1
|
|
|
""" |
2
|
|
|
Processor base class and helper functions. |
3
|
|
|
""" |
4
|
|
|
|
5
|
|
|
__all__ = [ |
6
|
|
|
'Processor', |
7
|
|
|
'generate_processor_help', |
8
|
|
|
'run_cli', |
9
|
|
|
'run_processor' |
10
|
|
|
] |
11
|
|
|
|
12
|
|
|
from functools import cached_property |
13
|
|
|
from os.path import exists, join |
14
|
|
|
from shutil import copyfileobj |
15
|
|
|
import json |
16
|
|
|
import os |
17
|
|
|
from os import getcwd |
18
|
|
|
from pathlib import Path |
19
|
|
|
from typing import List, Optional, Union, get_args |
20
|
|
|
import sys |
21
|
|
|
import inspect |
22
|
|
|
import tarfile |
23
|
|
|
import io |
24
|
|
|
import weakref |
25
|
|
|
from frozendict import frozendict |
26
|
|
|
from concurrent.futures import ThreadPoolExecutor, TimeoutError |
27
|
|
|
|
28
|
|
|
from click import wrap_text |
29
|
|
|
from deprecated import deprecated |
30
|
|
|
from requests import HTTPError |
31
|
|
|
|
32
|
|
|
from ..workspace import Workspace |
33
|
|
|
from ..mets_server import ClientSideOcrdMets |
34
|
|
|
from ocrd_models.ocrd_file import OcrdFileType |
35
|
|
|
from .ocrd_page_result import OcrdPageResult |
36
|
|
|
from ocrd_utils import ( |
37
|
|
|
VERSION as OCRD_VERSION, |
38
|
|
|
MIMETYPE_PAGE, |
39
|
|
|
MIME_TO_EXT, |
40
|
|
|
config, |
41
|
|
|
getLogger, |
42
|
|
|
list_resource_candidates, |
43
|
|
|
pushd_popd, |
44
|
|
|
list_all_resources, |
45
|
|
|
get_processor_resource_types, |
46
|
|
|
resource_filename, |
47
|
|
|
parse_json_file_with_comments, |
48
|
|
|
make_file_id, |
49
|
|
|
deprecation_warning |
50
|
|
|
) |
51
|
|
|
from ocrd_validators import ParameterValidator |
52
|
|
|
from ocrd_models.ocrd_page import ( |
53
|
|
|
PageType, |
54
|
|
|
AlternativeImageType, |
55
|
|
|
MetadataItemType, |
56
|
|
|
LabelType, |
57
|
|
|
LabelsType, |
58
|
|
|
OcrdPage, |
59
|
|
|
to_xml, |
60
|
|
|
) |
61
|
|
|
from ocrd_modelfactory import page_from_file |
62
|
|
|
from ocrd_validators.ocrd_tool_validator import OcrdToolValidator |
63
|
|
|
|
64
|
|
|
# XXX imports must remain for backwards-compatibility |
65
|
|
|
from .helpers import run_cli, run_processor # pylint: disable=unused-import |
66
|
|
|
|
67
|
|
|
|
68
|
|
|
class ResourceNotFoundError(FileNotFoundError): |
69
|
|
|
""" |
70
|
|
|
An exception signifying the requested processor resource |
71
|
|
|
cannot be resolved. |
72
|
|
|
""" |
73
|
|
|
def __init__(self, name, executable): |
74
|
|
|
self.name = name |
75
|
|
|
self.executable = executable |
76
|
|
|
self.message = (f"Could not find resource '{name}' for executable '{executable}'. " |
77
|
|
|
f"Try 'ocrd resmgr download {executable} {name}' to download this resource.") |
78
|
|
|
super().__init__(self.message) |
79
|
|
|
|
80
|
|
|
class NonUniqueInputFile(ValueError): |
81
|
|
|
""" |
82
|
|
|
An exception signifying the specified fileGrp / pageId / mimetype |
83
|
|
|
selector yields multiple PAGE files, or no PAGE files but multiple images, |
84
|
|
|
or multiple files of that mimetype. |
85
|
|
|
""" |
86
|
|
|
def __init__(self, fileGrp, pageId, mimetype): |
87
|
|
|
self.fileGrp = fileGrp |
88
|
|
|
self.pageId = pageId |
89
|
|
|
self.mimetype = mimetype |
90
|
|
|
self.message = (f"Could not determine unique input file for fileGrp {fileGrp} " |
91
|
|
|
f"and pageId {pageId} under mimetype {mimetype or 'PAGE+image(s)'}") |
92
|
|
|
super().__init__(self.message) |
93
|
|
|
|
94
|
|
|
class MissingInputFile(ValueError): |
95
|
|
|
""" |
96
|
|
|
An exception signifying the specified fileGrp / pageId / mimetype |
97
|
|
|
selector yields no PAGE files, or no PAGE and no image files, |
98
|
|
|
or no files of that mimetype. |
99
|
|
|
""" |
100
|
|
|
def __init__(self, fileGrp, pageId, mimetype): |
101
|
|
|
self.fileGrp = fileGrp |
102
|
|
|
self.pageId = pageId |
103
|
|
|
self.mimetype = mimetype |
104
|
|
|
self.message = (f"Could not find input file for fileGrp {fileGrp} " |
105
|
|
|
f"and pageId {pageId} under mimetype {mimetype or 'PAGE+image(s)'}") |
106
|
|
|
super().__init__(self.message) |
107
|
|
|
|
108
|
|
|
class Processor(): |
109
|
|
|
""" |
110
|
|
|
A processor is a tool that implements the uniform OCR-D |
111
|
|
|
`command-line interface for run-time data processing <https://ocr-d.de/en/spec/cli>`_. |
112
|
|
|
|
113
|
|
|
That is, it executes a single workflow step, or a combination of workflow steps, |
114
|
|
|
on the workspace (represented by local METS). It reads input files for all or selected |
115
|
|
|
physical pages of the input fileGrp(s), computes additional annotation, and writes output |
116
|
|
|
files for them into the output fileGrp(s). It may take a number of optional or mandatory |
117
|
|
|
parameters. |
118
|
|
|
""" |
119
|
|
|
|
120
|
|
|
max_instances : int = -1 |
121
|
|
|
""" |
122
|
|
|
maximum number of cached instances (ignored if negative), to be applied on top of |
123
|
|
|
:py:data:`~ocrd_utils.config.OCRD_MAX_PROCESSOR_CACHE` (i.e. whatever is smaller). |
124
|
|
|
|
125
|
|
|
(Override this if you know how many instances fit into memory - GPU / CPU RAM - at once.) |
126
|
|
|
""" |
127
|
|
|
|
128
|
|
|
max_workers : int = -1 |
129
|
|
|
""" |
130
|
|
|
maximum number of processor threads for page-parallel processing (ignored if negative), |
131
|
|
|
to be applied on top of :py:data:`~ocrd_utils.config.OCRD_MAX_PARALLEL_PAGES` (i.e. |
132
|
|
|
whatever is smaller). |
133
|
|
|
|
134
|
|
|
(Override this if you know how many pages fit into processing units - GPU shaders / CPU cores |
135
|
|
|
- at once, or if your class is not thread-safe.) |
136
|
|
|
""" |
137
|
|
|
|
138
|
|
|
max_page_seconds : int = -1 |
139
|
|
|
""" |
140
|
|
|
maximum number of seconds may be spent processing a single page (ignored if negative), |
141
|
|
|
to be applied on top of :py:data:`~ocrd_utils.config.OCRD_PROCESSING_PAGE_TIMEOUT` |
142
|
|
|
(i.e. whatever is smaller). |
143
|
|
|
|
144
|
|
|
(Override this if you know how costly this processor may be, irrespective of image size |
145
|
|
|
or complexity of the page.) |
146
|
|
|
""" |
147
|
|
|
|
148
|
|
|
@property |
149
|
|
|
def metadata_filename(self) -> str: |
150
|
|
|
""" |
151
|
|
|
Relative location of the ``ocrd-tool.json`` file inside the package. |
152
|
|
|
|
153
|
|
|
Used by :py:data:`metadata_location`. |
154
|
|
|
|
155
|
|
|
