data.py 51.8 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
'''
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
Alvin Noe Ladines's avatar
Alvin Noe Ladines committed
37
38
39
import yaml
import json
from cachetools import cached, LRUCache
Markus Scheidgen's avatar
Markus Scheidgen committed
40

41
from nomad import utils, config, infrastructure, search, datamodel
Markus Scheidgen's avatar
Markus Scheidgen committed
42
43
44
from nomad.files import (
    PathObject, UploadFiles, ExtractError, ArchiveBasedStagingUploadFiles,
    PublicUploadFiles, StagingUploadFiles)
45
from nomad.processing.base import Proc, process, task, PENDING, SUCCESS, FAILURE
Markus Scheidgen's avatar
Markus Scheidgen committed
46
from nomad.parsing.parsers import parser_dict, match_parser
Markus Scheidgen's avatar
Markus Scheidgen committed
47
from nomad.normalizing import normalizers
Alvin Noe Ladines's avatar
Alvin Noe Ladines committed
48
from nomad.datamodel import EntryArchive, EditableUserMetadata
49
from nomad.archive import query_archive
Markus Scheidgen's avatar
Markus Scheidgen committed
50
from nomad.datamodel.encyclopedia import EncyclopediaMetadata
Markus Scheidgen's avatar
Markus Scheidgen committed
51
52


53
54
55
56
section_metadata = datamodel.EntryArchive.section_metadata.name
section_workflow = datamodel.EntryArchive.section_workflow.name


Alvin Noe Ladines's avatar
Alvin Noe Ladines committed
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
_editable_metadata = {
    quantity.name: quantity for quantity in EditableUserMetadata.m_def.definitions}


@cached(cache=LRUCache(maxsize=100))
def metadata_cached(path):
    for ext in config.aux_metadata_exts:
        full_path = '%s.%s' % (path, ext)
        if os.path.isfile(full_path):
            with open(full_path) as f:
                if full_path.endswith('json'):
                    return json.load(f)
                elif full_path.endswith('yaml'):
                    return yaml.load(f, Loader=getattr(yaml, 'FullLoader'))
                else:
                    return {}
    return {}


76
77
78
79
80
81
82
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']})
83
        log_data.update(logger=logger.name)
84
85
86
87
88
89
90
91
92
93
94

        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,
95
    TimeStamper(fmt="%Y-%m-%d %H:%M.%S", utc=False)]
96
97


Markus Scheidgen's avatar
Markus Scheidgen committed
98
class Calc(Proc):
99
    '''
Markus Scheidgen's avatar
Markus Scheidgen committed
100
101
102
103
104
    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.

105
106
    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
107
108

    Attributes:
109
        calc_id: the calc_id of this calc
Markus Scheidgen's avatar
Markus Scheidgen committed
110
111
112
        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
113

114
115
        metadata: the metadata record wit calc and user metadata, see :class:`datamodel.EntryMetadata`
    '''
116
    calc_id = StringField(primary_key=True)
Markus Scheidgen's avatar
Markus Scheidgen committed
117
118
119
120
    upload_id = StringField()
    mainfile = StringField()
    parser = StringField()

121
122
    metadata = DictField()

Markus Scheidgen's avatar
Markus Scheidgen committed
123
    meta: Any = {
Markus Scheidgen's avatar
Markus Scheidgen committed
124
        'indexes': [
125
            'upload_id',
126
            'parser',
127
128
129
            ('upload_id', 'mainfile'),
            ('upload_id', 'parser'),
            ('upload_id', 'tasks_status'),
130
            ('upload_id', 'process_status'),
131
            ('upload_id', 'metadata.nomad_version'),
132
133
            'metadata.processed',
            'metadata.last_processing',
134
            'metadata.published',
135
            'metadata.datasets',
Markus Scheidgen's avatar
Markus Scheidgen committed
136
            'metadata.pid'
Markus Scheidgen's avatar
Markus Scheidgen committed
137
138
139
140
141
        ]
    }

    def __init__(self, *args, **kwargs):
        super().__init__(*args, **kwargs)
Markus Scheidgen's avatar
Markus Scheidgen committed
142
        self._parser_results: EntryArchive = None
143
144
        self._upload: Upload = None
        self._upload_files: ArchiveBasedStagingUploadFiles = None
145
        self._calc_proc_logs: List[Any] = None
146

