data.py 45.8 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, Tuple, Iterator, 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
from structlog.processors import StackInfoRenderer, format_exc_info, TimeStamper
Markus Scheidgen's avatar
Markus Scheidgen committed
37

38
from nomad import utils, config, infrastructure, search, datamodel
39
from nomad.files import PathObject, UploadFiles, ExtractError, ArchiveBasedStagingUploadFiles, PublicUploadFiles, StagingUploadFiles
40
from nomad.processing.base import Proc, process, task, PENDING, SUCCESS, FAILURE
41
from nomad.parsing import parser_dict, match_parser, Backend
Markus Scheidgen's avatar
Markus Scheidgen committed
42
43
44
from nomad.normalizing import normalizers


45
46
47
48
49
50
51
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']})
52
        log_data.update(logger=logger.name)
53
54
55
56
57
58
59
60
61
62
63

        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,
64
    TimeStamper(fmt="%Y-%m-%d %H:%M.%S", utc=False)]
65
66


Markus Scheidgen's avatar
Markus Scheidgen committed
67
class Calc(Proc):
68
    '''
Markus Scheidgen's avatar
Markus Scheidgen committed
69
70
71
72
73
    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.

74
75
    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
76
77

    Attributes:
78
        calc_id: the calc_id of this calc
Markus Scheidgen's avatar
Markus Scheidgen committed
79
80
81
        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
82

83
84
        metadata: the metadata record wit calc and user metadata, see :class:`datamodel.EntryMetadata`
    '''
85
    calc_id = StringField(primary_key=True)
Markus Scheidgen's avatar
Markus Scheidgen committed
86
87
88
89
    upload_id = StringField()
    mainfile = StringField()
    parser = StringField()

90
91
    metadata = DictField()

Markus Scheidgen's avatar
Markus Scheidgen committed
92
    meta: Any = {
Markus Scheidgen's avatar
Markus Scheidgen committed
93
        'indexes': [
94
95
96
97
            'upload_id',
            ('upload_id', 'mainfile'),
            ('upload_id', 'parser'),
            ('upload_id', 'tasks_status'),
98
            ('upload_id', 'process_status'),
99
            ('upload_id', 'metadata.nomad_version'),
100
            'parser',
101
102
            'metadata.processed',
            'metadata.last_processing',
103
            'metadata.published',
104
            'metadata.datasets',
Markus Scheidgen's avatar
Markus Scheidgen committed
105
            'metadata.pid'
Markus Scheidgen's avatar
Markus Scheidgen committed
106
107
108
109
110
        ]
    }

    def __init__(self, *args, **kwargs):
        super().__init__(*args, **kwargs)
111
        self._parser_backend: Backend = None
112
113
        self._upload: Upload = None
        self._upload_files: ArchiveBasedStagingUploadFiles = None
114
        self._calc_proc_logs: List[Any] = None
Markus Scheidgen's avatar
Markus Scheidgen committed
115

116
        self._entry_metadata = None
117

Markus Scheidgen's avatar
Markus Scheidgen committed
118
119
    @classmethod
    def get(cls, id):
120
        return cls.get_by_id(id, 'calc_id')
Markus Scheidgen's avatar
Markus Scheidgen committed
121

Markus Scheidgen's avatar
Markus Scheidgen committed
122
    @property
123
124
    def mainfile_file(self) -> PathObject:
        return self.upload_files.raw_file_object(self.mainfile)
Markus Scheidgen's avatar
Markus Scheidgen committed
125

126
127
128
129
    @property
    def upload(self) -> 'Upload':
        if not self._upload:
            self._upload = Upload.get(self.upload_id)
130
            self._upload.worker_hostname = self.worker_hostname
131
132
        return self._upload

133
134
135
136
137
138
139
140
141
142
143
144
    def apply_entry_metadata(self, entry_metadata: datamodel.EntryMetadata):
        self.metadata = entry_metadata.m_to_dict(
            include_defaults=True,
            categories=[datamodel.MongoMetadata])  # TODO use embedded doc?