(Override if ``ocrd-tool.json`` is not in the root of the module, |
156
|
|
|
e.g. ``namespace/ocrd-tool.json`` or ``data/ocrd-tool.json``). |
157
|
|
|
""" |
158
|
|
|
return 'ocrd-tool.json' |
159
|
|
|
|
160
|
|
|
@cached_property |
161
|
|
|
def metadata_location(self) -> Path: |
162
|
|
|
""" |
163
|
|
|
Absolute path of the ``ocrd-tool.json`` file as distributed with the package. |
164
|
|
|
|
165
|
|
|
Used by :py:data:`metadata_rawdict`. |
166
|
|
|
|
167
|
|
|
(Override if ``ocrd-tool.json`` is not distributed with the Python package.) |
168
|
|
|
""" |
169
|
|
|
module = inspect.getmodule(self) |
170
|
|
|
module_tokens = module.__package__.split('.') |
171
|
|
|
# for namespace packages, we cannot just use the first token |
172
|
|
|
for i in range(len(module_tokens)): |
173
|
|
|
prefix = '.'.join(module_tokens[:i + 1]) |
174
|
|
|
if sys.modules[prefix].__spec__.has_location: |
175
|
|
|
return resource_filename(prefix, self.metadata_filename) |
176
|
|
|
raise Exception("cannot find top-level module prefix for %s", module.__package__) |
177
|
|
|
|
178
|
|
|
@cached_property |
179
|
|
|
def metadata_rawdict(self) -> dict: |
180
|
|
|
""" |
181
|
|
|
Raw (unvalidated, unexpanded) ``ocrd-tool.json`` dict contents of the package. |
182
|
|
|
|
183
|
|
|
Used by :py:data:`metadata`. |
184
|
|
|
|
185
|
|
|
(Override if ``ocrd-tool.json`` is not in a file.) |
186
|
|
|
""" |
187
|
|
|
return parse_json_file_with_comments(self.metadata_location) |
188
|
|
|
|
189
|
|
|
@cached_property |
190
|
|
|
def metadata(self) -> dict: |
191
|
|
|
""" |
192
|
|
|
The ``ocrd-tool.json`` dict contents of the package, according to the OCR-D |
193
|
|
|
`spec <https://ocr-d.de/en/spec/ocrd_tool>`_ for processor tools. |
194
|
|
|
|
195
|
|
|
After deserialisation, it also gets validated against the |
196
|
|
|
`schema <https://ocr-d.de/en/spec/ocrd_tool#definition>`_ with all defaults |
197
|
|
|
expanded. |
198
|
|
|
|
199
|
|
|
Used by :py:data:`ocrd_tool` and :py:data:`version`. |
200
|
|
|
|
201
|
|
|
(Override if you want to provide metadata programmatically instead of a |
202
|
|
|
JSON file.) |
203
|
|
|
""" |
204
|
|
|
metadata = self.metadata_rawdict |
205
|
|
|
report = OcrdToolValidator.validate(metadata) |
206
|
|
|
if not report.is_valid: |
207
|
|
|
self.logger.error(f"The ocrd-tool.json of this processor is {'problematic' if not report.errors else 'invalid'}:\n" |
208
|
|
|
f"{report.to_xml()}.\nPlease open an issue at {metadata.get('git_url', 'the website')}.") |
209
|
|
|
return metadata |
210
|
|
|
|
211
|
|
|
@cached_property |
212
|
|
|
def version(self) -> str: |
213
|
|
|
""" |
214
|
|
|
The program version of the package. |
215
|
|
|
Usually the ``version`` part of :py:data:`metadata`. |
216
|
|
|
|
217
|
|
|
(Override if you do not want to use :py:data:`metadata` lookup |
218
|
|
|
mechanism.) |
219
|
|
|
""" |
220
|
|
|
return self.metadata['version'] |
221
|
|
|
|
222
|
|
|
@cached_property |
223
|
|
|
def executable(self) -> str: |
224
|
|
|
""" |
225
|
|
|
The executable name of this processor tool. Taken from the runtime |
226
|
|
|
filename. |
227
|
|
|
|
228
|
|
|
Used by :py:data:`ocrd_tool` for lookup in :py:data:`metadata`. |
229
|
|
|
|
230
|
|
|
(Override if your entry-point name deviates from the ``executable`` |
231
|
|
|
name, or the processor gets instantiated from another runtime.) |
232
|
|
|
""" |
233
|
|
|
return os.path.basename(inspect.stack()[-1].filename) |
234
|
|
|
|
235
|
|
|
@cached_property |
236
|
|
|
def ocrd_tool(self) -> dict: |
237
|
|
|
""" |
238
|
|
|
The ``ocrd-tool.json`` dict contents of this processor tool. |
239
|
|
|
Usually the :py:data:`executable` key of the ``tools`` part |
240
|
|
|
of :py:data:`metadata`. |
241
|
|
|
|
242
|
|
|
(Override if you do not want to use :py:data:`metadata` lookup |
243
|
|
|
mechanism.) |
244
|
|
|
""" |
245
|
|
|
return self.metadata['tools'][self.executable] |
246
|
|
|
|
247
|
|
|
@property |
248
|
|
|
def parameter(self) -> Optional[dict]: |
249
|
|
|
"""the runtime parameter dict to be used by this processor""" |
250
|
|
|
if hasattr(self, '_parameter'): |
251
|
|
|
return self._parameter |
252
|
|
|
return None |
253
|
|
|
|
254
|
|
|
@parameter.setter |
255
|
|
|
def parameter(self, parameter : dict) -> None: |
256
|
|
|
if self.parameter is not None: |
257
|
|
|
self.shutdown() |
258
|
|
|
parameterValidator = ParameterValidator(self.ocrd_tool) |
259
|
|
|
report = parameterValidator.validate(parameter) |
260
|
|
|
if not report.is_valid: |
261
|
|
|
raise ValueError(f'Invalid parameters:\n{report.to_xml()}') |
262
|
|
|
# make parameter dict read-only |
263
|
|
|
self._parameter = frozendict(parameter) |
264
|
|
|
# (re-)run setup to load models etc |
265
|
|
|
self.setup() |
266
|
|
|
|
267
|
|
|
def __init__( |
268
|
|
|
self, |
269
|
|
|
# FIXME: remove in favor of process_workspace(workspace) |
270
|
|
|
workspace : Optional[Workspace], |
271
|
|
|
ocrd_tool=None, |
272
|
|
|
parameter=None, |
273
|
|
|
input_file_grp=None, |
274
|
|
|
output_file_grp=None, |
275
|
|
|
page_id=None, |
276
|
|
|
download_files=config.OCRD_DOWNLOAD_INPUT, |
277
|
|
|
version=None |
278
|
|
|
): |
279
|
|
|
""" |
280
|
|
|
Instantiate, but do not setup (neither for processing nor other usage). |
281
|
|
|
If given, do parse and validate :py:data:`.parameter`. |
282
|
|
|
|
283
|
|
|
Args: |
284
|
|
|
workspace (:py:class:`~ocrd.Workspace`): The workspace to process. \ |
285
|
|
|
If not ``None``, then `chdir` to that directory. |
286
|
|
|
Deprecated since version 3.0: Should be ``None`` here, but then needs to be set \ |
287
|
|
|
before processing. |
288
|
|
|
Keyword Args: |
289
|
|
|
parameter (string): JSON of the runtime choices for ocrd-tool ``parameters``. \ |
290
|
|
|
Can be ``None`` even for processing, but then needs to be set before running. |
291
|
|
|
input_file_grp (string): comma-separated list of METS ``fileGrp`` used for input. \ |
292
|
|
|
Deprecated since version 3.0: Should be ``None`` here, but then needs to be set \ |
293
|
|
|
before processing. |
294
|
|
|
output_file_grp (string): comma-separated list of METS ``fileGrp`` used for output. \ |
295
|
|
|
Deprecated since version 3.0: Should be ``None`` here, but then needs to be set \ |
296
|
|
|
before processing. |
297
|
|
|
page_id (string): comma-separated list of METS physical ``page`` IDs to process \ |
