BZOJ #4771. 七彩树 [主席树，链并，差分]

乱搞即可。
不强制在线就可以dsu on tree+bit，强制在线只能主席树了（摊手）。

// powered by c++11
// by Isaunoya
#include <bits/stdc++.h>

#define rep(i, x, y) for (register int i = (x); i <= (y); ++i)
#define Rep(i, x, y) for (register int i = (x); i >= (y); --i)

using namespace std;
using db = double;
using ll = long long;
using uint = unsigned int;
using ull = unsigned long long;

using pii = pair<int, int>;

#define fir first
#define sec second

template <class T>

void cmax(T& x, const T& y) {
  if (x < y) x = y;
}

template <class T>

void cmin(T& x, const T& y) {
  if (x > y) x = y;
}

#define all(v) v.begin(), v.end()
#define sz(v) ((int)v.size())
#define pb emplace_back

template <class T>

void sort(vector<T>& v) {
  sort(all(v));
}

template <class T>

void reverse(vector<T>& v) {
  reverse(all(v));
}

template <class T>

void unique(vector<T>& v) {
  sort(all(v)), v.erase(unique(all(v)), v.end());
}

void reverse(string& s) { reverse(s.begin(), s.end()); }

const int io_size = 1 << 23 | 233;
const int io_limit = 1 << 22;
struct io_in {
  char ch;
#ifndef __WIN64
  char getchar() {
    static char buf[io_size], *p1 = buf, *p2 = buf;

    return (p1 == p2) && (p2 = (p1 = buf) + fread(buf, 1, io_size, stdin), p1 == p2) ? EOF : *p1++;
  }
#endif
  io_in& operator>>(char& c) {
    for (c = getchar(); isspace(c); c = getchar())
      ;

    return *this;
  }
  io_in& operator>>(string& s) {
    for (s.clear(); isspace(ch = getchar());)
      ;

    if (!~ch) return *this;

    for (s = ch; !isspace(ch = getchar()) && ~ch; s += ch)
      ;

    return *this;
  }

  io_in& operator>>(char* str) {
    char* cur = str;
    while (*cur) *cur++ = 0;

    for (cur = str; isspace(ch = getchar());)
      ;
    if (!~ch) return *this;

    for (*cur = ch; !isspace(ch = getchar()) && ~ch; *++cur = ch)
      ;

    return *++cur = 0, *this;
  }

  template <class T>

  void read(T& x) {
    bool f = 0;
    while ((ch = getchar()) < 48 && ~ch) f ^= (ch == 45);

    x = ~ch ? (ch ^ 48) : 0;
    while ((ch = getchar()) > 47) x = x * 10 + (ch ^ 48);
    x = f ? -x : x;
  }

  io_in& operator>>(int& x) { return read(x), *this; }

  io_in& operator>>(ll& x) { return read(x), *this; }

  io_in& operator>>(uint& x) { return read(x), *this; }

  io_in& operator>>(ull& x) { return read(x), *this; }

  io_in& operator>>(db& x) {
    read(x);
    bool f = x < 0;
    x = f ? -x : x;
    if (ch ^ '.') return *this;

    double d = 0.1;
    while ((ch = getchar()) > 47) x += d * (ch ^ 48), d *= .1;
    return x = f ? -x : x, *this;
  }
} in;

struct io_out {
  char buf[io_size], *s = buf;
  int pw[233], st[233];

  io_out() {
    set(7);
    rep(i, pw[0] = 1, 9) pw[i] = pw[i - 1] * 10;
  }

  ~io_out() { flush(); }

  void io_chk() {
    if (s - buf > io_limit) flush();
  }

  void flush() { fwrite(buf, 1, s - buf, stdout), fflush(stdout), s = buf; }

  io_out& operator<<(char c) { return *s++ = c, *this; }

  io_out& operator<<(string str) {
    for (char c : str) *s++ = c;
    return io_chk(), *this;
  }

  io_out& operator<<(char* str) {
    char* cur = str;
    while (*cur) *s++ = *cur++;
    return io_chk(), *this;
  }

  template <class T>

  void write(T x) {
    if (x < 0) *s++ = '-', x = -x;

    do {
      st[++st[0]] = x % 10, x /= 10;
    } while (x);

    while (st[0]) *s++ = st[st[0]--] ^ 48;
  }

  io_out& operator<<(int x) { return write(x), io_chk(), *this; }

  io_out& operator<<(ll x) { return write(x), io_chk(), *this; }

  io_out& operator<<(uint x) { return write(x), io_chk(), *this; }

  io_out& operator<<(ull x) { return write(x), io_chk(), *this; }

  int len, lft, rig;

  void set(int _length) { len = _length; }

  io_out& operator<<(db x) {
    bool f = x < 0;
    x = f ? -x : x, lft = x, rig = 1. * (x - lft) * pw[len];
    return write(f ? -lft : lft), *s++ = '.', write(rig), io_chk(), *this;
  }
} out;
#define int long long

template <int sz, int mod>

struct math_t {
  math_t() {
    fac.resize(sz + 1), ifac.resize(sz + 1);
    rep(i, fac[0] = 1, sz) fac[i] = fac[i - 1] * i % mod;

    ifac[sz] = inv(fac[sz]);
    Rep(i, sz - 1, 0) ifac[i] = ifac[i + 1] * (i + 1) % mod;
  }

  vector<int> fac, ifac;

