Total Complexity | 180 |
Total Lines | 1302 |
Duplicated Lines | 1.54 % |
Changes | 0 |
Duplicate code is one of the most pungent code smells. A rule that is often used is to re-structure code once it is duplicated in three or more places.
Common duplication problems, and corresponding solutions are:
Complex classes like ocrd.processor.base often do a lot of different things. To break such a class down, we need to identify a cohesive component within that class. A common approach to find such a component is to look for fields/methods that share the same prefixes, or suffixes.
Once you have determined the fields that belong together, you can apply the Extract Class refactoring. If the component makes sense as a sub-class, Extract Subclass is also a candidate, and is often faster.
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 Any, Dict, List, Optional, Tuple, Union, get_args |
||
20 | import sys |
||
21 | import logging |
||
22 | import logging.handlers |
||
23 | import inspect |
||
24 | import tarfile |
||
25 | import io |
||
26 | from collections import defaultdict |
||
27 | from frozendict import frozendict |
||
28 | # concurrent.futures is buggy in py38, |
||
29 | # this is where the fixes came from: |
||
30 | from loky import Future, ProcessPoolExecutor |
||
31 | import multiprocessing as mp |
||
32 | from threading import Timer |
||
33 | from _thread import interrupt_main |
||
34 | |||
35 | from click import wrap_text |
||
36 | from deprecated import deprecated |
||
37 | from requests import HTTPError |
||
38 | |||
39 | from ..workspace import Workspace |
||
40 | from ..mets_server import ClientSideOcrdMets |
||
41 | from ocrd_models.ocrd_file import OcrdFileType |
||
42 | from .ocrd_page_result import OcrdPageResult |
||
43 | from ocrd_utils import ( |
||
44 | VERSION as OCRD_VERSION, |
||
45 | MIMETYPE_PAGE, |
||
46 | MIME_TO_EXT, |
||
47 | config, |
||
48 | getLogger, |
||
49 | list_resource_candidates, |
||
50 | pushd_popd, |
||
51 | list_all_resources, |
||
52 | get_processor_resource_types, |
||
53 | resource_filename, |
||
54 | parse_json_file_with_comments, |
||
55 | make_file_id, |
||
56 | deprecation_warning |
||
57 | ) |
||
58 | from ocrd_validators import ParameterValidator |
||
59 | from ocrd_models.ocrd_page import ( |
||
60 | PageType, |
||
61 | AlternativeImageType, |
||
62 | MetadataItemType, |
||
63 | LabelType, |
||
64 | LabelsType, |
||
65 | OcrdPage, |
||
66 | to_xml, |
||
67 | ) |
||
68 | from ocrd_modelfactory import page_from_file |
||
69 | from ocrd_validators.ocrd_tool_validator import OcrdToolValidator |
||
70 | |||
71 | # XXX imports must remain for backwards-compatibility |
||
72 | from .helpers import run_cli, run_processor # pylint: disable=unused-import |
||
73 | |||
74 | |||
75 | class ResourceNotFoundError(FileNotFoundError): |
||
76 | """ |
||
77 | An exception signifying the requested processor resource |
||
78 | cannot be resolved. |
||
79 | """ |
||
80 | def __init__(self, name, executable): |
||
81 | self.name = name |
||
82 | self.executable = executable |
||
83 | self.message = (f"Could not find resource '{name}' for executable '{executable}'. " |
||
84 | f"Try 'ocrd resmgr download {executable} {name}' to download this resource.") |
||
85 | super().__init__(self.message) |
||
86 | |||
87 | class NonUniqueInputFile(ValueError): |
||
88 | """ |
||
89 | An exception signifying the specified fileGrp / pageId / mimetype |
||
90 | selector yields multiple PAGE files, or no PAGE files but multiple images, |
||
91 | or multiple files of that mimetype. |
||
92 | """ |
||
93 | def __init__(self, fileGrp, pageId, mimetype): |
||
94 | self.fileGrp = fileGrp |
||
95 | self.pageId = pageId |
||
96 | self.mimetype = mimetype |
||
97 | self.message = (f"Could not determine unique input file for fileGrp {fileGrp} " |
||
98 | f"and pageId {pageId} under mimetype {mimetype or 'PAGE+image(s)'}") |
||
99 | super().__init__(self.message) |
||
100 | |||
101 | class MissingInputFile(ValueError): |
||
102 | """ |
||
103 | An exception signifying the specified fileGrp / pageId / mimetype |
||
104 | selector yields no PAGE files, or no PAGE and no image files, |
||
105 | or no files of that mimetype. |
||
106 | """ |
||
107 | def __init__(self, fileGrp, pageId, mimetype): |
||
108 | self.fileGrp = fileGrp |
||
109 | self.pageId = pageId |
||
110 | self.mimetype = mimetype |
||
111 | self.message = (f"Could not find input file for fileGrp {fileGrp} " |
||
112 | f"and pageId {pageId} under mimetype {mimetype or 'PAGE+image(s)'}") |
||
113 | super().__init__(self.message) |
||
114 | |||
115 | class DummyFuture: |
||
116 | """ |
||
117 | Mimics some of `concurrent.futures.Future` but runs immediately. |
||
118 | """ |
||
119 | def __init__(self, fn, *args, **kwargs): |
||
120 | self.fn = fn |
||
121 | self.args = args |
||
122 | self.kwargs = kwargs |
||
123 | def result(self): |
||
124 | return self.fn(*self.args, **self.kwargs) |
||
125 | class DummyExecutor: |
||
126 | """ |
||
127 | Mimics some of `concurrent.futures.ProcessPoolExecutor` but runs |
||
128 | everything immediately in this process. |
||
129 | """ |
||
130 | def __init__(self, initializer=None, initargs=(), **kwargs): |
||
131 | initializer(*initargs) |
||
132 | def shutdown(self, **kwargs): |
||
133 | pass |
||
134 | def submit(self, fn, *args, **kwargs) -> DummyFuture: |
||
135 | return DummyFuture(fn, *args, **kwargs) |
||
136 | |||
137 | TFuture = Union[DummyFuture, Future] |
||
138 | TExecutor = Union[DummyExecutor, ProcessPoolExecutor] |
||
139 | |||
140 | class Processor(): |
||
141 | """ |
||
142 | A processor is a tool that implements the uniform OCR-D |
||
143 | `command-line interface for run-time data processing <https://ocr-d.de/en/spec/cli>`_. |
||
144 | |||
145 | That is, it executes a single workflow step, or a combination of workflow steps, |
||
146 | on the workspace (represented by local METS). It reads input files for all or selected |
||
147 | physical pages of the input fileGrp(s), computes additional annotation, and writes output |
||
148 | files for them into the output fileGrp(s). It may take a number of optional or mandatory |
||
149 | parameters. |
||
150 | """ |
||
151 | |||
152 | max_instances : int = -1 |
||
153 | """ |
||
154 | maximum number of cached instances (ignored if negative), to be applied on top of |
||
155 | :py:data:`~ocrd_utils.config.OCRD_MAX_PROCESSOR_CACHE` (i.e. whatever is smaller). |
||
156 | |||
157 | (Override this if you know how many instances fit into memory - GPU / CPU RAM - at once.) |
||
158 | """ |
||
159 | |||
160 | max_workers : int = -1 |
||
161 | """ |
||
162 | maximum number of processor forks for page-parallel processing (ignored if negative), |
||
163 | to be applied on top of :py:data:`~ocrd_utils.config.OCRD_MAX_PARALLEL_PAGES` (i.e. |
||
164 | whatever is smaller). |
||
165 | |||
166 | (Override this if you know how many pages fit into processing units - GPU shaders / CPU cores |
||
167 | - at once, or if your class already creates threads prior to forking, e.g. during ``setup``.) |
||
168 | """ |
||
169 | |||
170 | max_page_seconds : int = -1 |
||
171 | """ |
||
172 | maximum number of seconds may be spent processing a single page (ignored if negative), |
||
173 | to be applied on top of :py:data:`~ocrd_utils.config.OCRD_PROCESSING_PAGE_TIMEOUT` |
||
174 | (i.e. whatever is smaller). |
||
175 | |||
176 | (Override this if you know how costly this processor may be, irrespective of image size |
||
177 | or complexity of the page.) |
||
178 | """ |
||
179 | |||
180 | @property |
||
181 | def metadata_filename(self) -> str: |
||
182 | """ |
||
