bed-reader 0.2.37__cp312-none-win_amd64.whl
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- bed_reader/__init__.py +30 -0
- bed_reader/_open_bed.py +1290 -0
- bed_reader/_sample_data.py +99 -0
- bed_reader/_to_bed.py +235 -0
- bed_reader/bed_reader.cp312-win_amd64.pyd +0 -0
- bed_reader/tests/__init__.py +1 -0
- bed_reader/tests/benchmark/benchmark.py +129 -0
- bed_reader/tests/data/500x400_o640_array.memmap +0 -0
- bed_reader/tests/data/badfile.bed +1 -0
- bed_reader/tests/data/empty.bed +0 -0
- bed_reader/tests/data/plink_sim_10s_100v_10pmiss.bed +0 -0
- bed_reader/tests/data/plink_sim_10s_100v_10pmiss.bim +100 -0
- bed_reader/tests/data/plink_sim_10s_100v_10pmiss.fam +10 -0
- bed_reader/tests/data/small.bad_bim +4 -0
- bed_reader/tests/data/small.bed +1 -0
- bed_reader/tests/data/small.bim +4 -0
- bed_reader/tests/data/small.bim_bad_positions.bim +4 -0
- bed_reader/tests/data/small.deb +1 -0
- bed_reader/tests/data/small.fam +3 -0
- bed_reader/tests/data/small.fam_bad +4 -0
- bed_reader/tests/data/small.maf +3 -0
- bed_reader/tests/data/small.mib +4 -0
- bed_reader/tests/data/small2_array.memmap +0 -0
- bed_reader/tests/data/small_array.memmap +0 -0
- bed_reader/tests/data/small_no_bim.bed +1 -0
- bed_reader/tests/data/small_no_bim.fam +3 -0
- bed_reader/tests/data/small_no_fam.bed +1 -0
- bed_reader/tests/data/small_no_fam.bim +4 -0
- bed_reader/tests/data/small_too_short.bed +1 -0
- bed_reader/tests/data/small_too_short.bim +4 -0
- bed_reader/tests/data/small_too_short.fam +3 -0
- bed_reader/tests/data/smallmode0.bed +0 -0
- bed_reader/tests/data/smallmode0.bim +3 -0
- bed_reader/tests/data/smallmode0.fam +4 -0
- bed_reader/tests/data/smallmodebad.bed +1 -0
- bed_reader/tests/data/some_missing.bed +0 -0
- bed_reader/tests/data/some_missing.bim +100 -0
- bed_reader/tests/data/some_missing.fam +100 -0
- bed_reader/tests/data/some_missing.properties.npz +0 -0
- bed_reader/tests/data/some_missing.val.npy +0 -0
- bed_reader/tests/data/toydata.5chrom.bed +0 -0
- bed_reader/tests/data/toydata.5chrom.bim +10000 -0
- bed_reader/tests/data/toydata.5chrom.fam +500 -0
- bed_reader/tests/registry.txt +36 -0
- bed_reader/tests/test_file_dot.py +368 -0
- bed_reader/tests/test_open_bed.py +806 -0
- bed_reader-0.2.37.dist-info/METADATA +112 -0
- bed_reader-0.2.37.dist-info/RECORD +51 -0
- bed_reader-0.2.37.dist-info/WHEEL +4 -0
- bed_reader-0.2.37.dist-info/license_files/AUTHORS.txt +2 -0
- bed_reader-0.2.37.dist-info/license_files/LICENSE.md +202 -0
|
@@ -0,0 +1,368 @@
|
|
|
1
|
+
import logging
|
|
2
|
+
