data.py 44.4 KB
Newer Older
Markus Scheidgen's avatar
Markus Scheidgen committed
1
2
3
4
5
6
7
8
9
10
11
12
13
14
# Copyright 2018 Markus Scheidgen
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#   http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an"AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

15
'''
Markus Scheidgen's avatar
Markus Scheidgen committed
16
17
18
19
20
21
This module comprises a set of persistent document classes that hold all user related
data. These are information about users, their uploads and datasets, the associated
calculations, and files


.. autoclass:: Calc
Markus Scheidgen's avatar
Markus Scheidgen committed
22

Markus Scheidgen's avatar
Markus Scheidgen committed
23
.. autoclass:: Upload
Markus Scheidgen's avatar
Markus Scheidgen committed
24

25
'''
Markus Scheidgen's avatar
Markus Scheidgen committed
26

27
from typing import cast, List, Any, ContextManager, Tuple, Generator, Dict, cast, Iterable
28
from mongoengine import StringField, DateTimeField, DictField, BooleanField, IntField
Markus Scheidgen's avatar
Markus Scheidgen committed
29
import logging
30
from structlog import wrap_logger
31
from contextlib import contextmanager
32
import os.path
33
34
from datetime import datetime
from pymongo import UpdateOne
35
import hashlib
36
37
from structlog.processors import StackInfoRenderer, format_exc_info, TimeStamper
import json
Markus Scheidgen's avatar
Markus Scheidgen committed
38

39
from nomad import utils, config, infrastructure, search, datamodel
40
from nomad.files import PathObject, UploadFiles, ExtractError, ArchiveBasedStagingUploadFiles, PublicUploadFiles, StagingUploadFiles
41
from nomad.processing.base import Proc, process, task, PENDING, SUCCESS, FAILURE
42
from nomad.parsing import parser_dict, match_parser, Backend
Markus Scheidgen's avatar
Markus Scheidgen committed
43
44
45
from nomad.normalizing import normalizers


46
47
48
49
50
51
52
def _pack_log_event(logger, method_name, event_dict):
    try:
        log_data = dict(event_dict)
        log_data.update(**{
            key: value
            for key, value in getattr(logger, '_context', {}).items()
            if key not in ['service', 'release', 'upload_id', 'calc_id', 'mainfile', 'process_status']})
53
        log_data.update(logger=logger.name)
54
55
56
57
58
59
60
61
62
63
64

        return log_data
    except Exception:
        # raising an exception would cause an indefinite loop
        return event_dict


_log_processors = [
    StackInfoRenderer(),
    _pack_log_event,
    format_exc_info,
65
    TimeStamper(fmt="%Y-%m-%d %H:%M.%S", utc=False)]
66
67


Markus Scheidgen's avatar
Markus Scheidgen committed
68
class Calc(Proc):
69
    '''
Markus Scheidgen's avatar
Markus Scheidgen committed
70
71
72
73
74
    Instances of this class represent calculations. This class manages the elastic
    search index entry, files, and archive for the respective calculation.

    It also contains the calculations processing and its state.

75
76
    The attribute list, does not include the various metadata properties generated
    while parsing, including ``code_name``, ``code_version``, etc.
Markus Scheidgen's avatar
Markus Scheidgen committed
77
78

    Attributes:
79
        calc_id: the calc_id of this calc
Markus Scheidgen's avatar
Markus Scheidgen committed
80
81
82
        parser: the name of the parser used to process this calc
        upload_id: the id of the upload used to create this calculation
        mainfile: the mainfile (including path in upload) that was used to create this calc
83

84
85
        metadata: the metadata record wit calc and user metadata, see :class:`datamodel.EntryMetadata`
    '''
86
    calc_id = StringField(primary_key=True)
Markus Scheidgen's avatar
Markus Scheidgen committed
87
88
89
90
    upload_id = StringField()
    mainfile = StringField()
    parser = StringField()

91
92
    metadata = DictField()

Markus Scheidgen's avatar
Markus Scheidgen committed
93
    meta: Any = {
Markus Scheidgen's avatar
Markus Scheidgen committed
94
        'indexes': [
95
96
97
98
            'upload_id',
            ('upload_id', 'mainfile'),
            ('upload_id', 'parser'),
            ('upload_id', 'tasks_status'),
99
            ('upload_id', 'process_status'),
100
101
102
            ('upload_id', 'metadata.nomad_version'),
            'metadata.published',
            'metadata.datasets'
Markus Scheidgen's avatar
Markus Scheidgen committed
103
            'metadata.pid'
Markus Scheidgen's avatar
Markus Scheidgen committed
104
105
106
107
108
        ]
    }

    def __init__(self, *args, **kwargs):
        super().__init__(*args, **kwargs)
109
        self._parser_backend: Backend = None
110
111
        self._upload: Upload = None
        self._upload_files: ArchiveBasedStagingUploadFiles = None
112
        self._calc_proc_logwriter = None
113
        self._calc_proc_logwriter_ctx: ContextManager = None
Markus Scheidgen's avatar
Markus Scheidgen committed
114