183 | Relative location of the ``ocrd-tool.json`` file inside the package. |
||
184 | |||
185 | Used by :py:data:`metadata_location`. |
||
186 | |||
187 | (Override if ``ocrd-tool.json`` is not in the root of the module, |
||
188 | e.g. ``namespace/ocrd-tool.json`` or ``data/ocrd-tool.json``). |
||
189 | """ |
||
190 | return 'ocrd-tool.json' |
||
191 | |||
192 | @cached_property |
||
193 | def metadata_location(self) -> Path: |
||
194 | """ |
||
195 | Absolute path of the ``ocrd-tool.json`` file as distributed with the package. |
||
196 | |||
197 | Used by :py:data:`metadata_rawdict`. |
||
198 | |||
199 | (Override if ``ocrd-tool.json`` is not distributed with the Python package.) |
||
200 | """ |
||
201 | module = inspect.getmodule(self) |
||
202 | module_tokens = module.__package__.split('.') |
||
203 | # for namespace packages, we cannot just use the first token |
||
204 | for i in range(len(module_tokens)): |
||
205 | prefix = '.'.join(module_tokens[:i + 1]) |
||
206 | if sys.modules[prefix].__spec__.has_location: |
||
207 | return resource_filename(prefix, self.metadata_filename) |
||
208 | raise Exception("cannot find top-level module prefix for %s", module.__package__) |
||
209 | |||
210 | @cached_property |
||
211 | def metadata_rawdict(self) -> dict: |
||
212 | """ |
||
213 | Raw (unvalidated, unexpanded) ``ocrd-tool.json`` dict contents of the package. |
||
214 | |||
215 | Used by :py:data:`metadata`. |
||
216 | |||
217 | (Override if ``ocrd-tool.json`` is not in a file.) |
||
218 | """ |
||
219 | return parse_json_file_with_comments(self.metadata_location) |
||
220 | |||
221 | @cached_property |
||
222 | def metadata(self) -> dict: |
||
223 | """ |
||
224 | The ``ocrd-tool.json`` dict contents of the package, according to the OCR-D |
||
225 | `spec <https://ocr-d.de/en/spec/ocrd_tool>`_ for processor tools. |
||
226 | |||
227 | After deserialisation, it also gets validated against the |
||
228 | `schema <https://ocr-d.de/en/spec/ocrd_tool#definition>`_ with all defaults |
||
229 | expanded. |
||
230 | |||
231 | Used by :py:data:`ocrd_tool` and :py:data:`version`. |
||
232 | |||
233 | (Override if you want to provide metadata programmatically instead of a |
||
234 | JSON file.) |
||
235 | """ |
||
236 | metadata = self.metadata_rawdict |
||
237 | report = OcrdToolValidator.validate(metadata) |
||
238 | if not report.is_valid: |
||
239 | self.logger.error(f"The ocrd-tool.json of this processor is {'problematic' if not report.errors else 'invalid'}:\n" |
||
240 | f"{report.to_xml()}.\nPlease open an issue at {metadata.get('git_url', 'the website')}.") |
||
241 | return metadata |
||
242 | |||
243 | @cached_property |
||
244 | def version(self) -> str: |
||
245 | """ |
||
246 | The program version of the package. |
||
247 | Usually the ``version`` part of :py:data:`metadata`. |
||
248 | |||
249 | (Override if you do not want to use :py:data:`metadata` lookup |
||
250 | mechanism.) |
||
251 | """ |
||
252 | return self.metadata['version'] |
||
253 | |||
254 | @cached_property |
||
255 | def executable(self) -> str: |
||
256 | """ |
||
257 | The executable name of this processor tool. Taken from the runtime |
||
258 | filename. |
||
259 | |||
260 | Used by :py:data:`ocrd_tool` for lookup in :py:data:`metadata`. |
||
261 | |||
262 | (Override if your entry-point name deviates from the ``executable`` |
||
263 | name, or the processor gets instantiated from another runtime.) |
||
264 | """ |
||
265 | return os.path.basename(inspect.stack()[-1].filename) |
||
266 | |||
267 | @cached_property |
||
268 | def ocrd_tool(self) -> dict: |
||
269 | """ |
||
270 | The ``ocrd-tool.json`` dict contents of this processor tool. |
||
271 | Usually the :py:data:`executable` key of the ``tools`` part |
||
272 | of :py:data:`metadata`. |
||
273 | |||
274 | (Override if you do not want to use :py:data:`metadata` lookup |
||
275 | mechanism.) |
||
276 | """ |
||
277 | return self.metadata['tools'][self.executable] |
||
278 | |||
279 | @property |
||
280 | def parameter(self) -> Optional[dict]: |
||
281 | """the runtime parameter dict to be used by this processor""" |
||
282 | if hasattr(self, '_parameter'): |
||
283 | return self._parameter |
||
284 | return None |
||
285 | |||
286 | @parameter.setter |
||
287 | def parameter(self, parameter : dict) -> None: |
||
288 | if self.parameter is not None: |
||
289 | self.shutdown() |
||
290 | parameterValidator = ParameterValidator(self.ocrd_tool) |
||
291 | report = parameterValidator.validate(parameter) |
||
292 | if not report.is_valid: |
||
293 | raise ValueError(f'Invalid parameters:\n{report.to_xml()}') |
||
294 | # make parameter dict read-only |
||
295 | self._parameter = frozendict(parameter) |
||
296 | # (re-)run setup to load models etc |
||
297 | self.setup() |
||
298 | |||
299 | def __init__( |
||
300 | self, |
||
301 | # FIXME: remove in favor of process_workspace(workspace) |
||
302 | workspace : Optional[Workspace], |
||
303 | ocrd_tool=None, |
||
304 | parameter=None, |
||
305 | input_file_grp=None, |
||
306 | output_file_grp=None, |
||
307 | page_id=None, |
||
308 | download_files=config.OCRD_DOWNLOAD_INPUT, |
||
309 | version=None |
||
310 | ): |
||
311 | """ |
||
312 | Instantiate, but do not setup (neither for processing nor other usage). |
||
313 | If given, do parse and validate :py:data:`.parameter`. |
||
314 | |||
315 | Args: |
||
316 | workspace (:py:class:`~ocrd.Workspace`): The workspace to process. \ |
||
317 | If not ``None``, then `chdir` to that directory. |
||
318 | Deprecated since version 3.0: Should be ``None`` here, but then needs to be set \ |
||
319 | before processing. |
||
320 | Keyword Args: |
||
321 | parameter (string): JSON of the runtime choices for ocrd-tool ``parameters``. \ |
||
322 | Can be ``None`` even for processing, but then needs to be set before running. |
||
323 | input_file_grp (string): comma-separated list of METS ``fileGrp`` used for input. \ |
||
324 | Deprecated since version 3.0: Should be ``None`` here, but then needs to be set \ |
||
325 | before processing. |
||
326 | output_file_grp (string): comma-separated list of METS ``fileGrp`` used for output. \ |
||
327 | Deprecated since version 3.0: Should be ``None`` here, but then needs to be set \ |
||
328 | before processing. |
||
329 | page_id (string): comma-separated list of METS physical ``page`` IDs to process \ |
||
330 | (or empty for all pages). \ |
||
331 | Deprecated since version 3.0: Should be ``None`` here, but then needs to be set \ |
||
332 | before processing. |
||
333 | download_files (boolean): Whether input files will be downloaded prior to processing, \ |
||
334 | defaults to :py:attr:`ocrd_utils.config.OCRD_DOWNLOAD_INPUT` which is ``True`` by default |
||
335 | """ |
||
336 | if ocrd_tool is not None: |
||
337 | deprecation_warning("Passing 'ocrd_tool' as keyword argument to Processor is deprecated - " |
||
338 | "use or override metadata/executable/ocrd-tool properties instead") |
||
339 | self.ocrd_tool = ocrd_tool |
||
340 | self.executable = ocrd_tool['executable'] |
||
341 | if version is not None: |
||
342 | deprecation_warning("Passing 'version' as keyword argument to Processor is deprecated - " |
||
343 | "use or override metadata/version properties instead") |
||
344 | self.version = version |
||
345 | if workspace is not None: |
||
346 | deprecation_warning("Passing a workspace argument other than 'None' to Processor " |
||
347 | "is deprecated - pass as argument to process_workspace instead") |
||
348 | self.workspace = workspace |
||
349 | self.old_pwd = getcwd() |
||
350 | os.chdir(self.workspace.directory) |
||
351 | if input_file_grp is not None: |
||
