|
1
|
|
|
""" |
|
2
|
|
|
Helper methods for running and documenting processors |
|
3
|
|
|
""" |
|
4
|
|
|
from os import chdir, getcwd |
|
5
|
|
|
from time import perf_counter, process_time |
|
6
|
|
|
from functools import lru_cache |
|
7
|
|
|
import json |
|
8
|
|
|
import inspect |
|
9
|
|
|
from subprocess import run |
|
10
|
|
|
from typing import List |
|
11
|
|
|
|
|
12
|
|
|
from click import wrap_text |
|
13
|
|
|
from ocrd.workspace import Workspace |
|
14
|
|
|
from ocrd_utils import freeze_args, getLogger, config, setOverrideLogLevel, getLevelName |
|
15
|
|
|
|
|
16
|
|
|
|
|
17
|
|
|
__all__ = [ |
|
18
|
|
|
'generate_processor_help', |
|
19
|
|
|
'run_cli', |
|
20
|
|
|
'run_processor' |
|
21
|
|
|
] |
|
22
|
|
|
|
|
23
|
|
|
|
|
24
|
|
|
def _get_workspace(workspace=None, resolver=None, mets_url=None, working_dir=None, mets_server_url=None): |
|
25
|
|
|
if workspace is None: |
|
26
|
|
|
if resolver is None: |
|
27
|
|
|
raise Exception("Need to pass a resolver to create a workspace") |
|
28
|
|
|
if mets_url is None: |
|
29
|
|
|
raise Exception("Need to pass mets_url to create a workspace") |
|
30
|
|
|
workspace = resolver.workspace_from_url(mets_url, dst_dir=working_dir, mets_server_url=mets_server_url) |
|
31
|
|
|
return workspace |
|
32
|
|
|
|
|
33
|
|
|
def run_processor( |
|
34
|
|
|
processorClass, |
|
35
|
|
|
mets_url=None, |
|
36
|
|
|
resolver=None, |
|
37
|
|
|
workspace=None, |
|
38
|
|
|
page_id=None, |
|
39
|
|
|
log_level=None, |
|
40
|
|
|
input_file_grp=None, |
|
41
|
|
|
output_file_grp=None, |
|
42
|
|
|
show_resource=None, |
|
43
|
|
|
list_resources=False, |
|
44
|
|
|
parameter=None, |
|
45
|
|
|
parameter_override=None, |
|
46
|
|
|
working_dir=None, |
|
47
|
|
|
mets_server_url=None, |
|
48
|
|
|
instance_caching=False # TODO don't set this yet! |
|
49
|
|
|
): # pylint: disable=too-many-locals |
|
50
|
|
|
""" |
|
51
|
|
|
Instantiate a Pythonic processor, open a workspace, run the processor and save the workspace. |
|
52
|
|
|
|
|
53
|
|
|
If :py:attr:`workspace` is not none, reuse that. Otherwise, instantiate an |
|
54
|
|
|
:py:class:`~ocrd.Workspace` for :py:attr:`mets_url` (and :py:attr:`working_dir`) |
|
55
|
|
|
by using :py:meth:`ocrd.Resolver.workspace_from_url` (i.e. open or clone local workspace). |
|
56
|
|
|
|
|
57
|
|
|
Instantiate a Python object for :py:attr:`processorClass`, passing: |
|
58
|
|
|
- the workspace, |
|
59
|
|
|
- :py:attr:`page_id` |
|
60
|
|
|
- :py:attr:`input_file_grp` |
|
61
|
|
|
- :py:attr:`output_file_grp` |
|
62
|
|
|
- :py:attr:`parameter` (after applying any :py:attr:`parameter_override` settings) |
|
63
|
|
|
|
|
64
|
|
|
Warning: Avoid setting the `instance_caching` flag to True. It may have unexpected side effects. |
|
65
|
|
|
This flag is used for an experimental feature we would like to adopt in future. |
|
66
|
|
|
|
|
67
|
|
|
Run the processor on the workspace (creating output files in the filesystem). |
|
68
|
|
|
|
|
69
|
|
|
