|
|
這個題目立意很好，但是就是數(shù)據(jù)比較無語了。
O(N^3)的算法都能跑得比 O(N^2LgN)的快。
代碼寫很長，這份應該算比較爛了，連自己都不知道代碼的時間復雜度是多少！
最終170ms AC，速度不快，比幾百行的O(N^3)代碼要慢很多。。。
注意：
用分數(shù)保存斜率，可以不觸及精度問題，而且還比較方便。
提交了很多次都WA，不知道問題所在，于是還寫了一個腳本來測數(shù)據(jù)，然后發(fā)現(xiàn)了問題所在。 附帶腳本和測試數(shù)據(jù)，呵呵。 代碼：
 #include <stdio.h>
#include <stdlib.h>
#define MAX_N 780
  struct frac  {
 int up, down;
 };
struct point {
int x, y;
};
struct slash_node {
int from, to;
struct frac k;
};
struct ans_node {
int id[MAX_N], len;
};
struct seg_node {
int start, end;
};
struct slash_node slash[MAX_N * MAX_N];
struct point in[MAX_N];
struct ans_node ans[MAX_N];
char visited[MAX_N][MAX_N];
int N, ans_cnt, slash_cnt;
__inline int gcd(int a, int b)
{
int r;
  if (a < b) {
r = a;
a = b;
b = r;
 }
while (1) {
r = a % b;
if (!r)
return b;
a = b;
b = r;
}
}
__inline struct frac frac_init(int up, int down)
{
int r, s;
struct frac f;
if (!down) {
f.down = 1;
f.up = 200000;
return f;
}
if (!up) {
f.down = 1;
f.up = 0;
return f;
}
if (up < 0) {
up = -up;
s = -1;
} else
s = 1;
r = gcd(up, down);
f.up = up / r;
f.up *= s;
f.down = down / r;
return f;
}
__inline int frac_cmp(struct frac fa, struct frac fb)
{
return fa.up * fb.down - fa.down * fb.up;
}
int cmp_slash(const void *a, const void *b)
{
struct slash_node *p, *q;
int r;
p = (struct slash_node *)a;
q = (struct slash_node *)b;
if (p->from != q->from)
return p->from - q->from;
r = frac_cmp(p->k, q->k);
if (!r)
return p->to - q->to;
return r;
}
__inline void add_slash(int from, int to)
{
struct slash_node *s;
if (visited[from][to])
return ;
s = &slash[slash_cnt++];
s->from = from;
s->to = to;
if (in[from].x < in[to].x)
s->k = frac_init(in[to].y - in[from].y, in[to].x - in[from].x);
else
s->k = frac_init(in[from].y - in[to].y, in[from].x - in[to].x);
}
__inline void update(int start, int end)
{
int i;
struct ans_node *a;
a = &ans[ans_cnt++];
a->len = 0;
a->id[a->len++] = slash[start].from;
for (i = start; i <= end; i++)
a->id[a->len++] = slash[i].to;
for ( ; start <= end - 1; start++)
for (i = start + 1; i <= end; i++)
visited[slash[start].to][slash[i].to] = 1;
}
__inline int cmp_seg(struct seg_node *a, struct seg_node *b)
{
return slash[a->start].to - slash[b->start].to;
}
__inline void swap_seg(struct seg_node *a, struct seg_node *b)
{
struct seg_node t = *a;
*a = *b;
*b = t;
}
__inline void bsort(struct seg_node *arr, int len)
{
int i, j;
for (i = 0; i < len - 1; i++)
for (j = i; j < len - 1; j++)
if (cmp_seg(&arr[j], &arr[j + 1]) > 0)
swap_seg(&arr[j], &arr[j + 1]);
}
__inline int calc(int i)
{
int from, start, cnt, j;
static struct seg_node arr[MAX_N];
cnt = 0;
for (from = slash[i].from; slash[i].from == from; i++) {
start = i;
while (i + 1 < slash_cnt &&
slash[i + 1].from == from &&
!frac_cmp(slash[i + 1].k, slash[i].k)
)
i++;
if (i - start >= 1) {
arr[cnt].start = start;
arr[cnt].end = i;
cnt++;
}
}
if (cnt) {
bsort(arr, cnt);
for (j = 0; j < cnt; j++)
update(arr[j].start, arr[j].end);
}
return i;
}
__inline void dump_ans(struct ans_node *a)
{
int i, j, k;
for (i = 0; i < a->len - 2; i++)
for (j = i + 1; j < a->len - 1; j++)
for (k = j + 1; k < a->len; k++)
printf("%d %d %d\n", a->id[i], a->id[j], a->id[k]);
}
int main()
{
int i, j, cnt;
scanf("%d", &N);
for (i = 1; i <= N; i++)
scanf("%d%d", &in[i].x, &in[i].y);
slash_cnt = 0;
for (i = 1; i <= N; i++)
for (j = i + 1; j <= N; j++)
add_slash(i, j);
qsort(slash, slash_cnt, sizeof(slash[0]), cmp_slash);
for (i = 0; i < slash_cnt; i++)
i = calc(i);
cnt = 0;
for (i = 0; i < ans_cnt; i++)
cnt += (ans[i].len) * (ans[i].len - 1) * (ans[i].len - 2) / 6;
printf("%d\n", cnt);
for (i = 0; i < ans_cnt; i++)
dump_ans(&ans[i]);
return 0;
}
測試腳本（Python under windows）：
import subprocess, os
def dump_cow(data, i):
print '%d -> (%d, %d)' % (i, data[i][0], data[i][1])
def dump_ans(data, l):
dump_cow(data, l[0])
dump_cow(data, l[1])
dump_cow(data, l[2])
str2int = lambda l: [ (lambda x: [ int(xi, 10) for xi in x ])(j.split(' ')) for j in l ]
exe = 'e:/src/poj/Debug/poj.exe'
path = 'e:/doc/usaco/testdata/USACO 2005/USACO 2005 December/align.'
for i in range(1, 11):
l_in = open(path + ('%d.in' % i), 'r').readlines()
l_data = str2int(l_in)
p = subprocess.Popen('e:/src/poj/Debug/poj.exe',
shell = True,
stdin = subprocess.PIPE,
stdout = subprocess.PIPE
)
p.stdin.writelines(l_in)
l_out = str2int(p.stdout.readlines())
l_ans = str2int(open(path + ('%d.out' % i), 'r').readlines())
print 'data %d' % i
if l_out != l_ans:
print 'error'
for j in range(0, max(len(l_out), len(l_ans))):
if l_out[j] != l_ans[j]:
print 'idx: ', j
print 'mine:'
dump_ans(l_data, l_out[j])
print 'ans:'
dump_ans(l_data, l_ans[j])
else:
print 'ok'
os.system('pause')
測試數(shù)據(jù)： /Files/varg-vikernes/3174_align.rar
|