from datetime import datetime
|
|
3
|
+
from pathlib import Path
|
|
4
|
+
|
|
5
|
+
import numpy as np
|
|
6
|
+
import pytest
|
|
7
|
+
|
|
8
|
+
from bed_reader import (
|
|
9
|
+
file_aat_piece_f64_orderf,
|
|
10
|
+
file_ata_piece_f64_orderf,
|
|
11
|
+
file_b_less_aatbx,
|
|
12
|
+
file_dot_piece,
|
|
13
|
+
)
|
|
14
|
+
from bed_reader._open_bed import get_num_threads, open_bed # noqa
|
|
15
|
+
|
|
16
|
+
|
|
17
|
+
def file_ata(filename, offset, iid_count, sid_count, sid_step):
|
|
18
|
+
ata = np.full((sid_count, sid_count), np.nan)
|
|
19
|
+
for sid_index in range(0, sid_count, sid_step):
|
|
20
|
+
sid_range_len = min(sid_step, sid_count - sid_index)
|
|
21
|
+
ata_piece = np.full((sid_count - sid_index, sid_range_len), np.nan)
|
|
22
|
+
if sid_index % 2 == 0: # test new and old method
|
|
23
|
+
file_ata_piece_f64_orderf(
|
|
24
|
+
str(filename),
|
|
25
|
+
offset,
|
|
26
|
+
iid_count,
|
|
27
|
+
sid_count,
|
|
28
|
+
sid_index,
|
|
29
|
+
ata_piece,
|
|
30
|
+
num_threads=get_num_threads(None),
|
|
31
|
+
log_frequency=sid_range_len,
|
|
32
|
+
)
|
|
33
|
+
else:
|
|
34
|
+
file_dot_piece(
|
|
35
|
+
str(filename),
|
|
36
|
+
offset,
|
|
37
|
+
iid_count,
|
|
38
|
+
sid_index,
|
|
39
|
+
ata_piece,
|
|
40
|
+
num_threads=get_num_threads(None),
|
|
41
|
+
log_frequency=sid_range_len,
|
|
42
|
+
)
|
|
43
|
+
ata[sid_index:, sid_index : sid_index + sid_range_len] = ata_piece
|
|
44
|
+
for sid_index in range(sid_count):
|
|
45
|
+
ata[sid_index, sid_index + 1 :] = ata[sid_index + 1 :, sid_index]
|
|
46
|
+
return ata
|
|
47
|
+
|
|
48
|
+
|
|
49
|
+
def file_aat(filename, offset, iid_count, sid_count, iid_step):
|
|
50
|
+
aat = np.full((iid_count, iid_count), np.nan)
|
|
51
|
+
for iid_index in range(0, iid_count, iid_step):
|
|
52
|
+
iid_range_len = min(iid_step, iid_count - iid_index)
|
|
53
|
+
aat_piece = np.full((iid_count - iid_index, iid_range_len), np.nan)
|
|
54
|
+
file_aat_piece_f64_orderf(
|
|
55
|
+
str(filename),
|
|
56
|
+
offset,
|
|
57
|
+
iid_count,
|
|
58
|
+
sid_count,
|
|
59
|
+
iid_index,
|
|
60
|
+
aat_piece,
|
|
61
|
+
num_threads=get_num_threads(None),
|
|
62
|
+
log_frequency=iid_range_len,
|
|
63
|
+
)
|
|
64
|
+
aat[iid_index:, iid_index : iid_index + iid_range_len] = aat_piece
|
|
65
|
+
for iid_index in range(iid_count):
|
|
66
|
+
aat[iid_index, iid_index + 1 :] = aat[iid_index + 1 :, iid_index]
|
|
67
|
+
|
|
68
|
+
return aat
|
|
69
|
+
|
|
70
|
+
|
|
71
|
+
def write_read_test_file_ata(iid_count, sid_count, sid_step, tmp_path):
|
|
72
|
+
offset = 640
|
|
73
|
+
