回答編集履歴
2
修正
test
CHANGED
|
@@ -58,6 +58,22 @@
|
|
|
58
58
|
|
|
59
59
|
```
|
|
60
60
|
|
|
61
|
+
|
|
62
|
+
|
|
63
|
+
In [51]: slen = 1000
|
|
64
|
+
|
|
65
|
+
...: tslen = 1200
|
|
66
|
+
|
|
67
|
+
...: N = 1000
|
|
68
|
+
|
|
69
|
+
...:
|
|
70
|
+
|
|
71
|
+
...: random.seed(0)
|
|
72
|
+
|
|
73
|
+
...: s_org = [1]*(tslen-slen) + [random.randint(0, 1) for _ in range(1000)]
|
|
74
|
+
|
|
75
|
+
...:
|
|
76
|
+
|
|
61
77
|
In [52]: with timer("numpy"):
|
|
62
78
|
|
|
63
79
|
...: a = np.array(s_org)
|
|
@@ -78,4 +94,6 @@
|
|
|
78
94
|
|
|
79
95
|
[numpy] done in 0.00026 s
|
|
80
96
|
|
|
97
|
+
In [53]: assert len11 == 202
|
|
98
|
+
|
|
81
99
|
```
|
1
追記
test
CHANGED
|
@@ -35,3 +35,47 @@
|
|
|
35
35
|
|
|
36
36
|
|
|
37
37
|
pythonのfor-loopが遅いので行列のインデックスを駆使した方法です。理解してコーディングしないと収拾がつかなくなるので、適度な用法用量を守ってご参考にしてください。
|
|
38
|
+
|
|
39
|
+
|
|
40
|
+
|
|
41
|
+
----
|
|
42
|
+
|
|
43
|
+
追記
|
|
44
|
+
|
|
45
|
+
|
|
46
|
+
|
|
47
|
+
mkgreyさんのコードそのまま回したケース
|
|
48
|
+
|
|
49
|
+
[groupby] done in 0.39933 s
|
|
50
|
+
|
|
51
|
+
[forward-search] done in 0.12787 s
|
|
52
|
+
|
|
53
|
+
[backward-search] done in 0.05486 s
|
|
54
|
+
|
|
55
|
+
|
|
56
|
+
|
|
57
|
+
最大の長さを返すよう若干修正したバージョン
|
|
58
|
+
|
|
59
|
+
```
|
|
60
|
+
|
|
61
|
+
In [52]: with timer("numpy"):
|
|
62
|
+
|
|
63
|
+
...: a = np.array(s_org)
|
|
64
|
+
|
|
65
|
+
...: offset = 1
|
|
66
|
+
|
|
67
|
+
...: diff_a = a[1:] - a[:-1]
|
|
68
|
+
|
|
69
|
+
...: pos1to0 = np.where(diff_a ==-1)[0] + offset
|
|
70
|
+
|
|
71
|
+
...: pos0to1 = np.where(diff_a ==+1)[0] + offset
|
|
72
|
+
|
|
73
|
+
...: pos0to1 = np.append([0], pos0to1)
|
|
74
|
+
|
|
75
|
+
...: len11 = (pos1to0 - pos0to1).max()
|
|
76
|
+
|
|
77
|
+
...:
|
|
78
|
+
|
|
79
|
+
[numpy] done in 0.00026 s
|
|
80
|
+
|
|
81
|
+
```
|