結果
| 問題 | No.3618 Omega Cat（Making ver.） |
| コンテスト | |
| ユーザー |
 lif4635
|
| 提出日時 | 2026-08-06 16:17:12 |
| 言語 | PyPy3 (7.3.17) |
| 結果 |
AC
|
| 実行時間 | 852 ms / 2,000 ms |
| + 580µs | |
| コード長 | 14,603 bytes |
| 記録 | |
| コンパイル時間 | 226 ms |
| コンパイル使用メモリ | 95,980 KB |
| 実行使用メモリ | 170,880 KB |
| 最終ジャッジ日時 | 2026-08-06 16:17:35 |
| 合計ジャッジ時間 | 15,379 ms |
|
ジャッジサーバーID （参考情報） |
judge2_0 / judge3_0 |
（要ログイン）
| サブタスク | 配点 | 結果 |
|---|---|---|
| サンプル | 0 % | AC * 1 |
| 小課題1 | 20 % | AC * 5 |
| 小課題2 | 20 % | AC * 12 |
| 小課題3 | 40 % | AC * 24 |
| 小課題4 | 20 % | AC * 38 |
| 合計 | 3.5 * 100% = 350 点 |
ソースコード
# BEGIN Yura Desk bundle: library_codex.data_structure.SegmentTree
"""Point updates and range products for an arbitrary monoid.
Use this when values change one position at a time and a half-open interval
must be folded with an associative operation. ``max_right`` and ``min_left``
also find the first boundary where a monotone predicate stops holding.
"""
class SegmentTree:
__slots__ = ("n", "size", "log", "data", "op", "identity")
def __init__(self, values, op, identity):
if isinstance(values, int):
n = values
values = [identity] * n
else:
values = list(values)
n = len(values)
size = 1 << (n - 1).bit_length() if n else 1
data = [identity] * (size << 1)
data[size : size + n] = values
for node in range(size - 1, 0, -1):
data[node] = op(data[node << 1], data[node << 1 | 1])
self.n = n
self.size = size
self.log = size.bit_length() - 1
self.data = data
self.op = op
self.identity = identity
def set(self, index, value):
node = index + self.size
data = self.data
data[node] = value
op = self.op
node >>= 1
while node:
data[node] = op(data[node << 1], data[node << 1 | 1])
node >>= 1
def add(self, index, value):
"""indexの現在値をop(value, current)で置き換える。O(log N)。"""
node = index + self.size
data = self.data
op = self.op
data[node] = op(value, data[node])
node >>= 1
while node:
data[node] = op(data[node << 1], data[node << 1 | 1])
node >>= 1
def get(self, index):
return self.data[index + self.size]
def tolist(self):
"""現在の要素列をlistで返す。O(N)。"""
return self.data[self.size:self.size + self.n]
def __str__(self):
return str(self.tolist())
def __repr__(self):
return "SegmentTree(%r)" % self.tolist()
def prod(self, left, right):
left += self.size
right += self.size
first = self.identity
second = self.identity
data = self.data
op = self.op
while left < right:
if left & 1:
first = op(first, data[left])
left += 1
if right & 1:
right -= 1
second = op(data[right], second)
left >>= 1
right >>= 1
return op(first, second)
query = prod
def all_prod(self):
return self.data[1]
def max_right(self, left, predicate):
if left == self.n:
return self.n
left += self.size
value = self.identity
data = self.data
op = self.op
while True:
while not left & 1:
left >>= 1
merged = op(value, data[left])
if not predicate(merged):
while left < self.size:
left <<= 1
merged = op(value, data[left])
if predicate(merged):
value = merged
left += 1
return min(left - self.size, self.n)
value = merged
left += 1
if left & -left == left:
break
return self.n
def min_left(self, right, predicate):
if right == 0:
return 0
right += self.size
value = self.identity
data = self.data
op = self.op
while True:
right -= 1
while right > 1 and right & 1:
right >>= 1
merged = op(data[right], value)
if not predicate(merged):
while right < self.size:
right = right << 1 | 1
merged = op(data[right], value)
if predicate(merged):
value = merged
right -= 1
return max(0, right + 1 - self.size)
value = merged
if right & -right == right:
break
return 0
def __getitem__(self, index):
return self.get(index)
# END Yura Desk bundle: library_codex.data_structure.SegmentTree
# BEGIN Yura Desk bundle: library_codex.data_structure.LazySegmentTree
"""Range actions and range products with lazy propagation.