file_path = tmp_path / f"{iid_count}x{sid_count}_o{offset}_array.memmap"
|
|
74
|
+
|
|
75
|
+
mm = np.memmap(
|
|
76
|
+
file_path,
|
|
77
|
+
dtype="float64",
|
|
78
|
+
mode="w+",
|
|
79
|
+
offset=offset,
|
|
80
|
+
shape=(iid_count, sid_count),
|
|
81
|
+
order="F",
|
|
82
|
+
)
|
|
83
|
+
mm[:] = np.linspace(0, 1, mm.size).reshape(mm.shape)
|
|
84
|
+
mm.flush()
|
|
85
|
+
|
|
86
|
+
out_val = file_ata(file_path, offset, iid_count, sid_count, sid_step)
|
|
87
|
+
expected = mm.T.dot(mm)
|
|
88
|
+
assert np.allclose(expected, out_val, equal_nan=True)
|
|
89
|
+
|
|
90
|
+
|
|
91
|
+
def write_read_test_file_aat(
|
|
92
|
+
iid_count, sid_count, iid_step, tmp_path, skip_if_there=False
|
|
93
|
+
):
|
|
94
|
+
offset = 640
|
|
95
|
+
file_path = tmp_path / f"{iid_count}x{sid_count}_o{offset}_array.memmap"
|
|
96
|
+
if skip_if_there and file_path.exists():
|
|
97
|
+
logging.info(f"'{file_path}' exists")
|
|
98
|
+
mm = np.memmap(
|
|
99
|
+
file_path,
|
|
100
|
+
dtype="float64",
|
|
101
|
+
mode="r",
|
|
102
|
+
offset=offset,
|
|
103
|
+
shape=(iid_count, sid_count),
|
|
104
|
+
order="F",
|
|
105
|
+
)
|
|
106
|
+
else:
|
|
107
|
+
logging.info(f"Creating '{file_path}'")
|
|
108
|
+
mm = np.memmap(
|
|
109
|
+
file_path,
|
|
110
|
+
dtype="float64",
|
|
111
|
+
mode="w+",
|
|
112
|
+
offset=offset,
|
|
113
|
+
shape=(iid_count, sid_count),
|
|
114
|
+
order="F",
|
|
115
|
+
)
|
|
116
|
+
mm[:] = np.linspace(0, 1, mm.size).reshape(mm.shape)
|
|
117
|
+
mm.flush()
|
|
118
|
+
logging.info(f"Finished creating '{file_path}'")
|
|
119
|
+
|
|
120
|
+
start_time = datetime.now()
|
|
121
|
+
out_val = file_aat(file_path, offset, iid_count, sid_count, iid_step)
|
|
122
|
+
delta = datetime.now() - start_time
|
|
123
|
+
expected = mm.dot(mm.T)
|
|
124
|
+
assert np.allclose(expected, out_val, equal_nan=True)
|
|
125
|
+
return delta
|
|
126
|
+
|
|
127
|
+
|
|
128
|
+
def test_file_ata_medium(tmp_path):
|
|
129
|
+
write_read_test_file_ata(100, 1000, 33, tmp_path)
|
|
130
|
+
|
|
131
|
+
|
|
132
|
+
def test_file_aat_medium(tmp_path):
|
|
133
|
+
write_read_test_file_aat(100, 1000, 34, tmp_path)
|
|
134
|
+
|
|
135
|
+
|
|
136
|
+
# # Too slow
|
|
137
|
+
# def test_file_ata_giant(tmp_path):
|
|
138
|
+
# write_read_test_file_ata(100_000, 10_000, 1000, tmp_path)
|
|
139
|
+
|
|
140
|
+
# # Too slow
|
|
141
|
+
# def test_file_aat_giant(tmp_path):
|
|
142
|
+
# write_read_test_file_ata(1_000, 10_000, 100, tmp_path)
|
|
143
|
+
|
|
144
|
+
|
|
145
|
+
def test_file_ata_small(shared_datadir):
|
|
146
|
+
filename = shared_datadir / "small_array.memmap"
