1
|
|
|
# -*- coding: utf-8 -*- |
2
|
|
|
|
3
|
|
|
from contextlib import contextmanager |
4
|
|
|
from time import time |
5
|
|
|
|
6
|
|
|
from pytest import main, fixture, mark |
7
|
|
|
|
8
|
|
|
from ocrd import Resolver |
9
|
|
|
from ocrd_utils import MIME_TO_EXT, getLogger |
10
|
|
|
from ocrd_models import OcrdMets |
11
|
|
|
|
12
|
|
|
logger = getLogger('ocrd.benchmark.mets') |
13
|
|
|
|
14
|
|
|
GRPS_REG = ['SEG-REG', 'SEG-REPAIR', 'SEG-REG-DESKEW', 'SEG-REG-DESKEW-CLIP', 'SEG-LINE', 'SEG-REPAIR-LINE', 'SEG-LINE-RESEG-DEWARP'] |
15
|
|
|
GRPS_IMG = ['FULL', 'PRESENTATION', 'BIN', 'CROP', 'BIN2', 'BIN-DENOISE', 'BIN-DENOISE-DESKEW', 'OCR'] |
16
|
|
|
|
17
|
|
|
REGIONS_PER_PAGE = 2 |
18
|
|
|
LINES_PER_REGION = 2 |
19
|
|
|
FILES_PER_PAGE = len(GRPS_IMG) * LINES_PER_REGION + len(GRPS_REG) * REGIONS_PER_PAGE |
20
|
|
|
|
21
|
|
|
# Caching is disabled by default |
22
|
|
View Code Duplication |
def _build_mets(number_of_pages, force=False, cache_flag=False): |
|
|
|
|
23
|
|
|
mets = OcrdMets.empty_mets(cache_flag=cache_flag) |
24
|
|
|
mets._number_of_pages = number_of_pages |
25
|
|
|
|
26
|
|
|
for n in ['%04d' % (n + 1) for n in range(number_of_pages)]: |
27
|
|
|
_add_file = lambda n, fileGrp, mimetype, ID=None: mets.add_file( |
28
|
|
|
fileGrp, |
29
|
|
|
mimetype=mimetype, |
30
|
|
|
pageId='PHYS_%s' % n, |
31
|
|
|
ID=ID if ID else '%s_%s_%s' % (fileGrp, n, MIME_TO_EXT.get(mimetype)[1:].upper()), |
32
|
|
|
url='%s/%s%s' % (fileGrp, ID if ID else '%s_%s_%s' % (fileGrp, n, MIME_TO_EXT.get(mimetype)[1:].upper()), MIME_TO_EXT.get(mimetype)) |
33
|
|
|
) |
34
|
|
|
for grp in GRPS_IMG: |
35
|
|
|
# LINES_PER_REGION = 2 |
36
|
|
|
_add_file(n, grp, 'image/tiff') |
37
|
|
|
_add_file(n, grp, 'application/vnd.prima.page+xml') |
38
|
|
|
for grp in GRPS_REG: |
39
|
|
|
# REGIONS_PER_PAGE = 2 |
40
|
|
|
for region_n in range(REGIONS_PER_PAGE): |
41
|
|
|
_add_file(n, grp, 'image/png', '%s_%s_region%s' % (grp, n, region_n)) |
42
|
|
|
|
43
|
|
|
return mets |
44
|
|
|
|
45
|
|
|
def assert_len(expected_len, mets, kwargs): |
46
|
|
|
test_list = mets.find_all_files(**kwargs) |
47
|
|
|
assert expected_len == len(test_list) |
48
|
|
|
|
49
|
|
|
def benchmark_find_files(number_of_pages, mets): |
50
|
|
|
benchmark_find_files_filegrp(number_of_pages, mets) |
51
|
|
|
benchmark_find_files_fileid(number_of_pages, mets) |
52
|
|
|
benchmark_find_files_physical_page(number_of_pages, mets) |
53
|
|
|
# This is not really useful to measure. |
54
|
|
|
# We iterate all files in both cached and non-cached in the same routine |
55
|
|
|
# When no specific search parameters are provided |
56
|
|
|
# benchmark_find_files_all(number_of_pages, mets) |
57
|
|
|
|
58
|
|
|
def benchmark_find_files_filegrp(number_of_pages, mets): |
59
|
|
|
# Best case - first fileGrp |
60
|
|
|
assert_len((number_of_pages * REGIONS_PER_PAGE), mets, dict(fileGrp='SEG-REG')) |
61
|
|
|
# Worst case - does not exist |
62
|
|
|