298
|
|
|
(or empty for all pages). \ |
299
|
|
|
Deprecated since version 3.0: Should be ``None`` here, but then needs to be set \ |
300
|
|
|
before processing. |
301
|
|
|
download_files (boolean): Whether input files will be downloaded prior to processing, \ |
302
|
|
|
defaults to :py:attr:`ocrd_utils.config.OCRD_DOWNLOAD_INPUT` which is ``True`` by default |
303
|
|
|
""" |
304
|
|
|
if ocrd_tool is not None: |
305
|
|
|
deprecation_warning("Passing 'ocrd_tool' as keyword argument to Processor is deprecated - " |
306
|
|
|
"use or override metadata/executable/ocrd-tool properties instead") |
307
|
|
|
self.ocrd_tool = ocrd_tool |
308
|
|
|
self.executable = ocrd_tool['executable'] |
309
|
|
|
if version is not None: |
310
|
|
|
deprecation_warning("Passing 'version' as keyword argument to Processor is deprecated - " |
311
|
|
|
"use or override metadata/version properties instead") |
312
|
|
|
self.version = version |
313
|
|
|
if workspace is not None: |
314
|
|
|
deprecation_warning("Passing a workspace argument other than 'None' to Processor " |
315
|
|
|
"is deprecated - pass as argument to process_workspace instead") |
316
|
|
|
self.workspace = workspace |
317
|
|
|
self.old_pwd = getcwd() |
318
|
|
|
os.chdir(self.workspace.directory) |
319
|
|
|
if input_file_grp is not None: |
320
|
|
|
deprecation_warning("Passing an input_file_grp kwarg other than 'None' to Processor " |
321
|
|
|
"is deprecated - pass as argument to process_workspace instead") |
322
|
|
|
self.input_file_grp = input_file_grp |
323
|
|
|
if output_file_grp is not None: |
324
|
|
|
deprecation_warning("Passing an output_file_grp kwarg other than 'None' to Processor " |
325
|
|
|
"is deprecated - pass as argument to process_workspace instead") |
326
|
|
|
self.output_file_grp = output_file_grp |
327
|
|
|
if page_id is not None: |
328
|
|
|
deprecation_warning("Passing a page_id kwarg other than 'None' to Processor " |
329
|
|
|
"is deprecated - pass as argument to process_workspace instead") |
330
|
|
|
self.page_id = page_id or None |
331
|
|
|
self.download = download_files |
332
|
|
|
#: The logger to be used by processor implementations. |
333
|
|
|
# `ocrd.processor.base` internals should use :py:attr:`self._base_logger` |
334
|
|
|
self.logger = getLogger(f'ocrd.processor.{self.__class__.__name__}') |
335
|
|
|
self._base_logger = getLogger('ocrd.processor.base') |
336
|
|
|
if parameter is not None: |
337
|
|
|
self.parameter = parameter |
338
|
|
|
# ensure that shutdown gets called at destruction |
339
|
|
|
self._finalizer = weakref.finalize(self, self.shutdown) |
340
|
|
|
# workaround for deprecated#72 (@deprecated decorator does not work for subclasses): |
341
|
|
|
setattr(self, 'process', |
342
|
|
|
deprecated(version='3.0', reason='process() should be replaced with process_page() and process_workspace()')(getattr(self, 'process'))) |
343
|
|
|
|
344
|
|
|
def show_help(self, subcommand=None): |
345
|
|
|
""" |
346
|
|
|
Print a usage description including the standard CLI and all of this processor's ocrd-tool |
347
|
|
|
parameters and docstrings. |
348
|
|
|
""" |
349
|
|
|
print(generate_processor_help(self.ocrd_tool, processor_instance=self, subcommand=subcommand)) |
350
|
|
|
|
351
|
|
|
def show_version(self): |
352
|
|
|
""" |
353
|
|
|
Print information on this processor's version and OCR-D version. |
354
|
|
|
""" |
355
|
|
|
print("Version %s, ocrd/core %s" % (self.version, OCRD_VERSION)) |
356
|
|
|
|
357
|
|
|
def verify(self): |
358
|
|
|
""" |
359
|
|
|
Verify that :py:attr:`input_file_grp` and :py:attr:`output_file_grp` fulfill the processor's requirements. |
360
|
|
|
""" |
361
|
|
|
assert self.input_file_grp is not None |
362
|
|
|
assert self.output_file_grp is not None |
363
|
|
|
input_file_grps = self.input_file_grp.split(',') |
364
|
|
|
output_file_grps = self.output_file_grp.split(',') |
365
|
|
|
def assert_file_grp_cardinality(grps : List[str], spec : Union[int, List[int]], msg): |
366
|
|
|
if isinstance(spec, int): |
367
|
|
|
if spec > 0: |
368
|
|
|
assert len(grps) == spec, msg % (len(grps), str(spec)) |
369
|
|
|
else: |
370
|
|
|
assert isinstance(spec, list) |
371
|
|
|
minimum = spec[0] |
372
|
|
|
maximum = spec[1] |
373
|
|
|
if minimum > 0: |
374
|
|
|
assert len(grps) >= minimum, msg % (len(grps), str(spec)) |
375
|
|
|
if maximum > 0: |
376
|
|
|
assert len(grps) <= maximum, msg % (len(grps), str(spec)) |
377
|
|
|
assert_file_grp_cardinality(input_file_grps, self.ocrd_tool['input_file_grp_cardinality'], |
378
|
|
|
"Unexpected number of input file groups %d vs %s") |
379
|
|
|
assert_file_grp_cardinality(output_file_grps, self.ocrd_tool['output_file_grp_cardinality'], |
380
|
|
|
"Unexpected number of output file groups %d vs %s") |
381
|
|
|
for input_file_grp in input_file_grps: |
382
|
|
|
assert input_file_grp in self.workspace.mets.file_groups |
383
|
|
|
# keep this for backwards compatibility: |
384
|
|
|
return True |
385
|
|
|
|
386
|
|
|
def dump_json(self): |
387
|
|
|
""" |
388
|
|
|
Print :py:attr:`ocrd_tool` on stdout. |
389
|
|
|
""" |
390
|
|
|
print(json.dumps(self.ocrd_tool, indent=True)) |
391
|
|
|
|
392
|
|
|
def dump_module_dir(self): |
393
|
|
|
""" |
394
|
|
|
Print :py:attr:`moduledir` on stdout. |
395
|
|
|
""" |
396
|
|
|
print(self.moduledir) |
397
|
|
|
|
398
|
|
|
def list_resources(self): |
399
|
|
|
""" |
400
|
|
|
Find all installed resource files in the search paths and print their path names. |
401
|
|
|
""" |
402
|
|
|
for res in self.list_all_resources(): |
403
|
|
|
print(res) |
404
|
|
|
|
405
|
|
|
def setup(self) -> None: |
406
|
|
|
""" |
407
|
|
|
Prepare the processor for actual data processing, |
408
|
|
|
prior to changing to the workspace directory but |
409
|
|
|
after parsing parameters. |
410
|
|
|
|
411
|
|
|
(Override this to load models into memory etc.) |
412
|
|
|
""" |
413
|
|
|
pass |
414
|
|
|
|
415
|
|
|
def shutdown(self) -> None: |
416
|
|
|
""" |
417
|
|
|
Bring down the processor after data processing, |
418
|
|
|
after to changing back from the workspace directory but |
419
|
|
|
before exiting (or setting up with different parameters). |
420
|
|
|
|
421
|
|
|
(Override this to unload models from memory etc.) |
422
|
|
|
""" |
423
|
|
|
pass |
424
|
|
|
|
425
|
|
|