Finally, write back the workspace (updating the METS in the filesystem). |
|
70
|
|
|
|
|
71
|
|
|
Args: |
|
72
|
|
|
processorClass (object): Python class of the module processor. |
|
73
|
|
|
""" |
|
74
|
|
|
if log_level: |
|
75
|
|
|
setOverrideLogLevel(log_level) |
|
76
|
|
|
workspace = _get_workspace( |
|
77
|
|
|
workspace, |
|
78
|
|
|
resolver, |
|
79
|
|
|
mets_url, |
|
80
|
|
|
working_dir, |
|
81
|
|
|
mets_server_url |
|
82
|
|
|
) |
|
83
|
|
|
log = getLogger('ocrd.processor.helpers.run_processor') |
|
84
|
|
|
log.debug("Running processor %s", processorClass) |
|
85
|
|
|
|
|
86
|
|
|
old_cwd = getcwd() |
|
87
|
|
|
processor = get_processor( |
|
88
|
|
|
processor_class=processorClass, |
|
89
|
|
|
parameter=parameter, |
|
90
|
|
|
workspace=None, |
|
91
|
|
|
page_id=page_id, |
|
92
|
|
|
input_file_grp=input_file_grp, |
|
93
|
|
|
output_file_grp=output_file_grp, |
|
94
|
|
|
instance_caching=instance_caching |
|
95
|
|
|
) |
|
96
|
|
|
processor.workspace = workspace |
|
97
|
|
|
chdir(processor.workspace.directory) |
|
98
|
|
|
|
|
99
|
|
|
ocrd_tool = processor.ocrd_tool |
|
100
|
|
|
name = '%s v%s' % (ocrd_tool['executable'], processor.version) |
|
101
|
|
|
otherrole = ocrd_tool['steps'][0] |
|
102
|
|
|
logProfile = getLogger('ocrd.process.profile') |
|
103
|
|
|
log.debug("Processor instance %s (%s doing %s)", processor, name, otherrole) |
|
104
|
|
|
t0_wall = perf_counter() |
|
105
|
|
|
t0_cpu = process_time() |
|
106
|
|
|
if any(x in config.OCRD_PROFILE for x in ['RSS', 'PSS']): |
|
107
|
|
|
backend = 'psutil_pss' if 'PSS' in config.OCRD_PROFILE else 'psutil' |
|
108
|
|
|
from memory_profiler import memory_usage |
|
109
|
|
|
from sparklines import sparklines |
|
110
|
|
|
try: |
|
111
|
|
|
mem_usage = memory_usage(proc=processor.process, |
|
112
|
|
|
# only run process once |
|
113
|
|
|
max_iterations=1, |
|
114
|
|
|
interval=.1, timeout=None, timestamps=True, |
|
115
|
|
|
# include sub-processes |
|
116
|
|
|
multiprocess=True, include_children=True, |
|
117
|
|
|
# get proportional set size instead of RSS |
|
118
|
|
|
backend=backend) |
|
119
|
|
|
except Exception as err: |
|
120
|
|
|
log.exception("Failure in processor '%s'" % ocrd_tool['executable']) |
|
121
|
|
|
raise err |
|
122
|
|
|
finally: |
|
123
|
|
|
chdir(old_cwd) |
|
124
|
|
|
mem_usage_values = [mem for mem, _ in mem_usage] |
|
125
|
|
|
mem_output = 'memory consumption: ' |
|
126
|
|
|
mem_output += ''.join(sparklines(mem_usage_values)) |
|
127
|
|
|
mem_output += ' max: %.2f MiB min: %.2f MiB' % (max(mem_usage_values), min(mem_usage_values)) |
|
128
|
|
|
logProfile.info(mem_output) |
|
129
|
|
|
else: |
|
130
|
|
|
try: |
|
131
|
|
|
processor.process() |
|
132
|
|
|
except Exception as err: |
|
133
|
|
|
log.exception("Failure in processor '%s'" % ocrd_tool['executable']) |
|
134
|
|
|
raise err |
|
135
|
|
|
finally: |
|
136
|
|
|
chdir(old_cwd) |
|
137
|
|
|
|
|
138
|
|
|
t1_wall = perf_counter() - t0_wall |
|
139
|
|
|
t1_cpu = process_time() - t0_cpu |