|
|
147
|
+
|
|
148
|
+
out_val = file_ata(filename, 0, 2, 3, 2)
|
|
149
|
+
print(out_val)
|
|
150
|
+
|
|
151
|
+
expected = np.array([[17.0, 22.0, 27.0], [22.0, 29.0, 36.0], [27.0, 36.0, 45.0]])
|
|
152
|
+
print(expected)
|
|
153
|
+
assert np.allclose(expected, out_val, equal_nan=True)
|
|
154
|
+
|
|
155
|
+
|
|
156
|
+
def test_file_aat_small(shared_datadir):
|
|
157
|
+
filename = shared_datadir / "small_array.memmap"
|
|
158
|
+
|
|
159
|
+
out_val = file_aat(filename, 0, iid_count=2, sid_count=3, iid_step=1)
|
|
160
|
+
print(out_val)
|
|
161
|
+
|
|
162
|
+
expected = np.array([[14.0, 32.0], [32.0, 77.0]])
|
|
163
|
+
print(expected)
|
|
164
|
+
assert np.allclose(expected, out_val, equal_nan=True)
|
|
165
|
+
|
|
166
|
+
|
|
167
|
+
def mmultfile_b_less_aatb(a_snp_mem_map, b, log_frequency=0, force_python_only=False):
|
|
168
|
+
# Without memory efficiency
|
|
169
|
+
# a=a_snp_mem_map.val
|
|
170
|
+
# aTb = np.dot(a.T,b)
|
|
171
|
+
# aaTb = b-np.dot(a,aTb)
|
|
172
|
+
# return aTb, aaTb
|
|
173
|
+
|
|
174
|
+
if force_python_only:
|
|
175
|
+
aTb = np.zeros(
|
|
176
|
+
(a_snp_mem_map.shape[1], b.shape[1])
|
|
177
|
+
) # b can be destroyed. Is everything is in best order, i.e. F vs C
|
|
178
|
+
aaTb = b.copy()
|
|
179
|
+
b_mem = np.array(b, order="F")
|
|
180
|
+
with open(a_snp_mem_map.filename, "rb") as U_fp:
|
|
181
|
+
U_fp.seek(a_snp_mem_map.offset)
|
|
182
|
+
for i in range(a_snp_mem_map.shape[1]):
|
|
183
|
+
a_mem = np.fromfile(
|
|
184
|
+
U_fp, dtype=np.float64, count=a_snp_mem_map.shape[0]
|
|
185
|
+
)
|
|
186
|
+
if log_frequency > 0 and i % log_frequency == 0:
|
|
187
|
+
logging.info("{0}/{1}".format(i, a_snp_mem_map.shape[1]))
|
|
188
|
+
aTb[i, :] = np.dot(a_mem, b_mem)
|
|
189
|
+
aaTb -= np.dot(a_mem.reshape(-1, 1), aTb[i : i + 1, :])
|
|
190
|
+
else:
|
|
191
|
+
b1 = np.array(b, order="F")
|
|
192
|
+
aTb = np.zeros((a_snp_mem_map.shape[1], b.shape[1]))
|
|
193
|
+
aaTb = np.array(b1, order="F")
|
|
194
|
+
|
|
195
|
+
file_b_less_aatbx(
|
|
196
|
+
str(a_snp_mem_map.filename),
|
|
197
|
+
a_snp_mem_map.offset,
|
|
198
|
+
a_snp_mem_map.shape[0], # row count
|
|
199
|
+
b1, # B copy 1 in "F" order
|
|
200
|
+
aaTb, # B copy 2 in "F" order
|
|
201
|
+
aTb, # result
|
|
202
|
+
num_threads=get_num_threads(None),
|
|
203
|
+
log_frequency=log_frequency,
|
|
204
|
+
)
|
|
205
|
+
|
|
206
|
+
return aTb, aaTb
|
|
207
|
+
|
|
208
|
+
|
|
209
|
+
def write_read_test_file_b_less_aatbx(
|
|
210
|
+
iid_count, a_sid_count, b_sid_count, log_frequency, tmp_path, do_both=True