    def create_metadata(self) -> datamodel.EntryMetadata:
        '''
        Returns a :class:`nomad.datamodel.EntryMetadata` with values from this
        processing object, not necessarely the user metadata nor the metadata from
        the archive.
        '''
        entry_metadata = datamodel.EntryMetadata()
145
146
        if self.parser is not None:
            entry_metadata.domain = parser_dict[self.parser].domain
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
        entry_metadata.upload_id = self.upload_id
        entry_metadata.calc_id = self.calc_id
        entry_metadata.mainfile = self.mainfile
        entry_metadata.nomad_version = config.version
        entry_metadata.nomad_commit = config.commit
        entry_metadata.uploader = self.upload.user_id
        entry_metadata.upload_time = self.upload.upload_time
        entry_metadata.upload_name = self.upload.name

        return entry_metadata

    def entry_metadata(self, upload_files: UploadFiles) -> datamodel.EntryMetadata:
        '''
        Returns a complete set of :class:`nomad.datamodel.EntryMetadata` including
        the user metadata and metadata from the archive.

        Arguments:
            upload_files:
                The :class:`nomad.files.UploadFiles` instance to read the archive from.
            cache:
                A boolean that indicates if the archive file should be left unclosed,
                e.g. if this method is called for many entries of the same upload.
        '''
        archive = upload_files.read_archive(self.calc_id)
171
172
173
174
175
176
177
178
179
180
181
182
        try:
            entry_metadata = datamodel.EntryMetadata.m_from_dict(
                archive[self.calc_id][datamodel.EntryArchive.section_metadata.name].to_dict())

        except KeyError:
            # Due hard processing failures, it might be possible that an entry might not
            # have an archive
            if self._entry_metadata is not None:
                entry_metadata = self._entry_metadata

            else:
                entry_metadata = self.create_metadata()
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198

        entry_metadata.m_update_from_dict(self.metadata)

        return entry_metadata

    def user_metadata(self) -> datamodel.EntryMetadata:
        '''
        Returns a :class:`nomad.datamodel.EntryMetadata` with values from this
        processing object and the user metadata, not necessarely the metadata from
        the archive.
        '''
        entry_metadata = self.create_metadata()
        entry_metadata.m_update_from_dict(self.metadata)

        return entry_metadata

199
200
201
    @property
    def upload_files(self) -> ArchiveBasedStagingUploadFiles:
        if not self._upload_files:
202
203
            self._upload_files = ArchiveBasedStagingUploadFiles(
                self.upload_id, is_authorized=lambda: True, upload_path=self.upload.upload_path)
204
205
        return self._upload_files

Markus Scheidgen's avatar
Markus Scheidgen committed
206
    def get_logger(self, **kwargs):
207
        '''
208
209
        Returns a wrapped logger that additionally saves all entries to the calculation
        processing log in the archive.
210
        '''
211
212
213
        logger = super().get_logger()
        logger = logger.bind(
            upload_id=self.upload_id, mainfile=self.mainfile, calc_id=self.calc_id, **kwargs)
214

215
216
        if self._calc_proc_logs is None:
            self._calc_proc_logs = []
217

218
219
        def save_to_calc_log(logger, method_name, event_dict):
            try:
220
221
                # sanitize the event_dict, because all kinds of values might have been added
                dump_dict = {key: str(value) for key, value in event_dict.items()}
222
223
                dump_dict.update(level=method_name.upper())
                self._calc_proc_logs.append(dump_dict)
224

225
226
227
            except Exception:
                # Exceptions here will cause indefinite loop
                pass
228

229
            return event_dict
230

231
        return wrap_logger(logger, processors=_log_processors + [save_to_calc_log])
232

233
234
    @process
    def re_process_calc(self):
235
        '''
236
237
238
        Processes a calculation again. This means there is already metadata and
        instead of creating it initially, we are just updating the existing
        records.
239
        '''
240
        parser = match_parser(self.upload_files.raw_file_object(self.mainfile).os_path, strict=False)
241
        logger = self.get_logger()
242
243
244

        if parser is None and not config.reprocess_unmatched:
            self.errors = ['no parser matches during re-process, will not re-process this calc']
245