assert_len(0, mets, dict(fileGrp='SEG-REG-NOTEXIST')) |
63
|
|
|
|
64
|
|
|
def benchmark_find_files_fileid(number_of_pages, mets): |
65
|
|
|
# Best case - first file ID |
66
|
|
|
assert_len(1, mets, dict(ID='FULL_0001_TIF', fileGrp='FULL')) |
67
|
|
|
# Worst case - does not exist |
68
|
|
|
assert_len(0, mets, dict(ID='FULL_0001_TIF-NOTEXISTS', fileGrp='FULL-NOTEXIST')) |
69
|
|
|
|
70
|
|
|
def benchmark_find_files_physical_page(number_of_pages, mets): |
71
|
|
|
# Best case - first physical page |
72
|
|
|
assert_len(FILES_PER_PAGE, mets, dict(pageId='PHYS_0001')) |
73
|
|
|
# Worst case - does not exist |
74
|
|
|
assert_len(0, mets, dict(pageId='PHYS_0001-NOTEXISTS')) |
75
|
|
|
|
76
|
|
|
# Get all files, i.e., pass an empty search parameter -> dict() |
77
|
|
|
def benchmark_find_files_all(number_of_pages, mets): |
78
|
|
|
assert_len((number_of_pages * FILES_PER_PAGE), mets, dict()) |
79
|
|
|
|
80
|
|
|
|
81
|
|
|
|
82
|
|
|
|
83
|
|
|
# ---- BENCHMARKING for 50-500-1000-2000-5000 pages ---- # |
84
|
|
|
|
85
|
|
|
# ----- 50 pages -> build, search, build (cached), search (cached) ----- # |
86
|
|
|
mets_50 = None |
87
|
|
|
@mark.benchmark(group="build", max_time=0.1, min_rounds=1, disable_gc=False, warmup=False) |
88
|
|
|
def test_b50(benchmark): |
89
|
|
|
@benchmark |
90
|
|
|
def result(): |
91
|
|
|
global mets_50 |
92
|
|
|
mets_50 = _build_mets(50, force=True) |
93
|
|
|
|
94
|
|
|
@mark.benchmark(group="search", max_time=0.1, min_rounds=1, disable_gc=False, warmup=False) |
95
|
|
|
def test_s50(benchmark): |
96
|
|
|
@benchmark |
97
|
|
|
def ret(): |
98
|
|
|
global mets_50 |
99
|
|
|
benchmark_find_files(50, mets_50) |
100
|
|
|
del mets_50 |
101
|
|
|
|
102
|
|
|
mets_c_50 = None |
103
|
|
|
@mark.benchmark(group="build_cached", max_time=0.1, min_rounds=1, disable_gc=False, warmup=False) |
104
|
|
|
def test_b50_c(benchmark): |
105
|
|
|
@benchmark |
106
|
|
|
def result(): |
107
|
|
|
global mets_c_50 |
108
|
|
|
mets_c_50 = _build_mets(50, force=True, cache_flag=True) |
109
|
|
|
|
110
|
|
|
@mark.benchmark(group="search_cached", max_time=0.1, min_rounds=1, disable_gc=False, warmup=False) |
111
|
|
|
def test_s50_c(benchmark): |
112
|
|
|
@benchmark |
113
|
|
|
def ret(): |
114
|
|
|
global mets_c_50 |
115
|
|
|
benchmark_find_files(50, mets_c_50) |
116
|
|
|
del mets_c_50 |
117
|
|
|
# ----------------------------------------------------------------------- # |
118
|
|
|
|
119
|
|
|
|
120
|
|
|
|
121
|
|
|
# ----- 500 pages -> build, search, build (cached), search (cached) ----- # |
122
|
|
|
mets_500 = None |
123
|
|
|
@mark.benchmark(group="build", max_time=0.1, min_rounds=1, disable_gc=False, warmup=False) |
124
|
|
|
def test_b500(benchmark): |
125
|
|
|
@benchmark |
126
|
|
|
def result(): |
127
|
|
|
global mets_500 |
128
|
|
|
mets_500 = _build_mets(500, force=True) |
129
|
|
|
|
130
|
|
|
@mark.benchmark(group="search", max_time=0.1, min_rounds=1, disable_gc=False, warmup=False) |