@deprecated(version='3.0', reason='process() should be replaced with process_page_pcgts() or process_page_file() or process_workspace()') |
426
|
|
|
def process(self) -> None: |
427
|
|
|
""" |
428
|
|
|
Process all files of the :py:data:`workspace` |
429
|
|
|
from the given :py:data:`input_file_grp` |
430
|
|
|
to the given :py:data:`output_file_grp` |
431
|
|
|
for the given :py:data:`page_id` (or all pages) |
432
|
|
|
under the given :py:data:`parameter`. |
433
|
|
|
|
434
|
|
|
(This contains the main functionality and needs to be |
435
|
|
|
overridden by subclasses.) |
436
|
|
|
""" |
437
|
|
|
raise NotImplementedError() |
438
|
|
|
|
439
|
|
|
def process_workspace(self, workspace: Workspace) -> None: |
440
|
|
|
""" |
441
|
|
|
Process all files of the given ``workspace``, |
442
|
|
|
from the given :py:data:`input_file_grp` |
443
|
|
|
to the given :py:data:`output_file_grp` |
444
|
|
|
for the given :py:data:`page_id` (or all pages) |
445
|
|
|
under the given :py:data:`parameter`. |
446
|
|
|
|
447
|
|
|
(This will iterate over pages and files, calling |
448
|
|
|
:py:meth:`.process_page_file` and handling exceptions. |
449
|
|
|
It should be overridden by subclasses to handle cases |
450
|
|
|
like post-processing or computation across pages.) |
451
|
|
|
""" |
452
|
|
|
with pushd_popd(workspace.directory): |
453
|
|
|
self.workspace = workspace |
454
|
|
|
self.verify() |
455
|
|
|
try: |
456
|
|
|
nr_succeeded = 0 |
457
|
|
|
nr_skipped = 0 |
458
|
|
|
nr_copied = 0 |
459
|
|
|
|
460
|
|
|
# set up multithreading |
461
|
|
|
max_workers = max(0, config.OCRD_MAX_PARALLEL_PAGES) |
462
|
|
|
if self.max_workers > 0 and self.max_workers < config.OCRD_MAX_PARALLEL_PAGES: |
463
|
|
|
self._base_logger.info("limiting number of threads from %d to %d", max_workers, self.max_workers) |
464
|
|
|
max_workers = self.max_workers |
465
|
|
|
if max_workers > 1: |
466
|
|
|
assert isinstance(workspace.mets, ClientSideOcrdMets), \ |
467
|
|
|
"OCRD_MAX_PARALLEL_PAGES>1 requires also using --mets-server-url" |
468
|
|
|
max_seconds = max(0, config.OCRD_PROCESSING_PAGE_TIMEOUT) |
469
|
|
|
if self.max_page_seconds > 0 and self.max_page_seconds < config.OCRD_PROCESSING_PAGE_TIMEOUT: |
470
|
|
|
self._base_logger.info("limiting page timeout from %d to %d sec", max_seconds, self.max_page_seconds) |
471
|
|
|
max_seconds = self.max_page_seconds |
472
|
|
|
executor = ThreadPoolExecutor( |
473
|
|
|
max_workers=max_workers or 1, |
474
|
|
|
thread_name_prefix=f"pagetask.{workspace.mets.unique_identifier}" |
475
|
|
|
) |
476
|
|
|
self._base_logger.debug("started executor %s with %d workers", str(executor), max_workers or 1) |
477
|
|
|
tasks = {} |
478
|
|
|
|
479
|
|
|
for input_file_tuple in self.zip_input_files(on_error='abort', require_first=False): |
480
|
|
|
input_files : List[Optional[OcrdFileType]] = [None] * len(input_file_tuple) |
481
|
|
|
page_id = next(input_file.pageId |
482
|
|
|
for input_file in input_file_tuple |
483
|
|
|
if input_file) |
484
|
|
|
self._base_logger.info(f"preparing page {page_id}") |
485
|
|
|
for i, input_file in enumerate(input_file_tuple): |
486
|
|
|
if input_file is None: |
487
|
|
|
# file/page not found in this file grp |
488
|
|
|
continue |
489
|
|
|
input_files[i] = input_file |
490
|
|
|
if not self.download: |
491
|
|
|
continue |
492
|
|
|
try: |
493
|
|
|
input_files[i] = self.workspace.download_file(input_file) |
494
|
|
|
except (ValueError, FileNotFoundError, HTTPError) as e: |
495
|
|
|
self._base_logger.error(repr(e)) |
496
|
|
|
self._base_logger.warning(f"failed downloading file {input_file} for page {page_id}") |
497
|
|
|
# process page |
498
|
|
|
tasks[executor.submit(self.process_page_file, *input_files)] = (page_id, input_files) |
499
|
|
|
self._base_logger.debug("submitted %d processing tasks", len(tasks)) |
500
|
|
|
|
501
|
|
|
for task in tasks: |
502
|
|
|
# wait for results, handle errors |
503
|
|
|
page_id, input_files = tasks[task] |
504
|
|
|
# FIXME: differentiate error cases in various ways: |
505
|
|
|
# - ResourceNotFoundError → use ResourceManager to download (once), then retry |
506
|
|
|
# - transient (I/O or OOM) error → maybe sleep, retry |
507
|
|
|
# - persistent (data) error → skip / dummy / raise |
508
|
|
|
try: |
509
|
|
|
self._base_logger.debug("waiting for output of task %s (page %s) max_seconds=%d", task, page_id, max_seconds) |
510
|
|
|
task.result(timeout=max_seconds or None) |
511
|
|
|
nr_succeeded += 1 |
512
|
|
|
# exclude NotImplementedError, so we can try process() below |
513
|
|
|
except NotImplementedError: |
514
|
|
|
raise |
515
|
|
|
# handle input failures separately |
516
|
|
|
except FileExistsError as err: |
517
|
|
|
if config.OCRD_EXISTING_OUTPUT == 'ABORT': |
518
|
|
|
raise err |
519
|
|
|
if config.OCRD_EXISTING_OUTPUT == 'SKIP': |
520
|
|
|
continue |
521
|
|
|
if config.OCRD_EXISTING_OUTPUT == 'OVERWRITE': |
522
|
|
|
# too late here, must not happen |
523
|
|
|
raise Exception(f"got {err} despite OCRD_EXISTING_OUTPUT==OVERWRITE") |
524
|
|
|
# broad coverage of output failures (including TimeoutError) |
525
|
|
|
except (Exception, TimeoutError) as err: |
526
|
|
|
# FIXME: add re-usable/actionable logging |
527
|
|
|
if config.OCRD_MISSING_OUTPUT == 'ABORT': |
528
|
|
|
self._base_logger.error(f"Failure on page {page_id}: {str(err) or err.__class__.__name__}") |
529
|
|
|
raise err |
530
|
|
|
self._base_logger.exception(f"Failure on page {page_id}: {str(err) or err.__class__.__name__}") |
531
|
|
|
if config.OCRD_MISSING_OUTPUT == 'SKIP': |
532
|
|
|
nr_skipped += 1 |
533
|
|
|
continue |
534
|
|
|
if config.OCRD_MISSING_OUTPUT == 'COPY': |
535
|
|
|
self._copy_page_file(input_files[0]) |
536
|
|
|
nr_copied += 1 |
537
|
|
|
else: |
538
|
|
|
desc = config.describe('OCRD_MISSING_OUTPUT', wrap_text=False, indent_text=False) |
539
|
|
|
raise ValueError(f"unknown configuration value {config.OCRD_MISSING_OUTPUT} - {desc}") |
540
|
|
|
|
541
|
|
|
if nr_skipped > 0 and nr_succeeded / nr_skipped < config.OCRD_MAX_MISSING_OUTPUTS: |
542
|
|
|
raise Exception(f"too many failures with skipped output ({nr_skipped})") |
543
|
|
|
if nr_copied > 0 and nr_succeeded / nr_copied < config.OCRD_MAX_MISSING_OUTPUTS: |
544
|
|
|
raise Exception(f"too many failures with fallback output ({nr_skipped})") |
545
|
|
|
executor.shutdown() |
546
|
|
|
|
547
|
|
|
except NotImplementedError: |