352 | deprecation_warning("Passing an input_file_grp kwarg other than 'None' to Processor " |
||
353 | "is deprecated - pass as argument to process_workspace instead") |
||
354 | self.input_file_grp = input_file_grp |
||
355 | if output_file_grp is not None: |
||
356 | deprecation_warning("Passing an output_file_grp kwarg other than 'None' to Processor " |
||
357 | "is deprecated - pass as argument to process_workspace instead") |
||
358 | self.output_file_grp = output_file_grp |
||
359 | if page_id is not None: |
||
360 | deprecation_warning("Passing a page_id kwarg other than 'None' to Processor " |
||
361 | "is deprecated - pass as argument to process_workspace instead") |
||
362 | self.page_id = page_id or None |
||
363 | self.download = download_files |
||
364 | #: The logger to be used by processor implementations. |
||
365 | # `ocrd.processor.base` internals should use :py:attr:`self._base_logger` |
||
366 | self.logger = getLogger(f'ocrd.processor.{self.__class__.__name__}') |
||
367 | self._base_logger = getLogger('ocrd.processor.base') |
||
368 | if parameter is not None: |
||
369 | self.parameter = parameter |
||
370 | # workaround for deprecated#72 (@deprecated decorator does not work for subclasses): |
||
371 | setattr(self, 'process', |
||
372 | deprecated(version='3.0', reason='process() should be replaced with process_page_pcgts() or process_page_file() or process_workspace()')(getattr(self, 'process'))) |
||
373 | |||
374 | def __del__(self): |
||
375 | self._base_logger.debug("shutting down") |
||
376 | self.shutdown() |
||
377 | |||
378 | def show_help(self, subcommand=None): |
||
379 | """ |
||
380 | Print a usage description including the standard CLI and all of this processor's ocrd-tool |
||
381 | parameters and docstrings. |
||
382 | """ |
||
383 | print(generate_processor_help(self.ocrd_tool, processor_instance=self, subcommand=subcommand)) |
||
384 | |||
385 | def show_version(self): |
||
386 | """ |
||
387 | Print information on this processor's version and OCR-D version. |
||
388 | """ |
||
389 | print("Version %s, ocrd/core %s" % (self.version, OCRD_VERSION)) |
||
390 | |||
391 | def verify(self): |
||
392 | """ |
||
393 | Verify that :py:attr:`input_file_grp` and :py:attr:`output_file_grp` fulfill the processor's requirements. |
||
394 | """ |
||
395 | # verify input and output file groups in parameters |
||
396 | assert self.input_file_grp is not None |
||
397 | assert self.output_file_grp is not None |
||
398 | input_file_grps = self.input_file_grp.split(',') |
||
399 | output_file_grps = self.output_file_grp.split(',') |
||
400 | def assert_file_grp_cardinality(grps : List[str], spec : Union[int, List[int]], msg): |
||
401 | if isinstance(spec, int): |
||
402 | if spec > 0: |
||
403 | assert len(grps) == spec, msg % (len(grps), str(spec)) |
||
404 | else: |
||
405 | assert isinstance(spec, list) |
||
406 | minimum = spec[0] |
||
407 | maximum = spec[1] |
||
408 | if minimum > 0: |
||
409 | assert len(grps) >= minimum, msg % (len(grps), str(spec)) |
||
410 | if maximum > 0: |
||
411 | assert len(grps) <= maximum, msg % (len(grps), str(spec)) |
||
412 | # FIXME: enforce unconditionally as soon as grace period for deprecation is over |
||
413 | if 'input_file_grp_cardinality' in self.ocrd_tool: |
||
414 | assert_file_grp_cardinality(input_file_grps, self.ocrd_tool['input_file_grp_cardinality'], |
||
415 | "Unexpected number of input file groups %d vs %s") |
||
416 | if 'output_file_grp_cardinality' in self.ocrd_tool: |
||
417 | assert_file_grp_cardinality(output_file_grps, self.ocrd_tool['output_file_grp_cardinality'], |
||
418 | "Unexpected number of output file groups %d vs %s") |
||
419 | # verify input and output file groups in METS |
||
420 | for input_file_grp in input_file_grps: |
||
421 | assert input_file_grp in self.workspace.mets.file_groups, \ |
||
422 | f"input fileGrp {input_file_grp} does not exist in workspace {self.workspace}" |
||
423 | for output_file_grp in output_file_grps: |
||
424 | assert output_file_grp not in self.workspace.mets.file_groups \ |
||
425 | or config.OCRD_EXISTING_OUTPUT in ['OVERWRITE', 'SKIP'] \ |
||
426 | or not any(self.workspace.mets.find_files( |
||
427 | pageId=self.page_id, fileGrp=output_file_grp)), \ |
||
428 | f"output fileGrp {output_file_grp} already exists in workspace {self.workspace}" |
||
429 | # keep this for backwards compatibility: |
||
430 | return True |
||
431 | |||
432 | def dump_json(self): |
||
433 | """ |
||
434 | Print :py:attr:`ocrd_tool` on stdout. |
||
435 | """ |
||
436 | print(json.dumps(self.ocrd_tool, indent=True)) |
||
437 | |||
438 | def dump_module_dir(self): |
||
439 | """ |
||
440 | Print :py:attr:`moduledir` on stdout. |
||
441 | """ |
||
442 | print(self.moduledir) |
||
443 | |||
444 | def list_resources(self): |
||
445 | """ |
||
446 | Find all installed resource files in the search paths and print their path names. |
||
447 | """ |
||
448 | for res in self.list_all_resources(): |
||
449 | print(res) |
||
450 | |||
451 | def setup(self) -> None: |
||
452 | """ |
||
453 | Prepare the processor for actual data processing, |
||
454 | prior to changing to the workspace directory but |
||
455 | after parsing parameters. |
||
456 | |||
457 | (Override this to load models into memory etc.) |
||
458 | """ |
||
459 | pass |
||
460 | |||
461 | def shutdown(self) -> None: |
||
462 | """ |
||
463 | Bring down the processor after data processing, |
||
464 | after to changing back from the workspace directory but |
||
465 | before exiting (or setting up with different parameters). |
||
466 | |||
467 | (Override this to unload models from memory etc.) |
||
468 | """ |
||
469 | pass |
||
470 | |||
471 | @deprecated(version='3.0', reason='process() should be replaced with process_page_pcgts() or process_page_file() or process_workspace()') |
||
472 | def process(self) -> None: |
||
473 | """ |
||
474 | Process all files of the :py:data:`workspace` |
||
475 | from the given :py:data:`input_file_grp` |
||
476 | to the given :py:data:`output_file_grp` |
||
477 | for the given :py:data:`page_id` (or all pages) |
||
478 | under the given :py:data:`parameter`. |
||
479 | |||
480 | (This contains the main functionality and needs to be |
||
481 | overridden by subclasses.) |
||
482 | """ |
||
483 | raise NotImplementedError() |
||
484 | |||
485 | def process_workspace(self, workspace: Workspace) -> None: |
||
486 | """ |
||
487 | Process all files of the given ``workspace``, |
||
488 | from the given :py:data:`input_file_grp` |
||
489 | to the given :py:data:`output_file_grp` |
||
490 | for the given :py:data:`page_id` (or all pages) |
||
491 | under the given :py:data:`parameter`. |
||
492 | |||
493 | Delegates to :py:meth:`.process_workspace_submit_tasks` |
||
494 | and :py:meth:`.process_workspace_handle_tasks`. |
||
495 | |||
496 | (This will iterate over pages and files, calling |
||
497 | :py:meth:`.process_page_file` and handling exceptions. |
||
498 | It should be overridden by subclasses to handle cases |
||
499 | like post-processing or computation across pages.) |
||
500 | """ |
||
501 | with pushd_popd(workspace.directory): |
||
502 | self.workspace = workspace |
||
503 | self.verify() |
||
504 | try: |
||
505 | # set up multitasking |
||
506 | max_workers = max(0, config.OCRD_MAX_PARALLEL_PAGES) |
||
507 | if self.max_workers > 0 and self.max_workers < config.OCRD_MAX_PARALLEL_PAGES: |
||
508 | self._base_logger.info("limiting number of threads from %d to %d", max_workers, self.max_workers) |
||
509 | max_workers = self.max_workers |
||
510 | if max_workers > 1: |
||
511 | assert isinstance(workspace.mets, ClientSideOcrdMets), \ |
||