131
|
|
|
def test_s500(benchmark): |
132
|
|
|
@benchmark |
133
|
|
|
def ret(): |
134
|
|
|
global mets_500 |
135
|
|
|
benchmark_find_files(500, mets_500) |
136
|
|
|
del mets_500 |
137
|
|
|
|
138
|
|
|
|
139
|
|
|
mets_c_500 = None |
140
|
|
|
@mark.benchmark(group="build_cached", max_time=0.1, min_rounds=1, disable_gc=False, warmup=False) |
141
|
|
|
def test_b500_c(benchmark): |
142
|
|
|
@benchmark |
143
|
|
|
def result(): |
144
|
|
|
global mets_c_500 |
145
|
|
|
mets_c_500 = _build_mets(500, force=True, cache_flag=True) |
146
|
|
|
|
147
|
|
|
@mark.benchmark(group="search_cached", max_time=0.1, min_rounds=1, disable_gc=False, warmup=False) |
148
|
|
|
def test_s500_c(benchmark): |
149
|
|
|
@benchmark |
150
|
|
|
def ret(): |
151
|
|
|
global mets_c_500 |
152
|
|
|
benchmark_find_files(500, mets_c_500) |
153
|
|
|
del mets_c_500 |
154
|
|
|
|
155
|
|
|
# ----------------------------------------------------------------------- # |
156
|
|
|
|
157
|
|
|
|
158
|
|
|
|
159
|
|
|
# ----- 1000 pages -> build, search, build (cached), search (cached) ----- # |
160
|
|
|
mets_1000 = None |
161
|
|
|
@mark.benchmark(group="build", max_time=0.1, min_rounds=1, disable_gc=False, warmup=False) |
162
|
|
|
def test_b1000(benchmark): |
163
|
|
|
@benchmark |
164
|
|
|
def result(): |
165
|
|
|
global mets_1000 |
166
|
|
|
mets_1000 = _build_mets(1000, force=True) |
167
|
|
|
|
168
|
|
|
@mark.benchmark(group="search", max_time=0.1, min_rounds=1, disable_gc=False, warmup=False) |
169
|
|
|
def test_s1000(benchmark): |
170
|
|
|
@benchmark |
171
|
|
|
def ret(): |
172
|
|
|
global mets_1000 |
173
|
|
|
benchmark_find_files(1000, mets_1000) |
174
|
|
|
del mets_1000 |
175
|
|
|
|
176
|
|
|
mets_c_1000 = None |
177
|
|
|
@mark.benchmark(group="build_cached", max_time=0.1, min_rounds=1, disable_gc=False, warmup=False) |
178
|
|
|
def test_b1000_c(benchmark): |
179
|
|
|
@benchmark |
180
|
|
|
def result(): |
181
|
|
|
global mets_c_1000 |
182
|
|
|
mets_c_1000 = _build_mets(1000, force=True, cache_flag=True) |
183
|
|
|
|
184
|
|
|
@mark.benchmark(group="search_cached", max_time=0.1, min_rounds=1, disable_gc=False, warmup=False) |
185
|
|
|
def test_s1000_c(benchmark): |
186
|
|
|
@benchmark |
187
|
|
|
def ret(): |
188
|
|
|
global mets_c_1000 |
189
|
|
|
benchmark_find_files(1000, mets_c_1000) |
190
|
|
|
del mets_c_1000 |
191
|
|
|
|
192
|
|
|
# ------------------------------------------------------------------------ # |
193
|
|
|
|
194
|
|
|
|
195
|
|
|
|
196
|
|
|
# ----- 2000 pages -> build, search, build (cached), search (cached) ----- # |
197
|
|
|
mets_2000 = None |
198
|
|
|
@mark.benchmark(group="build", max_time=0.1, min_rounds=1, disable_gc=False, warmup=False) |
199
|
|
|
def test_b2000(benchmark): |
200
|
|
|
@benchmark |
201
|
|
|
def result(): |
202
|
|
|
global mets_2000 |
203
|
|
|
mets_2000 = _build_mets(2000, force=True) |
204
|
|
|
|
205
|
|
|
@mark.benchmark(group="search", max_time=0.1, min_rounds=1, disable_gc=False, warmup=False) |
