由于 w > 0 即 w >= 1, 则无论怎样dis[e] 都大于 L, ans = ”NO“;
然后其它时候都是有答案的
也是把 w == 0的边都赋值为1,然后跑一遍 Dijkstra,得到最短路 dis[e],但 dis[e] == L,则ok, 否则 对于 这条最短路径上任意一个赋值非1的variabel edgae += L - dise[e],然后再跑 Dijkstra,如此循环,由于每一个 varible edge最多在 variabel edgae += L - dise[e]的一次, 所以这样的循环最多执行n次就会得到答案
复杂度 O(nmlogn)
#include <iostream>
#include <cstdio>
#include <vector>
#include <queue>
#include <map>
using namespace std;
typedef long long LL;
typedef pair<LL,LL> ii;
const int maxn = 1e6 + 8;
const LL INF = 9e18;
vector<ii> G[maxn];
LL dis[maxn], n, src, L;
//O(nlogn)
void dijkstra()
{
for(int i = 0; i < n; i++) dis[i] = INF;
dis[src] = 0;
priority_queue<ii, vector<ii>, greater<ii> > pq;
pq.push(ii(0, src));
LL u, d, sz, v, w;
while(!pq.empty()){
u = pq.top().second; d = pq.top().first; pq.pop();
sz = G[u].size();
for(int i = 0; i < sz; i++){
v = G[u][i].first; w = G[u][i].second;
if(d + w < dis[v]){
dis[v] = d + w;
pq.push(ii(dis[v], v));
}
}
}
}
vector<ii> g2[maxn];
LL dis2[maxn];
//O(nlogn)
void dijkstra2()
{
for(int i = 0; i < n; i++) dis2[i] = INF;
dis2[src] = 0;
priority_queue<ii, vector<ii>, greater<ii> > pq;
pq.push(ii(0, src));
LL u, d, sz, v, w;
while(!pq.empty()){
u = pq.top().second; d = pq.top().first; pq.pop();
sz = g2[u].size();
for(int i = 0; i < sz; i++){
v = g2[u][i].first; w = g2[u][i].second;
if(d + w < dis2[v]){
dis2[v] = d + w;
pq.push(ii(dis2[v], v));
}
}
}
}
map<ii, int> variable, haveprint;
void update(const LL& x)
{
priority_queue<ii, vector<ii>, greater<ii> > pq;
pq.push(ii(0, src));
LL u, d, sz, v, w;
while(!pq.empty()){
u = pq.top().second; d = pq.top().first; pq.pop();
sz = g2[u].size();
for(int i = 0; i < sz; i++){
v = g2[u][i].first; w = g2[u][i].second;
if(d + w == dis2[v]){
if(w == 1 && variable[ii(u, v)] == 1){ //w == 1 表示之前没有被更改过
g2[u][i].second += x;
int sz2 = g2[v].size(); //无向边是双向读入,所以2个方向的边权都要更新
for(int k = 0; k < sz2; k++){
if(g2[v][k].first == u){
g2[v][k].second += x;
break;
}
}
return;
}
else{
pq.push(ii(dis2[v], v));
}
}
}
}
}
int main()
{
#ifdef LOCAL
freopen("d.txt", "r", stdin);
//freopen("o.txt", "w", stdout);
int T = 1;
while(T--){
#endif // LOCAL
ios::sync_with_stdio(false); cin.tie(0);
LL m, e, u, v, w;
cin >> n >> m >> L >> src >> e;
while(m--){
cin >> u >> v >> w;
if(w != 0){
G[u].push_back(ii(v, w));
G[v].push_back(ii(u, w));
g2[u].push_back(ii(v, w));
g2[v].push_back(ii(u, w));
}
else{
variable[ii(u, v)] = 1;
variable[ii(v, u)] = 1;
g2[u].push_back(ii(v, 1));
g2[v].push_back(ii(u, 1));
}
}
dijkstra();
if(dis[e] < L){
cout << "NO" << endl;
}
else{
dijkstra2();
if(dis2[e] > L){
cout << "NO" << endl;
}
else{
while(dis2[e] != L){
update(L - dis2[e]);
dijkstra2();
}
cout << "YES" << endl;
int sz;
for(int i = 0; i < n; i++){
sz = g2[i].size();
for(int j = 0; j < sz; j++){
if(haveprint[ii(i, g2[i][j].first)] == 0 && g2[i][j].second != 0){
cout << i << " " << g2[i][j].first << " " << g2[i][j].second << "\n";
haveprint[ii(i, g2[i][j].first)] = 1;
haveprint[ii(g2[i][j].first, i)] = 1;
}
}
}
}
}
#ifdef LOCAL
cout << endl;
}
#endif // LOCAL
return 0;
}
Thank you!
------from ProLights