512 | "OCRD_MAX_PARALLEL_PAGES>1 requires also using --mets-server-url" |
||
513 | max_seconds = max(0, config.OCRD_PROCESSING_PAGE_TIMEOUT) |
||
514 | if self.max_page_seconds > 0 and self.max_page_seconds < config.OCRD_PROCESSING_PAGE_TIMEOUT: |
||
515 | self._base_logger.info("limiting page timeout from %d to %d sec", max_seconds, self.max_page_seconds) |
||
516 | max_seconds = self.max_page_seconds |
||
517 | |||
518 | if max_workers > 1: |
||
519 | executor_cls = ProcessPoolExecutor |
||
520 | log_queue = mp.Queue() |
||
521 | # forward messages from log queue (in subprocesses) to all root handlers |
||
522 | log_listener = logging.handlers.QueueListener(log_queue, *logging.root.handlers, respect_handler_level=True) |
||
523 | else: |
||
524 | executor_cls = DummyExecutor |
||
525 | log_queue = None |
||
526 | log_listener = None |
||
527 | executor = executor_cls( |
||
528 | max_workers=max_workers or 1, |
||
529 | # only forking method avoids pickling |
||
530 | context=mp.get_context('fork'), |
||
531 | # share processor instance as global to avoid pickling |
||
532 | initializer=_page_worker_set_ctxt, |
||
533 | initargs=(self, log_queue), |
||
534 | ) |
||
535 | if max_workers > 1: |
||
536 | log_listener.start() |
||
537 | try: |
||
538 | self._base_logger.debug("started executor %s with %d workers", str(executor), max_workers or 1) |
||
539 | tasks = self.process_workspace_submit_tasks(executor, max_seconds) |
||
540 | stats = self.process_workspace_handle_tasks(tasks) |
||
541 | finally: |
||
542 | executor.shutdown(kill_workers=True, wait=False) |
||
543 | if max_workers > 1: |
||
544 | log_listener.stop() |
||
545 | |||
546 | except NotImplementedError: |
||
547 | # fall back to deprecated method |
||
548 | try: |
||
549 | self.process() |
||
550 | except Exception as err: |
||
551 | # suppress the NotImplementedError context |
||
552 | raise err from None |
||
553 | |||
554 | def process_workspace_submit_tasks(self, executor : TExecutor, max_seconds : int) -> Dict[TFuture, Tuple[str, List[Optional[OcrdFileType]]]]: |
||
555 | """ |
||
556 | Look up all input files of the given ``workspace`` |
||
557 | from the given :py:data:`input_file_grp` |
||
558 | for the given :py:data:`page_id` (or all pages), |
||
559 | and schedules calling :py:meth:`.process_page_file` |
||
560 | on them for each page via `executor` (enforcing |
||
561 | a per-page time limit of `max_seconds`). |
||
562 | |||
563 | When running with `OCRD_MAX_PARALLEL_PAGES>1` and |
||
564 | the workspace via METS Server, the executor will fork |
||
565 | this many worker parallel subprocesses each processing |
||
566 | one page at a time. (Interprocess communication is |
||
567 | done via task and result queues.) |
||
568 | |||
569 | Otherwise, tasks are run sequentially in the |
||
570 | current process. |
||
571 | |||
572 | Delegates to :py:meth:`.zip_input_files` to get |
||
573 | the input files for each page, and then calls |
||
574 | :py:meth:`.process_workspace_submit_page_task`. |
||
575 | |||
576 | Returns a dict mapping the per-page tasks |
||
577 | (i.e. futures submitted to the executor) |
||
578 | to their corresponding pageId and input files. |
||
579 | """ |
||
580 | tasks = {} |
||
581 | for input_file_tuple in self.zip_input_files(on_error='abort', require_first=False): |
||
582 | task, page_id, input_files = self.process_workspace_submit_page_task(executor, max_seconds, input_file_tuple) |
||
583 | tasks[task] = (page_id, input_files) |
||
584 | self._base_logger.debug("submitted %d processing tasks", len(tasks)) |
||
585 | return tasks |
||
586 | |||
587 | def process_workspace_submit_page_task(self, executor : TExecutor, max_seconds : int, input_file_tuple : List[Optional[OcrdFileType]]) -> Tuple[TFuture, str, List[Optional[OcrdFileType]]]: |
||
588 | """ |
||
589 | Ensure all input files for a single page are |
||
590 | downloaded to the workspace, then schedule |
||
591 | :py:meth:`.process_process_file` to be run on |
||
592 | them via `executor` (enforcing a per-page time |
||
593 | limit of `max_seconds`). |
||
594 | |||
595 | Delegates to :py:meth:`.process_page_file` |
||
596 | (wrapped in :py:func:`_page_worker` to share |
||
597 | the processor instance across forked processes). |
||
598 | |||
599 | \b |
||
600 | Returns a tuple of: |
||
601 | - the scheduled future object, |
||
602 | - the corresponding pageId, |
||
603 | - the corresponding input files. |
||
604 | """ |
||
605 | input_files : List[Optional[OcrdFileType]] = [None] * len(input_file_tuple) |
||
606 | page_id = next(input_file.pageId |
||
607 | for input_file in input_file_tuple |
||
608 | if input_file) |
||
609 | self._base_logger.info(f"preparing page {page_id}") |
||
610 | for i, input_file in enumerate(input_file_tuple): |
||
611 | if input_file is None: |
||
612 | # file/page not found in this file grp |
||
613 | continue |
||
614 | input_files[i] = input_file |
||
615 | if not self.download: |
||
616 | continue |
||
617 | try: |
||
618 | input_files[i] = self.workspace.download_file(input_file) |
||
619 | except (ValueError, FileNotFoundError, HTTPError) as e: |
||
620 | self._base_logger.error(repr(e)) |
||
621 | self._base_logger.warning(f"failed downloading file {input_file} for page {page_id}") |
||
622 | # process page |
||
623 | #executor.submit(self.process_page_file, *input_files) |
||
624 | return executor.submit(_page_worker, max_seconds, *input_files), page_id, input_files |
||
625 | |||
626 | def process_workspace_handle_tasks(self, tasks : Dict[TFuture, Tuple[str, List[Optional[OcrdFileType]]]]) -> Tuple[int, int, Dict[str, int], int]: |
||
627 | """ |
||
628 | Look up scheduled per-page futures one by one, |
||
629 | handle errors (exceptions) and gather results. |
||
630 | |||
631 | \b |
||
632 | Enforces policies configured by the following |
||
633 | environment variables: |
||
634 | - `OCRD_EXISTING_OUTPUT` (abort/skip/overwrite) |
||
635 | - `OCRD_MISSING_OUTPUT` (abort/skip/fallback-copy) |
||
636 | - `OCRD_MAX_MISSING_OUTPUTS` (abort after all). |
||
637 | |||
638 | \b |
||
639 | Returns a tuple of: |
||
640 | - the number of successfully processed pages |
||
641 | - the number of failed (i.e. skipped or copied) pages |
||
642 | - a dict of the type and corresponding number of exceptions seen |
||
643 | - the number of total requested pages (i.e. success+fail+existing). |
||
644 | |||
645 | Delegates to :py:meth:`.process_workspace_handle_page_task` |
||
646 | for each page. |
||
647 | """ |
||
648 | # aggregate info for logging: |
||
649 | nr_succeeded = 0 |
||
650 | nr_failed = 0 |
||
651 | nr_errors = defaultdict(int) # count causes |
||
652 | if config.OCRD_MISSING_OUTPUT == 'SKIP': |
||
653 | reason = "skipped" |
||
654 | elif config.OCRD_MISSING_OUTPUT == 'COPY': |
||
655 | reason = "fallback-copied" |
||
656 | for task in tasks: |
||
657 | # wait for results, handle errors |
||
658 | page_id, input_files = tasks[task] |
||
659 | result = self.process_workspace_handle_page_task(page_id, input_files, task) |
||
660 | if isinstance(result, Exception): |
||
661 | nr_errors[result.__class__.__name__] += 1 |
||
662 | nr_failed += 1 |
||
663 | # FIXME: this is just prospective, because len(tasks)==nr_failed+nr_succeeded is not guaranteed |
||
664 | if config.OCRD_MAX_MISSING_OUTPUTS > 0 and nr_failed / len(tasks) > config.OCRD_MAX_MISSING_OUTPUTS: |
||
665 | # already irredeemably many failures, stop short |
||
666 | nr_errors = dict(nr_errors) |
||
667 | raise Exception(f"too many failures with {reason} output ({nr_failed} of {nr_failed+nr_succeeded}, {str(nr_errors)})") |
||
668 | elif result: |
||
669 | nr_succeeded += 1 |
||
670 | # else skipped - already exists |