548
|
|
|
# fall back to deprecated method |
549
|
|
|
self.process() |
550
|
|
|
|
551
|
|
|
def _copy_page_file(self, input_file : OcrdFileType) -> None: |
552
|
|
|
""" |
553
|
|
|
Copy the given ``input_file`` of the :py:data:`workspace`, |
554
|
|
|
representing one physical page (passed as one opened |
555
|
|
|
:py:class:`~ocrd_models.OcrdFile` per input fileGrp) |
556
|
|
|
and add it as if it was a processing result. |
557
|
|
|
""" |
558
|
|
|
input_pcgts : OcrdPage |
559
|
|
|
assert isinstance(input_file, get_args(OcrdFileType)) |
560
|
|
|
self._base_logger.debug(f"parsing file {input_file.ID} for page {input_file.pageId}") |
561
|
|
|
try: |
562
|
|
|
input_pcgts = page_from_file(input_file) |
563
|
|
|
except ValueError as err: |
564
|
|
|
# not PAGE and not an image to generate PAGE for |
565
|
|
|
self._base_logger.error(f"non-PAGE input for page {input_file.pageId}: {err}") |
566
|
|
|
return |
567
|
|
|
output_file_id = make_file_id(input_file, self.output_file_grp) |
568
|
|
|
input_pcgts.set_pcGtsId(output_file_id) |
569
|
|
|
self.add_metadata(input_pcgts) |
570
|
|
|
self.workspace.add_file( |
571
|
|
|
file_id=output_file_id, |
572
|
|
|
file_grp=self.output_file_grp, |
573
|
|
|
page_id=input_file.pageId, |
574
|
|
|
local_filename=os.path.join(self.output_file_grp, output_file_id + '.xml'), |
575
|
|
|
mimetype=MIMETYPE_PAGE, |
576
|
|
|
content=to_xml(input_pcgts), |
577
|
|
|
force=config.OCRD_EXISTING_OUTPUT == 'OVERWRITE', |
578
|
|
|
) |
579
|
|
|
|
580
|
|
|
def process_page_file(self, *input_files : Optional[OcrdFileType]) -> None: |
581
|
|
|
""" |
582
|
|
|
Process the given ``input_files`` of the :py:data:`workspace`, |
583
|
|
|
representing one physical page (passed as one opened |
584
|
|
|
:py:class:`.OcrdFile` per input fileGrp) |
585
|
|
|
under the given :py:data:`.parameter`, and make sure the |
586
|
|
|
results get added accordingly. |
587
|
|
|
|
588
|
|
|
(This uses :py:meth:`.process_page_pcgts`, but should be overridden by subclasses |
589
|
|
|
to handle cases like multiple output fileGrps, non-PAGE input etc.) |
590
|
|
|
""" |
591
|
|
|
input_pcgts : List[Optional[OcrdPage]] = [None] * len(input_files) |
592
|
|
|
assert isinstance(input_files[0], get_args(OcrdFileType)) |
593
|
|
|
page_id = input_files[0].pageId |
594
|
|
|
self._base_logger.info("processing page %s", page_id) |
595
|
|
|
for i, input_file in enumerate(input_files): |
596
|
|
|
assert isinstance(input_file, get_args(OcrdFileType)) |
597
|
|
|
self._base_logger.debug(f"parsing file {input_file.ID} for page {page_id}") |
598
|
|
|
try: |
599
|
|
|
page_ = page_from_file(input_file) |
600
|
|
|
assert isinstance(page_, OcrdPage) |
601
|
|
|
input_pcgts[i] = page_ |
602
|
|
|
except ValueError as err: |
603
|
|
|
# not PAGE and not an image to generate PAGE for |
604
|
|
|
self._base_logger.error(f"non-PAGE input for page {page_id}: {err}") |
605
|
|
|
output_file_id = make_file_id(input_files[0], self.output_file_grp) |
606
|
|
|
result = self.process_page_pcgts(*input_pcgts, page_id=page_id) |
607
|
|
|
for image_result in result.images: |
608
|
|
|
image_file_id = f'{output_file_id}_{image_result.file_id_suffix}' |
609
|
|
|
image_file_path = join(self.output_file_grp, f'{image_file_id}.png') |
610
|
|
|
if isinstance(image_result.alternative_image, PageType): |
611
|
|
|
# special case: not an alternative image, but replacing the original image |
612
|
|
|
# (this is needed by certain processors when the original's coordinate system |
613
|
|
|
# cannot or must not be kept) |
614
|
|
|
image_result.alternative_image.set_imageFilename(image_file_path) |
615
|
|
|
image_result.alternative_image.set_imageWidth(image_result.pil.width) |
616
|
|
|
image_result.alternative_image.set_imageHeight(image_result.pil.height) |
617
|
|
|
elif isinstance(image_result.alternative_image, AlternativeImageType): |
618
|
|
|
image_result.alternative_image.set_filename(image_file_path) |
619
|
|
|
else: |
620
|
|
|
raise ValueError(f"process_page_pcgts returned an OcrdPageResultImage of unknown type " |
621
|
|
|
f"{type(image_result.alternative_image)}") |
622
|
|
|
self.workspace.save_image_file( |
623
|
|
|
image_result.pil, |
624
|
|
|
image_file_id, |
625
|
|
|
self.output_file_grp, |
626
|
|
|
page_id=page_id, |
627
|
|
|
file_path=image_file_path, |
628
|
|
|
force=config.OCRD_EXISTING_OUTPUT == 'OVERWRITE', |
629
|
|
|
) |
630
|
|
|
result.pcgts.set_pcGtsId(output_file_id) |
631
|
|
|
self.add_metadata(result.pcgts) |
632
|
|
|
self.workspace.add_file( |
633
|
|
|
file_id=output_file_id, |
634
|
|
|
file_grp=self.output_file_grp, |
635
|
|
|
page_id=page_id, |
636
|
|
|
local_filename=os.path.join(self.output_file_grp, output_file_id + '.xml'), |
637
|
|
|
mimetype=MIMETYPE_PAGE, |
638
|
|
|
content=to_xml(result.pcgts), |
639
|
|
|
force=config.OCRD_EXISTING_OUTPUT == 'OVERWRITE', |
640
|
|
|
) |
641
|
|
|
|
642
|
|
|
def process_page_pcgts(self, *input_pcgts : Optional[OcrdPage], page_id : Optional[str] = None) -> OcrdPageResult: |
643
|
|
|
""" |
644
|
|
|
Process the given ``input_pcgts`` of the :py:data:`.workspace`, |
645
|
|
|
representing one physical page (passed as one parsed |
646
|
|
|
:py:class:`.OcrdPage` per input fileGrp) |
647
|
|
|
under the given :py:data:`.parameter`, and return the |
648
|
|
|
resulting :py:class:`.OcrdPageResult`. |
649
|
|
|
|
650
|
|
|
Optionally, add to the ``images`` attribute of the resulting |
651
|
|
|
:py:class:`.OcrdPageResult` instances of :py:class:`.OcrdPageResultImage`, |
652
|
|
|
which have required fields for ``pil`` (:py:class:`PIL.Image` image data), |
653
|
|
|
``file_id_suffix`` (used for generating IDs of the saved image) and |
654
|
|
|
``alternative_image`` (reference of the :py:class:`ocrd_models.ocrd_page.AlternativeImageType` |
655
|
|
|
for setting the filename of the saved image). |
656
|
|
|
|
657
|
|
|
(This contains the main functionality and must be overridden by subclasses, |
658
|
|
|
unless it does not get called by some overriden :py:meth:`.process_page_file`.) |
659
|
|
|
""" |
660
|
|
|
raise NotImplementedError() |
661
|
|
|
|
662
|
|
|
def add_metadata(self, pcgts: OcrdPage) -> None: |
663
|
|
|
""" |
664
|
|
|
Add PAGE-XML :py:class:`~ocrd_models.ocrd_page.MetadataItemType` ``MetadataItem`` describing |
665
|
|
|