|
|
211
|
+
):
|
|
212
|
+
offset = 640
|
|
213
|
+
file_path = tmp_path / f"{iid_count}x{a_sid_count}_o{offset}_array.memmap"
|
|
214
|
+
mm = np.memmap(
|
|
215
|
+
file_path,
|
|
216
|
+
dtype="float64",
|
|
217
|
+
mode="w+",
|
|
218
|
+
offset=offset,
|
|
219
|
+
shape=(iid_count, a_sid_count),
|
|
220
|
+
order="F",
|
|
221
|
+
)
|
|
222
|
+
mm[:] = np.linspace(0, 1, mm.size).reshape(mm.shape)
|
|
223
|
+
mm.flush()
|
|
224
|
+
|
|
225
|
+
b = np.array(
|
|
226
|
+
np.linspace(0, 1, iid_count * b_sid_count).reshape(
|
|
227
|
+
(iid_count, b_sid_count), order="F"
|
|
228
|
+
)
|
|
229
|
+
)
|
|
230
|
+
b_again = b.copy()
|
|
231
|
+
|
|
232
|
+
logging.info("Calling Rust")
|
|
233
|
+
aTb, aaTb = mmultfile_b_less_aatb(
|
|
234
|
+
mm, b_again, log_frequency, force_python_only=False
|
|
235
|
+
)
|
|
236
|
+
|
|
237
|
+
if do_both:
|
|
238
|
+
logging.info("Calling Python")
|
|
239
|
+
aTb_python, aaTb_python = mmultfile_b_less_aatb(
|
|
240
|
+
mm, b, log_frequency, force_python_only=True
|
|
241
|
+
)
|
|
242
|
+
|
|
243
|
+
if (
|
|
244
|
+
not np.abs(aTb_python - aTb).max() < 1e-8
|
|
245
|
+
or not np.abs(aaTb_python - aaTb).max() < 1e-8
|
|
246
|
+
):
|
|
247
|
+
raise AssertionError(
|
|
248
|
+
"Expect Python and Rust to get the same mmultfile_b_less_aatb answer"
|
|
249
|
+
)
|
|
250
|
+
|
|
251
|
+
|
|
252
|
+
def test_file_b_less_aatbx_medium(tmp_path):
|
|
253
|
+
write_read_test_file_b_less_aatbx(500, 400, 100, 10, tmp_path, do_both=True)
|
|
254
|
+
|
|
255
|
+
|
|
256
|
+
def test_file_b_less_aatbx_medium2(tmp_path):
|
|
257
|
+
write_read_test_file_b_less_aatbx(5_000, 400, 100, 100, tmp_path, do_both=True)
|
|
258
|
+
|
|
259
|
+
|
|
260
|
+
# Slow and doesn't check answer
|
|
261
|
+
# def test_file_b_less_aatbx_2(tmp_path):
|
|
262
|
+
# write_read_test_file_b_less_aatbx(50_000, 4000, 1000, 100, tmp_path, do_both=False)
|
|
263
|
+
|
|
264
|
+
|
|
265
|
+
if __name__ == "__main__":
|
|
266
|
+
logging.basicConfig(level=logging.INFO)
|
|
267
|
+
shared_datadir = Path(r"D:\OneDrive\programs\bed-reader\bed_reader\tests\data")
|
|
268
|
+
tmp_path = Path(r"m:/deldir/tests")
|
|
269
|
+
|
|
270
|
+
if False:
|
|
271
|
+
small2_array = np.array(
|
|
272
|
+
[[1.0, 2.0, 3.0, 4.0], [5.0, 6.0, 7.0, 8.0], [9.0, 10.0, 11.0, 12.0]],
|
|
273
|
+
order="F",
|
|
274
|
+
)
|
|
275
|
+
mm = np.memmap(
|
|
276
|
+
tmp_path / "small2_array.memmap",
|
|
277
|
+
dtype="float64",
|
|
278
|
+
mode="w+",
|
|
279
|
+
shape=(3, 4),
|
|
280
|
+
order="F",
|
|
281
|
+
)
|
|
282
|
+
mm[:] = small2_array[:]
|
|
283
