data.py

#
# Copyright The NOMAD Authors.
#
# This file is part of NOMAD. See https://nomad-lab.eu for further info.
#
# 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.
#

'''
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

.. autoclass:: Upload

'''

from typing import cast, List, Any, Tuple, Iterator, Dict, cast, Iterable
from mongoengine import (
    StringField, DateTimeField, DictField, BooleanField, IntField, ListField)
import logging
from structlog import wrap_logger
from contextlib import contextmanager
import os.path
from datetime import datetime
from pymongo import UpdateOne
import hashlib
from structlog.processors import StackInfoRenderer, format_exc_info, TimeStamper
import yaml
import json
from functools import lru_cache
import urllib.parse
import requests

from nomad import utils, config, infrastructure, search, datamodel, metainfo
from nomad.files import (
    PathObject, UploadFiles, ExtractError, ArchiveBasedStagingUploadFiles,
    PublicUploadFiles, StagingUploadFiles)
from nomad.processing.base import Proc, process, task, PENDING, SUCCESS, FAILURE
from nomad.parsing.parsers import parser_dict, match_parser
from nomad.normalizing import normalizers
from nomad.datamodel import (
    EntryArchive, EditableUserMetadata, OasisMetadata, UserProvidableMetadata)
from nomad.archive import (
    query_archive, write_partial_archive_to_mongo, delete_partial_archives_from_mongo)
from nomad.datamodel.encyclopedia import EncyclopediaMetadata


section_metadata = datamodel.EntryArchive.section_metadata.name
section_workflow = datamodel.EntryArchive.section_workflow.name


_editable_metadata: Dict[str, metainfo.Definition] = {}
_editable_metadata.update(**{
    quantity.name: quantity for quantity in UserProvidableMetadata.m_def.definitions})
_editable_metadata.update(**{
    quantity.name: quantity for quantity in EditableUserMetadata.m_def.definitions})

_oasis_metadata = {
    quantity.name: quantity for quantity in OasisMetadata.m_def.definitions}


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']})
        log_data.update(logger=logger.name)

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


def _normalize_oasis_upload_metadata(upload_id, metadata):
    # This is overwritten by the tests to do necessary id manipulations
    return upload_id, metadata


class Calc(Proc):
    '''
    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.

    The attribute list, does not include the various metadata properties generated
    while parsing, including ``code_name``, ``code_version``, etc.

    Attributes:
        calc_id: the calc_id of this calc
        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

        metadata: the metadata record wit calc and user metadata, see :class:`datamodel.EntryMetadata`
    '''
    calc_id = StringField(primary_key=True)
    upload_id = StringField()
    mainfile = StringField()
    parser = StringField()

    metadata = DictField()

    meta: Any = {
        'strict': False,
        'indexes': [
            'upload_id',
            'parser',
            ('upload_id', 'mainfile'),
            ('upload_id', 'parser'),
            ('upload_id', 'tasks_status'),
            ('upload_id', 'process_status'),
            ('upload_id', 'metadata.nomad_version'),
            'metadata.processed',
            'metadata.last_processing',
            'metadata.published',
            'metadata.datasets',
            'metadata.pid'
        ]
    }

    def __init__(self, *args, **kwargs):
        super().__init__(*args, **kwargs)
        self._parser_results: EntryArchive = None
        self._upload: Upload = None
        self._upload_files: ArchiveBasedStagingUploadFiles = None
        self._calc_proc_logs: List[Any] = None

        self._entry_metadata = None

    @classmethod
    def get(cls, id):
        return cls.get_by_id(id, 'calc_id')

    @property
    def mainfile_file(self) -> PathObject:
        return self.upload_files.raw_file_object(self.mainfile)

    @property
    def upload(self) -> 'Upload':
        if not self._upload:
            self._upload = Upload.get(self.upload_id)
            self._upload.worker_hostname = self.worker_hostname
        return self._upload

    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 necessarily the user metadata nor the metadata from
        the archive.
        '''
        entry_metadata = datamodel.EntryMetadata()
        if self.parser is not None:
            parser = parser_dict[self.parser]
            if parser.domain:
                entry_metadata.domain = parser_dict[self.parser].domain
        entry_metadata.upload_id = self.upload_id
        entry_metadata.calc_id = self.calc_id
        entry_metadata.mainfile = self.mainfile
        entry_metadata.nomad_version = config.meta.version
        entry_metadata.nomad_commit = config.meta.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)
        try:
            # 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]

        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()

