#include<bits/stdc++.h>
using namespace std;
#pragma GCC optimize("O3,unroll-loops")
#pragma GCC target("avx2,bmi,bmi2,sse4,sse4.2,lzcnt,popcnt")
#define pb push_back
#define all(x) x.begin(), x.end()
const int MXN = 2e5+1;
struct node {
int val;
node *lc, *rc;
node(int val=0, node *lc=NULL, node *rc=NULL): val(val), lc(lc), rc(rc) {}
};
int n, a[MXN];
vector<int> vals[MXN];
vector<node*> dp[MXN];
void dfs(int v) {
vals[v].pb(a[v]);
dp[v].resize(vals[v].size(), NULL);
if((v<<1)<=n) {
vals[v<<1] = vals[v];
if((v<<1|1)<=n) vals[v<<1].pb(a[v<<1|1]);
dfs(v<<1);
}
if((v<<1|1)<=n) {
vals[v<<1|1] = vals[v];
vals[v<<1|1].pb(a[v<<1]);
dfs(v<<1|1);
}
}
inline vector<node*> arr(node *v) {
vector<node*> q;
q.pb(v);
int ptr=0;
while(ptr<q.size()) {
v = q[ptr++];
if(v->lc) q.pb(v->lc);
if(v->rc) q.pb(v->rc);
}
return q;
}
bool operator<(vector<node*> &A, vector<node*> &B) {
for(int i=0; i<A.size(); i++)
if(A[i]->val<B[i]->val) return 1;
else if(A[i]->val>B[i]->val) return 0;
}
node *DP(int v, int x) {
int p = lower_bound(all(vals[v]), x)-vals[v].begin();
if(dp[v][p]) return dp[v][p];
if((v<<1)>n) return dp[v][p] = new node(x);
if((v<<1|1)>n) {
if(x<a[v<<1]) return dp[v][p] = new node(x, DP(v<<1, a[v<<1]));
else return dp[v][p] = new node(a[v<<1], DP(v<<1, x));
}
if(x<a[v<<1] && x<a[v<<1|1])
return dp[v][p] = new node(x, DP(v<<1, a[v<<1]), DP(v<<1|1, a[v<<1|1]));
if(a[v<<1] < a[v<<1|1])
return dp[v][p] = new node(a[v<<1], DP(v<<1, x), DP(v<<1|1, a[v<<1|1]));
dp[v][p] = new node(a[v<<1|1], DP(v<<1, x), DP(v<<1|1, a[v<<1]));
vector<node*> v1 = arr(dp[v][p]);
dp[v][p]->lc = DP(v<<1, a[v<<1]); dp[v][p]->rc = DP(v<<1|1, x);
vector<node*> v2 = arr(dp[v][p]);
if(v1<v2) dp[v][p]->lc = DP(v<<1, x), dp[v][p]->rc = DP(v<<1|1, a[v<<1]);
return dp[v][p];
}
int32_t main() {
cin.tie(0); cout.tie(0); ios_base::sync_with_stdio(0);
cin >> n;
for(int i=1; i<=n; i++) cin >> a[i];
dfs(1);
for(int i=1; i<=n; i++)
sort(all(vals[i]));
vector<node*> ans = arr(DP(1, a[1]));
for(node *v : ans) cout << v->val << ' ';
cout << '\n';
return 0;
}