246
247
248
249
250
251
252
            try:
                upload_files = PublicUploadFiles(self.upload_id, is_authorized=lambda: True)
                with upload_files.read_archive(self.calc_id) as archive:
                    self.upload_files.write_archive(self.calc_id, archive[self.calc_id].to_dict())

            except Exception as e:
                logger.error('could not copy archive for non matching, non reprocessed entry', exc_info=e)
253
                raise e
254

255
256
257
258
259
            # mock the steps of actual processing
            self._continue_with('parsing')
            self._continue_with('normalizing')
            self._continue_with('archiving')
            self._complete()
260
261
            return

262
        if parser is None:
263
264
265
266
            self.get_logger().warn('no parser matches during re-process, use the old parser')
            self.warnings = ['no matching parser found during re-processing']

        elif self.parser != parser.name:
267
268
269
270
271
272
273
274
275
            if parser_dict[self.parser].name == parser.name:
                # parser was just renamed
                self.parser = parser.name

            else:
                self.parser = parser.name
                logger.info(
                    'different parser matches during re-process, use new parser',
                    parser=parser.name)
276

277
        try:
278
279
280
281
282
283
            self._entry_metadata = self.user_metadata()
            self._entry_metadata.calc_hash = self.upload_files.calc_hash(self.mainfile)
            self._entry_metadata.nomad_version = config.version
            self._entry_metadata.nomad_commit = config.commit
            self._entry_metadata.last_processing = datetime.utcnow()
            self._entry_metadata.files = self.upload_files.calc_files(self.mainfile)
284
285
286
287
288
289

            self.parsing()
            self.normalizing()
            self.archiving()
        finally:
            # close loghandler that was not closed due to failures
290
291
292
293
294
295
            try:
                if self._parser_backend and self._parser_backend.resource:
                    self._parser_backend.resource.unload()
            except Exception as e:
                logger.error('could unload processing results', exc_info=e)

Markus Scheidgen's avatar
Markus Scheidgen committed
296
    @process
297
    def process_calc(self):
298
        '''
299
300
        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.
301
        '''
302
        logger = self.get_logger()
303
        if self.upload is None:
304
            logger.error('calculation upload does not exist')
Markus Scheidgen's avatar
Markus Scheidgen committed
305
306

        try:
307
308
            # save preliminary minimum calc metadata in case processing fails
            # successful processing will replace it with the actual metadata
309
310
311
312
313
314
            self._entry_metadata = self.create_metadata()
            self._entry_metadata.calc_hash = self.upload_files.calc_hash(self.mainfile)
            self._entry_metadata.last_processing = datetime.utcnow()
            self._entry_metadata.files = self.upload_files.calc_files(self.mainfile)

            if len(self._entry_metadata.files) >= config.auxfile_cutoff:
315
316
317
318
                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
319
320
321
322
            self.parsing()
            self.normalizing()
            self.archiving()
        finally:
Markus Scheidgen's avatar
Markus Scheidgen committed
323
            # close loghandler that was not closed due to failures
324
325
326
327
328
329
            try:
                if self._parser_backend and self._parser_backend.resource:
                    self._parser_backend.resource.unload()
            except Exception as e:
                logger.error('could unload processing results', exc_info=e)

330
    def on_fail(self):
331
332
333
        # in case of failure, index a minimum set of metadata and mark
        # processing failure
        try:
334
            self._entry_metadata.processed = False
335

336
            self.apply_entry_metadata(self._entry_metadata)
337
338
339
340
341
            if self._parser_backend and self._parser_backend.resource:
                backend = self._parser_backend
            else:
                backend = None
            self._entry_metadata.apply_domain_metadata(backend)
342

343
            self._entry_metadata.a_elastic.index()
344
        except Exception as e:
345
346
            self.get_logger().error(
                'could not index after processing failure', exc_info=e)
347

348
        try:
349
            self.write_archive(None)
350
        except Exception as e:
351
352
            self.get_logger().error(
                'could not write archive after processing failure', exc_info=e)
353

354
355
356
357
    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
358
        if process_name == 'process_calc' or process_name == 're_process_calc' or process_name is None:
359
360
            self.upload.reload()
            self.upload.check_join()
Markus Scheidgen's avatar
Markus Scheidgen committed
361
362
363