147
        self._entry_metadata = None
148

Markus Scheidgen's avatar
Markus Scheidgen committed
149
150
    @classmethod
    def get(cls, id):
151
        return cls.get_by_id(id, 'calc_id')
Markus Scheidgen's avatar
Markus Scheidgen committed
152

Markus Scheidgen's avatar
Markus Scheidgen committed
153
    @property
154
155
    def mainfile_file(self) -> PathObject:
        return self.upload_files.raw_file_object(self.mainfile)
Markus Scheidgen's avatar
Markus Scheidgen committed
156

157
158
159
160
    @property
    def upload(self) -> 'Upload':
        if not self._upload:
            self._upload = Upload.get(self.upload_id)
161
            self._upload.worker_hostname = self.worker_hostname
162
163
        return self._upload

164
165
166
167
168
169
170
171
172
173
174
175
    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()
176
177
        if self.parser is not None:
            entry_metadata.domain = parser_dict[self.parser].domain
178
179
180
        entry_metadata.upload_id = self.upload_id
        entry_metadata.calc_id = self.calc_id
        entry_metadata.mainfile = self.mainfile
181
182
        entry_metadata.nomad_version = config.meta.version
        entry_metadata.nomad_commit = config.meta.commit
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
        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)
202
        try:
203
204
205
206
207
208
209
210
211
            # instead of loading the whole archive, it should be enough to load the
            # parts that are referenced by section_metadata/EntryMetadata
            # TODO somehow it should determine which root setions too load from the metainfo
            # or configuration
            calc_archive = archive[self.calc_id]
            entry_archive_dict = {section_metadata: calc_archive[section_metadata].to_dict()}
            if section_workflow in calc_archive:
                entry_archive_dict[section_workflow] = calc_archive[section_workflow].to_dict()
            entry_metadata = datamodel.EntryArchive.m_from_dict(entry_archive_dict)[section_metadata]
212
213
214
215
216
217
218
219
220

        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()
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236

        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

237
238
239
    @property
    def upload_files(self) -> ArchiveBasedStagingUploadFiles:
        if not self._upload_files:
240
241
            self._upload_files = ArchiveBasedStagingUploadFiles(
                self.upload_id, is_authorized=lambda: True, upload_path=self.upload.upload_path)
242
243
        return self._upload_files

Markus Scheidgen's avatar
Markus Scheidgen committed
244
    def get_logger(self, **kwargs):
245
        '''
246
247
        Returns a wrapped logger that additionally saves all entries to the calculation
        processing log in the archive.
248
        '''
249
250
251
        logger = super().get_logger()
        logger = logger.bind(
            upload_id=self.upload_id, mainfile=self.mainfile, calc_id=self.calc_id, **kwargs)
252

253
254
        if self._calc_proc_logs is None:
            self._calc_proc_logs = []
255

256
257
        def save_to_calc_log(logger, method_name, event_dict):
            try:
258
259
                # 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()}
260
261
                dump_dict.update(level=method_name.upper())
                self._calc_proc_logs.append(dump_dict)
262

263
264
265
            except Exception:
                # Exceptions here will cause indefinite loop
                pass
266

267
            return event_dict
268

269
        return wrap_logger(logger, processors=_log_processors + [save_to_calc_log])
270

271
    @process
272
    def re_process_calc(self):
273
        '''
274
275
276
        Processes a calculation again. This means there is already metadata and
        instead of creating it initially, we are just updating the existing
        records.
277
        '''
278
        parser = match_parser(self.upload_files.raw_file_object(self.mainfile).os_path, strict=False)
279
        logger = self.get_logger()
280
281
282

        if parser is None and not config.reprocess_unmatched:
            self.errors = ['no parser matches during re-process, will not re-process this calc']
283

284
285
286
287
288
289
290
            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)
291
                raise e
292