||
671 | nr_errors = dict(nr_errors) |
||
672 | if nr_failed > 0: |
||
673 | nr_all = nr_succeeded + nr_failed |
||
674 | if config.OCRD_MAX_MISSING_OUTPUTS > 0 and nr_failed / nr_all > config.OCRD_MAX_MISSING_OUTPUTS: |
||
675 | raise Exception(f"too many failures with {reason} output ({nr_failed} of {nr_all}, {str(nr_errors)})") |
||
676 | self._base_logger.warning("%s %d of %d pages due to %s", reason, nr_failed, nr_all, str(nr_errors)) |
||
|
|||
677 | return nr_succeeded, nr_failed, nr_errors, len(tasks) |
||
678 | |||
679 | def process_workspace_handle_page_task(self, page_id : str, input_files : List[Optional[OcrdFileType]], task : TFuture) -> Union[bool, Exception]: |
||
680 | """ |
||
681 | \b |
||
682 | Await a single page result and handle errors (exceptions), |
||
683 | enforcing policies configured by the following |
||
684 | environment variables: |
||
685 | - `OCRD_EXISTING_OUTPUT` (abort/skip/overwrite) |
||
686 | - `OCRD_MISSING_OUTPUT` (abort/skip/fallback-copy) |
||
687 | - `OCRD_MAX_MISSING_OUTPUTS` (abort after all). |
||
688 | |||
689 | \b |
||
690 | Returns |
||
691 | - true in case of success |
||
692 | - false in case the output already exists |
||
693 | - the exception in case of failure |
||
694 | """ |
||
695 | # FIXME: differentiate error cases in various ways: |
||
696 | # - ResourceNotFoundError → use ResourceManager to download (once), then retry |
||
697 | # - transient (I/O or OOM) error → maybe sleep, retry |
||
698 | # - persistent (data) error → skip / dummy / raise |
||
699 | try: |
||
700 | self._base_logger.debug("waiting for output of task %s (page %s)", task, page_id) |
||
701 | # timeout kwarg on future is useless: it only raises TimeoutError here, |
||
702 | # but does not stop the running process/thread, and executor itself |
||
703 | # offers nothing to that effect: |
||
704 | # task.result(timeout=max_seconds or None) |
||
705 | # so we instead applied the timeout within the worker function |
||
706 | task.result() |
||
707 | return True |
||
708 | except NotImplementedError: |
||
709 | # exclude NotImplementedError, so we can try process() below |
||
710 | raise |
||
711 | # handle input failures separately |
||
712 | except FileExistsError as err: |
||
713 | if config.OCRD_EXISTING_OUTPUT == 'ABORT': |
||
714 | raise err |
||
715 | if config.OCRD_EXISTING_OUTPUT == 'SKIP': |
||
716 | return False |
||
717 | if config.OCRD_EXISTING_OUTPUT == 'OVERWRITE': |
||
718 | # too late here, must not happen |
||
719 | raise Exception(f"got {err} despite OCRD_EXISTING_OUTPUT==OVERWRITE") |
||
720 | except KeyboardInterrupt: |
||
721 | raise |
||
722 | # broad coverage of output failures (including TimeoutError) |
||
723 | except Exception as err: |
||
724 | # FIXME: add re-usable/actionable logging |
||
725 | if config.OCRD_MISSING_OUTPUT == 'ABORT': |
||
726 | self._base_logger.error(f"Failure on page {page_id}: {str(err) or err.__class__.__name__}") |
||
727 | raise err |
||
728 | self._base_logger.exception(f"Failure on page {page_id}: {str(err) or err.__class__.__name__}") |
||
729 | if config.OCRD_MISSING_OUTPUT == 'SKIP': |
||
730 | pass |
||
731 | elif config.OCRD_MISSING_OUTPUT == 'COPY': |
||
732 | self._copy_page_file(input_files[0]) |
||
733 | else: |
||
734 | desc = config.describe('OCRD_MISSING_OUTPUT', wrap_text=False, indent_text=False) |
||
735 | raise ValueError(f"unknown configuration value {config.OCRD_MISSING_OUTPUT} - {desc}") |
||
736 | return err |
||
737 | |||
738 | def _copy_page_file(self, input_file : OcrdFileType) -> None: |
||
739 | """ |
||
740 | Copy the given ``input_file`` of the :py:data:`workspace`, |
||
741 | representing one physical page (passed as one opened |
||
742 | :py:class:`~ocrd_models.OcrdFile` per input fileGrp) |
||
743 | and add it as if it was a processing result. |
||
744 | """ |
||
745 | input_pcgts : OcrdPage |
||
746 | assert isinstance(input_file, get_args(OcrdFileType)) |
||
747 | self._base_logger.debug(f"parsing file {input_file.ID} for page {input_file.pageId}") |
||
748 | try: |
||
749 | input_pcgts = page_from_file(input_file) |
||
750 | except ValueError as err: |
||
751 | # not PAGE and not an image to generate PAGE for |
||
752 | self._base_logger.error(f"non-PAGE input for page {input_file.pageId}: {err}") |
||
753 | return |
||
754 | output_file_id = make_file_id(input_file, self.output_file_grp) |
||
755 | input_pcgts.set_pcGtsId(output_file_id) |
||
756 | self.add_metadata(input_pcgts) |
||
757 | self.workspace.add_file( |
||
758 | file_id=output_file_id, |
||
759 | file_grp=self.output_file_grp, |
||
760 | page_id=input_file.pageId, |
||
761 | local_filename=os.path.join(self.output_file_grp, output_file_id + '.xml'), |
||
762 | mimetype=MIMETYPE_PAGE, |
||
763 | content=to_xml(input_pcgts), |
||
764 | ) |
||
765 | |||
766 | def process_page_file(self, *input_files : Optional[OcrdFileType]) -> None: |
||
767 | """ |
||
768 | Process the given ``input_files`` of the :py:data:`workspace`, |
||
769 | representing one physical page (passed as one opened |
||
770 | :py:class:`.OcrdFile` per input fileGrp) |
||
771 | under the given :py:data:`.parameter`, and make sure the |
||
772 | results get added accordingly. |
||
773 | |||
774 | (This uses :py:meth:`.process_page_pcgts`, but should be overridden by subclasses |
||
775 | to handle cases like multiple output fileGrps, non-PAGE input etc.) |
||
776 | """ |
||
777 | input_pcgts : List[Optional[OcrdPage]] = [None] * len(input_files) |
||
778 | assert isinstance(input_files[0], get_args(OcrdFileType)) |
||
779 | page_id = input_files[0].pageId |
||
780 | self._base_logger.info("processing page %s", page_id) |
||
781 | for i, input_file in enumerate(input_files): |
||
782 | assert isinstance(input_file, get_args(OcrdFileType)) |
||
783 | self._base_logger.debug(f"parsing file {input_file.ID} for page {page_id}") |
||
784 | try: |
||
785 | page_ = page_from_file(input_file) |
||
786 | assert isinstance(page_, OcrdPage) |
||
787 | input_pcgts[i] = page_ |
||
788 | except ValueError as err: |
||
789 | # not PAGE and not an image to generate PAGE for |
||
790 | self._base_logger.error(f"non-PAGE input for page {page_id}: {err}") |
||
791 | output_file_id = make_file_id(input_files[0], self.output_file_grp) |
||
792 | output_file = next(self.workspace.mets.find_files(ID=output_file_id), None) |
||
793 | if output_file and config.OCRD_EXISTING_OUTPUT != 'OVERWRITE': |
||
794 | # short-cut avoiding useless computation: |
||
795 | raise FileExistsError( |
||
796 | f"A file with ID=={output_file_id} already exists {output_file} and neither force nor ignore are set" |
||
797 | ) |
||
798 | result = self.process_page_pcgts(*input_pcgts, page_id=page_id) |
||
799 | for image_result in result.images: |
||
800 | image_file_id = f'{output_file_id}_{image_result.file_id_suffix}' |
||
801 | image_file_path = join(self.output_file_grp, f'{image_file_id}.png') |
||
802 | if isinstance(image_result.alternative_image, PageType): |
||
803 | # special case: not an alternative image, but replacing the original image |
||
804 | # (this is needed by certain processors when the original's coordinate system |
||
805 | # cannot or must not be kept) |
||
806 | image_result.alternative_image.set_imageFilename(image_file_path) |
||
807 | image_result.alternative_image.set_imageWidth(image_result.pil.width) |
||
808 | image_result.alternative_image.set_imageHeight(image_result.pil.height) |
||
809 | elif isinstance(image_result.alternative_image, AlternativeImageType): |
||
810 | image_result.alternative_image.set_filename(image_file_path) |
||
811 | elif image_result.alternative_image is None: |