|
140
|
|
|
logProfile.info("Executing processor '%s' took %fs (wall) %fs (CPU)( [--input-file-grp='%s' --output-file-grp='%s' --parameter='%s' --page-id='%s']" % ( |
|
141
|
|
|
ocrd_tool['executable'], |
|
142
|
|
|
t1_wall, |
|
143
|
|
|
t1_cpu, |
|
144
|
|
|
processor.input_file_grp or '', |
|
145
|
|
|
processor.output_file_grp or '', |
|
146
|
|
|
json.dumps(processor.parameter) or '', |
|
147
|
|
|
processor.page_id or '' |
|
148
|
|
|
)) |
|
149
|
|
|
workspace.mets.add_agent( |
|
150
|
|
|
name=name, |
|
151
|
|
|
_type='OTHER', |
|
152
|
|
|
othertype='SOFTWARE', |
|
153
|
|
|
role='OTHER', |
|
154
|
|
|
otherrole=otherrole, |
|
155
|
|
|
notes=[({'option': 'input-file-grp'}, processor.input_file_grp or ''), |
|
156
|
|
|
({'option': 'output-file-grp'}, processor.output_file_grp or ''), |
|
157
|
|
|
({'option': 'parameter'}, json.dumps(processor.parameter or '')), |
|
158
|
|
|
({'option': 'page-id'}, processor.page_id or '')] |
|
159
|
|
|
) |
|
160
|
|
|
workspace.save_mets() |
|
161
|
|
|
return processor |
|
162
|
|
|
|
|
163
|
|
|
|
|
164
|
|
|
def run_cli( |
|
165
|
|
|
executable, |
|
166
|
|
|
mets_url=None, |
|
167
|
|
|
resolver=None, |
|
168
|
|
|
workspace=None, |
|
169
|
|
|
page_id=None, |
|
170
|
|
|
overwrite=None, |
|
171
|
|
|
log_level=None, |
|
172
|
|
|
input_file_grp=None, |
|
173
|
|
|
output_file_grp=None, |
|
174
|
|
|
parameter=None, |
|
175
|
|
|
working_dir=None, |
|
176
|
|
|
mets_server_url=None, |
|
177
|
|
|
): |
|
178
|
|
|
""" |
|
179
|
|
|
Open a workspace and run a processor on the command line. |
|
180
|
|
|
|
|
181
|
|
|
If :py:attr:`workspace` is not none, reuse that. Otherwise, instantiate an |
|
182
|
|
|
:py:class:`~ocrd.Workspace` for :py:attr:`mets_url` (and :py:attr:`working_dir`) |
|
183
|
|
|
by using :py:meth:`ocrd.Resolver.workspace_from_url` (i.e. open or clone local workspace). |
|
184
|
|
|
|
|
185
|
|
|
Run the processor CLI :py:attr:`executable` on the workspace, passing: |
|
186
|
|
|
- the workspace, |
|
187
|
|
|
- :py:attr:`page_id` |
|
188
|
|
|
- :py:attr:`input_file_grp` |
|
189
|
|
|
- :py:attr:`output_file_grp` |
|
190
|
|
|
- :py:attr:`parameter` (after applying any :py:attr:`parameter_override` settings) |
|
191
|
|
|
|
|
192
|
|
|
(Will create output files and update the in the filesystem). |
|
193
|
|
|
|
|
194
|
|
|
Args: |
|
195
|
|
|
executable (string): Executable name of the module processor. |
|
196
|
|
|
""" |
|
197
|
|
|
workspace = _get_workspace(workspace, resolver, mets_url, working_dir) |
|
198
|
|
|
args = [executable, '--working-dir', workspace.directory] |
|
199
|
|
|
args += ['--mets', mets_url] |
|
200
|
|
|
if log_level: |
|
201
|
|
|
args += ['--log-level', log_level if isinstance(log_level, str) else getLevelName(log_level)] |
|
202
|
|
|
if page_id: |
|
203
|
|
|
args += ['--page-id', page_id] |
|
204
|
|
|
if input_file_grp: |
|
205
|
|
|
args += ['--input-file-grp', input_file_grp] |
|
206
|
|
|
if output_file_grp: |
|
207
|
|
|
