forked from pandas-dev/pandas
/
join.pyx
284 lines (219 loc) · 8.16 KB
/
join.pyx
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
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
# cython: profile=False
from numpy cimport *
cimport numpy as np
import numpy as np
cimport cython
import_array()
cimport util
from numpy cimport NPY_INT8 as NPY_int8
from numpy cimport NPY_INT16 as NPY_int16
from numpy cimport NPY_INT32 as NPY_int32
from numpy cimport NPY_INT64 as NPY_int64
from numpy cimport NPY_FLOAT16 as NPY_float16
from numpy cimport NPY_FLOAT32 as NPY_float32
from numpy cimport NPY_FLOAT64 as NPY_float64
from numpy cimport (int8_t, int16_t, int32_t, int64_t, uint8_t, uint16_t,
uint32_t, uint64_t, float16_t, float32_t, float64_t)
int8 = np.dtype(np.int8)
int16 = np.dtype(np.int16)
int32 = np.dtype(np.int32)
int64 = np.dtype(np.int64)
float16 = np.dtype(np.float16)
float32 = np.dtype(np.float32)
float64 = np.dtype(np.float64)
cdef double NaN = <double> np.NaN
cdef double nan = NaN
from pandas._libs.algos import groupsort_indexer, ensure_platform_int
from pandas.core.algorithms import take_nd
include "join_func_helper.pxi"
def inner_join(ndarray[int64_t] left, ndarray[int64_t] right,
Py_ssize_t max_groups):
cdef:
Py_ssize_t i, j, k, count = 0
ndarray[int64_t] left_count, right_count, left_sorter, right_sorter
ndarray[int64_t] left_indexer, right_indexer
int64_t lc, rc
# NA group in location 0
left_sorter, left_count = groupsort_indexer(left, max_groups)
right_sorter, right_count = groupsort_indexer(right, max_groups)
# First pass, determine size of result set, do not use the NA group
for i in range(1, max_groups + 1):
lc = left_count[i]
rc = right_count[i]
if rc > 0 and lc > 0:
count += lc * rc
# group 0 is the NA group
cdef:
Py_ssize_t loc, left_pos = 0, right_pos = 0, position = 0
Py_ssize_t offset
# exclude the NA group
left_pos = left_count[0]
right_pos = right_count[0]
left_indexer = np.empty(count, dtype=np.int64)
right_indexer = np.empty(count, dtype=np.int64)
for i in range(1, max_groups + 1):
lc = left_count[i]
rc = right_count[i]
if rc > 0 and lc > 0:
for j in range(lc):
offset = position + j * rc
for k in range(rc):
left_indexer[offset + k] = left_pos + j
right_indexer[offset + k] = right_pos + k
position += lc * rc
left_pos += lc
right_pos += rc
return (_get_result_indexer(left_sorter, left_indexer),
_get_result_indexer(right_sorter, right_indexer))
def left_outer_join(ndarray[int64_t] left, ndarray[int64_t] right,
Py_ssize_t max_groups, sort=True):
cdef:
Py_ssize_t i, j, k, count = 0
ndarray[int64_t] left_count, right_count
ndarray left_sorter, right_sorter, rev
ndarray[int64_t] left_indexer, right_indexer
int64_t lc, rc
# NA group in location 0
left_sorter, left_count = groupsort_indexer(left, max_groups)
right_sorter, right_count = groupsort_indexer(right, max_groups)
# First pass, determine size of result set, do not use the NA group
for i in range(1, max_groups + 1):
if right_count[i] > 0:
count += left_count[i] * right_count[i]
else:
count += left_count[i]
# group 0 is the NA group
cdef:
Py_ssize_t loc, left_pos = 0, right_pos = 0, position = 0
Py_ssize_t offset
# exclude the NA group
left_pos = left_count[0]
right_pos = right_count[0]
left_indexer = np.empty(count, dtype=np.int64)
right_indexer = np.empty(count, dtype=np.int64)
for i in range(1, max_groups + 1):
lc = left_count[i]
rc = right_count[i]
if rc == 0:
for j in range(lc):
left_indexer[position + j] = left_pos + j
