nopaque/app/models.py

from datetime import datetime, timedelta
from enum import Enum, IntEnum
import re
from flask import current_app, url_for
from flask_hashids import HashidMixin
from flask_login import UserMixin
from time import sleep
from tqdm import tqdm
from werkzeug.security import generate_password_hash, check_password_hash
from werkzeug.utils import secure_filename
import json
import jwt
import os
import requests
import secrets
import shutil
import xml.etree.ElementTree as ET
import yaml
from app import db, hashids, login, mail, socketio
from app.converters.vrt import normalize_vrt_file
from app.email import create_message


##############################################################################
# enums                                                                      #
##############################################################################
# region enums
class CorpusStatus(IntEnum):
    UNPREPARED = 1
    SUBMITTED = 2
    QUEUED = 3
    BUILDING = 4
    BUILT = 5
    FAILED = 6
    STARTING_ANALYSIS_SESSION = 7
    RUNNING_ANALYSIS_SESSION = 8
    CANCELING_ANALYSIS_SESSION = 9


class JobStatus(IntEnum):
    INITIALIZING = 1
    SUBMITTED = 2
    QUEUED = 3
    RUNNING = 4
    CANCELING = 5
    CANCELED = 6
    COMPLETED = 7
    FAILED = 8


class Permission(IntEnum):
    '''
    Defines User permissions as integers by the power of 2. User permission
    can be evaluated using the bitwise operator &.
    '''
    ADMINISTRATE = 1
    CONTRIBUTE = 2
    USE_API = 4


class UserSettingJobStatusMailNotificationLevel(IntEnum):
    NONE = 1
    END = 2
    ALL = 3
# endregion enums


##############################################################################
# mixins                                                                     #
##############################################################################
# region mixins
class FileMixin:
    '''
    Mixin for db.Model classes. All file related models should use this.
    '''
    creation_date = db.Column(db.DateTime, default=datetime.utcnow)
    filename = db.Column(db.String(255))
    last_edited_date = db.Column(db.DateTime)
    mimetype = db.Column(db.String(255))

    def file_mixin_to_json(self, backrefs=False, relationships=False):
        return {
            'creation_date': f'{self.creation_date.isoformat()}Z',
            'filename': self.filename,
            'last_edited_date': (
                None if self.last_edited_date is None
                else f'{self.last_edited_date.isoformat()}Z'
            ),
            'mimetype': self.mimetype
        }
    
    @classmethod
    def create(cls, file_storage, **kwargs):
        filename = kwargs.pop('filename', file_storage.filename)
        mimetype = kwargs.pop('mimetype', file_storage.mimetype)
        obj = cls(
            filename=secure_filename(filename),
            mimetype=mimetype,
            **kwargs
        )
        db.session.add(obj)
        db.session.flush(objects=[obj])
        db.session.refresh(obj)
        try:
            file_storage.save(obj.path)
        except (AttributeError, OSError) as e:
            current_app.logger.error(e)
            db.session.rollback()
            raise e
        return obj
# endregion mixins


##############################################################################
# type_decorators                                                            #
##############################################################################
# region type_decorators
class IntEnumColumn(db.TypeDecorator):
    impl = db.Integer

    def __init__(self, enum_type, *args, **kwargs):
        super().__init__(*args, **kwargs)
        self.enum_type = enum_type

    def process_bind_param(self, value, dialect):
        if isinstance(value, self.enum_type) and isinstance(value.value, int):
            return value.value
        elif isinstance(value, int):
            return self.enum_type(value).value
        else:
            return TypeError()

    def process_result_value(self, value, dialect):
        return self.enum_type(value)


class ContainerColumn(db.TypeDecorator):
    impl = db.String

    def __init__(self, container_type, *args, **kwargs):
        super().__init__(*args, **kwargs)
        self.container_type = container_type

    def process_bind_param(self, value, dialect):
        if isinstance(value, self.container_type):
            return json.dumps(value)
        elif (isinstance(value, str)
                  and isinstance(json.loads(value), self.container_type)):
            return value
        else:
            return TypeError()

    def process_result_value(self, value, dialect):
        return json.loads(value)
# endregion type_decorators