    @task
    def parsing(self):
364
        ''' The *task* that encapsulates all parsing related actions. '''
365
        context = dict(parser=self.parser, step=self.parser)
366
        logger = self.get_logger(**context)
367
        parser = parser_dict[self.parser]
368
        self._entry_metadata.parser_name = self.parser
Markus Scheidgen's avatar
Markus Scheidgen committed
369

370
        with utils.timer(logger, 'parser executed', input_size=self.mainfile_file.size):
371
372
373
374
            try:
                self._parser_backend = parser.run(
                    self.upload_files.raw_file_object(self.mainfile).os_path, logger=logger)
            except Exception as e:
375
                self.fail('parser failed with exception', exc_info=e, error=str(e), **context)
376
                return
377
            except SystemExit:
378
                self.fail('parser raised system exit', error='system exit', **context)
379
                return
Markus Scheidgen's avatar
Markus Scheidgen committed
380

Markus Scheidgen's avatar
Markus Scheidgen committed
381
382
        if self._parser_backend.status[0] != 'ParseSuccess':
            error = self._parser_backend.status[1]
383
            self.fail('parser failed', error=error, **context)
384
385
386
387
388
389
390
391

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

        if self._parser_backend.status[0] == 'ParseSuccess':
            warnings = getattr(self._parser_backend, '_warnings', [])
392

393
            if len(warnings) > 0:
394
395
396
397
                self.get_logger().warn(
                    'processor completed successful with warnings',
                    processor=processor_name, warnings=[str(warning) for warning in warnings])

398
            else:
399
400
401
402
                self.get_logger().info(
                    'processor completed successful',
                    processor=processor_name)

403
        else:
404
            errors = self._parser_backend.status[1]
405
406
407
            self.get_logger().error(
                'processor completed with failure',
                processor=processor_name, errors=str(errors))
Markus Scheidgen's avatar
Markus Scheidgen committed
408
409
410

    @task
    def normalizing(self):
411
        ''' The *task* that encapsulates all normalizing related actions. '''
412
413
414
415
416

        # allow normalizer to access and add data to the entry metadata
        self._parser_backend.entry_archive.m_add_sub_section(
            datamodel.EntryArchive.section_metadata, self._entry_metadata)

Markus Scheidgen's avatar
Markus Scheidgen committed
417
        for normalizer in normalizers:
418
            if normalizer.domain != parser_dict[self.parser].domain:
419
420
                continue

Markus Scheidgen's avatar
Markus Scheidgen committed
421
            normalizer_name = normalizer.__name__
422
            context = dict(normalizer=normalizer_name, step=normalizer_name)
423
            logger = self.get_logger(**context)
Markus Scheidgen's avatar
Markus Scheidgen committed
424
425

            with utils.timer(
426
                    logger, 'normalizer executed', input_size=self.mainfile_file.size):
427
                with self.use_parser_backend(normalizer_name) as backend:
428
429
430
                    try:
                        normalizer(backend).normalize(logger=logger)
                    except Exception as e:
431
                        self._parser_backend.finishedParsingSession('ParseFailure', [str(e)])
432
433
434
                        logger.error(
                            'normalizer failed with exception', exc_info=e,
                            error=str(e), **context)
435
436
437
438
439
440
441
442
443
                        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
444
445
446

    @task
    def archiving(self):
447
        ''' The *task* that encapsulates all archival related actions. '''
448
449
        logger = self.get_logger()

450
451
        self._entry_metadata.apply_domain_metadata(self._parser_backend)
        self._entry_metadata.processed = True
452

453
454
        # persist the calc metadata
        with utils.timer(logger, 'saved calc metadata', step='metadata'):
455
            self.apply_entry_metadata(self._entry_metadata)
456
457

        # index in search
458
        with utils.timer(logger, 'indexed', step='index'):
459
            self._entry_metadata.a_elastic.index()
460

461
        # persist the archive
Markus Scheidgen's avatar
Markus Scheidgen committed
462
        with utils.timer(
463
                logger, 'archived', step='archive',
Markus Scheidgen's avatar
Markus Scheidgen committed
464
                input_size=self.mainfile_file.size) as log_data:
465