115
    @classmethod
116
    def from_entry_metadata(cls, entry_metadata):
117
        calc = Calc.create(
118
119
120
121
            calc_id=entry_metadata.calc_id,
            upload_id=entry_metadata.upload_id,
            mainfile=entry_metadata.mainfile,
            metadata=entry_metadata.m_to_dict(include_defaults=True))
122
123
124

        return calc

Markus Scheidgen's avatar
Markus Scheidgen committed
125
126
    @classmethod
    def get(cls, id):
127
        return cls.get_by_id(id, 'calc_id')
Markus Scheidgen's avatar
Markus Scheidgen committed
128

Markus Scheidgen's avatar
Markus Scheidgen committed
129
    @property
130
131
    def mainfile_file(self) -> PathObject:
        return self.upload_files.raw_file_object(self.mainfile)
Markus Scheidgen's avatar
Markus Scheidgen committed
132

133
134
135
136
    @property
    def upload(self) -> 'Upload':
        if not self._upload:
            self._upload = Upload.get(self.upload_id)
137
            self._upload.worker_hostname = self.worker_hostname
138
139
        return self._upload

140
141
142
    @property
    def upload_files(self) -> ArchiveBasedStagingUploadFiles:
        if not self._upload_files:
143
144
            self._upload_files = ArchiveBasedStagingUploadFiles(
                self.upload_id, is_authorized=lambda: True, upload_path=self.upload.upload_path)
145
146
        return self._upload_files

Markus Scheidgen's avatar
Markus Scheidgen committed
147
    def get_logger(self, **kwargs):
148
        '''
149
150
        Returns a wrapped logger that additionally saves all entries to the calculation
        processing log in the archive.
151
        '''
152
153
154
        logger = super().get_logger()
        logger = logger.bind(
            upload_id=self.upload_id, mainfile=self.mainfile, calc_id=self.calc_id, **kwargs)
155

156
        if self._calc_proc_logwriter_ctx is None:
157
158
159
160
161
162
163
164
165
166
167
168
            try:
                self._calc_proc_logwriter_ctx = self.upload_files.archive_log_file(self.calc_id, 'wt')
                self._calc_proc_logwriter = self._calc_proc_logwriter_ctx.__enter__()  # pylint: disable=E1101
            except KeyError:
                # cannot open log file
                pass

        if self._calc_proc_logwriter_ctx is None:
            return logger
        else:
            def save_to_calc_log(logger, method_name, event_dict):
                if self._calc_proc_logwriter is not None:
169
                    try:
170
171
172
                        dump_dict = dict(event_dict)
                        dump_dict.update(level=method_name.upper())
                        json.dump(dump_dict, self._calc_proc_logwriter, sort_keys=True)
173
                        self._calc_proc_logwriter.write('\n')
174

175
176
177
                    except Exception:
                        # Exceptions here will cause indefinite loop
                        pass
178
179
180

                return event_dict

181
            return wrap_logger(logger, processors=_log_processors + [save_to_calc_log])
182

183
184
    @process
    def re_process_calc(self):
185
        '''
186
187
188
        Processes a calculation again. This means there is already metadata and
        instead of creating it initially, we are just updating the existing
        records.
189
        '''
190
        parser = match_parser(self.upload_files.raw_file_object(self.mainfile).os_path, strict=False)
191
192

        if parser is None and not config.reprocess_unmatched:
193
194
195
196
197
198
199
200
201
            # Remove the logsfile and set a fake logwriter to avoid creating a log file,
            # because we will probably remove this calc and don't want to have ghost logfiles.
            if self._calc_proc_logwriter_ctx is not None:
                self._calc_proc_logwriter_ctx.__exit__(None, None, None)
                self.upload_files.archive_log_file_object(self.calc_id).delete()

            self._calc_proc_logwriter_ctx = open('/dev/null', 'wt')
            self._calc_proc_logwriter = self._calc_proc_logwriter_ctx.__enter__()
            self.get_logger().error(
202
203
204
                'no parser matches during re-process, will not re-process this calc')

            self.errors = ['no parser matches during re-process, will not re-process this calc']
205
206
207
208
209
210

            # mock the steps of actual processing
            self._continue_with('parsing')
            self._continue_with('normalizing')
            self._continue_with('archiving')
            self._complete()
211
212
            return

213
        logger = self.get_logger()
214
        if parser is None:
215
216
            self.get_logger().error('no parser matches during re-process, use the old parser')
            self.errors = ['no matching parser found during re-processing']
217
        if self.parser != parser.name:
218
219
220
            self.parser = parser.name
            logger.info(
                'different parser matches during re-process, use new parser',
221
                parser=parser.name)
222