args += ['--output-file-grp', output_file_grp] |
|
208
|
|
|
if parameter: |
|
209
|
|
|
args += ['--parameter', parameter] |
|
210
|
|
|
if overwrite: |
|
211
|
|
|
args += ['--overwrite'] |
|
212
|
|
|
if mets_server_url: |
|
213
|
|
|
args += ['--mets-server-url', mets_server_url] |
|
214
|
|
|
log = getLogger('ocrd.processor.helpers.run_cli') |
|
215
|
|
|
log.debug("Running subprocess '%s'", ' '.join(args)) |
|
216
|
|
|
result = run(args, check=False) |
|
217
|
|
|
return result.returncode |
|
218
|
|
|
|
|
219
|
|
|
def generate_processor_help(ocrd_tool, processor_instance=None, subcommand=None): |
|
220
|
|
|
"""Generate a string describing the full CLI of this processor including params. |
|
221
|
|
|
|
|
222
|
|
|
Args: |
|
223
|
|
|
ocrd_tool (dict): this processor's ``tools`` section of the module's ``ocrd-tool.json`` |
|
224
|
|
|
processor_instance (object, optional): the processor implementation |
|
225
|
|
|
(for adding any module/class/function docstrings) |
|
226
|
|
|
subcommand (string): 'worker' or 'server' |
|
227
|
|
|
""" |
|
228
|
|
|
doc_help = '' |
|
229
|
|
|
if processor_instance: |
|
230
|
|
|
module = inspect.getmodule(processor_instance) |
|
231
|
|
|
if module and module.__doc__: |
|
232
|
|
|
doc_help += '\n' + inspect.cleandoc(module.__doc__) + '\n' |
|
233
|
|
|
if processor_instance.__doc__: |
|
234
|
|
|
doc_help += '\n' + inspect.cleandoc(processor_instance.__doc__) + '\n' |
|
235
|
|
|
if processor_instance.process.__doc__: |
|
236
|
|
|
doc_help += '\n' + inspect.cleandoc(processor_instance.process.__doc__) + '\n' |
|
237
|
|
|
if doc_help: |
|
238
|
|
|
doc_help = '\n\n' + wrap_text(doc_help, width=72, |
|
239
|
|
|
initial_indent=' > ', |
|
240
|
|
|
subsequent_indent=' > ', |
|
241
|
|
|
preserve_paragraphs=True) |
|
242
|
|
|
subcommands = '''\ |
|
243
|
|
|
worker Start a processing worker rather than do local processing |
|
244
|
|
|
server Start a processor server rather than do local processing |
|
245
|
|
|
''' |
|
246
|
|
|
|
|
247
|
|
|
processing_worker_options = '''\ |
|
248
|
|
|
--queue The RabbitMQ server address in format |
|
249
|
|
|
"amqp://{user}:{pass}@{host}:{port}/{vhost}" |
|
250
|
|
|
[amqp://admin:admin@localhost:5672] |
|
251
|
|
|
--database The MongoDB server address in format |
|
252
|
|
|
"mongodb://{host}:{port}" |
|
253
|
|
|
[mongodb://localhost:27018] |
|
254
|
|
|
--log-filename Filename to redirect STDOUT/STDERR to, |
|
255
|
|
|
if specified. |
|
256
|
|
|
''' |
|
257
|
|
|
|
|
258
|
|
|
processing_server_options = '''\ |
|
259
|
|
|
--address The Processor server address in format |
|
260
|
|
|
"{host}:{port}" |
|
261
|
|
|
--database The MongoDB server address in format |
|
262
|
|
|
"mongodb://{host}:{port}" |
|
263
|
|
|
[mongodb://localhost:27018] |
|
264
|
|
|
''' |
|
265
|
|
|
|
|
266
|
|
|
processing_options = '''\ |
|
267
|
|
|
-m, --mets URL-PATH URL or file path of METS to process [./mets.xml] |
|
268
|
|
|
-w, --working-dir PATH Working directory of local workspace [dirname(URL-PATH)] |