        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 necessarily the metadata from
        the archive.
        '''
        entry_metadata = self.create_metadata()
        entry_metadata.m_update_from_dict(self.metadata)

        return entry_metadata

    @property
    def upload_files(self) -> ArchiveBasedStagingUploadFiles:
        if not self._upload_files:
            self._upload_files = ArchiveBasedStagingUploadFiles(
                self.upload_id, is_authorized=lambda: True, upload_path=self.upload.upload_path)
        return self._upload_files

    def get_logger(self, **kwargs):
        '''
        Returns a wrapped logger that additionally saves all entries to the calculation
        processing log in the archive.
        '''
        logger = super().get_logger()
        logger = logger.bind(
            upload_id=self.upload_id, mainfile=self.mainfile, calc_id=self.calc_id, **kwargs)

        if self._calc_proc_logs is None:
            self._calc_proc_logs = []

        def save_to_calc_log(logger, method_name, event_dict):
            try:
                # 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()}
                dump_dict.update(level=method_name.upper())
                self._calc_proc_logs.append(dump_dict)

            except Exception:
                # Exceptions here will cause indefinite loop
                pass

            return event_dict

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

    @process
    def re_process_calc(self):
        '''
        Processes a calculation again. This means there is already metadata and
        instead of creating it initially, we are just updating the existing
        records.
        '''
        parser = match_parser(self.upload_files.raw_file_object(self.mainfile).os_path, strict=False)
        logger = self.get_logger()

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

            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)
                raise e

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

        if parser is None:
            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:
            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)

        try:
            self._entry_metadata = self.user_metadata()
            self._entry_metadata.calc_hash = self.upload_files.calc_hash(self.mainfile)
            self._entry_metadata.nomad_version = config.meta.version
            self._entry_metadata.nomad_commit = config.meta.commit
            self._entry_metadata.last_processing = datetime.utcnow()
            self._entry_metadata.files = self.upload_files.calc_files(self.mainfile)

            self.parsing()
            self.normalizing()
            self.archiving()
        finally:
            # close loghandler that was not closed due to failures
            try:
                if self._parser_results and self._parser_results.m_resource:
                    self._parser_results.section_metadata = None
                    self._parser_results.m_resource.unload()
            except Exception as e:
                logger.error('could not unload processing results', exc_info=e)

    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)
        self._entry_metadata.parser_name = self.parser

    @process
    def process_calc(self):
        '''
        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.
        '''
        logger = self.get_logger()
        if self.upload is None:
            logger.error('calculation upload does not exist')

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

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

            self.parsing()
            self.normalizing()
            self.archiving()
        finally:
            # close loghandler that was not closed due to failures
            try:
                if self._parser_results and self._parser_results.m_resource:
                    self._parser_results.section_metadata = None
                    self._parser_results.m_resource.unload()
            except Exception as e:
                logger.error('could unload processing results', exc_info=e)

    def on_fail(self):
        # in case of failure, index a minimum set of metadata and mark
        # processing failure
        try:
            if self._entry_metadata is None:
                self._setup_fallback_metadata()

            self._entry_metadata.processed = False

            try:
                self.apply_entry_metadata(self._entry_metadata)
            except Exception as e:
                self.get_logger().error(
                    'could not apply entry metadata to entry', exc_info=e)

            try:
                self._entry_metadata.apply_domain_metadata(self._parser_results)
            except Exception as e:
                self.get_logger().error(
                    'could not apply domain metadata to entry', exc_info=e)

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

        try:
            self.write_archive(self._parser_results)
        except Exception as e:
            self.get_logger().error(
                'could not write archive after processing failure', exc_info=e)

    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()

    @task
    def parsing(self):
        ''' The *task* that encapsulates all parsing related actions. '''
        context = dict(parser=self.parser, step=self.parser)
        logger = self.get_logger(**context)
        parser = parser_dict[self.parser]
        self._entry_metadata.parser_name = self.parser

        with utils.timer(logger, 'parser executed', input_size=self.mainfile_file.size):
            try:
                self._parser_results = EntryArchive()
                # allow parsers to read/write metadata
                self._parser_results.m_add_sub_section(EntryArchive.section_metadata, self._entry_metadata)
                parser.parse(
                    self.upload_files.raw_file_object(self.mainfile).os_path,
                    self._parser_results, logger=logger)

            except Exception as e:
                self.fail('parser failed with exception', exc_info=e, error=str(e), **context)
                return
            except SystemExit:
                self.fail('parser raised system exit', error='system exit', **context)
                return

    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:
            logger = self.get_logger(parser=self.parser, step=self.parser)

            # 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)
            self._entry_metadata = phonon_archive.section_metadata
            self._calc_proc_logs = phonon_archive.processing_logs

            # Re-create the parse results
            self._parser_results = phonon_archive

            # 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.
            scc = self._parser_results.section_run[0].section_single_configuration_calculation[0]
            relative_ref = scc.section_calculation_to_calculation_refs[0].calculation_to_calculation_external_url
            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.
            ref_enc_method = ref_archive.section_metadata.encyclopedia.method
            if ref_enc_method is None or len(ref_enc_method) == 0 or ref_enc_method.functional_type is None:
                raise ValueError("No method information available in referenced calculation.")
            self._parser_results.section_metadata.encyclopedia.method = ref_enc_method