223
        try:
224
225
226
227
228
229
230
231
232
233
            entry_metadata = datamodel.EntryMetadata.m_from_dict(self.metadata)
            entry_metadata.upload_id = self.upload_id
            entry_metadata.calc_id = self.calc_id
            entry_metadata.calc_hash = self.upload_files.calc_hash(self.mainfile)
            entry_metadata.mainfile = self.mainfile
            entry_metadata.nomad_version = config.version
            entry_metadata.nomad_commit = config.commit
            entry_metadata.last_processing = datetime.utcnow()
            entry_metadata.files = self.upload_files.calc_files(self.mainfile)
            self.metadata = entry_metadata.m_to_dict(include_defaults=True)
234
235
236
237
238
239
240
241
242
243
244
245
246

            self.parsing()
            self.normalizing()
            self.archiving()
        finally:
            # close loghandler that was not closed due to failures
            try:
                if self._calc_proc_logwriter is not None:
                    self._calc_proc_logwriter.close()
                    self._calc_proc_logwriter = None
            except Exception as e:
                logger.error('could not close calculation proc log', exc_info=e)

Markus Scheidgen's avatar
Markus Scheidgen committed
247
    @process
248
    def process_calc(self):
249
        '''
250
251
        Processes a new calculation that has no prior records in the mongo, elastic,
        or filesystem storage. It will create an initial set of (user) metadata.
252
        '''
253
        logger = self.get_logger()
254
        if self.upload is None:
255
            logger.error('calculation upload does not exist')
Markus Scheidgen's avatar
Markus Scheidgen committed
256
257

        try:
258
259
            # save preliminary minimum calc metadata in case processing fails
            # successful processing will replace it with the actual metadata
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
            calc_metadata = datamodel.EntryMetadata()
            calc_metadata.domain = parser_dict[self.parser].domain
            calc_metadata.upload_id = self.upload_id
            calc_metadata.calc_id = self.calc_id
            calc_metadata.calc_hash = self.upload_files.calc_hash(self.mainfile)
            calc_metadata.mainfile = self.mainfile
            calc_metadata.calc_hash = self.upload_files.calc_hash(self.mainfile)
            calc_metadata.nomad_version = config.version
            calc_metadata.nomad_commit = config.commit
            calc_metadata.last_processing = datetime.utcnow()
            calc_metadata.files = self.upload_files.calc_files(self.mainfile)
            calc_metadata.uploader = self.upload.user_id
            calc_metadata.upload_time = self.upload.upload_time
            calc_metadata.upload_name = self.upload.name
            self.metadata = calc_metadata.m_to_dict(include_defaults=True)  # TODO use embedded doc?

            if len(calc_metadata.files) >= config.auxfile_cutoff:
277
278
279
280
                self.warning(
                    'This calc has many aux files in its directory. '
                    'Have you placed many calculations in the same directory?')

Markus Scheidgen's avatar
Markus Scheidgen committed
281
282
283
284
            self.parsing()
            self.normalizing()
            self.archiving()
        finally:
Markus Scheidgen's avatar
Markus Scheidgen committed
285
            # close loghandler that was not closed due to failures
286
            try:
287
288
289
                if self._calc_proc_logwriter is not None:
                    self._calc_proc_logwriter.close()
                    self._calc_proc_logwriter = None
290
291
292
            except Exception as e:
                logger.error('could not close calculation proc log', exc_info=e)

293
294
295
296
    def fail(self, *errors, log_level=logging.ERROR, **kwargs):
        # in case of failure, index a minimum set of metadata and mark
        # processing failure
        try:
297
            entry_metadata = datamodel.EntryMetadata.m_from_dict(self.metadata)
298
299
300
            if self.parser is not None:
                parser = parser_dict[self.parser]
                if hasattr(parser, 'code_name'):
301
                    entry_metadata.code_name = parser.code_name
302

303
304
            entry_metadata.processed = False
            self.metadata = entry_metadata.m_to_dict(include_defaults=True)
305
            entry_metadata.a_elastic.index()
306
307
308
309
310
        except Exception as e:
            self.get_logger().error('could not index after processing failure', exc_info=e)

        super().fail(*errors, log_level=log_level, **kwargs)

311
312
313
314
    def on_process_complete(self, process_name):
        # the save might be necessary to correctly read the join condition from the db
        self.save()
        # in case of error, the process_name might be unknown
315
        if process_name == 'process_calc' or process_name == 're_process_calc' or process_name is None:
316
317
            self.upload.reload()
            self.upload.check_join()
Markus Scheidgen's avatar
Markus Scheidgen committed
318
319
320

    @task
    def parsing(self):
321
        ''' The *task* that encapsulates all parsing related actions. '''
322
        context = dict(parser=self.parser, step=self.parser)
323
        logger = self.get_logger(**context)
324
        parser = parser_dict[self.parser]
325
        self.metadata['parser_name'] = self.parser
Markus Scheidgen's avatar
Markus Scheidgen committed
326

327
        with utils.timer(logger, 'parser executed', input_size=self.mainfile_file.size):
328
329
330
331
            try:
                self._parser_backend = parser.run(
                    self.upload_files.raw_file_object(self.mainfile).os_path, logger=logger)
            except Exception as e:
332
                self.fail('parser failed with exception', exc_info=e, error=str(e), **context)
333
                return
334
            except SystemExit:
335
                self.fail('parser raised system exit', error='system exit', **context)
336
                return
Markus Scheidgen's avatar
Markus Scheidgen committed
337