466
            archive_size = self.write_archive(self._parser_backend)
467
            log_data.update(archive_size=archive_size)
Markus Scheidgen's avatar
Markus Scheidgen committed
468

469
    def write_archive(self, backend: Backend):
470
471
472
473
474
475
476
        def filter_processing_logs(logs):
            if len(logs) > 100:
                return [
                    log for log in logs
                    if log.get('level') != 'DEBUG']
            return logs

477
478
479
480
481
482
483
484
        if self._calc_proc_logs is None:
            self._calc_proc_logs = []

        if backend is not None:
            entry_archive = backend.entry_archive.m_copy()
        else:
            entry_archive = datamodel.EntryArchive()

485
486
487
        if entry_archive.section_metadata is None:
            entry_archive.m_add_sub_section(datamodel.EntryArchive.section_metadata, self._entry_metadata)

488
        entry_archive.processing_logs = filter_processing_logs(self._calc_proc_logs)
489
490
491
492
493
494
495
496
497
498

        try:
            return self.upload_files.write_archive(self.calc_id, entry_archive.m_to_dict())
        except Exception as e:
            if backend is None:
                raise e

            # most likely failed due to domain data, try to write metadata and processing logs
            entry_archive = datamodel.EntryArchive()
            entry_archive.m_add_sub_section(datamodel.EntryArchive.section_metadata, self._entry_metadata)
499
            entry_archive.processing_logs = filter_processing_logs(self._calc_proc_logs)
500
501
502
            self.upload_files.write_archive(self.calc_id, entry_archive.m_to_dict())
            raise e

503
    def __str__(self):
504
        return 'calc %s calc_id=%s upload_id%s' % (super().__str__(), self.calc_id, self.upload_id)
505

506

507
class Upload(Proc):
508
    '''
Markus Scheidgen's avatar
Markus Scheidgen committed
509
510
511
512
513
    Represents uploads in the databases. Provides persistence access to the files storage,
    and processing state.

    Attributes:
        name: optional user provided upload name
514
515
        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
516
517
        upload_id: the upload id generated by the database
        upload_time: the timestamp when the system realised the upload
518
        user_id: the id of the user that created this upload
Markus Scheidgen's avatar
Markus Scheidgen committed
519
520
        published: Boolean that indicates the publish status
        publish_time: Date when the upload was initially published
521
        last_update: Date of the last publishing/re-processing
Markus Scheidgen's avatar
Markus Scheidgen committed
522
        joined: Boolean indicates if the running processing has joined (:func:`check_join`)
523
    '''
Markus Scheidgen's avatar
Markus Scheidgen committed
524
525
526
    id_field = 'upload_id'

    upload_id = StringField(primary_key=True)
527
528
    upload_path = StringField(default=None)
    temporary = BooleanField(default=False)
529
    embargo_length = IntField(default=36)
Markus Scheidgen's avatar
Markus Scheidgen committed
530
531
532

    name = StringField(default=None)
    upload_time = DateTimeField()
533
    user_id = StringField(required=True)
534
535
    published = BooleanField(default=False)
    publish_time = DateTimeField()
536
    last_update = DateTimeField()
Markus Scheidgen's avatar
Markus Scheidgen committed
537

538
539
    joined = BooleanField(default=False)

Markus Scheidgen's avatar
Markus Scheidgen committed
540
541
    meta: Any = {
        'indexes': [
542
            'user_id', 'tasks_status', 'process_status', 'published', 'upload_time'
Markus Scheidgen's avatar
Markus Scheidgen committed
543
544
545
546
547
        ]
    }

    def __init__(self, **kwargs):
        super().__init__(**kwargs)
548
        self._upload_files: ArchiveBasedStagingUploadFiles = None
Markus Scheidgen's avatar
Markus Scheidgen committed
549

550
551
    @property
    def metadata(self) -> dict:
552
        '''
Markus Scheidgen's avatar
Markus Scheidgen committed
553
554
555
556
557
        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.
558
        '''
559
560
561
562
563
        try:
            upload_files = PublicUploadFiles(self.upload_id, is_authorized=lambda: True)
        except KeyError:
            return None
        return upload_files.user_metadata
564
565
566