|
269
|
|
|
-I, --input-file-grp USE File group(s) used as input |
|
270
|
|
|
-O, --output-file-grp USE File group(s) used as output |
|
271
|
|
|
-g, --page-id ID Physical page ID(s) to process instead of full document [] |
|
272
|
|
|
--overwrite Remove existing output pages/images |
|
273
|
|
|
(with "--page-id", remove only those) |
|
274
|
|
|
--profile Enable profiling |
|
275
|
|
|
--profile-file PROF-PATH Write cProfile stats to PROF-PATH. Implies "--profile" |
|
276
|
|
|
-p, --parameter JSON-PATH Parameters, either verbatim JSON string |
|
277
|
|
|
or JSON file path |
|
278
|
|
|
-P, --param-override KEY VAL Override a single JSON object key-value pair, |
|
279
|
|
|
taking precedence over --parameter |
|
280
|
|
|
-U, --mets-server-url URL URL of a METS Server for parallel incremental access to METS |
|
281
|
|
|
If URL starts with http:// start an HTTP server there, |
|
282
|
|
|
otherwise URL is a path to an on-demand-created unix socket |
|
283
|
|
|
-l, --log-level [OFF|ERROR|WARN|INFO|DEBUG|TRACE] |
|
284
|
|
|
Override log level globally [INFO] |
|
285
|
|
|
''' |
|
286
|
|
|
|
|
287
|
|
|
information_options = '''\ |
|
288
|
|
|
-C, --show-resource RESNAME Dump the content of processor resource RESNAME |
|
289
|
|
|
-L, --list-resources List names of processor resources |
|
290
|
|
|
-J, --dump-json Dump tool description as JSON |
|
291
|
|
|
-D, --dump-module-dir Show the 'module' resource location path for this processor |
|
292
|
|
|
-h, --help Show this message |
|
293
|
|
|
-V, --version Show version |
|
294
|
|
|
''' |
|
295
|
|
|
|
|
296
|
|
|
parameter_help = '' |
|
297
|
|
|
if 'parameters' not in ocrd_tool or not ocrd_tool['parameters']: |
|
298
|
|
|
parameter_help = ' NONE\n' |
|
299
|
|
|
else: |
|
300
|
|
|
def wrap(s): |
|
301
|
|
|
return wrap_text(s, initial_indent=' '*3, |
|
302
|
|
|
subsequent_indent=' '*4, |
|
303
|
|
|
width=72, preserve_paragraphs=True) |
|
304
|
|
|
for param_name, param in ocrd_tool['parameters'].items(): |
|
305
|
|
|
parameter_help += wrap('"%s" [%s%s]' % ( |
|
306
|
|
|
param_name, |
|
307
|
|
|
param['type'], |
|
308
|
|
|
' - REQUIRED' if 'required' in param and param['required'] else |
|
309
|
|
|
' - %s' % json.dumps(param['default']) if 'default' in param else '')) |
|
310
|
|
|
parameter_help += '\n ' + wrap(param['description']) |
|
311
|
|
|
if 'enum' in param: |
|
312
|
|
|
parameter_help += '\n ' + wrap('Possible values: %s' % json.dumps(param['enum'])) |
|
313
|
|
|
parameter_help += "\n" |
|
314
|
|
|
|
|
315
|
|
|
if not subcommand: |
|
316
|
|
|
return f'''\ |
|
317
|
|
|
Usage: {ocrd_tool['executable']} [worker|server] [OPTIONS] |
|
318
|
|
|
|
|
319
|
|
|
{ocrd_tool['description']}{doc_help} |
|
320
|
|
|
|
|
321
|
|
|
Subcommands: |
|
322
|
|
|
{subcommands} |
|
323
|
|
|
Options for processing: |
|
324
|
|
|
{processing_options} |
|
325
|
|
|
Options for information: |
|
326
|
|
|
{information_options} |
|
327
|
|
|
Parameters: |
|
328
|
|
|