the processing step and runtime parameters to :py:class:`.OcrdPage` ``pcgts``. |
666
|
|
|
""" |
667
|
|
|
metadata_obj = pcgts.get_Metadata() |
668
|
|
|
assert metadata_obj is not None |
669
|
|
|
metadata_obj.add_MetadataItem( |
670
|
|
|
MetadataItemType(type_="processingStep", |
671
|
|
|
name=self.ocrd_tool['steps'][0], |
672
|
|
|
value=self.ocrd_tool['executable'], |
673
|
|
|
Labels=[LabelsType( |
674
|
|
|
externalModel="ocrd-tool", |
675
|
|
|
externalId="parameters", |
676
|
|
|
Label=[LabelType(type_=name, |
677
|
|
|
value=self.parameter[name]) |
678
|
|
|
for name in self.parameter.keys()]), |
679
|
|
|
LabelsType( |
680
|
|
|
externalModel="ocrd-tool", |
681
|
|
|
externalId="version", |
682
|
|
|
Label=[LabelType(type_=self.ocrd_tool['executable'], |
683
|
|
|
value=self.version), |
684
|
|
|
LabelType(type_='ocrd/core', |
685
|
|
|
value=OCRD_VERSION)]) |
686
|
|
|
])) |
687
|
|
|
|
688
|
|
|
def resolve_resource(self, val): |
689
|
|
|
""" |
690
|
|
|
Resolve a resource name to an absolute file path with the algorithm in |
691
|
|
|
`spec <https://ocr-d.de/en/spec/ocrd_tool#file-parameters>`_ |
692
|
|
|
|
693
|
|
|
Args: |
694
|
|
|
val (string): resource value to resolve |
695
|
|
|
""" |
696
|
|
|
executable = self.ocrd_tool['executable'] |
697
|
|
|
if exists(val): |
698
|
|
|
self._base_logger.debug("Resolved to absolute path %s" % val) |
699
|
|
|
return val |
700
|
|
|
# FIXME: remove once workspace arg / old_pwd is gone: |
701
|
|
|
if hasattr(self, 'old_pwd'): |
702
|
|
|
cwd = self.old_pwd |
703
|
|
|
else: |
704
|
|
|
cwd = getcwd() |
705
|
|
|
ret = [cand for cand in list_resource_candidates(executable, val, |
706
|
|
|
cwd=cwd, moduled=self.moduledir) |
707
|
|
|
if exists(cand)] |
708
|
|
|
if ret: |
709
|
|
|
self._base_logger.debug("Resolved %s to absolute path %s" % (val, ret[0])) |
710
|
|
|
return ret[0] |
711
|
|
|
raise ResourceNotFoundError(val, executable) |
712
|
|
|
|
713
|
|
|
def show_resource(self, val): |
714
|
|
|
""" |
715
|
|
|
Resolve a resource name to a file path with the algorithm in |
716
|
|
|
`spec <https://ocr-d.de/en/spec/ocrd_tool#file-parameters>`_, |
717
|
|
|
then print its contents to stdout. |
718
|
|
|
|
719
|
|
|
Args: |
720
|
|
|
val (string): resource value to show |
721
|
|
|
""" |
722
|
|
|
res_fname = self.resolve_resource(val) |
723
|
|
|
fpath = Path(res_fname) |
724
|
|
|
if fpath.is_dir(): |
725
|
|
|
with pushd_popd(fpath): |
726
|
|
|
fileobj = io.BytesIO() |
727
|
|
|
with tarfile.open(fileobj=fileobj, mode='w:gz') as tarball: |
728
|
|
|
tarball.add('.') |
729
|
|
|
fileobj.seek(0) |
730
|
|
|
copyfileobj(fileobj, sys.stdout.buffer) |
731
|
|
|
else: |
732
|
|
|
sys.stdout.buffer.write(fpath.read_bytes()) |
733
|
|
|
|
734
|
|
|
def list_all_resources(self): |
735
|
|
|
""" |
736
|
|
|
List all resources found in the filesystem and matching content-type by filename suffix |
737
|
|
|
""" |
738
|
|
|
mimetypes = get_processor_resource_types(None, self.ocrd_tool) |
739
|
|
|
for res in list_all_resources(self.ocrd_tool['executable'], moduled=self.moduledir): |
740
|
|
|
res = Path(res) |
741
|
|
|
if not '*/*' in mimetypes: |
742
|
|
|
if res.is_dir() and not 'text/directory' in mimetypes: |
743
|
|
|
continue |
744
|
|
|
# if we do not know all MIME types, then keep the file, otherwise require suffix match |
745
|
|
|
if res.is_file() and not any(res.suffix == MIME_TO_EXT.get(mime, res.suffix) |
746
|
|
|
for mime in mimetypes): |
747
|
|
|
continue |
748
|
|
|
yield res |
749
|
|
|
|
750
|
|
|
@property |
751
|
|
|
def module(self): |
752
|
|
|
""" |
753
|
|
|
The top-level module this processor belongs to. |
754
|
|
|
""" |
755
|
|
|
# find shortest prefix path that is not just a namespace package |
756
|
|
|
fqname = '' |
757
|
|
|
for name in self.__module__.split('.'): |
758
|
|
|
if fqname: |
759
|
|
|
fqname += '.' |
760
|
|
|
fqname += name |
761
|
|
|
if getattr(sys.modules[fqname], '__file__', None): |
762
|
|
|
return fqname |
763
|
|
|
# fall-back |
764
|
|
|
return self.__module__ |
765
|
|
|
|
766
|
|
|
@property |
767
|
|
|
def moduledir(self): |
768
|
|
|
""" |
769
|
|
|
The filesystem path of the module directory. |
770
|
|
|
""" |
771
|
|
|
return resource_filename(self.module, '.') |
772
|
|
|
|
773
|
|
|
@property |
774
|
|
|
def input_files(self): |
775
|
|
|
""" |
776
|
|
|
List the input files (for single-valued :py:attr:`input_file_grp`). |
777
|
|
|
|
778
|
|
|
For each physical page: |
779
|
|
|
|
780
|
|
|
- If there is a single PAGE-XML for the page, take it (and forget about all |
781
|
|
|
other files for that page) |
782
|
|
|
- Else if there is a single image file, take it (and forget about all other |
783
|
|
|
files for that page) |
784
|
|
|
- Otherwise raise an error (complaining that only PAGE-XML warrants |
785
|
|
|
having multiple images for a single page) |
786
|
|
|
|
787
|
|
|
See `algorithm <https://github.com/cisocrgroup/ocrd_cis/pull/57#issuecomment-656336593>`_ |
788
|
|
|
|
789
|
|
|
Returns: |
790
|
|
|
A list of :py:class:`ocrd_models.ocrd_file.OcrdFile` objects. |
791
|
|
|
""" |
792
|
|
|
if not self.input_file_grp: |
793
|
|
|
raise ValueError("Processor is missing input fileGrp") |
794
|
|
|
ret = self.zip_input_files(mimetype=None, on_error='abort') |
795
|
|
|
if not ret: |
796
|
|
|
return [] |
797
|
|
|
assert len(ret[0]) == 1, 'Use zip_input_files() instead of input_files when processing multiple input fileGrps' |
798
|
|
|
return [tuples[0] for tuples in ret] |
799
|
|
|
|
800
|
|
|
def zip_input_files(self, require_first=True, mimetype=None, on_error='skip'): |
801
|
|
|
""" |
802
|
|
|
List tuples of input files (for multi-valued :py:attr:`input_file_grp`). |
803
|
|
|
|
804
|
|
|
Processors that expect/need multiple input file groups, |
805
|
|
|
cannot use :py:data:`input_files`. They must align (zip) input files |
806
|
|
|
across pages. This includes the case where not all pages |
807
|
|
|
are equally present in all file groups. It also requires |
808
|
|
|
making a consistent selection if there are multiple files |
809
|
|
|
per page. |
810
|
|
|
|
811
|
|
|
Following the OCR-D functional model, this function tries to |
812
|
|
|
find a single PAGE file per page, or fall back to a single |
813
|
|
|
image file per page. In either case, multiple matches per page |
814
|
|
|
are an error (see error handling below). |