    @metadata.setter
    def metadata(self, data: dict) -> None:
567
568
        upload_files = PublicUploadFiles(self.upload_id, is_authorized=lambda: True, create=True)
        upload_files.user_metadata = data
569

Markus Scheidgen's avatar
Markus Scheidgen committed
570
    @classmethod
571
    def get(cls, id: str, include_published: bool = True) -> 'Upload':
572
        return cls.get_by_id(id, 'upload_id')
Markus Scheidgen's avatar
Markus Scheidgen committed
573
574

    @classmethod
575
    def user_uploads(cls, user: datamodel.User, **kwargs) -> List['Upload']:
576
        ''' Returns all uploads for the given user. Kwargs are passed to mongo query. '''
577
        return cls.objects(user_id=str(user.user_id), **kwargs)
Markus Scheidgen's avatar
Markus Scheidgen committed
578

579
580
    @property
    def uploader(self):
581
        return datamodel.User.get(self.user_id)
582

Markus Scheidgen's avatar
Markus Scheidgen committed
583
584
    def get_logger(self, **kwargs):
        logger = super().get_logger()
585
586
        user = self.uploader
        user_name = '%s %s' % (user.first_name, user.last_name)
587
        # We are not using 'user_id' because logstash (?) will filter these entries ?!
588
589
        logger = logger.bind(
            upload_id=self.upload_id, upload_name=self.name, user_name=user_name,
590
            user=user.user_id, **kwargs)
Markus Scheidgen's avatar
Markus Scheidgen committed
591
592
593
594
        return logger

    @classmethod
    def create(cls, **kwargs) -> 'Upload':
595
        '''
Markus Scheidgen's avatar
Markus Scheidgen committed
596
597
598
        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.
599
600

        Arguments:
601
            user: The user that created the upload.
602
        '''
603
604
        # use kwargs to keep compatibility with super method
        user: datamodel.User = kwargs['user']
605
        del(kwargs['user'])
606

607
608
        if 'upload_id' not in kwargs:
            kwargs.update(upload_id=utils.create_uuid())
609
        kwargs.update(user_id=user.user_id)
Markus Scheidgen's avatar
Markus Scheidgen committed
610
        self = super().create(**kwargs)
611

Markus Scheidgen's avatar
Markus Scheidgen committed
612
        self._continue_with('uploading')
613

Markus Scheidgen's avatar
Markus Scheidgen committed
614
615
        return self

616
    def delete(self):
617
        ''' Deletes this upload process state entry and its calcs. '''
618
619
620
        Calc.objects(upload_id=self.upload_id).delete()
        super().delete()

621
    def delete_upload_local(self):
622
        '''
623
        Deletes the upload, including its processing state and
624
        staging files. Local version without celery processing.
625
        '''
626
627
628
        logger = self.get_logger()

        with utils.lnr(logger, 'staged upload delete failed'):
629
            with utils.timer(
630
                    logger, 'upload deleted from index', step='index',
631
                    upload_size=self.upload_files.size):
632
                search.delete_upload(self.upload_id)
633

634
            with utils.timer(
635
                    logger, 'staged upload deleted', step='files',
636
637
                    upload_size=self.upload_files.size):
                self.upload_files.delete()
638
639

            self.delete()
640

641
    @process
642
    def delete_upload(self):
643
        '''
644
645
        Deletes of the upload, including its processing state and
        staging files. This starts the celery process of deleting the upload.
646
        '''
647
        self.delete_upload_local()
648

649
        return True  # do not save the process status on the delete upload
650

651
    @process
652
    def publish_upload(self):
653
        '''
654
655
        Moves the upload out of staging to the public area. It will
        pack the staging upload files in to public upload files.
656
        '''
657
658
        assert self.processed_calcs > 0

659
        logger = self.get_logger()
660
        logger.info('started to publish')
661

662
        with utils.lnr(logger, 'publish failed'):
663
            with self.entries_metadata(self.metadata) as calcs:
664