##############################################################################
# Models                                                                     #
##############################################################################
# region models
class Role(HashidMixin, db.Model):
    __tablename__ = 'roles'
    # Primary key
    id = db.Column(db.Integer, primary_key=True)
    # Fields
    name = db.Column(db.String(64), unique=True)
    default = db.Column(db.Boolean, default=False, index=True)
    permissions = db.Column(db.Integer, default=0)
    # Relationships
    users = db.relationship('User', backref='role', lazy='dynamic')

    def __repr__(self):
        return f'<Role {self.name}>'

    def add_permission(self, permission):
        if not self.has_permission(permission):
            self.permissions += permission

    def has_permission(self, permission):
        return self.permissions & permission == permission

    def remove_permission(self, permission):
        if self.has_permission(permission):
            self.permissions -= permission

    def reset_permissions(self):
        self.permissions = 0

    def to_json(self, backrefs=False, relationships=False):
        _json = {
            'id': self.hashid,
            'default': self.default,
            'name': self.name,
            'permissions': self.permissions
        }
        if relationships:
            _json['users'] = {
                x.hashid: x.to_json(relationships=True)
                for x in self.users
            }
        return _json

    @staticmethod
    def insert_defaults():
        roles = {
            'User': [],
            'API user': [Permission.USE_API],
            'Contributor': [Permission.CONTRIBUTE],
            'Administrator': [
                Permission.ADMINISTRATE,
                Permission.CONTRIBUTE,
                Permission.USE_API
            ],
            'System user': []
        }
        default_role_name = 'User'
        for role_name, permissions in roles.items():
            role = Role.query.filter_by(name=role_name).first()
            if role is None:
                role = Role(name=role_name)
            role.reset_permissions()
            for permission in permissions:
                role.add_permission(permission)
            role.default = role.name == default_role_name
            db.session.add(role)
        db.session.commit()


class Token(db.Model):
    __tablename__ = 'tokens'
    # Primary key
    id = db.Column(db.Integer, primary_key=True)
    # Foreign keys
    user_id = db.Column(db.Integer, db.ForeignKey('users.id'))
    # Fields
    access_token = db.Column(db.String(64), index=True)
    access_expiration = db.Column(db.DateTime)
    refresh_token = db.Column(db.String(64), index=True)
    refresh_expiration = db.Column(db.DateTime)
    # Backrefs: user: User

    def expire(self):
        self.access_expiration = datetime.utcnow()
        self.refresh_expiration = datetime.utcnow()

    @staticmethod
    def clean():
        """Remove any tokens that have been expired for more than a day."""
        yesterday = datetime.utcnow() - timedelta(days=1)
        Token.query.filter(Token.refresh_expiration < yesterday).delete()


class User(HashidMixin, UserMixin, db.Model):
    __tablename__ = 'users'
    # Primary key
    id = db.Column(db.Integer, primary_key=True)
    # Foreign keys
    role_id = db.Column(db.Integer, db.ForeignKey('roles.id'))
    # Fields
    email = db.Column(db.String(254), index=True, unique=True)
    username = db.Column(db.String(64), index=True, unique=True)
    password_hash = db.Column(db.String(128))
    confirmed = db.Column(db.Boolean, default=False)
    member_since = db.Column(db.DateTime(), default=datetime.utcnow)
    setting_dark_mode = db.Column(db.Boolean, default=False)
    setting_job_status_mail_notification_level = db.Column(
        IntEnumColumn(UserSettingJobStatusMailNotificationLevel),
        default=UserSettingJobStatusMailNotificationLevel.END
    )
    last_seen = db.Column(db.DateTime())
    # Backrefs: role: Role
    # Relationships
    tesseract_ocr_pipeline_models = db.relationship(
        'TesseractOCRPipelineModel',
        backref='user',
        cascade='all, delete-orphan',
        lazy='dynamic'
    )
    spacy_nlp_pipeline_models = db.relationship(
        'SpaCyNLPPipelineModel',
        backref='user',
        cascade='all, delete-orphan',
        lazy='dynamic'
    )
    corpora = db.relationship(
        'Corpus',
        backref='user',
        cascade='all, delete-orphan',
        lazy='dynamic'
    )
    jobs = db.relationship(
        'Job',
        backref='user',
        cascade='all, delete-orphan',
        lazy='dynamic'
    )
    tokens = db.relationship(
        'Token',
        backref='user',
        cascade='all, delete-orphan',
        lazy='dynamic'
    )

    def __init__(self, **kwargs):
        super().__init__(**kwargs)
        if self.role is not None:
            return
        if self.email == current_app.config['NOPAQUE_ADMIN']:
            self.role = Role.query.filter_by(name='Administrator').first()
        else:
            self.role = Role.query.filter_by(default=True).first()

    def __repr__(self):
        return f'<User {self.username}>'