338
339
        # add the non code specific calc metadata to the backend
        # all other quantities have been determined by parsers/normalizers
340
        self._parser_backend.openNonOverlappingSection('section_entry_info')
341
        self._parser_backend.addValue('upload_id', self.upload_id)
342
        self._parser_backend.addValue('calc_id', self.calc_id)
343
        self._parser_backend.addValue('calc_hash', self.metadata['calc_hash'])
344
        self._parser_backend.addValue('mainfile', self.mainfile)
345
        self._parser_backend.addValue('parser_name', self.parser)
346
347
348
349
350
351
352
353
354
355
356
357
        filepaths = self.metadata['files']
        self._parser_backend.addValue('number_of_files', len(filepaths))
        self._parser_backend.addValue('filepaths', filepaths)
        uploader = self.upload.uploader
        self._parser_backend.addValue(
            'entry_uploader_name', '%s, %s' % (uploader.first_name, uploader.last_name))
        self._parser_backend.addValue(
            'entry_uploader_id', str(uploader.user_id))
        self._parser_backend.addValue('entry_upload_time', int(self.upload.upload_time.timestamp()))
        self._parser_backend.closeNonOverlappingSection('section_entry_info')

        self.add_processor_info(self.parser)
358

Markus Scheidgen's avatar
Markus Scheidgen committed
359
360
        if self._parser_backend.status[0] != 'ParseSuccess':
            error = self._parser_backend.status[1]
361
            self.fail('parser failed', error=error, **context)
362
363
364
365
366
367
368
369

    @contextmanager
    def use_parser_backend(self, processor_name):
        self._parser_backend.reset_status()
        yield self._parser_backend
        self.add_processor_info(processor_name)

    def add_processor_info(self, processor_name: str) -> None:
370
        self._parser_backend.openContext('/section_entry_info/0')
371
372
373
374
375
376
377
        self._parser_backend.openNonOverlappingSection('section_archive_processing_info')
        self._parser_backend.addValue('archive_processor_name', processor_name)

        if self._parser_backend.status[0] == 'ParseSuccess':
            warnings = getattr(self._parser_backend, '_warnings', [])
            if len(warnings) > 0:
                self._parser_backend.addValue('archive_processor_status', 'WithWarnings')
378
                self._parser_backend.addValue('number_of_archive_processor_warnings', len(warnings))
379
380
381
382
                self._parser_backend.addArrayValues('archive_processor_warnings', [str(warning) for warning in warnings])
            else:
                self._parser_backend.addValue('archive_processor_status', 'Success')
        else:
383
            errors = self._parser_backend.status[1]
384
385
386
            self._parser_backend.addValue('archive_processor_error', str(errors))

        self._parser_backend.closeNonOverlappingSection('section_archive_processing_info')
387
        self._parser_backend.closeContext('/section_entry_info/0')
Markus Scheidgen's avatar
Markus Scheidgen committed
388
389
390

    @task
    def normalizing(self):
391
        ''' The *task* that encapsulates all normalizing related actions. '''
Markus Scheidgen's avatar
Markus Scheidgen committed
392
        for normalizer in normalizers:
393
            if normalizer.domain != parser_dict[self.parser].domain:
394
395
                continue

Markus Scheidgen's avatar
Markus Scheidgen committed
396
            normalizer_name = normalizer.__name__
397
            context = dict(normalizer=normalizer_name, step=normalizer_name)
398
            logger = self.get_logger(**context)
Markus Scheidgen's avatar
Markus Scheidgen committed
399
400

            with utils.timer(
401
                    logger, 'normalizer executed', input_size=self.mainfile_file.size):
402
                with self.use_parser_backend(normalizer_name) as backend:
403
404
405
                    try:
                        normalizer(backend).normalize(logger=logger)
                    except Exception as e:
406
                        self._parser_backend.finishedParsingSession('ParseFailure', [str(e)])
407
                        self.fail(
408
409
410
411
412
413
414
415
416
417
                            'normalizer failed with exception', exc_info=e, error=str(e), **context)
                        break
                    else:
                        if self._parser_backend.status[0] != 'ParseSuccess':
                            error = self._parser_backend.status[1]
                            self.fail('normalizer failed', error=error, **context)
                            break
                        else:
                            logger.debug(
                                'completed normalizer successfully', normalizer=normalizer_name)
Markus Scheidgen's avatar
Markus Scheidgen committed
418
419
420

    @task
    def archiving(self):
421
        ''' The *task* that encapsulates all archival related actions. '''
422
423
        logger = self.get_logger()

424
425
426
        entry_metadata = datamodel.EntryMetadata.m_from_dict(self.metadata)
        entry_metadata.apply_domain_metadata(self._parser_backend)
        entry_metadata.processed = True
427

428
429
        # persist the calc metadata
        with utils.timer(logger, 'saved calc metadata', step='metadata'):
430
            self.metadata = entry_metadata.m_to_dict(include_defaults=True)
431
432