right_indexer[position + j] = -1
position += lc
else:
for j in range(lc):
offset = position + j * rc
for k in range(rc):
left_indexer[offset + k] = left_pos + j
right_indexer[offset + k] = right_pos + k
position += lc * rc
left_pos += lc
right_pos += rc
left_indexer = _get_result_indexer(left_sorter, left_indexer)
right_indexer = _get_result_indexer(right_sorter, right_indexer)
if not sort: # if not asked to sort, revert to original order
if len(left) == len(left_indexer):
# no multiple matches for any row on the left
# this is a short-cut to avoid groupsort_indexer
# otherwise, the `else` path also works in this case
left_sorter = ensure_platform_int(left_sorter)
rev = np.empty(len(left), dtype=np.intp)
rev.put(left_sorter, np.arange(len(left)))
else:
rev, _ = groupsort_indexer(left_indexer, len(left))
rev = ensure_platform_int(rev)
right_indexer = right_indexer.take(rev)
left_indexer = left_indexer.take(rev)
return left_indexer, right_indexer
def full_outer_join(ndarray[int64_t] left, ndarray[int64_t] right,
Py_ssize_t max_groups):
cdef:
Py_ssize_t i, j, k, count = 0
ndarray[int64_t] left_count, right_count, left_sorter, right_sorter
ndarray[int64_t] left_indexer, right_indexer
int64_t lc, rc
# NA group in location 0
left_sorter, left_count = groupsort_indexer(left, max_groups)
right_sorter, right_count = groupsort_indexer(right, max_groups)
# First pass, determine size of result set, do not use the NA group
for i in range(1, max_groups + 1):
lc = left_count[i]
rc = right_count[i]
if rc > 0 and lc > 0:
count += lc * rc
else:
count += lc + rc
# group 0 is the NA group
cdef:
int64_t left_pos = 0, right_pos = 0
Py_ssize_t offset, position = 0
# exclude the NA group
left_pos = left_count[0]
right_pos = right_count[0]
left_indexer = np.empty(count, dtype=np.int64)
right_indexer = np.empty(count, dtype=np.int64)
for i in range(1, max_groups + 1):
lc = left_count[i]
rc = right_count[i]
if rc == 0:
for j in range(lc):
left_indexer[position + j] = left_pos + j
right_indexer[position + j] = -1
position += lc
elif lc == 0:
for j in range(rc):
left_indexer[position + j] = -1
right_indexer[position + j] = right_pos + j
position += rc
else:
for j in range(lc):
offset = position + j * rc
for k in range(rc):
left_indexer[offset + k] = left_pos + j
right_indexer[offset + k] = right_pos + k
position += lc * rc
left_pos += lc
right_pos += rc
return (_get_result_indexer(left_sorter, left_indexer),
_get_result_indexer(right_sorter, right_indexer))
def _get_result_indexer(sorter, indexer):
if len(sorter) > 0:
res = take_nd(sorter, indexer, fill_value=-1)
else:
# length-0 case
res = np.empty(len(indexer), dtype=np.int64)
res.fill(-1)
return res
def ffill_indexer(ndarray[int64_t] indexer):
cdef:
Py_ssize_t i, n = len(indexer)
ndarray[int64_t] result
int64_t val, last_obs
result = np.empty(n, dtype=np.int64)
last_obs = -1
for i in range(n):
val = indexer[i]
if val == -1:
result[i] = last_obs
else:
result[i] = val
last_obs = val
return result
def ffill_by_group(ndarray[int64_t] indexer, ndarray[int64_t] group_ids,
int64_t max_group):
cdef:
Py_ssize_t i, n = len(indexer)
ndarray[int64_t] result, last_obs
int64_t gid, val
result = np.empty(n, dtype=np.int64)
last_obs = np.empty(max_group, dtype=np.int64)
last_obs.fill(-1)
for i in range(n):
gid = group_ids[i]
val = indexer[i]
if val == -1:
result[i] = last_obs[gid]
else:
result[i] = val
last_obs[gid] = val
return result
include "join_helper.pxi"