206
|
|
|
def test_s2000(benchmark): |
207
|
|
|
@benchmark |
208
|
|
|
def ret(): |
209
|
|
|
global mets_2000 |
210
|
|
|
benchmark_find_files(2000, mets_2000) |
211
|
|
|
del mets_2000 |
212
|
|
|
|
213
|
|
|
mets_c_2000 = None |
214
|
|
|
@mark.benchmark(group="build_cached", max_time=0.1, min_rounds=1, disable_gc=False, warmup=False) |
215
|
|
|
def test_b2000_c(benchmark): |
216
|
|
|
@benchmark |
217
|
|
|
def result(): |
218
|
|
|
global mets_c_2000 |
219
|
|
|
mets_c_2000 = _build_mets(2000, force=True, cache_flag=True) |
220
|
|
|
|
221
|
|
|
@mark.benchmark(group="search_cached", max_time=0.1, min_rounds=1, disable_gc=False, warmup=False) |
222
|
|
|
def test_s2000_c(benchmark): |
223
|
|
|
@benchmark |
224
|
|
|
def ret(): |
225
|
|
|
global mets_c_2000 |
226
|
|
|
benchmark_find_files(2000, mets_c_2000) |
227
|
|
|
del mets_c_2000 |
228
|
|
|
|
229
|
|
|
# ------------------------------------------------------------------------ # |
230
|
|
|
|
231
|
|
|
|
232
|
|
|
|
233
|
|
|
# ----- 5000 pages -> build, search, build (cached), search (cached) ----- # |
234
|
|
|
mets_5000 = None |
235
|
|
|
@mark.benchmark(group="build", max_time=0.1, min_rounds=1, disable_gc=False, warmup=False) |
236
|
|
|
def test_b5000(benchmark): |
237
|
|
|
@benchmark |
238
|
|
|
def result(): |
239
|
|
|
global mets_5000 |
240
|
|
|
mets_5000 = _build_mets(5000, force=True) |
241
|
|
|
|
242
|
|
|
@mark.benchmark(group="search", max_time=0.1, min_rounds=1, disable_gc=False, warmup=False) |
243
|
|
|
def test_s5000(benchmark): |
244
|
|
|
@benchmark |
245
|
|
|
def ret(): |
246
|
|
|
global mets_5000 |
247
|
|
|
benchmark_find_files(5000, mets_5000) |
248
|
|
|
del mets_5000 |
249
|
|
|
|
250
|
|
|
mets_c_5000 = None |
251
|
|
|
@mark.benchmark(group="build_cached", max_time=0.1, min_rounds=1, disable_gc=False, warmup=False) |
252
|
|
|
def test_b5000_c(benchmark): |
253
|
|
|
@benchmark |
254
|
|
|
def result(): |
255
|
|
|
global mets_c_5000 |
256
|
|
|
mets_c_5000 = _build_mets(5000, force=True, cache_flag=True) |
257
|
|
|
|
258
|
|
|
@mark.benchmark(group="search_cached", max_time=0.1, min_rounds=1, disable_gc=False, warmup=False) |
259
|
|
|
def test_s5000_c(benchmark): |
260
|
|
|
@benchmark |
261
|
|
|
def ret(): |
262
|
|
|
global mets_c_5000 |
263
|
|
|
benchmark_find_files(5000, mets_c_5000) |
264
|
|
|
del mets_c_5000 |
265
|
|
|
|
266
|
|
|
# ------------------------------------------------------------------------ # |
267
|
|
|
|
268
|
|
|
if __name__ == '__main__': |
269
|
|
|
args = [''] |
270
|
|
|
# args.append('--benchmark-max-time=10') |
271
|
|
|
# args.append('--benchmark-min-time=0.1') |
272
|
|
|
# args.append('--benchmark-warmup=False') |
273
|
|
|
# args.append('--benchmark-disable-gc') |
274
|
|
|
args.append('--benchmark-verbose') |
275
|
|
|
args.append('--benchmark-min-rounds=1') |
276
|
|
|
args.append('--tb=short') |
277
|
|
|
main(args) |
278
|
|
|
|