    @property
    def jsonpatch_path(self):
        return f'/users/{self.hashid}'

    @property
    def password(self):
        raise AttributeError('password is not a readable attribute')

    @password.setter
    def password(self, password):
        self.password_hash = generate_password_hash(password)

    @property
    def path(self):
        return os.path.join(
            current_app.config.get('NOPAQUE_DATA_DIR'), 'users', str(self.id))

    @staticmethod
    def create(**kwargs):
        user = User(**kwargs)
        db.session.add(user)
        db.session.flush(objects=[user])
        db.session.refresh(user)
        try:
            os.mkdir(user.path)
            os.mkdir(os.path.join(user.path, 'spacy_nlp_pipeline_models'))
            os.mkdir(os.path.join(user.path, 'tesseract_ocr_pipeline_models'))
            os.mkdir(os.path.join(user.path, 'corpora'))
            os.mkdir(os.path.join(user.path, 'jobs'))
        except OSError as e:
            current_app.logger.error(e)
            db.session.rollback()
            raise e
        return user

    @staticmethod
    def insert_defaults():
        nopaque_user = User.query.filter_by(username='nopaque').first()
        system_user_role = Role.query.filter_by(name='System user').first()
        if nopaque_user is None:
            nopaque_user = User.create(
                username='nopaque',
                role=system_user_role
            )
            db.session.add(nopaque_user)
        elif nopaque_user.role != system_user_role:
            nopaque_user.role = system_user_role
        db.session.commit()

    @staticmethod
    def reset_password(token, new_password):
        try:
            payload = jwt.decode(
                token,
                current_app.config['SECRET_KEY'],
                algorithms=['HS256'],
                issuer=current_app.config['SERVER_NAME'],
                options={'require': ['exp', 'iat', 'iss', 'purpose', 'sub']}
            )
        except jwt.PyJWTError:
            return False
        if payload.get('purpose') != 'User.reset_password':
            return False
        user_hashid = payload.get('sub')
        user_id = hashids.decode(user_hashid)
        user = User.query.get(user_id)
        if user is None:
            return False
        user.password = new_password
        db.session.add(user)
        return True

    @staticmethod
    def verify_access_token(access_token, refresh_token=None):
        token = Token.query.filter(Token.access_token == access_token).first()
        if token is not None:
            if token.access_expiration > datetime.utcnow():
                token.user.ping()
                db.session.commit()
                if token.user.role.name != 'System user':
                    return token.user

    @staticmethod
    def verify_refresh_token(refresh_token, access_token):
        token = Token.query.filter((Token.refresh_token == refresh_token) & (Token.access_token == access_token)).first()
        if token is not None:
            if token.refresh_expiration > datetime.utcnow():
                return token
            # someone tried to refresh with an expired token
            # revoke all tokens from this user as a precaution
            token.user.revoke_auth_tokens()
            db.session.commit()

    def can(self, permission):
        return self.role.has_permission(permission)

    def confirm(self, confirmation_token):
        try:
            payload = jwt.decode(
                confirmation_token,
                current_app.config['SECRET_KEY'],
                algorithms=['HS256'],
                issuer=current_app.config['SERVER_NAME'],
                options={'require': ['exp', 'iat', 'iss', 'purpose', 'sub']}
            )
            current_app.logger.warning(payload)
        except jwt.PyJWTError:
            return False
        if payload.get('purpose') != 'user.confirm':
            return False
        if payload.get('sub') != self.hashid:
            return False
        self.confirmed = True
        db.session.add(self)
        return True

    def delete(self):
        shutil.rmtree(self.path, ignore_errors=True)
        db.session.delete(self)

    def generate_auth_token(self):
        return Token(
            access_token=secrets.token_urlsafe(),
            access_expiration=datetime.utcnow() + timedelta(minutes=15),
            refresh_token=secrets.token_urlsafe(),
            refresh_expiration=datetime.utcnow() + timedelta(days=7),
            user=self
        )

    def generate_confirm_token(self, expiration=3600):
        now = datetime.utcnow()
        payload = {
            'exp': now + timedelta(seconds=expiration),
            'iat': now,
            'iss': current_app.config['SERVER_NAME'],
            'purpose': 'user.confirm',
            'sub': self.hashid
        }
        return jwt.encode(
            payload,
            current_app.config['SECRET_KEY'],
            algorithm='HS256'
        )