293
294
295
296
297
            # mock the steps of actual processing
            self._continue_with('parsing')
            self._continue_with('normalizing')
            self._continue_with('archiving')
            self._complete()
298
299
            return

300
        if parser is None:
301
302
303
304
            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:
305
306
307
308
309
310
311
312
313
            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)
314

315
        try:
316
317
            self._entry_metadata = self.user_metadata()
            self._entry_metadata.calc_hash = self.upload_files.calc_hash(self.mainfile)
318
319
            self._entry_metadata.nomad_version = config.meta.version
            self._entry_metadata.nomad_commit = config.meta.commit
320
321
            self._entry_metadata.last_processing = datetime.utcnow()
            self._entry_metadata.files = self.upload_files.calc_files(self.mainfile)
322
323
324
325
326
327
328

            self.parsing()
            self.normalizing()
            self.archiving()
        finally:
            # close loghandler that was not closed due to failures
            try:
Markus Scheidgen's avatar
Markus Scheidgen committed
329
                if self._parser_results and self._parser_results.m_resource:
330
                    self._parser_results.section_metadata = None
Markus Scheidgen's avatar
Markus Scheidgen committed
331
                    self._parser_results.m_resource.unload()
332
            except Exception as e:
333
                logger.error('could not unload processing results', exc_info=e)
334

335
336
337
338
339
340
    def _setup_fallback_metadata(self):
        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)

341
    @process
342
    def process_calc(self):
343
        '''
344
345
        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.
346
        '''
347
        logger = self.get_logger()
348
        if self.upload is None:
349
            logger.error('calculation upload does not exist')
Markus Scheidgen's avatar
Markus Scheidgen committed
350
351

        try:
352
353
            # save preliminary minimum calc metadata in case processing fails
            # successful processing will replace it with the actual metadata
354
            self._setup_fallback_metadata()
355
356

            if len(self._entry_metadata.files) >= config.auxfile_cutoff:
357
358
359
360
                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
361
362
363
364
            self.parsing()
            self.normalizing()
            self.archiving()
        finally:
Markus Scheidgen's avatar
Markus Scheidgen committed
365
            # close loghandler that was not closed due to failures
366
            try:
Markus Scheidgen's avatar
Markus Scheidgen committed
367
                if self._parser_results and self._parser_results.m_resource:
368
                    self._parser_results.section_metadata = None
Markus Scheidgen's avatar
Markus Scheidgen committed
369
                    self._parser_results.m_resource.unload()
370
            except Exception as e:
371
                logger.error('could unload processing results', exc_info=e)
372

373
    def on_fail(self):
374
375
376
        # in case of failure, index a minimum set of metadata and mark
        # processing failure
        try:
377
378
379
            if self._entry_metadata is None:
                self._setup_fallback_metadata()

380
            self._entry_metadata.processed = False
381

382
            self.apply_entry_metadata(self._entry_metadata)
Markus Scheidgen's avatar
Markus Scheidgen committed
383
384
            self._entry_metadata.apply_domain_metadata(self._parser_results)
            if self._parser_results and self._parser_results.m_resource:
385
                self._parser_results.section_metadata = None
Markus Scheidgen's avatar
Markus Scheidgen committed
386
                self._parser_results.m_resource.unload()
387

388
            self._entry_metadata.a_elastic.index()
389
        except Exception as e:
390
391
            self.get_logger().error(
                'could not index after processing failure', exc_info=e)
392

393
        try:
394
            self.write_archive(None)
395
        except Exception as e:
396
397
            self.get_logger().error(
                'could not write archive after processing failure', exc_info=e)
398

399
400
401
402
403
404
405
406
    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
        if process_name == 'process_calc' or process_name == 're_process_calc' or process_name is None:
            self.upload.reload()
            self.upload.check_join()

Markus Scheidgen's avatar
Markus Scheidgen committed
407
408
    @task
    def parsing(self):
409
        ''' The *task* that encapsulates all parsing related actions. '''
410
        context = dict(parser=self.parser, step=self.parser)
411
        logger = self.get_logger(**context)
412
        parser = parser_dict[self.parser]
413
        self._entry_metadata.parser_name = self.parser
Markus Scheidgen's avatar
Markus Scheidgen committed
414