        # index in search
433
        with utils.timer(logger, 'indexed', step='index'):
434
            entry_metadata.a_elastic.index()
435

436
        # persist the archive
Markus Scheidgen's avatar
Markus Scheidgen committed
437
        with utils.timer(
438
                logger, 'archived', step='archive',
Markus Scheidgen's avatar
Markus Scheidgen committed
439
                input_size=self.mainfile_file.size) as log_data:
440

441
            with self.upload_files.archive_file(self.calc_id, 'wt') as out:
442
443
                json.dump(self._parser_backend.resource.m_to_dict(
                    lambda section: section.m_def.name in datamodel.root_sections), out, indent=2)
444

445
            log_data.update(archive_size=self.upload_files.archive_file_object(self.calc_id).size)
Markus Scheidgen's avatar
Markus Scheidgen committed
446
447
448
449

        # close loghandler
        if self._calc_proc_logwriter is not None:
            with utils.timer(
450
                    logger, 'archived log', step='logs',
Markus Scheidgen's avatar
Markus Scheidgen committed
451
                    input_size=self.mainfile_file.size) as log_data:
452
                self._calc_proc_logwriter_ctx.__exit__(None, None, None)  # pylint: disable=E1101
453
                self._calc_proc_logwriter = None
Markus Scheidgen's avatar
Markus Scheidgen committed
454

455
                log_data.update(log_size=self.upload_files.archive_log_file_object(self.calc_id).size)
Markus Scheidgen's avatar
Markus Scheidgen committed
456

457
    def __str__(self):
458
        return 'calc %s calc_id=%s upload_id%s' % (super().__str__(), self.calc_id, self.upload_id)
459

460

461
class Upload(Proc):
462
    '''
Markus Scheidgen's avatar
Markus Scheidgen committed
463
464
465
466
467
    Represents uploads in the databases. Provides persistence access to the files storage,
    and processing state.

    Attributes:
        name: optional user provided upload name
468
469
        upload_path: the path were the uploaded files was stored
        temporary: True if the uploaded file should be removed after extraction
Markus Scheidgen's avatar
Markus Scheidgen committed
470
471
        upload_id: the upload id generated by the database
        upload_time: the timestamp when the system realised the upload
472
        user_id: the id of the user that created this upload
Markus Scheidgen's avatar
Markus Scheidgen committed
473
474
        published: Boolean that indicates the publish status
        publish_time: Date when the upload was initially published
475
        last_update: Date of the last publishing/re-processing
Markus Scheidgen's avatar
Markus Scheidgen committed
476
        joined: Boolean indicates if the running processing has joined (:func:`check_join`)
477
    '''
Markus Scheidgen's avatar
Markus Scheidgen committed
478
479
480
    id_field = 'upload_id'

    upload_id = StringField(primary_key=True)
481
482
    upload_path = StringField(default=None)
    temporary = BooleanField(default=False)
483
    embargo_length = IntField(default=36)
Markus Scheidgen's avatar
Markus Scheidgen committed
484
485
486

    name = StringField(default=None)
    upload_time = DateTimeField()
487
    user_id = StringField(required=True)
488
489
    published = BooleanField(default=False)
    publish_time = DateTimeField()
490
    last_update = DateTimeField()
Markus Scheidgen's avatar
Markus Scheidgen committed
491

492
493
    joined = BooleanField(default=False)

Markus Scheidgen's avatar
Markus Scheidgen committed
494
495
    meta: Any = {
        'indexes': [
496
            'user_id', 'tasks_status', 'process_status', 'published', 'upload_time'
Markus Scheidgen's avatar
Markus Scheidgen committed
497
498
499
500
501
        ]
    }

    def __init__(self, **kwargs):
        super().__init__(**kwargs)
502
        self._upload_files: ArchiveBasedStagingUploadFiles = None
Markus Scheidgen's avatar
Markus Scheidgen committed
503

504
505
    @property
    def metadata(self) -> dict:
506
        '''
Markus Scheidgen's avatar
Markus Scheidgen committed
507
508
509
510
511
        Getter, setter for user metadata. Metadata is pickled to and from the public
        bucket to allow sharing among all processes. Usually uploads do not have (much)
        user defined metadata, but users provide all metadata per upload as part of
        the publish process. This will change, when we introduce editing functionality
        and metadata will be provided through different means.
512
        '''
513
514
515
516
517
        try:
            upload_files = PublicUploadFiles(self.upload_id, is_authorized=lambda: True)
        except KeyError:
            return None
        return upload_files.user_metadata
518
519
520

    @metadata.setter
    def metadata(self, data: dict) -> None:
521
522
        upload_files = PublicUploadFiles(self.upload_id, is_authorized=lambda: True, create=True)
        upload_files.user_metadata = data
523

Markus Scheidgen's avatar
Markus Scheidgen committed
524
    @classmethod
525
    def get(cls, id: str, include_published: bool = True) -> 'Upload':
526
        return cls.get_by_id(id, 'upload_id')
Markus Scheidgen's avatar
Markus Scheidgen committed
527
528