    def generate_reset_password_token(self, expiration=3600):
        now = datetime.utcnow()
        payload = {
            'exp': now + timedelta(seconds=expiration),
            'iat': now,
            'iss': current_app.config['SERVER_NAME'],
            'purpose': 'User.reset_password',
            'sub': self.hashid
        }
        return jwt.encode(
            payload,
            current_app.config['SECRET_KEY'],
            algorithm='HS256'
        )

    def is_administrator(self):
        return self.can(Permission.ADMINISTRATE)

    def ping(self):
        self.last_seen = datetime.utcnow()

    def revoke_auth_tokens(self):
        for token in self.tokens:
            db.session.delete(token)

    def verify_password(self, password):
        if self.role.name == 'System user':
            return False
        return check_password_hash(self.password_hash, password)

    def to_json(self, backrefs=False, relationships=False):
        _json = {
            'id': self.hashid,
            'confirmed': self.confirmed,
            'email': self.email,
            'last_seen': (
                None if self.last_seen is None
                else f'{self.last_seen.isoformat()}Z'
            ),
            'member_since': f'{self.member_since.isoformat()}Z',
            'username': self.username,
            'settings': {
                'dark_mode': self.setting_dark_mode,
                'job_status_mail_notification_level': \
                    self.setting_job_status_mail_notification_level.name
            }
        }
        if backrefs:
            _json['role'] = self.role.to_json(backrefs=True)
        if relationships:
            _json['corpora'] = {
                x.hashid: x.to_json(relationships=True)
                for x in self.corpora
            }
            _json['jobs'] = {
                x.hashid: x.to_json(relationships=True)
                for x in self.jobs
            }
            _json['tesseract_ocr_pipeline_models'] = {
                x.hashid: x.to_json(relationships=True)
                for x in self.tesseract_ocr_pipeline_models
            }
            _json['spacy_nlp_pipeline_models'] = {
                x.hashid: x.to_json(relationships=True)
                for x in self.spacy_nlp_pipeline_models
            }
        return _json

class TesseractOCRPipelineModel(FileMixin, HashidMixin, db.Model):
    __tablename__ = 'tesseract_ocr_pipeline_models'
    # Primary key
    id = db.Column(db.Integer, primary_key=True)
    # Foreign keys
    user_id = db.Column(db.Integer, db.ForeignKey('users.id'))
    # Fields
    title = db.Column(db.String(64))
    description = db.Column(db.String(255))
    version = db.Column(db.String(16))
    compatible_service_versions = db.Column(ContainerColumn(list, 255))
    publisher = db.Column(db.String(128))
    publisher_url = db.Column(db.String(512))
    publishing_url = db.Column(db.String(512))
    publishing_year = db.Column(db.Integer)
    shared = db.Column(db.Boolean, default=False)
    # Backrefs: user: User

    @property
    def path(self):
        return os.path.join(
            self.user.path,
            'tesseract_ocr_pipeline_models',
            str(self.id)
        )

    @staticmethod
    def insert_defaults():
        nopaque_user = User.query.filter_by(username='nopaque').first()
        defaults_file = os.path.join(
            os.path.dirname(os.path.abspath(__file__)),
            'TesseractOCRPipelineModel.defaults.yml'
        )
        with open(defaults_file, 'r') as f:
            defaults = yaml.safe_load(f)
        for m in defaults:
            model = TesseractOCRPipelineModel.query.filter_by(title=m['title'], version=m['version']).first()  # noqa
            if model is not None:
                model.compatible_service_versions = m['compatible_service_versions']
                model.description = m['description']
                model.publisher = m['publisher']
                model.publisher_url = m['publisher_url']
                model.publishing_url = m['publishing_url']
                model.publishing_year = m['publishing_year']
                model.shared = True
                model.title = m['title']
                model.version = m['version']
                continue
            model = TesseractOCRPipelineModel(
                compatible_service_versions=m['compatible_service_versions'],
                description=m['description'],
                publisher=m['publisher'],
                publisher_url=m['publisher_url'],
                publishing_url=m['publishing_url'],
                publishing_year=m['publishing_year'],
                shared=True,
                title=m['title'],
                user=nopaque_user,
                version=m['version']
            )
            db.session.add(model)
            db.session.flush(objects=[model])
            db.session.refresh(model)
            model.filename = f'{model.id}.traineddata'
            r = requests.get(m['url'], stream=True)
            pbar = tqdm(
                desc=f'{model.title} ({model.filename})',
                unit="B",
                unit_scale=True,
                unit_divisor=1024,
                total=int(r.headers['Content-Length'])
            )
            pbar.clear()
            with open(model.path, 'wb') as f:
                for chunk in r.iter_content(chunk_size=1024):
                    if chunk:  # filter out keep-alive new chunks
                        pbar.update(len(chunk))
                        f.write(chunk)
                pbar.close()
        db.session.commit()