||
812 | pass # do not reference in PAGE result |
||
813 | else: |
||
814 | raise ValueError(f"process_page_pcgts returned an OcrdPageResultImage of unknown type " |
||
815 | f"{type(image_result.alternative_image)}") |
||
816 | self.workspace.save_image_file( |
||
817 | image_result.pil, |
||
818 | image_file_id, |
||
819 | self.output_file_grp, |
||
820 | page_id=page_id, |
||
821 | file_path=image_file_path, |
||
822 | ) |
||
823 | result.pcgts.set_pcGtsId(output_file_id) |
||
824 | self.add_metadata(result.pcgts) |
||
825 | self.workspace.add_file( |
||
826 | file_id=output_file_id, |
||
827 | file_grp=self.output_file_grp, |
||
828 | page_id=page_id, |
||
829 | local_filename=os.path.join(self.output_file_grp, output_file_id + '.xml'), |
||
830 | mimetype=MIMETYPE_PAGE, |
||
831 | content=to_xml(result.pcgts), |
||
832 | ) |
||
833 | |||
834 | def process_page_pcgts(self, *input_pcgts : Optional[OcrdPage], page_id : Optional[str] = None) -> OcrdPageResult: |
||
835 | """ |
||
836 | Process the given ``input_pcgts`` of the :py:data:`.workspace`, |
||
837 | representing one physical page (passed as one parsed |
||
838 | :py:class:`.OcrdPage` per input fileGrp) |
||
839 | under the given :py:data:`.parameter`, and return the |
||
840 | resulting :py:class:`.OcrdPageResult`. |
||
841 | |||
842 | Optionally, add to the ``images`` attribute of the resulting |
||
843 | :py:class:`.OcrdPageResult` instances of :py:class:`.OcrdPageResultImage`, |
||
844 | which have required fields for ``pil`` (:py:class:`PIL.Image` image data), |
||
845 | ``file_id_suffix`` (used for generating IDs of the saved image) and |
||
846 | ``alternative_image`` (reference of the :py:class:`ocrd_models.ocrd_page.AlternativeImageType` |
||
847 | for setting the filename of the saved image). |
||
848 | |||
849 | (This contains the main functionality and must be overridden by subclasses, |
||
850 | unless it does not get called by some overriden :py:meth:`.process_page_file`.) |
||
851 | """ |
||
852 | raise NotImplementedError() |
||
853 | |||
854 | def add_metadata(self, pcgts: OcrdPage) -> None: |
||
855 | """ |
||
856 | Add PAGE-XML :py:class:`~ocrd_models.ocrd_page.MetadataItemType` ``MetadataItem`` describing |
||
857 | the processing step and runtime parameters to :py:class:`.OcrdPage` ``pcgts``. |
||
858 | """ |
||
859 | metadata_obj = pcgts.get_Metadata() |
||
860 | assert metadata_obj is not None |
||
861 | metadata_obj.add_MetadataItem( |
||
862 | MetadataItemType(type_="processingStep", |
||
863 | name=self.ocrd_tool['steps'][0], |
||
864 | value=self.ocrd_tool['executable'], |
||
865 | Labels=[LabelsType( |
||
866 | externalModel="ocrd-tool", |
||
867 | externalId="parameters", |
||
868 | Label=[LabelType(type_=name, |
||
869 | value=self.parameter[name]) |
||
870 | for name in self.parameter.keys()]), |
||
871 | LabelsType( |
||
872 | externalModel="ocrd-tool", |
||
873 | externalId="version", |
||
874 | Label=[LabelType(type_=self.ocrd_tool['executable'], |
||
875 | value=self.version), |
||
876 | LabelType(type_='ocrd/core', |
||
877 | value=OCRD_VERSION)]) |
||
878 | ])) |
||
879 | |||
880 | def resolve_resource(self, val): |
||
881 | """ |
||
882 | Resolve a resource name to an absolute file path with the algorithm in |
||
883 | `spec <https://ocr-d.de/en/spec/ocrd_tool#file-parameters>`_ |
||
884 | |||
885 | Args: |
||
886 | val (string): resource value to resolve |
||
887 | """ |
||
888 | executable = self.ocrd_tool['executable'] |
||
889 | if exists(val): |
||
890 | self._base_logger.debug("Resolved to absolute path %s" % val) |
||
891 | return val |
||
892 | # FIXME: remove once workspace arg / old_pwd is gone: |
||
893 | if hasattr(self, 'old_pwd'): |
||
894 | cwd = self.old_pwd |
||
895 | else: |
||
896 | cwd = getcwd() |
||
897 | ret = [cand for cand in list_resource_candidates(executable, val, |
||
898 | cwd=cwd, moduled=self.moduledir) |
||
899 | if exists(cand)] |
||
900 | if ret: |
||
901 | self._base_logger.debug("Resolved %s to absolute path %s" % (val, ret[0])) |
||
902 | return ret[0] |
||
903 | raise ResourceNotFoundError(val, executable) |
||
904 | |||
905 | def show_resource(self, val): |
||
906 | """ |
||
907 | Resolve a resource name to a file path with the algorithm in |
||
908 | `spec <https://ocr-d.de/en/spec/ocrd_tool#file-parameters>`_, |
||
909 | then print its contents to stdout. |
||
910 | |||
911 | Args: |
||
912 | val (string): resource value to show |
||
913 | """ |
||
914 | res_fname = self.resolve_resource(val) |
||
915 | fpath = Path(res_fname) |
||
916 | if fpath.is_dir(): |
||
917 | with pushd_popd(fpath): |
||
918 | fileobj = io.BytesIO() |
||
919 | with tarfile.open(fileobj=fileobj, mode='w:gz') as tarball: |
||
920 | tarball.add('.') |
||
921 | fileobj.seek(0) |
||
922 | copyfileobj(fileobj, sys.stdout.buffer) |
||
923 | else: |
||
924 | sys.stdout.buffer.write(fpath.read_bytes()) |
||
925 | |||
926 | def list_all_resources(self): |
||
927 | """ |
||
928 | List all resources found in the filesystem and matching content-type by filename suffix |
||
929 | """ |
||
930 | mimetypes = get_processor_resource_types(None, self.ocrd_tool) |
||
931 | for res in list_all_resources(self.ocrd_tool['executable'], moduled=self.moduledir): |
||
932 | res = Path(res) |
||
933 | if not '*/*' in mimetypes: |
||
934 | if res.is_dir() and not 'text/directory' in mimetypes: |
||
935 | continue |
||
936 | # if we do not know all MIME types, then keep the file, otherwise require suffix match |
||
937 | if res.is_file() and not any(res.suffix == MIME_TO_EXT.get(mime, res.suffix) |
||
938 | for mime in mimetypes): |
||
939 | continue |
||
940 | yield res |
||
941 | |||
942 | @property |
||
943 | def module(self): |
||
944 | """ |
||
945 | The top-level module this processor belongs to. |
||
946 | """ |
||
947 | # find shortest prefix path that is not just a namespace package |
||
948 | fqname = '' |
||
949 | for name in self.__module__.split('.'): |
||
950 | if fqname: |
||
951 | fqname += '.' |
||
952 | fqname += name |
||
953 | if getattr(sys.modules[fqname], '__file__', None): |
||
954 | return fqname |
||
955 | # fall-back |
||
956 | return self.__module__ |
||
957 | |||
958 | @property |
||
959 | def moduledir(self): |
||
960 | """ |
||
961 | The filesystem path of the module directory. |
||
962 | """ |
||
963 | return resource_filename(self.module, '.') |
||
964 | |||
965 | @property |
||
966 | def input_files(self): |
||
967 | """ |
||
968 | List the input files (for single-valued :py:attr:`input_file_grp`). |
||
969 | |||
970 | For each physical page: |
||
971 | |||
972 | - If there is a single PAGE-XML for the page, take it (and forget about all |
||
973 | other files for that page) |
||
974 | - Else if there is a single image file, take it (and forget about all other |
||
975 | files for that page) |
||
976 | - Otherwise raise an error (complaining that only PAGE-XML warrants |
||
977 | having multiple images for a single page) |
||
978 | |||
979 | See `algorithm <https://github.com/cisocrgroup/ocrd_cis/pull/57#issuecomment-656336593>`_ |
||
980 | |||
981 | Returns: |
||
982 | A list of :py:class:`ocrd_models.ocrd_file.OcrdFile` objects. |
||
983 | """ |
||
984 | if not self.input_file_grp: |
||
985 | raise ValueError("Processor is missing input fileGrp") |
||
986 | ret = self.zip_input_files(mimetype=None, on_error='abort') |
||
987 | if not ret: |
||
988 | return [] |
||
989 | assert len(ret[0]) == 1, 'Use zip_input_files() instead of input_files when processing multiple input fileGrps' |
||
990 | return [tuples[0] for tuples in ret] |
||
991 | |||
992 | def zip_input_files(self, require_first=True, mimetype=None, on_error='skip'): |