665
                with utils.timer(
666
                        logger, 'upload metadata updated', step='metadata',
667
                        upload_size=self.upload_files.size):
668

669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
                    def create_update(calc):
                        calc.published = True
                        calc.with_embargo = calc.with_embargo if calc.with_embargo is not None else False
                        return UpdateOne(
                            {'_id': calc.calc_id},
                            {'$set': {'metadata': calc.m_to_dict(
                                include_defaults=True, categories=[datamodel.MongoMetadata])}})

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

                if isinstance(self.upload_files, StagingUploadFiles):
                    with utils.timer(
                            logger, 'staged upload files packed', step='pack',
                            upload_size=self.upload_files.size):
                        self.upload_files.pack(calcs)
684

685
                with utils.timer(
686
                        logger, 'index updated', step='index',
687
                        upload_size=self.upload_files.size):
688
689
690
691
692
693
694
695
696
697
698
699
                    search.publish(calcs)

                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
                        self.publish_time = datetime.utcnow()
                        self.last_update = datetime.utcnow()
                        self.save()
                else:
700
                    self.last_update = datetime.utcnow()
701
                    self.save()
702

703
704
    @process
    def re_process_upload(self):
705
        '''
Markus Scheidgen's avatar
Markus Scheidgen committed
706
707
708
        A *process* that performs the re-processing of a earlier processed
        upload.

709
710
711
712
713
714
        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.
715
        '''
716
717
718
        logger = self.get_logger()
        logger.info('started to re-process')

719
720
721
722
723
        # 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')
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738

        if self.published:
            try:
                staging_upload_files = StagingUploadFiles(self.upload_id)
                # public files exist and there is a staging directory, it is probably old
                # and we delete it first
                staging_upload_files.delete()
                logger.warn('deleted old staging files')

            except KeyError as e:
                logger.info('reprocessing published files')
        else:
            logger.info('reprocessing staging files')

        staging_upload_files = self.upload_files.to_staging_upload_files(create=True)
739
740

        self._continue_with('parse_all')
741
        try:
742
            # check if a calc is already/still processing
743
744
745
746
747
748
749
750
            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)
751
752

            # reset all calcs
753
            Calc._get_collection().update_many(
754
755
756
                dict(upload_id=self.upload_id),
                {'$set': Calc.reset_pymongo_update(worker_hostname=self.worker_hostname)})

757
            # process call calcs
758
759
760
            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')
761
762
        except Exception as e:
            # try to remove the staging copy in failure case
763
764
            logger.error('failed to trigger re-process of all calcs', exc_info=e)

765
766
767
            if self.published:
                if staging_upload_files is not None and staging_upload_files.exists():
                    staging_upload_files.delete()
768
769

            raise e
770
771
772
773

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

774
775
    @process
    def re_pack(self):
776
        ''' A *process* that repacks the raw and archive data based on the current embargo data. '''
777
778
779
780
781
782
783
784
        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')

785
        self.upload_files.re_pack(self.user_metadata())
786
787
788
        self.joined = True
        self._complete()

Markus Scheidgen's avatar
Markus Scheidgen committed
789
    @process
790
    def process_upload(self):
791
        ''' A *process* that performs the initial upload processing. '''
Markus Scheidgen's avatar
Markus Scheidgen committed
792
793
794
795
796
        self.extracting()
        self.parse_all()

    @task
    def uploading(self):
797
        ''' A no-op *task* as a stand-in for receiving upload data. '''
Markus Scheidgen's avatar
Markus Scheidgen committed
798
799
        pass

800
    @property
801
802
    def upload_files(self) -> UploadFiles:
        upload_files_class = ArchiveBasedStagingUploadFiles if not self.published else PublicUploadFiles
803
        kwargs = dict(upload_path=self.upload_path) if not self.published else {}
804
805
806

        if not self._upload_files or not isinstance(self._upload_files, upload_files_class):
            self._upload_files = upload_files_class(
807
                self.upload_id, is_authorized=lambda: True, **kwargs)
808

809
        return self._upload_files
810

811
812
813
814
815
    @property
    def staging_upload_files(self) -> ArchiveBasedStagingUploadFiles:
        assert not self.published
        return cast(ArchiveBasedStagingUploadFiles, self.upload_files)