    def delete(self):
        try:
            os.remove(self.path)
        except OSError as e:
            current_app.logger.error(e)
        db.session.delete(self)

    def to_json(self, backrefs=False, relationships=False):
        _json = {
            'id': self.hashid,
            'compatible_service_versions': self.compatible_service_versions,
            'description': self.description,
            'publisher': self.publisher,
            'publisher_url': self.publisher_url,
            'publishing_url': self.publishing_url,
            'publishing_year': self.publishing_year,
            'shared': self.shared,
            'title': self.title,
            'version': self.version,
            **self.file_mixin_to_json()
        }
        if backrefs:
            _json['user'] = self.user.to_json(backrefs=True)
        return _json


class SpaCyNLPPipelineModel(FileMixin, HashidMixin, db.Model):
    __tablename__ = 'spacy_nlp_pipeline_models'
    # Primary key
    id = db.Column(db.Integer, primary_key=True)
    # Foreign keys
    user_id = db.Column(db.Integer, db.ForeignKey('users.id'))
    # Fields
    title = db.Column(db.String(64))
    description = db.Column(db.String(255))
    version = db.Column(db.String(16))
    compatible_service_versions = db.Column(ContainerColumn(list, 255))
    publisher = db.Column(db.String(128))
    publisher_url = db.Column(db.String(512))
    publishing_url = db.Column(db.String(512))
    publishing_year = db.Column(db.Integer)
    pipeline_name = db.Column(db.String(64))
    shared = db.Column(db.Boolean, default=False)
    # Backrefs: user: User

    @property
    def path(self):
        return os.path.join(
            self.user.path,
            'spacy_nlp_pipeline_models',
            str(self.id)
        )

    @staticmethod
    def insert_defaults():
        nopaque_user = User.query.filter_by(username='nopaque').first()
        defaults_file = os.path.join(
            os.path.dirname(os.path.abspath(__file__)),
            'SpaCyNLPPipelineModel.defaults.yml'
        )
        with open(defaults_file, 'r') as f:
            defaults = yaml.safe_load(f)
        for m in defaults:
            model = SpaCyNLPPipelineModel.query.filter_by(title=m['title'], version=m['version']).first()  # noqa
            if model is not None:
                model.compatible_service_versions = m['compatible_service_versions']
                model.description = m['description']
                model.publisher = m['publisher']
                model.publisher_url = m['publisher_url']
                model.publishing_url = m['publishing_url']
                model.publishing_year = m['publishing_year']
                model.shared = True
                model.title = m['title']
                model.version = m['version']
                model.pipeline_name = m['pipeline_name']
                continue
            model = SpaCyNLPPipelineModel(
                compatible_service_versions=m['compatible_service_versions'],
                description=m['description'],
                publisher=m['publisher'],
                publisher_url=m['publisher_url'],
                publishing_url=m['publishing_url'],
                publishing_year=m['publishing_year'],
                shared=True,
                title=m['title'],
                user=nopaque_user,
                version=m['version'],
                pipeline_name=m['pipeline_name']
            )
            db.session.add(model)
            db.session.flush(objects=[model])
            db.session.refresh(model)
            model.filename = m['url'].split('/')[-1]
            r = requests.get(m['url'], stream=True)
            pbar = tqdm(
                desc=f'{model.title} ({model.filename})',
                unit="B",
                unit_scale=True,
                unit_divisor=1024,
                total=int(r.headers['Content-Length'])
            )
            pbar.clear()
            with open(model.path, 'wb') as f:
                for chunk in r.iter_content(chunk_size=1024):
                    if chunk:  # filter out keep-alive new chunks
                        pbar.update(len(chunk))
                        f.write(chunk)
                pbar.close()
        db.session.commit()
    
    def delete(self):
        try:
            os.remove(self.path)
        except OSError as e:
            current_app.logger.error(e)
        db.session.delete(self)