    @classmethod
529
    def user_uploads(cls, user: datamodel.User, **kwargs) -> List['Upload']:
530
        ''' Returns all uploads for the given user. Kwargs are passed to mongo query. '''
531
        return cls.objects(user_id=str(user.user_id), **kwargs)
Markus Scheidgen's avatar
Markus Scheidgen committed
532

533
534
    @property
    def uploader(self):
535
        return datamodel.User.get(self.user_id)
536

Markus Scheidgen's avatar
Markus Scheidgen committed
537
538
    def get_logger(self, **kwargs):
        logger = super().get_logger()
539
540
        user = self.uploader
        user_name = '%s %s' % (user.first_name, user.last_name)
541
        # We are not using 'user_id' because logstash (?) will filter these entries ?!
542
543
        logger = logger.bind(
            upload_id=self.upload_id, upload_name=self.name, user_name=user_name,
544
            user=user.user_id, **kwargs)
Markus Scheidgen's avatar
Markus Scheidgen committed
545
546
547
548
        return logger

    @classmethod
    def create(cls, **kwargs) -> 'Upload':
549
        '''
Markus Scheidgen's avatar
Markus Scheidgen committed
550
551
552
        Creates a new upload for the given user, a user given name is optional.
        It will populate the record with a signed url and pending :class:`UploadProc`.
        The upload will be already saved to the database.
553
554

        Arguments:
555
            user: The user that created the upload.
556
        '''
557
558
        # use kwargs to keep compatibility with super method
        user: datamodel.User = kwargs['user']
559
        del(kwargs['user'])
560

561
562
        if 'upload_id' not in kwargs:
            kwargs.update(upload_id=utils.create_uuid())
563
        kwargs.update(user_id=user.user_id)
Markus Scheidgen's avatar
Markus Scheidgen committed
564
        self = super().create(**kwargs)
565

Markus Scheidgen's avatar
Markus Scheidgen committed
566
        self._continue_with('uploading')
567

Markus Scheidgen's avatar
Markus Scheidgen committed
568
569
        return self

570
    def delete(self):
571
        ''' Deletes this upload process state entry and its calcs. '''
572
573
574
        Calc.objects(upload_id=self.upload_id).delete()
        super().delete()

575
    def delete_upload_local(self):
576
        '''
577
        Deletes the upload, including its processing state and
578
        staging files. Local version without celery processing.
579
        '''
580
581
582
        logger = self.get_logger()

        with utils.lnr(logger, 'staged upload delete failed'):
583
            with utils.timer(
584
                    logger, 'upload deleted from index', step='index',
585
                    upload_size=self.upload_files.size):
586
                search.delete_upload(self.upload_id)
587

588
            with utils.timer(
589
                    logger, 'staged upload deleted', step='files',
590
591
                    upload_size=self.upload_files.size):
                self.upload_files.delete()
592
593

            self.delete()
594

595
    @process
596
    def delete_upload(self):
597
        '''
598
599
        Deletes of the upload, including its processing state and
        staging files. This starts the celery process of deleting the upload.
600
        '''
601
        self.delete_upload_local()
602

603
        return True  # do not save the process status on the delete upload
604

605
    @process
606
    def publish_upload(self):
607
        '''
608
609
        Moves the upload out of staging to the public area. It will
        pack the staging upload files in to public upload files.
610
        '''
611
612
        assert self.processed_calcs > 0

613
        logger = self.get_logger()
614
        logger.info('started to publish')
615

616
        with utils.lnr(logger, 'publish failed'):
617
            calcs = self.entries_metadata(self.metadata)
618

619
            with utils.timer(
620
                    logger, 'upload metadata updated', step='metadata',
621
                    upload_size=self.upload_files.size):
622
623

                def create_update(calc):
624
                    calc.published = True
625
                    calc.with_embargo = calc.with_embargo if calc.with_embargo is not None else False
626
627
                    return UpdateOne(
                        {'_id': calc.calc_id},
628
                        {'$set': {'metadata': calc.m_to_dict(include_defaults=True)}})
629
630

                Calc._get_collection().bulk_write([create_update(calc) for calc in calcs])
631

632
633
634
635
            if isinstance(self.upload_files, StagingUploadFiles):
                with utils.timer(
                        logger, 'staged upload files packed', step='pack',
                        upload_size=self.upload_files.size):
636
                    self.upload_files.pack(calcs)
637
638
639
640

            with utils.timer(
                    logger, 'index updated', step='index',
                    upload_size=self.upload_files.size):
641
                search.publish(calcs)
642

643
644
645
646
647
648
            if isinstance(self.upload_files, StagingUploadFiles):
                with utils.timer(
                        logger, 'staged upload deleted', step='delete staged',
                        upload_size=self.upload_files.size):
                    self.upload_files.delete()
                    self.published = True
649
650
                    self.publish_time = datetime.utcnow()
                    self.last_update = datetime.utcnow()
651
652
                    self.save()
            else:
653
                self.last_update = datetime.utcnow()
654
                self.save()
655

656
657
    @process
    def re_process_upload(self):
658
        '''
Markus Scheidgen's avatar
Markus Scheidgen committed
659
660
661
        A *process* that performs the re-processing of a earlier processed
        upload.