415
        with utils.timer(logger, 'parser executed', input_size=self.mainfile_file.size):
416
            try:
417
418
419
                self._parser_results = EntryArchive()
                # allow parsers to read/write metadata
                self._parser_results.m_add_sub_section(EntryArchive.section_metadata, self._entry_metadata)
Markus Scheidgen's avatar
Markus Scheidgen committed
420
421
422
                parser.parse(
                    self.upload_files.raw_file_object(self.mainfile).os_path,
                    self._parser_results, logger=logger)
423

424
            except Exception as e:
425
                self.fail('parser failed with exception', exc_info=e, error=str(e), **context)
426
                return
427
            except SystemExit:
428
                self.fail('parser raised system exit', error='system exit', **context)
429
                return
Markus Scheidgen's avatar
Markus Scheidgen committed
430

431
432
433
434
435
436
437
438
439
440
    def process_phonon(self):
        """Function that is run for phonon calculation before cleanup.
        This task is run by the celery process that is calling the join for the
        upload.

        This function re-opens the Archive for this calculation to add method
        information from another referenced archive. Updates the method
        information in section_encyclopedia as well as the DFT domain metadata.
        """
        try:
441
            logger = self.get_logger(parser=self.parser, step=self.parser)
442
443
444
445
446
447

            # Open the archive of the phonon calculation.
            upload_files = StagingUploadFiles(self.upload_id, is_authorized=lambda: True)
            with upload_files.read_archive(self.calc_id) as archive:
                arch = query_archive(archive, {self.calc_id: self.calc_id})[self.calc_id]
                phonon_archive = EntryArchive.m_from_dict(arch)
Markus Scheidgen's avatar
Markus Scheidgen committed
448
            self._entry_metadata = phonon_archive.section_metadata
449
            self._calc_proc_logs = phonon_archive.processing_logs
450

Markus Scheidgen's avatar
Markus Scheidgen committed
451
452
            # Re-create the parse results
            self._parser_results = phonon_archive
Markus Scheidgen's avatar
Markus Scheidgen committed
453

454
455
456
            # Read in the first referenced calculation. The reference is given as
            # an absolute path which needs to be converted into a path that is
            # relative to upload root.
Markus Scheidgen's avatar
Markus Scheidgen committed
457
            scc = self._parser_results.section_run[0].section_single_configuration_calculation[0]
458
            relative_ref = scc.section_calculation_to_calculation_refs[0].calculation_to_calculation_external_url
459
460
461
462
463
464
            ref_id = upload_files.calc_id(relative_ref)
            with upload_files.read_archive(ref_id) as archive:
                arch = query_archive(archive, {ref_id: ref_id})[ref_id]
                ref_archive = EntryArchive.m_from_dict(arch)

            # Get encyclopedia method information directly from the referenced calculation.
465
            ref_enc_method = ref_archive.section_metadata.encyclopedia.method
466
            if ref_enc_method is None or len(ref_enc_method) == 0 or ref_enc_method.functional_type is None:
467
                raise ValueError("No method information available in referenced calculation.")
Markus Scheidgen's avatar
Markus Scheidgen committed
468
            self._parser_results.section_metadata.encyclopedia.method = ref_enc_method
469
470
471
472
473
474
475
476

            # Overwrite old entry with new data. The metadata is updated with
            # new timestamp and method details taken from the referenced
            # archive.
            self._entry_metadata.last_processing = datetime.utcnow()
            self._entry_metadata.dft.xc_functional = ref_archive.section_metadata.dft.xc_functional
            self._entry_metadata.dft.basis_set = ref_archive.section_metadata.dft.basis_set
            self._entry_metadata.dft.update_group_hash()
477
            self._entry_metadata.encyclopedia.status = EncyclopediaMetadata.status.type.success
478
        except Exception as e:
479
            logger.error("Could not retrieve method information for phonon calculation.", exc_info=e)
Markus Scheidgen's avatar
Markus Scheidgen committed
480
481
482
483
            if self._entry_metadata is None:
                self._setup_fallback_metadata()
                self._entry_metadata.processed = False