    def to_json(self, backrefs=False, relationships=False):
        _json = {
            'id': self.hashid,
            'compatible_service_versions': self.compatible_service_versions,
            'description': self.description,
            'publisher': self.publisher,
            'publisher_url': self.publisher_url,
            'publishing_url': self.publishing_url,
            'publishing_year': self.publishing_year,
            'pipeline_name': self.pipeline_name,
            'shared': self.shared,
            'title': self.title,
            'version': self.version,
            **self.file_mixin_to_json()
        }
        if backrefs:
            _json['user'] = self.user.to_json(backrefs=True)
        return _json


class JobInput(FileMixin, HashidMixin, db.Model):
    __tablename__ = 'job_inputs'
    # Primary key
    id = db.Column(db.Integer, primary_key=True)
    # Foreign keys
    job_id = db.Column(db.Integer, db.ForeignKey('jobs.id'))
    # Backrefs: job: Job

    def __repr__(self):
        return f'<JobInput {self.filename}>'

    @property
    def content_url(self):
        return url_for(
            'jobs.download_job_input',
            job_id=self.job.id,
            job_input_id=self.id
        )

    @property
    def jsonpatch_path(self):
        return f'{self.job.jsonpatch_path}/inputs/{self.hashid}'

    @property
    def path(self):
        return os.path.join(self.job.path, 'inputs', str(self.id))

    @property
    def url(self):
        return url_for(
            'jobs.job',
            job_id=self.job_id,
            _anchor=f'job-{self.job.hashid}-input-{self.hashid}'
        )

    @property
    def user_hashid(self):
        return self.job.user.hashid

    @property
    def user_id(self):
        return self.job.user_id

    def to_json(self, backrefs=False, relationships=False):
        _json = {
            'id': self.hashid,
            **self.file_mixin_to_json()
        }
        if backrefs:
            _json['job'] = self.job.to_json(backrefs=True)
        return _json


class JobResult(FileMixin, HashidMixin, db.Model):
    __tablename__ = 'job_results'
    # Primary key
    id = db.Column(db.Integer, primary_key=True)
    # Foreign keys
    job_id = db.Column(db.Integer, db.ForeignKey('jobs.id'))
    # Fields
    description = db.Column(db.String(255))
    # Backrefs: job: Job

    def __repr__(self):
        return f'<JobResult {self.filename}>'

    @property
    def download_url(self):
        return url_for(
            'jobs.download_job_result',
            job_id=self.job_id,
            job_result_id=self.id
        )

    @property
    def jsonpatch_path(self):
        return f'{self.job.jsonpatch_path}/results/{self.hashid}'

    @property
    def path(self):
        return os.path.join(self.job.path, 'results', str(self.id))

    @property
    def url(self):
        return url_for(
            'jobs.job',
            job_id=self.job_id,
            _anchor=f'job-{self.job.hashid}-result-{self.hashid}'
        )

    @property
    def user_hashid(self):
        return self.job.user.hashid

    @property
    def user_id(self):
        return self.job.user_id

    def to_json(self, backrefs=False, relationships=False):
        _json = {
            'id': self.hashid,
            'description': self.description,
            **self.file_mixin_to_json(
                backrefs=backrefs,
                relationships=relationships
            )
        }
        if backrefs:
            _json['job'] = self.job.to_json(backrefs=True)
        return _json


class Job(HashidMixin, db.Model):
    '''
    Class to define Jobs.
    '''
    __tablename__ = 'jobs'
    # Primary key
    id = db.Column(db.Integer, primary_key=True)
    # Foreign keys
    user_id = db.Column(db.Integer, db.ForeignKey('users.id'))
    # Fields
    creation_date = \
        db.Column(db.DateTime(), default=datetime.utcnow)
    description = db.Column(db.String(255))
    end_date = db.Column(db.DateTime())
    service = db.Column(db.String(64))
    service_args = db.Column(ContainerColumn(dict, 255))
    service_version = db.Column(db.String(16))
    status = db.Column(
        IntEnumColumn(JobStatus),
        default=JobStatus.INITIALIZING
    )
    title = db.Column(db.String(32))
    # Backrefs: user: User
    # Relationships
    inputs = db.relationship(
        'JobInput',
        backref='job',
        cascade='all, delete-orphan',
        lazy='dynamic'
    )
    results = db.relationship(
        'JobResult',
        backref='job',
        cascade='all, delete-orphan',
        lazy='dynamic'
    )

    def __repr__(self):
        return f'<Job {self.title}>'