            # 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()
            self._entry_metadata.encyclopedia.status = EncyclopediaMetadata.status.type.success
        except Exception as e:
            logger.error("Could not retrieve method information for phonon calculation.", exc_info=e)
            if self._entry_metadata is None:
                self._setup_fallback_metadata()
                self._entry_metadata.processed = False

            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)

        finally:
            # 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:

                archive_size = self.write_archive(self._parser_results)
                log_data.update(archive_size=archive_size)

    @task
    def normalizing(self):
        ''' The *task* that encapsulates all normalizing related actions. '''

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

        for normalizer in normalizers:
            if normalizer.domain != parser_dict[self.parser].domain:
                continue

            normalizer_name = normalizer.__name__
            context = dict(normalizer=normalizer_name, step=normalizer_name)
            logger = self.get_logger(**context)

            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)

    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.metadata_file_name
        metadata_dir = os.path.dirname(self.mainfile_file.os_path)

        metadata = {}
        while True:
            # top-level nomad file can also contain an entries dict with entry
            # metadata per mainfile as key
            if metadata_dir == self.upload_files.os_path:
                entries = metadata_part.get('entries', {})
                metadata_part = entries.get(self.mainfile, {})
                for key, val in metadata_part.items():
                    metadata.setdefault(key, val)

            # consider the nomad file of the current directory
            metadata_part = self.upload.metadata_file_cached(
                os.path.join(metadata_dir, metadata_file))
            for key, val in metadata_part.items():
                if key in ['entries', 'oasis_datasets']:
                    continue
                metadata.setdefault(key, val)

            if metadata_dir == self.upload_files.os_path:
                break

            metadata_dir = os.path.dirname(metadata_dir)

        if len(metadata) > 0:
            logger.info('Apply user metadata from nomad.yaml/json file')

        for key, val in metadata.items():
            if key == 'entries':
                continue

            definition = _editable_metadata.get(key, None)
            if definition is None and self.upload.from_oasis:
                definition = _oasis_metadata.get(key, None)

            if key == 'uploader':
                if datamodel.User.get(self.upload.user_id).is_admin:
                    definition = datamodel.EntryMetadata.uploader

            if definition is None:
                logger.warn('Users cannot set metadata', quantity=key)
                continue

            try:
                self._entry_metadata.m_set(definition, val)
                if definition == datamodel.EntryMetadata.calc_id:
                    self.calc_id = val
            except Exception as e:
                logger.error(
                    'Could not apply user metadata from nomad.yaml/json file',
                    quantitiy=definition.name, exc_info=e)

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

        self._entry_metadata.apply_domain_metadata(self._parser_results)
        self._entry_metadata.processed = True

        if self.upload.publish_directly:
            self._entry_metadata.published |= True

        self._read_metadata_from_file(logger)

        # 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:

            archive_size = self.write_archive(self._parser_results)
            log_data.update(archive_size=archive_size)

    def write_archive(self, archive: EntryArchive):
        # save the archive mongo entry
        try:
            if self._entry_metadata.processed:
                write_partial_archive_to_mongo(archive)
        except Exception as e:
            self.get_logger().error('could not write mongodb archive entry', exc_info=e)

        # add the processing logs to the archive
        def filter_processing_logs(logs):
            if len(logs) > 100:
                return [
                    log for log in logs
                    if log.get('level') != 'DEBUG']
            return logs

        if self._calc_proc_logs is None:
            self._calc_proc_logs = []

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

        if archive.section_metadata is None:
            archive.m_add_sub_section(datamodel.EntryArchive.section_metadata, self._entry_metadata)

        archive.processing_logs = filter_processing_logs(self._calc_proc_logs)

        # save the archive msg-pack
        try:
            return self.upload_files.write_archive(self.calc_id, archive.m_to_dict())
        except Exception as e:
            # most likely failed due to domain data, try to write metadata and processing logs
            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())
            raise e

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


class Upload(Proc):
    '''
    Represents uploads in the databases. Provides persistence access to the files storage,
    and processing state.

