1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
|
"""
Cython utilities for numpy structured arrays.
join_inner(): Do the inner-loop cartesian product for np_utils.join() processing.
(The "inner" is about the inner loop, not inner join).
"""
import numpy as np
import numpy.ma as ma
from numpy.lib.recfunctions import drop_fields
cimport cython
cimport numpy as np
DTYPE = np.int
ctypedef np.intp_t DTYPE_t
@cython.wraparound(False)
@cython.boundscheck(False)
def join_inner(np.ndarray[DTYPE_t, ndim=1] idxs,
np.ndarray[DTYPE_t, ndim=1] idx_sort,
int len_left,
int jointype):
"""
Do the inner-loop cartesian product for np_utils.join() processing.
(The "inner" is about the inner loop, not inner join).
"""
cdef int n_out = 0
cdef int max_key_idxs = 0
cdef DTYPE_t ii, key_idxs, n_left, n_right, idx0, idx1, idx, i
cdef DTYPE_t i_left, i_right, i_out
cdef int masked
# First count the final number of rows and max number of indexes
# for a single key
masked = 0
for ii in range(idxs.shape[0] - 1):
idx0 = idxs[ii]
idx1 = idxs[ii + 1]
# Number of indexes for this key
key_idxs = idx1 - idx0
if key_idxs > max_key_idxs:
max_key_idxs = key_idxs
# Number of rows for this key
n_left = 0
n_right = 0
for idx in range(idx0, idx1):
i = idx_sort[idx]
if i < len_left:
n_left += 1
else:
n_right += 1
# Fix n_left and n_right for different join types
if jointype == 0:
pass
elif jointype == 1:
if n_left == 0:
masked = 1
n_left = 1
if n_right == 0:
masked = 1
n_right = 1
elif jointype == 2:
if n_right == 0:
masked = 1
n_right = 1
elif jointype == 3:
if n_left == 0:
masked = 1
n_left = 1
n_out += n_left * n_right
cdef np.ndarray left_out = np.empty(n_out, dtype=DTYPE)
cdef np.ndarray right_out = np.empty(n_out, dtype=DTYPE)
cdef np.ndarray left_mask = np.zeros(n_out, dtype=np.bool)
cdef np.ndarray right_mask = np.zeros(n_out, dtype=np.bool)
cdef np.ndarray left_idxs = np.empty(max_key_idxs, dtype=DTYPE)
cdef np.ndarray right_idxs = np.empty(max_key_idxs, dtype=DTYPE)
i_out = 0
for ii in range(idxs.shape[0] - 1):
idx0 = idxs[ii]
idx1 = idxs[ii + 1]
# Number of rows for this key
n_left = 0
n_right = 0
for idx in range(idx0, idx1):
i = idx_sort[idx]
if i < len_left:
left_idxs[n_left] = i
n_left += 1
else:
right_idxs[n_right] = i - len_left
n_right += 1
if jointype == 0:
pass
elif jointype == 1:
if n_left == 0:
left_idxs[0] = -1
n_left = 1
if n_right == 0:
right_idxs[0] = -1
n_right = 1
elif jointype == 2:
if n_right == 0:
right_idxs[0] = -1
n_right = 1
elif jointype == 3:
if n_left == 0:
left_idxs[0] = -1
n_left = 1
for i_left in range(n_left):
for i_right in range(n_right):
idx = left_idxs[i_left]
if idx < 0:
idx = 0
left_mask[i_out] = 1
left_out[i_out] = idx
idx = right_idxs[i_right]
if idx < 0:
idx = 0
right_mask[i_out] = 1
right_out[i_out] = idx
i_out += 1
return masked, n_out, left_out, left_mask, right_out, right_mask
|