662
663
664
665
666
667
        Runs the distributed process of fully reparsing/renormalizing an existing and
        already published upload. Will renew the archive part of the upload and update
        mongo and elastic search entries.

        TODO this implementation does not do any re-matching. This will be more complex
        due to handling of new or missing matches.
668
        '''
669
670
        assert self.published

671
672
673
        logger = self.get_logger()
        logger.info('started to re-process')

674
675
676
677
678
679
        # mock the steps of actual processing
        self._continue_with('uploading')

        # extract the published raw files into a staging upload files instance
        self._continue_with('extracting')
        public_upload_files = cast(PublicUploadFiles, self.upload_files)
680
        staging_upload_files = public_upload_files.to_staging_upload_files(create=True)
681
682

        self._continue_with('parse_all')
683
        try:
684
            # check if a calc is already/still processing
685
686
687
688
689
690
691
692
            processing = Calc.objects(
                upload_id=self.upload_id,
                **Calc.process_running_mongoengine_query()).count()

            if processing > 0:
                logger.warn(
                    'processes are still/already running on calc, they will be resetted',
                    count=processing)
693
694

            # reset all calcs
695
            Calc._get_collection().update_many(
696
697
698
                dict(upload_id=self.upload_id),
                {'$set': Calc.reset_pymongo_update(worker_hostname=self.worker_hostname)})

699
            # process call calcs
700
701
702
            Calc.process_all(Calc.re_process_calc, dict(upload_id=self.upload_id), exclude=['metadata'])

            logger.info('completed to trigger re-process of all calcs')
703
704
        except Exception as e:
            # try to remove the staging copy in failure case
705
706
            logger.error('failed to trigger re-process of all calcs', exc_info=e)

707
            if staging_upload_files is not None and staging_upload_files.exists():
708
709
710
                staging_upload_files.delete()

            raise e
711
712
713
714

        # the packing and removing of the staging upload files, will be trigged by
        # the 'cleanup' task after processing all calcs

715
716
    @process
    def re_pack(self):
717
        ''' A *process* that repacks the raw and archive data based on the current embargo data. '''
718
719
720
721
722
723
724
725
        assert self.published

        # mock the steps of actual processing
        self._continue_with('uploading')
        self._continue_with('extracting')
        self._continue_with('parse_all')
        self._continue_with('cleanup')

726
        self.upload_files.re_pack(self.entries_metadata())
727
728
729
        self.joined = True
        self._complete()

Markus Scheidgen's avatar
Markus Scheidgen committed
730
    @process
731
    def process_upload(self):
732
        ''' A *process* that performs the initial upload processing. '''
Markus Scheidgen's avatar
Markus Scheidgen committed
733
734
735
736
737
        self.extracting()
        self.parse_all()

    @task
    def uploading(self):
738
        ''' A no-op *task* as a stand-in for receiving upload data. '''
Markus Scheidgen's avatar
Markus Scheidgen committed
739
740
        pass

741
    @property
742
743
    def upload_files(self) -> UploadFiles:
        upload_files_class = ArchiveBasedStagingUploadFiles if not self.published else PublicUploadFiles
744
        kwargs = dict(upload_path=self.upload_path) if not self.published else {}
745
746
747

        if not self._upload_files or not isinstance(self._upload_files, upload_files_class):
            self._upload_files = upload_files_class(
748
                self.upload_id, is_authorized=lambda: True, **kwargs)
749

750
        return self._upload_files
751

752
753
754
755
756
    @property
    def staging_upload_files(self) -> ArchiveBasedStagingUploadFiles:
        assert not self.published
        return cast(ArchiveBasedStagingUploadFiles, self.upload_files)

Markus Scheidgen's avatar
Markus Scheidgen committed
757
758
    @task
    def extracting(self):
759
        '''
760
761
        The *task* performed before the actual parsing/normalizing: extracting
        the uploaded files.
762
        '''
763
764
765
766
767
        # extract the uploaded file
        self._upload_files = ArchiveBasedStagingUploadFiles(
            upload_id=self.upload_id, is_authorized=lambda: True, create=True,
            upload_path=self.upload_path)

Markus Scheidgen's avatar
Markus Scheidgen committed
768
769
        logger = self.get_logger()
        try:
Markus Scheidgen's avatar
Markus Scheidgen committed
770
771
            with utils.timer(
                    logger, 'upload extracted', step='extracting',
772
773
                    upload_size=self.upload_files.size):
                self.upload_files.extract()
774
775
776
777

            if self.temporary:
                os.remove(self.upload_path)
                self.upload_path = None
778

779
        except KeyError:
780
            self.fail('processing requested for non existing upload', log_level=logging.ERROR)
781
782
783
            return
        except ExtractError:
            self.fail('bad .zip/.tar file', log_level=logging.INFO)
Markus Scheidgen's avatar
Markus Scheidgen committed
784
785
            return