484
485
486
487
488
489
            try:
                if self._entry_metadata.encyclopedia is None:
                    self._entry_metadata.encyclopedia = EncyclopediaMetadata()
                self._entry_metadata.encyclopedia.status = EncyclopediaMetadata.status.type.failure
            except Exception as e:
                logger.error("Could set encyclopedia status.", exc_info=e)
490

491
        finally:
492
493
494
495
496
497
498
499
500
501
502
503
504
            # persist the calc metadata
            with utils.timer(logger, 'saved calc metadata', step='metadata'):
                self.apply_entry_metadata(self._entry_metadata)

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

            # persist the archive
            with utils.timer(
                    logger, 'archived', step='archive',
                    input_size=self.mainfile_file.size) as log_data:

Markus Scheidgen's avatar
Markus Scheidgen committed
505
                archive_size = self.write_archive(self._parser_results)
506
507
                log_data.update(archive_size=archive_size)

Markus Scheidgen's avatar
Markus Scheidgen committed
508
509
    @task
    def normalizing(self):
510
        ''' The *task* that encapsulates all normalizing related actions. '''
511
512

        # allow normalizer to access and add data to the entry metadata
513
514
515
        if self._parser_results.section_metadata is None:
            self._parser_results.m_add_sub_section(
                datamodel.EntryArchive.section_metadata, self._entry_metadata)
516

Markus Scheidgen's avatar
Markus Scheidgen committed
517
        for normalizer in normalizers:
518
            if normalizer.domain != parser_dict[self.parser].domain:
519
520
                continue

Markus Scheidgen's avatar
Markus Scheidgen committed
521
            normalizer_name = normalizer.__name__
522
            context = dict(normalizer=normalizer_name, step=normalizer_name)
523
            logger = self.get_logger(**context)
Markus Scheidgen's avatar
Markus Scheidgen committed
524

Markus Scheidgen's avatar
Markus Scheidgen committed
525
526
527
528
529
530
            with utils.timer(logger, 'normalizer executed', input_size=self.mainfile_file.size):
                try:
                    normalizer(self._parser_results).normalize(logger=logger)
                    logger.info('processor completed successfull', **context)
                except Exception as e:
                    self.fail('normalizer failed with exception', exc_info=e, error=str(e), **context)
Markus Scheidgen's avatar
Markus Scheidgen committed
531

Alvin Noe Ladines's avatar
Alvin Noe Ladines committed
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
    def _read_metadata_from_file(self, logger):
        # metadata file name defined in nomad.config nomad_metadata.yaml/json
        # which can be placed in the directory containing the mainfile or somewhere up
        # highest priority is directory with mainfile

        metadata_file = config.aux_metadata_file
        metadata_dir = os.path.dirname(self.mainfile_file.os_path)

        metadata = {}
        metadata_path = None
        while metadata_dir:
            metadata_part = metadata_cached(os.path.join(metadata_dir, metadata_file))

            for key, val in metadata_part.items():
                metadata.setdefault(key, val)

            metadata_dir = os.path.dirname(metadata_dir)
            if metadata_path is not None:
                break

        for key, val in metadata.items():
            definition = _editable_metadata.get(key, None)
            if not definition:
                logger.warn('Cannot set metadata %s' % key)
                continue
            self._entry_metadata.m_set(definition, val)

Markus Scheidgen's avatar
Markus Scheidgen committed
559
560
    @task
    def archiving(self):
561
        ''' The *task* that encapsulates all archival related actions. '''
562
563
        logger = self.get_logger()

Markus Scheidgen's avatar
Markus Scheidgen committed
564
        self._entry_metadata.apply_domain_metadata(self._parser_results)
565
        self._entry_metadata.processed = True
566

Alvin Noe Ladines's avatar
Alvin Noe Ladines committed
567
568
569
        # read metadata from file
        self._read_metadata_from_file(logger)

570
571
        # persist the calc metadata
        with utils.timer(logger, 'saved calc metadata', step='metadata'):
572
            self.apply_entry_metadata(self._entry_metadata)
573
574