||
993 | """ |
||
994 | List tuples of input files (for multi-valued :py:attr:`input_file_grp`). |
||
995 | |||
996 | Processors that expect/need multiple input file groups, |
||
997 | cannot use :py:data:`input_files`. They must align (zip) input files |
||
998 | across pages. This includes the case where not all pages |
||
999 | are equally present in all file groups. It also requires |
||
1000 | making a consistent selection if there are multiple files |
||
1001 | per page. |
||
1002 | |||
1003 | Following the OCR-D functional model, this function tries to |
||
1004 | find a single PAGE file per page, or fall back to a single |
||
1005 | image file per page. In either case, multiple matches per page |
||
1006 | are an error (see error handling below). |
||
1007 | This default behaviour can be changed by using a fixed MIME |
||
1008 | type filter via :py:attr:`mimetype`. But still, multiple matching |
||
1009 | files per page are an error. |
||
1010 | |||
1011 | Single-page multiple-file errors are handled according to |
||
1012 | :py:attr:`on_error`: |
||
1013 | |||
1014 | - if ``skip``, then the page for the respective fileGrp will be |
||
1015 | silently skipped (as if there was no match at all) |
||
1016 | - if ``first``, then the first matching file for the page will be |
||
1017 | silently selected (as if the first was the only match) |
||
1018 | - if ``last``, then the last matching file for the page will be |
||
1019 | silently selected (as if the last was the only match) |
||
1020 | - if ``abort``, then an exception will be raised. |
||
1021 | |||
1022 | Multiple matches for PAGE-XML will always raise an exception. |
||
1023 | |||
1024 | Keyword Args: |
||
1025 | require_first (boolean): If true, then skip a page entirely |
||
1026 | whenever it is not available in the first input `fileGrp`. |
||
1027 | on_error (string): How to handle multiple file matches per page. |
||
1028 | mimetype (string): If not `None`, filter by the specified MIME |
||
1029 | type (literal or regex prefixed by `//`). Otherwise prefer |
||
1030 | PAGE or image. |
||
1031 | Returns: |
||
1032 | A list of :py:class:`ocrd_models.ocrd_file.OcrdFile` tuples. |
||
1033 | """ |
||
1034 | if not self.input_file_grp: |
||
1035 | raise ValueError("Processor is missing input fileGrp") |
||
1036 | |||
1037 | ifgs = self.input_file_grp.split(",") |
||
1038 | # Iterating over all files repeatedly may seem inefficient at first sight, |
||
1039 | # but the unnecessary OcrdFile instantiations for posterior fileGrp filtering |
||
1040 | # can actually be much more costly than traversing the ltree. |
||
1041 | # This might depend on the number of pages vs number of fileGrps. |
||
1042 | |||
1043 | pages = {} |
||
1044 | for i, ifg in enumerate(ifgs): |
||
1045 | files_ = sorted(self.workspace.mets.find_all_files( |
||
1046 | pageId=self.page_id, fileGrp=ifg, mimetype=mimetype), |
||
1047 | # sort by MIME type so PAGE comes before images |
||
1048 | key=lambda file_: file_.mimetype) |
||
1049 | for file_ in files_: |
||
1050 | if not file_.pageId: |
||
1051 | # ignore document-global files |
||
1052 | continue |
||
1053 | ift = pages.setdefault(file_.pageId, [None]*len(ifgs)) |
||
1054 | if ift[i]: |
||
1055 | self._base_logger.debug(f"another file {file_.ID} for page {file_.pageId} in input file group {ifg}") |
||
1056 | # fileGrp has multiple files for this page ID |
||
1057 | if mimetype: |
||
1058 | # filter was active, this must not happen |
||
1059 | self._base_logger.warning(f"added file {file_.ID} for page {file_.pageId} in input file group {ifg} " |
||
1060 | f"conflicts with file {ift[i].ID} of same MIME type {mimetype} - on_error={on_error}") |
||
1061 | View Code Duplication | if on_error == 'skip': |
|
1062 | ift[i] = None |
||
1063 | elif on_error == 'first': |
||
1064 | pass # keep first match |
||
1065 | elif on_error == 'last': |
||
1066 | ift[i] = file_ |
||
1067 | elif on_error == 'abort': |
||
1068 | raise NonUniqueInputFile(ifg, file_.pageId, mimetype) |
||
1069 | else: |
||
1070 | raise Exception("Unknown 'on_error' strategy '%s'" % on_error) |
||
1071 | elif (ift[i].mimetype == MIMETYPE_PAGE and |
||
1072 | file_.mimetype != MIMETYPE_PAGE): |
||
1073 | pass # keep PAGE match |
||
1074 | elif (ift[i].mimetype == MIMETYPE_PAGE and |
||
1075 | file_.mimetype == MIMETYPE_PAGE): |
||
1076 | raise NonUniqueInputFile(ifg, file_.pageId, None) |
||
1077 | else: |
||
1078 | # filter was inactive but no PAGE is in control, this must not happen |
||
1079 | self._base_logger.warning(f"added file {file_.ID} for page {file_.pageId} in input file group {ifg} " |
||
1080 | f"conflicts with file {ift[i].ID} but no PAGE available - on_error={on_error}") |
||
1081 | View Code Duplication | if on_error == 'skip': |
|
1082 | ift[i] = None |
||
1083 | elif on_error == 'first': |
||
1084 | pass # keep first match |
||
1085 | elif on_error == 'last': |
||
1086 | ift[i] = file_ |
||
1087 | elif on_error == 'abort': |
||
1088 | raise NonUniqueInputFile(ifg, file_.pageId, None) |
||
1089 | else: |
||
1090 | raise Exception("Unknown 'on_error' strategy '%s'" % on_error) |
||
1091 | else: |
||
1092 | self._base_logger.debug(f"adding file {file_.ID} for page {file_.pageId} to input file group {ifg}") |
||
1093 | ift[i] = file_ |
||
1094 | # Warn if no files found but pageId was specified, because that might be due to invalid page_id (range) |
||
1095 | if self.page_id and not any(pages): |
||
1096 | self._base_logger.critical(f"Could not find any files for selected pageId {self.page_id}.\n" |
||
1097 | f"compare '{self.page_id}' with the output of 'orcd workspace list-page'.") |
||
1098 | ifts = [] |
||
1099 | for page, ifiles in pages.items(): |
||
1100 | for i, ifg in enumerate(ifgs): |
||
1101 | if not ifiles[i]: |
||
1102 | # could be from non-unique with on_error=skip or from true gap |
||
1103 | self._base_logger.error(f'Found no file for page {page} in file group {ifg}') |
||
1104 | if config.OCRD_MISSING_INPUT == 'abort': |
||
1105 | raise MissingInputFile(ifg, page, mimetype) |
||
1106 | if not any(ifiles): |
||
1107 | # must be from non-unique with on_error=skip |
||
1108 | self._base_logger.warning(f'Found no files for {page} - skipping') |
||
1109 | continue |
||
1110 | if ifiles[0] or not require_first: |
||
1111 | ifts.append(tuple(ifiles)) |
||
1112 | return ifts |
||
1113 | |||
1114 | _page_worker_processor = None |
||
1115 | """ |
||
1116 | This global binding for the processor is required to avoid |
||
1117 | squeezing the processor through a mp.Queue (which is impossible |
||
1118 | due to unpicklable attributes like .workspace.mets._tree anyway) |
||
1119 | when calling Processor.process_page_file as page worker processes |
||
1120 | in Processor.process_workspace. Forking allows inheriting global |
||
1121 | objects, and with the METS Server we do not mutate the local |
||
1122 | processor instance anyway. |
||
1123 | """ |
||
1124 | def _page_worker_set_ctxt(processor, log_queue): |
||
1125 | """ |
||
1126 | Overwrites `ocrd.processor.base._page_worker_processor` instance |
||
1127 | for sharing with subprocesses in ProcessPoolExecutor initializer. |
||
1128 | """ |
||
1129 | global _page_worker_processor |
||
1130 | _page_worker_processor = processor |
||
1131 | if log_queue: |
||
1132 | # replace all log handlers with just one queue handler |
||
1133 | logging.root.handlers = [logging.handlers.QueueHandler(log_queue)] |
||
1134 | |||
1135 | def _page_worker(timeout, *input_files): |
||
1136 | """ |
||
1137 | Wraps a `Processor.process_page_file` call as payload (call target) |
||
1138 | of the ProcessPoolExecutor workers, but also enforces the given timeout. |