    Attributes:
        name: Optional user provided upload name.
        upload_path: The fs path were the uploaded files was stored during upload.
        temporary: True if the uploaded file should be removed after extraction.

        upload_id: The upload id generated by the database or the uploaded NOMAD deployment.
        upload_time: Datetime of the original upload independent of the NOMAD deployment
            it was first uploaded to.
        user_id: The id of the user that created this upload.
        published: Boolean that indicates that the upload is published on this NOMAD deployment.
        publish_time: Datetime when the upload was initially published on this NOMAD deployment.
        last_update: Datetime of the last modifying process run (publish, re-processing, upload).

        publish_directly: Boolean indicating that this upload should be published after initial processing.
        from_oasis: Boolean indicating that this upload is coming from another NOMAD deployment.
        oasis_id: The deployment id of the NOMAD that uploaded the upload.
        published_to: A list of deployment ids where this upload has been successfully uploaded to.

        joined: Boolean indicates if the running processing has joined (:func:`check_join`).
    '''
    id_field = 'upload_id'

    upload_id = StringField(primary_key=True)
    upload_path = StringField(default=None)
    temporary = BooleanField(default=False)
    embargo_length = IntField(default=36)

    name = StringField(default=None)
    upload_time = DateTimeField()
    user_id = StringField(required=True)
    published = BooleanField(default=False)
    publish_time = DateTimeField()
    last_update = DateTimeField()

    publish_directly = BooleanField(default=False)
    from_oasis = BooleanField(default=False)
    oasis_deployment_id = StringField(default=None)
    published_to = ListField(StringField())

    joined = BooleanField(default=False)

    meta: Any = {
        'strict': False,
        'indexes': [
            'user_id', 'tasks_status', 'process_status', 'published', 'upload_time'
        ]
    }

    def __init__(self, **kwargs):
        super().__init__(**kwargs)
        self.publish_directly = self.publish_directly or self.from_oasis
        self._upload_files: ArchiveBasedStagingUploadFiles = None

    @lru_cache()
    def metadata_file_cached(self, path):
        for ext in config.metadata_file_extensions:
            full_path = '%s.%s' % (path, ext)
            if os.path.isfile(full_path):
                try:
                    with open(full_path) as f:
                        if full_path.endswith('.json'):
                            return json.load(f)
                        elif full_path.endswith('.yaml') or full_path.endswith('.yml'):
                            return yaml.load(f, Loader=getattr(yaml, 'FullLoader'))
                        else:
                            return {}
                except Exception as e:
                    self.get_logger().warn('could not parse nomad.yaml/json', path=path, exc_info=e)
                    # ignore the file contents if the file is not parsable
                    pass
        return {}

    @property
    def metadata(self) -> dict:
        '''
        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.
        '''
        try:
            upload_files = PublicUploadFiles(self.upload_id, is_authorized=lambda: True)
        except KeyError:
            return None
        return upload_files.user_metadata

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

    @classmethod
    def get(cls, id: str, include_published: bool = True) -> 'Upload':
        return cls.get_by_id(id, 'upload_id')

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

    @property
    def uploader(self):
        return datamodel.User.get(self.user_id)

    def get_logger(self, **kwargs):
        logger = super().get_logger()
        user = self.uploader
        user_name = '%s %s' % (user.first_name, user.last_name)
        # We are not using 'user_id' because logstash (?) will filter these entries ?!
        logger = logger.bind(
            upload_id=self.upload_id, upload_name=self.name, user_name=user_name,
            user=user.user_id, **kwargs)
        return logger

    @classmethod
    def create(cls, **kwargs) -> 'Upload':
        '''
        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.

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

        if 'upload_id' not in kwargs:
            kwargs.update(upload_id=utils.create_uuid())
        kwargs.update(user_id=user.user_id)
        self = super().create(**kwargs)

        self._continue_with('uploading')

        return self

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

    def delete_upload_local(self):
        '''
        Deletes the upload, including its processing state and
        staging files. Local version without celery processing.
        '''
        logger = self.get_logger()

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

            with utils.timer(
                    logger, 'upload partial archives', step='files',
                    upload_size=self.upload_files.size):

                calc_ids = [calc.calc_id for calc in Calc.objects(upload_id=self.upload_id)]
                delete_partial_archives_from_mongo(calc_ids)

            with utils.timer(
                    logger, 'upload deleted', step='files',
                    upload_size=self.upload_files.size):
                self.upload_files.delete()

            self.delete()

    @process
    def delete_upload(self):
        '''
        Deletes the upload, including its processing state and
        staging files. This starts the celery process of deleting the upload.
        '''
        self.delete_upload_local()

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

    @process
    def publish_upload(self):
        '''
        Moves the upload out of staging to the public area. It will
        pack the staging upload files in to public upload files.
        '''
        assert self.processed_calcs > 0

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

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

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

                    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)

                with utils.timer(
                        logger, 'index updated', step='index',
                        upload_size=self.upload_files.size):
                    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:
                    self.last_update = datetime.utcnow()
                    self.save()

    @process
    def publish_from_oasis(self):
        '''
        Uploads the already published upload to a different NOMAD deployment. This allows
        to push uploads from an OASIS to the central NOMAD.
        '''