        # index in search
575
        with utils.timer(logger, 'indexed', step='index'):
576
            self._entry_metadata.a_elastic.index()
577

578
        # persist the archive
Markus Scheidgen's avatar
Markus Scheidgen committed
579
        with utils.timer(
580
                logger, 'archived', step='archive',
Markus Scheidgen's avatar
Markus Scheidgen committed
581
                input_size=self.mainfile_file.size) as log_data:
582

Markus Scheidgen's avatar
Markus Scheidgen committed
583
            archive_size = self.write_archive(self._parser_results)
584
            log_data.update(archive_size=archive_size)
Markus Scheidgen's avatar
Markus Scheidgen committed
585

Markus Scheidgen's avatar
Markus Scheidgen committed
586
    def write_archive(self, archive: EntryArchive):
587
588
589
590
591
592
        def filter_processing_logs(logs):
            if len(logs) > 100:
                return [
                    log for log in logs
                    if log.get('level') != 'DEBUG']
            return logs
Markus Scheidgen's avatar
Markus Scheidgen committed
593

594
595
596
        if self._calc_proc_logs is None:
            self._calc_proc_logs = []

Markus Scheidgen's avatar
Markus Scheidgen committed
597
598
        if archive is not None:
            archive = archive.m_copy()
599
        else:
Markus Scheidgen's avatar
Markus Scheidgen committed
600
            archive = datamodel.EntryArchive()
601

Markus Scheidgen's avatar
Markus Scheidgen committed
602
603
        if archive.section_metadata is None:
            archive.m_add_sub_section(datamodel.EntryArchive.section_metadata, self._entry_metadata)
604

Markus Scheidgen's avatar
Markus Scheidgen committed
605
        archive.processing_logs = filter_processing_logs(self._calc_proc_logs)
606
607

        try:
Markus Scheidgen's avatar
Markus Scheidgen committed
608
            return self.upload_files.write_archive(self.calc_id, archive.m_to_dict())
609
610
        except Exception as e:
            # most likely failed due to domain data, try to write metadata and processing logs
Markus Scheidgen's avatar
Markus Scheidgen committed
611
612
613
614
            archive = datamodel.EntryArchive()
            archive.m_add_sub_section(datamodel.EntryArchive.section_metadata, self._entry_metadata)
            archive.processing_logs = filter_processing_logs(self._calc_proc_logs)
            self.upload_files.write_archive(self.calc_id, archive.m_to_dict())
615
            raise e
Markus Scheidgen's avatar
Markus Scheidgen committed
616

617
    def __str__(self):
618
        return 'calc %s calc_id=%s upload_id%s' % (super().__str__(), self.calc_id, self.upload_id)
619

620

621
class Upload(Proc):
622
    '''
Markus Scheidgen's avatar
Markus Scheidgen committed
623
624
625
626
627
    Represents uploads in the databases. Provides persistence access to the files storage,
    and processing state.

    Attributes:
        name: optional user provided upload name
628
629
        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
630
631
        upload_id: the upload id generated by the database
        upload_time: the timestamp when the system realised the upload
632
        user_id: the id of the user that created this upload
Markus Scheidgen's avatar
Markus Scheidgen committed
633
634
        published: Boolean that indicates the publish status
        publish_time: Date when the upload was initially published
635
        last_update: Date of the last publishing/re-processing
636
        joined: Boolean indicates if the running processing has joined (:func:`check_join`)
637
    '''
Markus Scheidgen's avatar
Markus Scheidgen committed
638
639
640
    id_field = 'upload_id'

    upload_id = StringField(primary_key=True)
641
642
    upload_path = StringField(default=None)
    temporary = BooleanField(default=False)
643
    embargo_length = IntField(default=36)
Markus Scheidgen's avatar
Markus Scheidgen committed
644
645
646

    name = StringField(default=None)
    upload_time = DateTimeField()
647
    user_id = StringField(required=True)
648
649
    published = BooleanField(default=False)
    publish_time = DateTimeField()
650
    last_update = DateTimeField()
Markus Scheidgen's avatar
Markus Scheidgen committed
651

652
653
    joined = BooleanField(default=False)