Markus Scheidgen's avatar
Markus Scheidgen committed
816
817
    @task
    def extracting(self):
818
        '''
819
820
        The *task* performed before the actual parsing/normalizing: extracting
        the uploaded files.
821
        '''
822
823
824
825
826
        # 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
827
828
        logger = self.get_logger()
        try:
Markus Scheidgen's avatar
Markus Scheidgen committed
829
830
            with utils.timer(
                    logger, 'upload extracted', step='extracting',
831
832
                    upload_size=self.upload_files.size):
                self.upload_files.extract()
833
834
835
836

            if self.temporary:
                os.remove(self.upload_path)
                self.upload_path = None
837

838
        except KeyError:
839
            self.fail('processing requested for non existing upload', log_level=logging.ERROR)
840
841
842
            return
        except ExtractError:
            self.fail('bad .zip/.tar file', log_level=logging.INFO)
Markus Scheidgen's avatar
Markus Scheidgen committed
843
844
            return

845
    def _preprocess_files(self, path):
846
        '''
847
848
        Some files need preprocessing. Currently we need to add a stripped POTCAR version
        and always restrict/embargo the original.
849
        '''
850
        if os.path.basename(path).startswith('POTCAR'):
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
            # 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))

874
    def match_mainfiles(self) -> Iterator[Tuple[str, object]]:
875
        '''
876
877
878
879
880
        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
881
        '''
882
        directories_with_match: Dict[str, str] = dict()
883
884
        upload_files = self.staging_upload_files
        for filename in upload_files.raw_file_manifest():
885
            self._preprocess_files(filename)
886
            try:
887
                parser = match_parser(upload_files.raw_file_object(filename).os_path)
888
                if parser is not None:
889
890
                    directory = os.path.dirname(filename)
                    if directory in directories_with_match:
891
892
893
                        # TODO this might give us the chance to store directory based relationship
                        # between calcs for the future?
                        pass
894
895
                    else:
                        directories_with_match[directory] = filename
896
897

                    yield filename, parser
898
899
900
901
            except Exception as e:
                self.get_logger().error(
                    'exception while matching pot. mainfile',
                    mainfile=filename, exc_info=e)
902

Markus Scheidgen's avatar
Markus Scheidgen committed
903
904
    @task
    def parse_all(self):
905
        '''
Markus Scheidgen's avatar
Markus Scheidgen committed
906
        The *task* used to identify mainfile/parser combinations among the upload's files, creates
907
        respective :class:`Calc` instances, and triggers their processing.
908
        '''
909
910
        logger = self.get_logger()

Markus Scheidgen's avatar
Markus Scheidgen committed
911
912
        with utils.timer(
                logger, 'upload extracted', step='matching',
913
914
                upload_size=self.upload_files.size):
            for filename, parser in self.match_mainfiles():
915
                calc = Calc.create(
916
                    calc_id=self.upload_files.calc_id(filename),
917
                    mainfile=filename, parser=parser.name,
918
                    worker_hostname=self.worker_hostname,
919
920
                    upload_id=self.upload_id)

921
                calc.process_calc()
Markus Scheidgen's avatar
Markus Scheidgen committed
922

923
    def on_process_complete(self, process_name):
924
        if process_name == 'process_upload' or process_name == 're_process_upload':
925
926
927
            self.check_join()

    def check_join(self):
928
        '''
Markus Scheidgen's avatar
Markus Scheidgen committed
929
930
931
932
933
934
        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).
935
        '''
936
937
938
939
        total_calcs = self.total_calcs
        processed_calcs = self.processed_calcs

        self.get_logger().debug('check join', processed_calcs=processed_calcs, total_calcs=total_calcs)
940
941
        # 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
942
        if not self.process_running and processed_calcs >= total_calcs:
943
            # this can easily be called multiple times, e.g. upload finished after all calcs finished
944
945
946
            modified_upload = self._get_collection().find_one_and_update(
                {'_id': self.upload_id, 'joined': {'$ne': True}},
                {'$set': {'joined': True}})
947
948
949