F. Copy or Prefix Sum(优化dp转移)

F. Copy or Prefix Sum

题目链接:Copy or Prefix Sum
题目大意:给定一个$b$数组，一个$a$是合法的指对于每一个$i$都有$b_i=a_i$ or $b_i=\sum _{j=1}^i{a}_j$。问合法的$a$有多少个。
数据范围:$1\le t\le 1e4,1\le \sum n\le 2e5,-1e9\le b_i\le 1e9$
题解:考虑$dp$，$dp[i][j]$表示$[1-i]$中有多少种情况$(\sum _{j=1}^i{a}_i)==j$。答案显然是$\sum _{j=-inf}^{inf}dp[n][j]$。现在来考虑转移，对于$a[i]$来说可以等于$b[i]orb[i]-sum[i-1]$
$a[i]$等于$b[i]$的情况，对于$j\in [-inf,inf]$都有$$dp[i][j]=dp[i-1][j-b[i]]$$$a[i]$等于$b[i]-sum[i-1]$的情况，可以发现$(\sum _{j=1}^{i}a[i])=b[i]$,所以$$dp[i][b[i]]=\sum _{j=-inf}^{inf}dp[i-1][j]$$。我们可以使用$map$来进行转移，每次$i->i+1$将$map$内的元素都修改一次。时间复杂度$O(n^{2}logn)$。不过对于这种全局修改的转移，可以使用一种"水位线技术"。参见博客。可以设置一个全局变量，让转移变成$O(nlogn)$。其中$log$是$map$自带的。
AC代码:

#include<bits/stdc++.h>

#define ld long double
#define ll long long
using namespace std;
template<class T>
void read(T& x)
{
	T res = 0, f = 1; char c = getchar();
	while (!isdigit(c)) {
		if (c == '-')f = -1; c = getchar();
	}
	while (isdigit(c)) {
		res = (res << 3) + (res << 1) + c - '0'; c = getchar();
	}
	x = res * f;
}
#define int long long
const ll N = 200000 + 10;
const int mod = 1e9 + 7;
int t, n, b[N];
map<int, int>mp;
signed main()
{
	//ios::sync_with_stdio(false);
#ifndef ONLINE_JUDGE
	freopen("test.in", "r", stdin);
#endif // ONLINE_JUDGE

	read(t);
	while (t--)
	{
		int wel = 0, tol = 0;
		read(n);
		mp.clear();
		mp[0] = 1, tol = 1;
		for (int i = 1; i <= n; i++)read(b[i]);
		for (int i = 1; i <= n; i++)
		{
			wel -= b[i];
			int change = tol - mp[b[i] + wel];
			mp[b[i] + wel] = tol;
			tol = (tol + change)%mod;
		}
		printf("%lld\n", (tol % mod + mod)%mod);
	}
	return 0;
}