Markus Scheidgen's avatar
Markus Scheidgen committed
654
655
    meta: Any = {
        'indexes': [
656
            'user_id', 'tasks_status', 'process_status', 'published', 'upload_time'
657
        ]
Markus Scheidgen's avatar
Markus Scheidgen committed
658
659
660
661
    }

    def __init__(self, **kwargs):
        super().__init__(**kwargs)
662
        self._upload_files: ArchiveBasedStagingUploadFiles = None
Markus Scheidgen's avatar
Markus Scheidgen committed
663

664
665
    @property
    def metadata(self) -> dict:
666
        '''
Markus Scheidgen's avatar
Markus Scheidgen committed
667
668
669
670
671
        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.
672
        '''
673
674
675
676
677
        try:
            upload_files = PublicUploadFiles(self.upload_id, is_authorized=lambda: True)
        except KeyError:
            return None
        return upload_files.user_metadata
678
679
680

    @metadata.setter
    def metadata(self, data: dict) -> None:
681
682
        upload_files = PublicUploadFiles(self.upload_id, is_authorized=lambda: True, create=True)
        upload_files.user_metadata = data
683

Markus Scheidgen's avatar
Markus Scheidgen committed
684
    @classmethod
685
    def get(cls, id: str, include_published: bool = True) -> 'Upload':
686
        return cls.get_by_id(id, 'upload_id')
Markus Scheidgen's avatar
Markus Scheidgen committed
687
688

    @classmethod
689
    def user_uploads(cls, user: datamodel.User, **kwargs) -> List['Upload']:
690
        ''' Returns all uploads for the given user. Kwargs are passed to mongo query. '''
691
        return cls.objects(user_id=str(user.user_id), **kwargs)
Markus Scheidgen's avatar
Markus Scheidgen committed
692

693
694
    @property
    def uploader(self):
695
        return datamodel.User.get(self.user_id)
696

Markus Scheidgen's avatar
Markus Scheidgen committed
697
698
    def get_logger(self, **kwargs):
        logger = super().get_logger()
699
700
        user = self.uploader
        user_name = '%s %s' % (user.first_name, user.last_name)
701
        # We are not using 'user_id' because logstash (?) will filter these entries ?!
702
703
        logger = logger.bind(
            upload_id=self.upload_id, upload_name=self.name, user_name=user_name,
704
            user=user.user_id, **kwargs)
Markus Scheidgen's avatar
Markus Scheidgen committed
705
706
707
708
        return logger

    @classmethod
    def create(cls, **kwargs) -> 'Upload':
709
        '''
Markus Scheidgen's avatar
Markus Scheidgen committed
710
711
712
        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.
713
714

        Arguments:
715
            user: The user that created the upload.
716
        '''
717
718
        # use kwargs to keep compatibility with super method
        user: datamodel.User = kwargs['user']
719
        del(kwargs['user'])
720

721
722
        if 'upload_id' not in kwargs:
            kwargs.update(upload_id=utils.create_uuid())
723
        kwargs.update(user_id=user.user_id)
Markus Scheidgen's avatar
Markus Scheidgen committed
724
        self = super().create(**kwargs)
725

Markus Scheidgen's avatar
Markus Scheidgen committed
726
        self._continue_with('uploading')
727

Markus Scheidgen's avatar
Markus Scheidgen committed
728
729
        return self

730
    def delete(self):
731
        ''' Deletes this upload process state entry and its calcs. '''
732
733
734
        Calc.objects(upload_id=self.upload_id).delete()
        super().delete()

735
    def delete_upload_local(self):
736
        '''
737
        Deletes the upload, including its processing state and
738
        staging files. Local version without celery processing.
739
        '''
740
741
        logger = self.get_logger()

742
        with utils.lnr(logger, 'upload delete failed'):
743
            with utils.timer(
744
                    logger, 'upload deleted from index', step='index',
745
                    upload_size=self.upload_files.size):
746
                search.delete_upload(self.upload_id)
747