  int qpow(int x, int y) {
    int ans = 1;
    for (; y; y >>= 1, x = x * x % mod)
      if (y & 1) ans = ans * x % mod;
    return ans;
  }

  int inv(int x) { return qpow(x, mod - 2); }

  int C(int n, int m) {
    if (n < 0 || m < 0 || n < m) return 0;
    return fac[n] * ifac[m] % mod * ifac[n - m] % mod;
  }
};

int gcd(int x, int y) { return !y ? x : gcd(y, x % y); }
int lcm(int x, int y) { return x * y / gcd(x, y); }

const int maxn = 5e5 + 55;
vector<int> g[maxn];
int dfn[maxn], idx = 0, dfo[maxn];
int dep[maxn];
vector<int> d[maxn];

struct qwqwq {
  int n;
  int f[maxn][22];
  
  void init(int _n) {
    n = _n;
    idx = 0;
    for (int i = 1; i <= n; i++) dfn[i] = dfo[i] = dep[i] = 0;
    for (int i = 1; i <= n; i++)
      for (int j = 0; j < 22; j++) f[i][j] = 0;
  }

  void dfs(int u) {
  	dep[u] = dep[f[u][0]] + 1;
    dfn[u] = ++idx;
    d[dep[u]].pb(u);
    for (int v : g[u]) {
      if (v ^ f[u][0]) {
        f[v][0] = u;
        dfs(v);
      }
    }
    dfo[u] = idx;
  }

  void solve() {
    dfs(1);
    for (int j = 1; j < 22; j++)
      for (int i = 1; i <= n; i++) f[i][j] = f[f[i][j - 1]][j - 1];
  }

  int lca(int x, int y) {
    if (dep[x] < dep[y]) x ^= y ^= x ^= y;
    for (int i = 20; ~i; i--)
      if (dep[f[x][i]] >= dep[y]) x = f[x][i];
    if (x == y) return x;
    for (int i = 20; ~i; i--)
      if (f[x][i] ^ f[y][i]) x = f[x][i], y = f[y][i];
    return f[x][0];
  }
} mt;

struct set_cmp {
  int x;
  bool operator()(const int x, const int y) { return dfn[x] < dfn[y]; }
};

set<int, set_cmp> s[maxn];
int c[maxn];

struct smt {
  int cnt;
  smt() { cnt = 0; }
  int rt[maxn];
  int val[maxn << 6];
  int ls[maxn << 6];
  int rs[maxn << 6];
	
	void init(int _n) {
		for(int i = 1 ; i <= _n ; i ++)
			rt[i] = 0;
	}
	
  int newnode() {
    ++cnt;
    val[cnt] = 0;
    ls[cnt] = 0;
    rs[cnt] = 0;
    return cnt;
  }

  void upd(int& p, int l, int r, int x, int v) {
    if (!p) p = newnode();
    val[p] += v;
    if (l == r) return;
    int mid = l + r >> 1;
    if (x <= mid)
      upd(ls[p], l, mid, x, v);
    else
      upd(rs[p], mid + 1, r, x, v);
  }

  int merge(int x, int y) {
    if (!x || !y) return x | y;
    val[x] += val[y];
    ls[x] = merge(ls[x], ls[y]);
    rs[x] = merge(rs[x], rs[y]);
    return x;
  }

  int qry(int p, int a, int b, int l, int r) {
    if (a <= l && r <= b) return val[p];
    int mid = l + r >> 1, ans = 0;
    if (a <= mid) ans += qry(ls[p], a, b, l, mid);
    if (b > mid) ans += qry(rs[p], a, b, mid + 1, r);
    return ans;
  }
} smt;

int n, m;
signed main() {
  // code begin.
  int _;
  in >> _;
  while (_--) {
    in >> n >> m;
    mt.init(n);
    smt.init(n);
    smt.cnt = 0;
    for (int i = 1; i <= n; i++) g[i].clear(), s[i].clear(), d[i].clear();
    for (int i = 1; i <= n; i++) in >> c[i];
    for (int i = 2; i <= n; i++) {
      int x;
      in >> x;
      g[x].pb(i);
    }
    mt.solve();
    int mx = 0;
    for (int i = 1; i <= n; i++) {
      if (sz(d[i])) mx = i;
      for (int x : d[i]) {
        auto lft = s[c[x]].lower_bound(x);
        auto rig = s[c[x]].lower_bound(x);
        int l, r;

        if (lft == s[c[x]].begin())
          l = -1;
        else
          --lft, l = *lft;

        if (rig == s[c[x]].end())
          r = -1;
        else
          r = *rig;

        smt.upd(smt.rt[i], 1, n, dfn[x], 1);
        if (~l) smt.upd(smt.rt[i], 1, n, dfn[mt.lca(l, x)], -1);
        if (~r) smt.upd(smt.rt[i], 1, n, dfn[mt.lca(x, r)], -1);
        if (~l && ~r) smt.upd(smt.rt[i], 1, n, dfn[mt.lca(l, r)], 1);
        s[c[x]].insert(x);
      }
      smt.rt[i] = smt.merge(smt.rt[i], smt.rt[i - 1]);
    }
    
    int ans = 0;
    while (m--) {
      int x, d;
      in >> x >> d;
      x ^= ans;
      d ^= ans;
      d = min(mx, dep[x] + d);
      ans = smt.qry(smt.rt[d], dfn[x], dfo[x], 1, n);
      out << ans << '
';
    }
  }
  return 0;
  // code end.
}