|
+
print(mm.offset)
|
|
284
|
+
mm.flush()
|
|
285
|
+
|
|
286
|
+
if False:
|
|
287
|
+
small_array = np.array([[1.0, 2.0, 3.0], [4.0, 5.0, 6.0]], order="F")
|
|
288
|
+
mm = np.memmap(
|
|
289
|
+
tmp_path / "small_array.memmap",
|
|
290
|
+
dtype="float64",
|
|
291
|
+
mode="w+",
|
|
292
|
+
shape=(2, 3),
|
|
293
|
+
order="F",
|
|
294
|
+
)
|
|
295
|
+
mm[:] = small_array[:]
|
|
296
|
+
print(mm.offset)
|
|
297
|
+
mm.flush()
|
|
298
|
+
|
|
299
|
+
if False:
|
|
300
|
+
mm = np.memmap(
|
|
301
|
+
tmp_path / "100x1000_o640_array.memmap",
|
|
302
|
+
dtype="float64",
|
|
303
|
+
mode="w+",
|
|
304
|
+
offset=640,
|
|
305
|
+
shape=(100, 1000),
|
|
306
|
+
order="F",
|
|
307
|
+
)
|
|
308
|
+
total = mm.shape[0] * mm.shape[1]
|
|
309
|
+
lin = np.linspace(0, 1, total).reshape(mm.shape)
|
|
310
|
+
|
|
311
|
+
mm[:] = lin[:]
|
|
312
|
+
print(mm.offset)
|
|
313
|
+
mm.flush()
|
|
314
|
+
|
|
315
|
+
if False:
|
|
316
|
+
mm = np.memmap(
|
|
317
|
+
tmp_path / "1000x10000_o640_array.memmap",
|
|
318
|
+
dtype="float64",
|
|
319
|
+
mode="w+",
|
|
320
|
+
offset=640,
|
|
321
|
+
shape=(1000, 10_000),
|
|
322
|
+
order="F",
|
|
323
|
+
)
|
|
324
|
+
total = mm.shape[0] * mm.shape[1]
|
|
325
|
+
lin = np.linspace(0, 1, total).reshape(mm.shape)
|
|
326
|
+
|
|
327
|
+
mm[:] = lin[:]
|
|
328
|
+
print(mm.offset)
|
|
329
|
+
mm.flush()
|
|
330
|
+
|
|
331
|
+
if False:
|
|
332
|
+
mm = np.memmap(
|
|
333
|
+
tmp_path / "10_000x100_000_o6400_array.memmap",
|
|
334
|
+
dtype="float64",
|
|
335
|
+
mode="w+",
|
|
336
|
+
offset=640,
|
|
337
|
+
shape=(10_000, 100_000),
|
|
338
|
+
order="F",
|
|
339
|
+
)
|
|
340
|
+
total = mm.shape[0] * mm.shape[1]
|
|
341
|
+
lin = np.linspace(0, 1, total).reshape(mm.shape)
|
|
342
|
+
|
|
343
|
+
mm[:] = lin[:]
|
|
344
|
+
print(mm.offset)
|
|
345
|
+
mm.flush()
|
|
346
|
+
|
|
347
|
+
if False:
|
|
348
|
+
mm = np.memmap(
|
|
349
|
+
tmp_path / "100_000x10_000_o6400_array.memmap",
|
|
350
|
+
dtype="float64",
|
|
351
|
+
mode="w+",
|
|
352
|
+
offset=640,
|
|
353
|
+
shape=(100_000, 10_000),
|
|
354
|
+
order="F",
|
|
355
|
+
)
|
|
356
|
+
total = mm.shape[0] * mm.shape[1]
|
|
357
|
+
lin = np.linspace(0, 1, total).reshape(mm.shape)
|
|
358
|
+
|
|
359
|
+
mm[:] = lin[:]
|
|
360
|
+
print(mm.offset)
|
|
361
|
+
mm.flush()
|
|
362
|
+
|
|
363
|
+
# test_file_b_less_aatbx_2(tmp_path)
|
|
364
|
+
# test_zero_files(tmp_path)
|
|
365
|
+
# test_index(shared_datadir)
|
|
366
|
+
# test_c_reader_bed(shared_datadir)
|
|
367
|
+
# test_read1(shared_datadir)
|
|
368
|
+
pytest.main([__file__])
|