815
|
|
|
This default behaviour can be changed by using a fixed MIME |
816
|
|
|
type filter via :py:attr:`mimetype`. But still, multiple matching |
817
|
|
|
files per page are an error. |
818
|
|
|
|
819
|
|
|
Single-page multiple-file errors are handled according to |
820
|
|
|
:py:attr:`on_error`: |
821
|
|
|
|
822
|
|
|
- if ``skip``, then the page for the respective fileGrp will be |
823
|
|
|
silently skipped (as if there was no match at all) |
824
|
|
|
- if ``first``, then the first matching file for the page will be |
825
|
|
|
silently selected (as if the first was the only match) |
826
|
|
|
- if ``last``, then the last matching file for the page will be |
827
|
|
|
silently selected (as if the last was the only match) |
828
|
|
|
- if ``abort``, then an exception will be raised. |
829
|
|
|
|
830
|
|
|
Multiple matches for PAGE-XML will always raise an exception. |
831
|
|
|
|
832
|
|
|
Keyword Args: |
833
|
|
|
require_first (boolean): If true, then skip a page entirely |
834
|
|
|
whenever it is not available in the first input `fileGrp`. |
835
|
|
|
on_error (string): How to handle multiple file matches per page. |
836
|
|
|
mimetype (string): If not `None`, filter by the specified MIME |
837
|
|
|
type (literal or regex prefixed by `//`). Otherwise prefer |
838
|
|
|
PAGE or image. |
839
|
|
|
Returns: |
840
|
|
|
A list of :py:class:`ocrd_models.ocrd_file.OcrdFile` tuples. |
841
|
|
|
""" |
842
|
|
|
if not self.input_file_grp: |
843
|
|
|
raise ValueError("Processor is missing input fileGrp") |
844
|
|
|
|
845
|
|
|
ifgs = self.input_file_grp.split(",") |
846
|
|
|
# Iterating over all files repeatedly may seem inefficient at first sight, |
847
|
|
|
# but the unnecessary OcrdFile instantiations for posterior fileGrp filtering |
848
|
|
|
# can actually be much more costly than traversing the ltree. |
849
|
|
|
# This might depend on the number of pages vs number of fileGrps. |
850
|
|
|
|
851
|
|
|
pages = {} |
852
|
|
|
for i, ifg in enumerate(ifgs): |
853
|
|
|
files_ = sorted(self.workspace.mets.find_all_files( |
854
|
|
|
pageId=self.page_id, fileGrp=ifg, mimetype=mimetype), |
855
|
|
|
# sort by MIME type so PAGE comes before images |
856
|
|
|
key=lambda file_: file_.mimetype) |
857
|
|
|
for file_ in files_: |
858
|
|
|
if not file_.pageId: |
859
|
|
|
# ignore document-global files |
860
|
|
|
continue |
861
|
|
|
ift = pages.setdefault(file_.pageId, [None]*len(ifgs)) |
862
|
|
|
if ift[i]: |
863
|
|
|
self._base_logger.debug(f"another file {file_.ID} for page {file_.pageId} in input file group {ifg}") |
864
|
|
|
# fileGrp has multiple files for this page ID |
865
|
|
|
if mimetype: |
866
|
|
|
# filter was active, this must not happen |
867
|
|
|
self._base_logger.warning(f"added file {file_.ID} for page {file_.pageId} in input file group {ifg} " |
868
|
|
|
f"conflicts with file {ift[i].ID} of same MIME type {mimetype} - on_error={on_error}") |
869
|
|
View Code Duplication |
if on_error == 'skip': |
|
|
|
|
870
|
|
|
ift[i] = None |
871
|
|
|
elif on_error == 'first': |
872
|
|
|
pass # keep first match |
873
|
|
|
elif on_error == 'last': |
874
|
|
|
ift[i] = file_ |
875
|
|
|
elif on_error == 'abort': |
876
|
|
|
raise NonUniqueInputFile(ifg, file_.pageId, mimetype) |
877
|
|
|
else: |
878
|
|
|
raise Exception("Unknown 'on_error' strategy '%s'" % on_error) |
879
|
|
|
elif (ift[i].mimetype == MIMETYPE_PAGE and |
880
|
|
|
file_.mimetype != MIMETYPE_PAGE): |
881
|
|
|
pass # keep PAGE match |
882
|
|
|
elif (ift[i].mimetype == MIMETYPE_PAGE and |
883
|
|
|
file_.mimetype == MIMETYPE_PAGE): |
884
|
|
|
raise NonUniqueInputFile(ifg, file_.pageId, None) |
885
|
|
|
else: |
886
|
|
|
# filter was inactive but no PAGE is in control, this must not happen |
887
|
|
|
self._base_logger.warning(f"added file {file_.ID} for page {file_.pageId} in input file group {ifg} " |
888
|
|
|
f"conflicts with file {ift[i].ID} but no PAGE available - on_error={on_error}") |
889
|
|
View Code Duplication |
if on_error == 'skip': |
|
|
|
|
890
|
|
|
ift[i] = None |
891
|
|
|
elif on_error == 'first': |
892
|
|
|
pass # keep first match |
893
|
|
|
elif on_error == 'last': |
894
|
|
|
ift[i] = file_ |
895
|
|
|
elif on_error == 'abort': |
896
|
|
|
raise NonUniqueInputFile(ifg, file_.pageId, None) |
897
|
|
|
else: |
898
|
|
|
raise Exception("Unknown 'on_error' strategy '%s'" % on_error) |
899
|
|
|
else: |
900
|
|
|
self._base_logger.debug(f"adding file {file_.ID} for page {file_.pageId} to input file group {ifg}") |
901
|
|
|
ift[i] = file_ |
902
|
|
|
# Warn if no files found but pageId was specified, because that might be due to invalid page_id (range) |
903
|
|
|
if self.page_id and not any(pages): |
904
|
|
|
self._base_logger.critical(f"Could not find any files for selected pageId {self.page_id}.\n" |
905
|
|
|
f"compare '{self.page_id}' with the output of 'orcd workspace list-page'.") |
906
|
|
|
ifts = [] |
907
|
|
|
for page, ifiles in pages.items(): |
908
|
|
|
for i, ifg in enumerate(ifgs): |
909
|
|
|
if not ifiles[i]: |
910
|
|
|
# could be from non-unique with on_error=skip or from true gap |
911
|
|
|
self._base_logger.error(f'Found no file for page {page} in file group {ifg}') |
912
|
|
|
if config.OCRD_MISSING_INPUT == 'abort': |
913
|
|
|
raise MissingInputFile(ifg, page, mimetype) |
914
|
|
|
if not any(ifiles): |
915
|
|
|
# must be from non-unique with on_error=skip |
916
|
|
|
self._base_logger.warning(f'Found no files for {page} - skipping') |
917
|
|
|
continue |
918
|
|
|
if ifiles[0] or not require_first: |
919
|
|
|
ifts.append(tuple(ifiles)) |
920
|
|
|
return ifts |
921
|
|
|
|
922
|
|
|
def generate_processor_help(ocrd_tool, processor_instance=None, subcommand=None): |
923
|
|
|
"""Generate a string describing the full CLI of this processor including params. |
924
|
|
|
|
925
|
|
|
Args: |
926
|
|
|
ocrd_tool (dict): this processor's ``tools`` section of the module's ``ocrd-tool.json`` |
927
|
|
|
processor_instance (object, optional): the processor implementation |
928
|
|
|
(for adding any module/class/function docstrings) |
929
|
|
|
subcommand (string): 'worker' or 'server' |
930
|
|
|
""" |
931
|
|
|
doc_help = '' |
932
|
|
|
if processor_instance: |
933
|
|
|
module = inspect.getmodule(processor_instance) |
934
|
|
|
if module and module.__doc__: |
935
|
|
|