786
    def _preprocess_files(self, path):
787
        '''
788
789
        Some files need preprocessing. Currently we need to add a stripped POTCAR version
        and always restrict/embargo the original.
790
        '''
791
        if os.path.basename(path).startswith('POTCAR'):
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
            # create checksum
            hash = hashlib.sha224()
            with open(self.staging_upload_files.raw_file_object(path).os_path, 'rb') as orig_f:
                for line in orig_f.readlines():
                    hash.update(line)

            checksum = hash.hexdigest()

            # created stripped POTCAR
            stripped_path = path + '.stripped'
            with open(self.staging_upload_files.raw_file_object(stripped_path).os_path, 'wt') as stripped_f:
                stripped_f.write('Stripped POTCAR file. Checksum of original file (sha224): %s\n' % checksum)
            os.system(
                '''
                    awk < %s >> %s '
                    BEGIN { dump=1 }
                    /End of Dataset/ { dump=1 }
                    dump==1 { print }
                    /END of PSCTR/ { dump=0 }'
                ''' % (
                    self.staging_upload_files.raw_file_object(path).os_path,
                    self.staging_upload_files.raw_file_object(stripped_path).os_path))

815
    def match_mainfiles(self) -> Generator[Tuple[str, object], None, None]:
816
        '''
817
818
819
820
821
        Generator function that matches all files in the upload to all parsers to
        determine the upload's mainfiles.

        Returns:
            Tuples of mainfile, filename, and parsers
822
        '''
823
        directories_with_match: Dict[str, str] = dict()
824
825
        upload_files = self.staging_upload_files
        for filename in upload_files.raw_file_manifest():
826
            self._preprocess_files(filename)
827
            try:
828
                parser = match_parser(upload_files.raw_file_object(filename).os_path)
829
                if parser is not None:
830
831
                    directory = os.path.dirname(filename)
                    if directory in directories_with_match:
832
833
834
                        # TODO this might give us the chance to store directory based relationship
                        # between calcs for the future?
                        pass
835
836
                    else:
                        directories_with_match[directory] = filename
837
838

                    yield filename, parser
839
840
841
842
            except Exception as e:
                self.get_logger().error(
                    'exception while matching pot. mainfile',
                    mainfile=filename, exc_info=e)
843

Markus Scheidgen's avatar
Markus Scheidgen committed
844
845
    @task
    def parse_all(self):
846
        '''
Markus Scheidgen's avatar
Markus Scheidgen committed
847
        The *task* used to identify mainfile/parser combinations among the upload's files, creates
848
        respective :class:`Calc` instances, and triggers their processing.
849
        '''
850
851
        logger = self.get_logger()

Markus Scheidgen's avatar
Markus Scheidgen committed
852
853
        with utils.timer(
                logger, 'upload extracted', step='matching',
854
855
                upload_size=self.upload_files.size):
            for filename, parser in self.match_mainfiles():
856
                calc = Calc.create(
857
                    calc_id=self.upload_files.calc_id(filename),
858
                    mainfile=filename, parser=parser.name,
859
                    worker_hostname=self.worker_hostname,
860
861
                    upload_id=self.upload_id)

862
                calc.process_calc()
Markus Scheidgen's avatar
Markus Scheidgen committed
863

864
    def on_process_complete(self, process_name):
865
        if process_name == 'process_upload' or process_name == 're_process_upload':
866
867
868
            self.check_join()

    def check_join(self):
869
        '''
Markus Scheidgen's avatar
Markus Scheidgen committed
870
871
872
873
874
875
        Performs an evaluation of the join condition and triggers the :func:`cleanup`
        task if necessary. The join condition allows to run the ``cleanup`` after
        all calculations have been processed. The upload processing stops after all
        calculation processings have been triggered (:func:`parse_all` or
        :func:`re_process_upload`). The cleanup task is then run within the last
        calculation process (the one that triggered the join by calling this method).
876
        '''
877
878
879
880
        total_calcs = self.total_calcs
        processed_calcs = self.processed_calcs

        self.get_logger().debug('check join', processed_calcs=processed_calcs, total_calcs=total_calcs)
881
882
        # check if process is not running anymore, i.e. not still spawining new processes to join
        # check the join condition, i.e. all calcs have been processed
883
        if not self.process_running and processed_calcs >= total_calcs:
884
            # this can easily be called multiple times, e.g. upload finished after all calcs finished
885
886
887
            modified_upload = self._get_collection().find_one_and_update(
                {'_id': self.upload_id, 'joined': {'$ne': True}},
                {'$set': {'joined': True}})
888
889
890
891
892
893
            if modified_upload is not None:
                self.get_logger().debug('join')
                self.cleanup()
            else:
                # the join was already done due to a prior call
                pass
894

895
896
897
    def reset(self):
        self.joined = False
        super().reset()
Markus Scheidgen's avatar
Markus Scheidgen committed
898

Markus Scheidgen's avatar