||
1139 | """ |
||
1140 | page_id = next((file.pageId for file in input_files |
||
1141 | if hasattr(file, 'pageId')), "") |
||
1142 | if timeout > 0: |
||
1143 | timer = Timer(timeout, interrupt_main) |
||
1144 | timer.start() |
||
1145 | try: |
||
1146 | _page_worker_processor.process_page_file(*input_files) |
||
1147 | _page_worker_processor.logger.debug("page worker completed for page %s", page_id) |
||
1148 | except KeyboardInterrupt: |
||
1149 | _page_worker_processor.logger.debug("page worker timed out for page %s", page_id) |
||
1150 | raise TimeoutError() |
||
1151 | finally: |
||
1152 | if timeout > 0: |
||
1153 | timer.cancel() |
||
1154 | |||
1155 | def generate_processor_help(ocrd_tool, processor_instance=None, subcommand=None): |
||
1156 | """Generate a string describing the full CLI of this processor including params. |
||
1157 | |||
1158 | Args: |
||
1159 | ocrd_tool (dict): this processor's ``tools`` section of the module's ``ocrd-tool.json`` |
||
1160 | processor_instance (object, optional): the processor implementation |
||
1161 | (for adding any module/class/function docstrings) |
||
1162 | subcommand (string): 'worker' or 'server' |
||
1163 | """ |
||
1164 | doc_help = '' |
||
1165 | if processor_instance: |
||
1166 | module = inspect.getmodule(processor_instance) |
||
1167 | if module and module.__doc__: |
||
1168 | doc_help += '\n' + inspect.cleandoc(module.__doc__) + '\n' |
||
1169 | if processor_instance.__doc__: |
||
1170 | doc_help += '\n' + inspect.cleandoc(processor_instance.__doc__) + '\n' |
||
1171 | # Try to find the most concrete docstring among the various methods that an implementation |
||
1172 | # could overload, first serving. |
||
1173 | # In doing so, compare with Processor to avoid a glitch in the way py>=3.5 inherits docstrings. |
||
1174 | # (They are supposed to only repeat information inspect.getdoc, rather than inherit __doc__ itself.) |
||
1175 | for method in ['process_page_pcgts', 'process_page_file', 'process_workspace', 'process']: |
||
1176 | instance_method = getattr(processor_instance, method) |
||
1177 | superclass_method = getattr(Processor, method) |
||
1178 | if instance_method.__doc__ and instance_method.__doc__ != superclass_method.__doc__: |
||
1179 | doc_help += '\n' + inspect.cleandoc(instance_method.__doc__) + '\n' |
||
1180 | break |
||
1181 | if doc_help: |
||
1182 | doc_help = '\n\n' + wrap_text(doc_help, width=72, |
||
1183 | initial_indent=' > ', |
||
1184 | subsequent_indent=' > ', |
||
1185 | preserve_paragraphs=True) |
||
1186 | subcommands = '''\ |
||
1187 | worker Start a processing worker rather than do local processing |
||
1188 | server Start a processor server rather than do local processing |
||
1189 | ''' |
||
1190 | |||
1191 | processing_worker_options = '''\ |
||
1192 | --queue The RabbitMQ server address in format |
||
1193 | "amqp://{user}:{pass}@{host}:{port}/{vhost}" |
||
1194 | [amqp://admin:admin@localhost:5672] |
||
1195 | --database The MongoDB server address in format |
||
1196 | "mongodb://{host}:{port}" |
||
1197 | [mongodb://localhost:27018] |
||
1198 | --log-filename Filename to redirect STDOUT/STDERR to, |
||
1199 | if specified. |
||
1200 | ''' |
||
1201 | |||
1202 | processing_server_options = '''\ |
||
1203 | --address The Processor server address in format |
||
1204 | "{host}:{port}" |
||
1205 | --database The MongoDB server address in format |
||
1206 | "mongodb://{host}:{port}" |
||
1207 | [mongodb://localhost:27018] |
||
1208 | ''' |
||
1209 | |||
1210 | processing_options = '''\ |
||
1211 | -m, --mets URL-PATH URL or file path of METS to process [./mets.xml] |
||
1212 | -w, --working-dir PATH Working directory of local workspace [dirname(URL-PATH)] |
||
1213 | -I, --input-file-grp USE File group(s) used as input |
||
1214 | -O, --output-file-grp USE File group(s) used as output |
||
1215 | -g, --page-id ID Physical page ID(s) to process instead of full document [] |
||
1216 | --overwrite Remove existing output pages/images |
||
1217 | (with "--page-id", remove only those). |
||
1218 | Short-hand for OCRD_EXISTING_OUTPUT=OVERWRITE |
||
1219 | --debug Abort on any errors with full stack trace. |
||
1220 | Short-hand for OCRD_MISSING_OUTPUT=ABORT |
||
1221 | --profile Enable profiling |
||
1222 | --profile-file PROF-PATH Write cProfile stats to PROF-PATH. Implies "--profile" |
||
1223 | -p, --parameter JSON-PATH Parameters, either verbatim JSON string |
||
1224 | or JSON file path |
||
1225 | -P, --param-override KEY VAL Override a single JSON object key-value pair, |
||
1226 | taking precedence over --parameter |
||
1227 | -U, --mets-server-url URL URL of a METS Server for parallel incremental access to METS |
||
1228 | If URL starts with http:// start an HTTP server there, |
||
1229 | otherwise URL is a path to an on-demand-created unix socket |
||
1230 | -l, --log-level [OFF|ERROR|WARN|INFO|DEBUG|TRACE] |
||
1231 | Override log level globally [INFO] |
||
1232 | --log-filename LOG-PATH File to redirect stderr logging to (overriding ocrd_logging.conf). |
||
1233 | ''' |
||
1234 | |||
1235 | information_options = '''\ |
||
1236 | -C, --show-resource RESNAME Dump the content of processor resource RESNAME |
||
1237 | -L, --list-resources List names of processor resources |
||
1238 | -J, --dump-json Dump tool description as JSON |
||
1239 | -D, --dump-module-dir Show the 'module' resource location path for this processor |
||
1240 | -h, --help Show this message |
||
1241 | -V, --version Show version |
||
1242 | ''' |
||
1243 | |||
1244 | parameter_help = '' |
||
1245 | if 'parameters' not in ocrd_tool or not ocrd_tool['parameters']: |
||
1246 | parameter_help = ' NONE\n' |
||
1247 | else: |
||
1248 | def wrap(s): |
||
1249 | return wrap_text(s, initial_indent=' '*3, |
||
1250 | subsequent_indent=' '*4, |
||
1251 | width=72, preserve_paragraphs=True) |
||
1252 | for param_name, param in ocrd_tool['parameters'].items(): |
||
1253 | parameter_help += wrap('"%s" [%s%s]' % ( |
||
1254 | param_name, |
||
1255 | param['type'], |
||
1256 | ' - REQUIRED' if 'required' in param and param['required'] else |
||
1257 | ' - %s' % json.dumps(param['default']) if 'default' in param else '')) |
||
1258 | parameter_help += '\n ' + wrap(param['description']) |
||
1259 | if 'enum' in param: |
||
1260 | parameter_help += '\n ' + wrap('Possible values: %s' % json.dumps(param['enum'])) |
||
1261 | parameter_help += "\n" |
||
1262 | |||
1263 | if not subcommand: |
||
1264 | return f'''\ |
||
1265 | Usage: {ocrd_tool['executable']} [worker|server] [OPTIONS] |
||
1266 | |||
1267 | {ocrd_tool['description']}{doc_help} |
||
1268 | |||
1269 | Subcommands: |
||
1270 | {subcommands} |
||
1271 | Options for processing: |
||
1272 | {processing_options} |
||
1273 | Options for information: |
||
1274 | {information_options} |
||
1275 | Parameters: |
||
1276 | {parameter_help} |
||
1277 | ''' |
||
1278 | elif subcommand == 'worker': |
||
1279 | return f'''\ |
||
1280 | Usage: {ocrd_tool['executable']} worker [OPTIONS] |
||
1281 | |||
1282 | Run {ocrd_tool['executable']} as a processing worker. |
||
1283 | |||
1284 | {ocrd_tool['description']}{doc_help} |
||
1285 | |||
1286 | Options: |
||
1287 | {processing_worker_options} |
||
1288 | ''' |
||
1289 | elif subcommand == 'server': |
||
1290 | return f'''\ |
||
1291 | Usage: {ocrd_tool['executable']} server [OPTIONS] |
||
1292 | |||
1293 | Run {ocrd_tool['executable']} as a processor sever. |
||
1294 | |||
1295 | {ocrd_tool['description']}{doc_help} |
||
1296 | |||
1297 | Options: |
||
1298 | {processing_server_options} |
||
1299 | ''' |
||
1300 | else: |
||
1301 | pass |
||
1302 |