Use this when one update affects every element in a half-open interval and
interval aggregates are also required. The caller supplies the monoid, how an
action changes an aggregate, and how newer and older actions compose.
"""
class LazySegmentTree:
__slots__ = (
"n", "size", "log", "data", "lazy", "pending", "length",
"op", "identity", "mapping", "composition"
)
def __init__(
self,
values,
op,
identity,
mapping,
composition,
):
if isinstance(values, int):
n = values
values = [identity] * n
else:
values = list(values)
n = len(values)
size = 1 << (n - 1).bit_length() if n else 1
data = [identity] * (size << 1)
data[size : size + n] = values
length = [0] * (size << 1)
for index in range(size, size + n):
length[index] = 1
for node in range(size - 1, 0, -1):
data[node] = op(data[node << 1], data[node << 1 | 1])
length[node] = length[node << 1] + length[node << 1 | 1]
self.n = n
self.size = size
self.log = size.bit_length() - 1
self.data = data
self.lazy = [None] * size
self.pending = bytearray(size)
self.length = length
self.op = op
self.identity = identity
self.mapping = mapping
self.composition = composition
def _update(self, node):
self.data[node] = self.op(
self.data[node << 1], self.data[node << 1 | 1]
)
def _all_apply(self, node, action):
self.data[node] = self.mapping(
action, self.data[node], self.length[node]
)
if node < self.size:
if self.pending[node]:
self.lazy[node] = self.composition(action, self.lazy[node])
else:
self.lazy[node] = action
self.pending[node] = 1
def _push(self, node):
if self.pending[node]:
action = self.lazy[node]
self._all_apply(node << 1, action)
self._all_apply(node << 1 | 1, action)
self.lazy[node] = None
self.pending[node] = 0
def set(self, index, value):
node = index + self.size
for shift in range(self.log, 0, -1):
self._push(node >> shift)
self.data[node] = value
for shift in range(1, self.log + 1):
self._update(node >> shift)
def add(self, index, value):
"""indexの現在値をop(value, current)で置き換える。O(log N)。"""
node = index + self.size
for shift in range(self.log, 0, -1):
self._push(node >> shift)
self.data[node] = self.op(value, self.data[node])
for shift in range(1, self.log + 1):
self._update(node >> shift)
def get(self, index):
node = index + self.size
for shift in range(self.log, 0, -1):
self._push(node >> shift)
return self.data[node]
def tolist(self):
"""遅延作用を反映した現在の要素列をlistで返す。O(N)。"""
for node in range(1, self.size):
self._push(node)
return self.data[self.size:self.size + self.n]
def __str__(self):
return str(self.tolist())
def __repr__(self):
return "LazySegmentTree(%r)" % self.tolist()
def prod(self, left, right):
if left == right:
return self.identity
left += self.size
right += self.size
for shift in range(self.log, 0, -1):
if (left >> shift) << shift != left:
self._push(left >> shift)
if (right >> shift) << shift != right:
self._push((right - 1) >> shift)
first = self.identity
second = self.identity
op = self.op
data = self.data
while left < right:
if left & 1:
first = op(first, data[left])
left += 1
if right & 1:
right -= 1
second = op(data[right], second)
left >>= 1
right >>= 1
return op(first, second)
query = prod
def apply(self, left, right=None, action=None):
if action is None:
index = left
action = right
node = index + self.size
for shift in range(self.log, 0, -1):
self._push(node >> shift)
self._all_apply(node, action)
for shift in range(1, self.log + 1):
self._update(node >> shift)
return
if left == right:
return
left += self.size
right += self.size
original_left = left
original_right = right
for shift in range(self.log, 0, -1):
if (left >> shift) << shift != left:
self._push(left >> shift)
if (right >> shift) << shift != right:
self._push((right - 1) >> shift)
while left < right:
if left & 1:
self._all_apply(left, action)
left += 1
if right & 1:
right -= 1
self._all_apply(right, action)
left >>= 1
right >>= 1
left = original_left
right = original_right
for shift in range(1, self.log + 1):
if (left >> shift) << shift != left:
self._update(left >> shift)
if (right >> shift) << shift != right:
self._update((right - 1) >> shift)
range_apply = apply
def all_prod(self):
return self.data[1]
def max_right(self, left, predicate):
if left == self.n:
return self.n
node = left + self.size
for shift in range(self.log, 0, -1):
self._push(node >> shift)
value = self.identity
while True:
while not node & 1:
node >>= 1
merged = self.op(value, self.data[node])
if not predicate(merged):
while node < self.size:
self._push(node)
node <<= 1
merged = self.op(value, self.data[node])
if predicate(merged):
value = merged
node += 1
return min(node - self.size, self.n)
value = merged
node += 1
if node & -node == node:
break
return self.n
def min_left(self, right, predicate):
if right == 0:
return 0
node = right + self.size
for shift in range(self.log, 0, -1):
self._push((node - 1) >> shift)
value = self.identity
while True:
node -= 1
while node > 1 and node & 1:
node >>= 1
merged = self.op(self.data[node], value)
if not predicate(merged):
while node < self.size:
self._push(node)
node = node << 1 | 1
merged = self.op(self.data[node], value)
if predicate(merged):
value = merged
node -= 1
return max(0, node + 1 - self.size)
value = merged
if node & -node == node:
break
return 0
# END Yura Desk bundle: library_codex.data_structure.LazySegmentTree
# input
import sys
input = sys.stdin.readline
II = lambda : int(input())
MI = lambda : map(int, input().split())
LI = lambda : [int(a) for a in input().split()]
SI = lambda : input().rstrip()
LLI = lambda n : [[int(a) for a in input().split()] for _ in range(n)]
LSI = lambda n : [input().rstrip() for _ in range(n)]
MI_1 = lambda : map(lambda x:int(x)-1, input().split())
LI_1 = lambda : [int(a)-1 for a in input().split()]
mod = 998244353
inf = 1001001001001001001
ordalp = lambda s : ord(s)-65 if s.isupper() else ord(s)-97
ordallalp = lambda s : ord(s)-39 if s.isupper() else ord(s)-97
yes = lambda : print("Yes")
no = lambda : print("No")
yn = lambda flag : print("Yes" if flag else "No")
prinf = lambda ans : print(ans if ans < 1000001001001001001 else -1)
alplow = "abcdefghijklmnopqrstuvwxyz"
alpup = "ABCDEFGHIJKLMNOPQRSTUVWXYZ"
alpall = "abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ"
URDL = {'U':(-1,0), 'R':(0,1), 'D':(1,0), 'L':(0,-1)}
DIR_4 = [[-1,0],[0,1],[1,0],[0,-1]]
DIR_8 = [[-1,0],[-1,1],[0,1],[1,1],[1,0],[1,-1],[0,-1],[-1,-1]]
DIR_BISHOP = [[-1,1],[1,1],[1,-1],[-1,-1]]
prime60 = [2,3,5,7,11,13,17,19,23,29,31,37,41,43,47,53,59]
sys.set_int_max_str_digits(0)
# sys.setrecursionlimit(10**6)
# import pypyjit
# pypyjit.set_param('max_unroll_recursion=-1')
from collections import defaultdict,deque
from heapq import heappop,heappush
from bisect import bisect_left,bisect_right
DD = defaultdict
BSL = bisect_left
BSR = bisect_right
def solve():
n = II()
h = LI()
xs = sorted(h)
idx = {x: i for i, x in enumerate(xs)}
seg = [SegmentTree([-inf] * n, max, -inf) for _ in range(4)]
# のこした猫がいない
emp = [0, -inf, -inf, -inf]
# 全体の最大値
mx = [-inf] * 4
for i, x in enumerate(h):
r = idx[x]
rs = [n-r-1, r, n-r-1, r]
keep = [-inf] * 4
# のこす
for p in range(4):
q = max(emp[p], seg[p].prod(0, rs[p]))
keep[p] = q + 1
for p in range(1, 4):
# p = 1 -> 1 ~ n-4
if p <= i <= n - 5 + p:
q = max(emp[p-1], seg[p-1].prod(0, rs[p-1]))
keep[p] = max(keep[p], q + 1)
pre = [max(emp[p], mx[p]) for p in range(4)]
for p in range(1, 4):
if p <= i <= n - 5 + p:
emp[p] = max(emp[p], pre[p-1])
for p in range(4):
seg[p].set(rs[p], keep[p])
mx[p] = max(mx[p], keep[p])
# print(mx)
ans = max(emp[3], mx[3])
print(n - ans)
t = II()
for _ in range(t):
solve()