748
            with utils.timer(
749
                    logger, 'upload deleted', step='files',
750
751
                    upload_size=self.upload_files.size):
                self.upload_files.delete()
752
753

            self.delete()
754

755
    @process
756
    def delete_upload(self):
757
        '''
758
759
        Deletes of the upload, including its processing state and
        staging files. This starts the celery process of deleting the upload.
760
        '''
761
        self.delete_upload_local()
762

763
        return True  # do not save the process status on the delete upload
764

765
    @process
766
    def publish_upload(self):
767
        '''
768
769
        Moves the upload out of staging to the public area. It will
        pack the staging upload files in to public upload files.
770
        '''
771
772
        assert self.processed_calcs > 0

773
        logger = self.get_logger()
774
        logger.info('started to publish')
775

776
        with utils.lnr(logger, 'publish failed'):
777
            with self.entries_metadata(self.metadata) as calcs:
778

779
                with utils.timer(
780
                        logger, 'upload metadata updated', step='metadata',
781
                        upload_size=self.upload_files.size):
782

783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
                    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)
798

799
                with utils.timer(
800
                        logger, 'index updated', step='index',
801
                        upload_size=self.upload_files.size):
802
803
804
805
806
807
808
809
810
811
812
813
                    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:
814
                    self.last_update = datetime.utcnow()
815
                    self.save()
816

817
818
    @process
    def re_process_upload(self):
819
        '''
Markus Scheidgen's avatar
Markus Scheidgen committed
820
821
822
        A *process* that performs the re-processing of a earlier processed
        upload.

823
824
825
826
827
828
        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.
829
        '''
830
831
832
        logger = self.get_logger()
        logger.info('started to re-process')

833
834
835
836
837
        # 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')
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852

        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)
853
854

        self._continue_with('parse_all')
855
        try:
856
            # check if a calc is already/still processing
857
858
859
860
861
862
863
864
            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)
865

866
            # reset all calcs
867
            Calc._get_collection().update_many(
868
869
870
                dict(upload_id=self.upload_id),
                {'$set': Calc.reset_pymongo_update(worker_hostname=self.worker_hostname)})

871
872
            # process call calcs
            Calc.process_all(Calc.re_process_calc, dict(upload_id=self.upload_id), exclude=['metadata'])
873
874

            logger.info('completed to trigger re-process of all calcs')
875
876
        except Exception as e:
            # try to remove the staging copy in failure case
877
878
            logger.error('failed to trigger re-process of all calcs', exc_info=e)

879
880
881
            if self.published:
                if staging_upload_files is not None and staging_upload_files.exists():
                    staging_upload_files.delete()
882
883

            raise e
884
885
886
887

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

888
889
    @process
    def re_pack(self):
890
        ''' A *process* that repacks the raw and archive data based on the current embargo data. '''
891
892
893
894
895
896
897
898
        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')

899
        self.upload_files.re_pack(self.user_metadata())
900
        self.joined = True
901
902
        self._complete()

Markus Scheidgen's avatar
Markus Scheidgen committed
903
    @process
904
    def process_upload(self):
905
        ''' A *process* that performs the initial upload processing. '''
Markus Scheidgen's avatar
Markus Scheidgen committed
906
907
908
909
910
        self.extracting()
        self.parse_all()

    @task
    def uploading(self):
911
        ''' A no-op *task* as a stand-in for receiving upload data. '''
Markus Scheidgen's avatar
Markus Scheidgen committed
912
913
        pass

914
    @property
915
916
    def upload_files(self) -> UploadFiles:
        upload_files_class = ArchiveBasedStagingUploadFiles if not self.published else PublicUploadFiles
917
        kwargs = dict(upload_path=self.upload_path) if not self.published else {}
918
919
920

        if not self._upload_files or not isinstance(self._upload_files, upload_files_class):
            self._upload_files = upload_files_class(
921
                self.upload_id, is_authorized=lambda: True, **kwargs)
922

923
        return self._upload_files
924

925
926
927
928
929
    @property
    def staging_upload_files(self) -> ArchiveBasedStagingUploadFiles:
        assert not self.published
        return cast(ArchiveBasedStagingUploadFiles, self.upload_files)

Markus Scheidgen's avatar
Markus Scheidgen committed
930
931
    @task
    def extracting(self):
932