doc_help += '\n' + inspect.cleandoc(module.__doc__) + '\n' |
936
|
|
|
if processor_instance.__doc__: |
937
|
|
|
doc_help += '\n' + inspect.cleandoc(processor_instance.__doc__) + '\n' |
938
|
|
|
# Try to find the most concrete docstring among the various methods that an implementation |
939
|
|
|
# could overload, first serving. |
940
|
|
|
# In doing so, compare with Processor to avoid a glitch in the way py>=3.5 inherits docstrings. |
941
|
|
|
# (They are supposed to only repeat information inspect.getdoc, rather than inherit __doc__ itself.) |
942
|
|
|
for method in ['process_page_pcgts', 'process_page_file', 'process_workspace', 'process']: |
943
|
|
|
instance_method = getattr(processor_instance, method) |
944
|
|
|
superclass_method = getattr(Processor, method) |
945
|
|
|
if instance_method.__doc__ and instance_method.__doc__ != superclass_method.__doc__: |
946
|
|
|
doc_help += '\n' + inspect.cleandoc(instance_method.__doc__) + '\n' |
947
|
|
|
break |
948
|
|
|
if doc_help: |
949
|
|
|
doc_help = '\n\n' + wrap_text(doc_help, width=72, |
950
|
|
|
initial_indent=' > ', |
951
|
|
|
subsequent_indent=' > ', |
952
|
|
|
preserve_paragraphs=True) |
953
|
|
|
subcommands = '''\ |
954
|
|
|
worker Start a processing worker rather than do local processing |
955
|
|
|
server Start a processor server rather than do local processing |
956
|
|
|
''' |
957
|
|
|
|
958
|
|
|
processing_worker_options = '''\ |
959
|
|
|
--queue The RabbitMQ server address in format |
960
|
|
|
"amqp://{user}:{pass}@{host}:{port}/{vhost}" |
961
|
|
|
[amqp://admin:admin@localhost:5672] |
962
|
|
|
--database The MongoDB server address in format |
963
|
|
|
"mongodb://{host}:{port}" |
964
|
|
|
[mongodb://localhost:27018] |
965
|
|
|
--log-filename Filename to redirect STDOUT/STDERR to, |
966
|
|
|
if specified. |
967
|
|
|
''' |
968
|
|
|
|
969
|
|
|
processing_server_options = '''\ |
970
|
|
|
--address The Processor server address in format |
971
|
|
|
"{host}:{port}" |
972
|
|
|
--database The MongoDB server address in format |
973
|
|
|
"mongodb://{host}:{port}" |
974
|
|
|
[mongodb://localhost:27018] |
975
|
|
|
''' |
976
|
|
|
|
977
|
|
|
processing_options = '''\ |
978
|
|
|
-m, --mets URL-PATH URL or file path of METS to process [./mets.xml] |
979
|
|
|
-w, --working-dir PATH Working directory of local workspace [dirname(URL-PATH)] |
980
|
|
|
-I, --input-file-grp USE File group(s) used as input |
981
|
|
|
-O, --output-file-grp USE File group(s) used as output |
982
|
|
|
-g, --page-id ID Physical page ID(s) to process instead of full document [] |
983
|
|
|
--overwrite Remove existing output pages/images |
984
|
|
|
(with "--page-id", remove only those). |
985
|
|
|
Short-hand for OCRD_EXISTING_OUTPUT=OVERWRITE |
986
|
|
|
--debug Abort on any errors with full stack trace. |
987
|
|
|
Short-hand for OCRD_MISSING_OUTPUT=ABORT |
988
|
|
|
--profile Enable profiling |
989
|
|
|
--profile-file PROF-PATH Write cProfile stats to PROF-PATH. Implies "--profile" |
990
|
|
|
-p, --parameter JSON-PATH Parameters, either verbatim JSON string |
991
|
|
|
or JSON file path |
992
|
|
|
-P, --param-override KEY VAL Override a single JSON object key-value pair, |
993
|
|
|
taking precedence over --parameter |
994
|
|
|
-U, --mets-server-url URL URL of a METS Server for parallel incremental access to METS |
995
|
|
|
If URL starts with http:// start an HTTP server there, |
996
|
|
|
otherwise URL is a path to an on-demand-created unix socket |
997
|
|
|
-l, --log-level [OFF|ERROR|WARN|INFO|DEBUG|TRACE] |
998
|
|
|
Override log level globally [INFO] |
999
|
|
|
--log-filename LOG-PATH File to redirect stderr logging to (overriding ocrd_logging.conf). |
1000
|
|
|
''' |
1001
|
|
|
|
1002
|
|
|
information_options = '''\ |
1003
|
|
|
-C, --show-resource RESNAME Dump the content of processor resource RESNAME |
1004
|
|
|
-L, --list-resources List names of processor resources |
1005
|
|
|
-J, --dump-json Dump tool description as JSON |
1006
|
|
|
-D, --dump-module-dir Show the 'module' resource location path for this processor |
1007
|
|
|
-h, --help Show this message |
1008
|
|
|
-V, --version Show version |
1009
|
|
|
''' |
1010
|
|
|
|
1011
|
|
|
parameter_help = '' |
1012
|
|
|
if 'parameters' not in ocrd_tool or not ocrd_tool['parameters']: |
1013
|
|
|
parameter_help = ' NONE\n' |
1014
|
|
|
else: |
1015
|
|
|
def wrap(s): |
1016
|
|
|
return wrap_text(s, initial_indent=' '*3, |
1017
|
|
|
subsequent_indent=' '*4, |
1018
|
|
|
width=72, preserve_paragraphs=True) |
1019
|
|
|
for param_name, param in ocrd_tool['parameters'].items(): |
1020
|
|
|
parameter_help += wrap('"%s" [%s%s]' % ( |
1021
|
|
|
param_name, |
1022
|
|
|
param['type'], |
1023
|
|
|
' - REQUIRED' if 'required' in param and param['required'] else |
1024
|
|
|
' - %s' % json.dumps(param['default']) if 'default' in param else '')) |
1025
|
|
|
parameter_help += '\n ' + wrap(param['description']) |
1026
|
|
|
if 'enum' in param: |
1027
|
|
|
parameter_help += '\n ' + wrap('Possible values: %s' % json.dumps(param['enum'])) |
1028
|
|
|
parameter_help += "\n" |
1029
|
|
|
|
1030
|
|
|
if not subcommand: |
1031
|
|
|
return f'''\ |
1032
|
|
|
Usage: {ocrd_tool['executable']} [worker|server] [OPTIONS] |
1033
|
|
|
|
1034
|
|
|
{ocrd_tool['description']}{doc_help} |
1035
|
|
|
|
1036
|
|
|
Subcommands: |
1037
|
|
|
{subcommands} |
1038
|
|
|
Options for processing: |
1039
|
|
|
{processing_options} |
1040
|
|
|
Options for information: |
1041
|
|
|
{information_options} |
1042
|
|
|
Parameters: |
1043
|
|
|
{parameter_help} |
1044
|
|
|
''' |
1045
|
|
|
elif subcommand == 'worker': |
1046
|
|
|
return f'''\ |
1047
|
|
|
Usage: {ocrd_tool['executable']} worker [OPTIONS] |
1048
|
|
|
|
1049
|
|
|
Run {ocrd_tool['executable']} as a processing worker. |
1050
|
|
|
|
1051
|
|
|
{ocrd_tool['description']}{doc_help} |
1052
|
|
|
|
1053
|
|
|
Options: |
1054
|
|
|
{processing_worker_options} |
1055
|
|
|
''' |
1056
|
|
|
elif subcommand == 'server': |
1057
|
|
|
return f'''\ |
1058
|
|
|
Usage: {ocrd_tool['executable']} server [OPTIONS] |
1059
|
|
|
|
1060
|
|
|
Run {ocrd_tool['executable']} as a processor sever. |
1061
|
|
|
|
1062
|
|
|
{ocrd_tool['description']}{doc_help} |
1063
|
|
|
|
1064
|
|
|
Options: |
1065
|
|
|
{processing_server_options} |
1066
|
|
|
''' |
1067
|
|
|
else: |
1068
|
|
|
pass |
1069
|
|
|
|