{parameter_help} |
|
329
|
|
|
''' |
|
330
|
|
|
elif subcommand == 'worker': |
|
331
|
|
|
return f'''\ |
|
332
|
|
|
Usage: {ocrd_tool['executable']} worker [OPTIONS] |
|
333
|
|
|
|
|
334
|
|
|
Run {ocrd_tool['executable']} as a processing worker. |
|
335
|
|
|
|
|
336
|
|
|
{ocrd_tool['description']}{doc_help} |
|
337
|
|
|
|
|
338
|
|
|
Options: |
|
339
|
|
|
{processing_worker_options} |
|
340
|
|
|
''' |
|
341
|
|
|
elif subcommand == 'server': |
|
342
|
|
|
return f'''\ |
|
343
|
|
|
Usage: {ocrd_tool['executable']} server [OPTIONS] |
|
344
|
|
|
|
|
345
|
|
|
Run {ocrd_tool['executable']} as a processor sever. |
|
346
|
|
|
|
|
347
|
|
|
{ocrd_tool['description']}{doc_help} |
|
348
|
|
|
|
|
349
|
|
|
Options: |
|
350
|
|
|
{processing_server_options} |
|
351
|
|
|
''' |
|
352
|
|
|
else: |
|
353
|
|
|
pass |
|
354
|
|
|
|
|
355
|
|
|
|
|
356
|
|
|
# Taken from https://github.com/OCR-D/core/pull/884 |
|
357
|
|
|
@freeze_args |
|
358
|
|
|
@lru_cache(maxsize=config.OCRD_MAX_PROCESSOR_CACHE) |
|
359
|
|
|
def get_cached_processor(parameter: dict, processor_class): |
|
360
|
|
|
""" |
|
361
|
|
|
Call this function to get back an instance of a processor. |
|
362
|
|
|
The results are cached based on the parameters. |
|
363
|
|
|
Args: |
|
364
|
|
|
parameter (dict): a dictionary of parameters. |
|
365
|
|
|
processor_class: the concrete `:py:class:~ocrd.Processor` class. |
|
366
|
|
|
Returns: |
|
367
|
|
|
When the concrete class of the processor is unknown, `None` is returned. |
|
368
|
|
|
Otherwise, an instance of the `:py:class:~ocrd.Processor` is returned. |
|
369
|
|
|
""" |
|
370
|
|
|
if processor_class: |
|
371
|
|
|
dict_params = dict(parameter) if parameter else None |
|
372
|
|
|
return processor_class(workspace=None, parameter=dict_params) |
|
373
|
|
|
return None |
|
374
|
|
|
|
|
375
|
|
|
|
|
376
|
|
|
def get_processor( |
|
377
|
|
|
processor_class, |
|
378
|
|
|
parameter: dict, |
|
379
|
|
|
workspace: Workspace = None, |
|
380
|
|
|
page_id: str = None, |
|
381
|
|
|
input_file_grp: List[str] = None, |
|
382
|
|
|
output_file_grp: List[str] = None, |
|
383
|
|
|
instance_caching: bool = False, |
|
384
|
|
|
): |
|
385
|
|
|
if processor_class: |
|
386
|
|
|
if instance_caching: |
|
387
|
|
|
cached_processor = get_cached_processor( |
|
388
|
|
|
parameter=parameter, |
|
389
|
|
|
processor_class=processor_class |
|
390
|
|
|
) |
|
391
|
|
|
cached_processor.workspace = workspace |
|
392
|
|
|
cached_processor.page_id = page_id |
|
393
|
|
|
cached_processor.input_file_grp = input_file_grp |
|
394
|
|
|
cached_processor.output_file_grp = output_file_grp |
|
395
|
|
|
return cached_processor |
|
396
|
|
|
return processor_class( |
|
397
|
|
|
workspace=workspace, |
|
398
|
|
|
page_id=page_id, |
|
399
|
|
|
input_file_grp=input_file_grp, |
|
400
|
|
|
output_file_grp=output_file_grp, |
|
401
|
|
|
parameter=parameter |
|
402
|
|
|
) |
|
403
|
|
|
raise ValueError("Processor class is not known") |
|
404
|
|
|
|
OCR-D /
core