    @property
    def jsonpatch_path(self):
        return f'{self.user.jsonpatch_path}/jobs/{self.hashid}'

    @property
    def path(self):
        return os.path.join(self.user.path, 'jobs', str(self.id))

    @property
    def url(self):
        return url_for('jobs.job', job_id=self.id)

    @property
    def user_hashid(self):
        return self.user.hashid

    @staticmethod
    def create(**kwargs):
        job = Job(**kwargs)
        db.session.add(job)
        db.session.flush(objects=[job])
        db.session.refresh(job)
        try:
            os.mkdir(job.path)
            os.mkdir(os.path.join(job.path, 'inputs'))
            os.mkdir(os.path.join(job.path, 'pipeline_data'))
            os.mkdir(os.path.join(job.path, 'results'))
        except OSError as e:
            current_app.logger.error(e)
            db.session.rollback()
            raise e
        return job


    def delete(self):
        ''' Delete the job and its inputs and results from the database. '''
        if self.status not in [JobStatus.COMPLETED, JobStatus.FAILED]:  # noqa
            self.status = JobStatus.CANCELING
            db.session.commit()
            while self.status != JobStatus.CANCELED:
                # In case the daemon handled a job in any way
                if self.status != JobStatus.CANCELING:
                    self.status = JobStatus.CANCELING
                    db.session.commit()
                sleep(1)
                db.session.refresh(self)
        try:
            shutil.rmtree(self.path)
        except OSError as e:
            current_app.logger.error(e)
            db.session.rollback()
            raise e
        db.session.delete(self)

    def restart(self):
        ''' Restart a job - only if the status is failed '''
        if self.status != JobStatus.FAILED:
            raise Exception('Job status is not "failed"')
        shutil.rmtree(os.path.join(self.path, 'results'), ignore_errors=True)
        shutil.rmtree(os.path.join(self.path, 'pyflow.data'), ignore_errors=True)
        for result in self.results:
            db.session.delete(result)
        self.end_date = None
        self.status = JobStatus.SUBMITTED

    def to_json(self, backrefs=False, relationships=False):
        _json = {
            'id': self.hashid,
            'creation_date': f'{self.creation_date.isoformat()}Z',
            'description': self.description,
            'end_date': (
                None if self.end_date is None
                else f'{self.end_date.isoformat()}Z'
            ),
            'service': self.service,
            'service_args': self.service_args,
            'service_version': self.service_version,
            'status': self.status.name,
            'title': self.title,
            'url': self.url
        }
        if backrefs:
            _json['user'] = self.user.to_json(backrefs=True)
        if relationships:
            _json['inputs'] = {
                x.hashid: x.to_json(relationships=True)
                for x in self.inputs
            }
            _json['results'] = {
                x.hashid: x.to_json(relationships=True)
                for x in self.results
            }
        return _json


class CorpusFile(FileMixin, HashidMixin, db.Model):
    __tablename__ = 'corpus_files'
    # Primary key
    id = db.Column(db.Integer, primary_key=True)
    # Foreign keys
    corpus_id = db.Column(db.Integer, db.ForeignKey('corpora.id'))
    # Fields
    author = db.Column(db.String(255))
    publishing_year = db.Column(db.Integer)
    title = db.Column(db.String(255))
    address = db.Column(db.String(255))
    booktitle = db.Column(db.String(255))
    chapter = db.Column(db.String(255))
    editor = db.Column(db.String(255))
    institution = db.Column(db.String(255))
    journal = db.Column(db.String(255))
    pages = db.Column(db.String(255))
    publisher = db.Column(db.String(255))
    school = db.Column(db.String(255))
    # Backrefs: corpus: Corpus

    @property
    def download_url(self):
        return url_for(
            'corpora.download_corpus_file',
            corpus_id=self.corpus_id,
            corpus_file_id=self.id
        )

    @property
    def jsonpatch_path(self):
        return f'{self.corpus.jsonpatch_path}/files/{self.hashid}'

    @property
    def path(self):
        return os.path.join(self.corpus.path, 'files', str(self.id))

    @property
    def url(self):
        return url_for(
            'corpora.corpus_file',
            corpus_id=self.corpus_id,
            corpus_file_id=self.id
        )

    @property
    def user_hashid(self):
        return self.corpus.user.hashid

    @property
    def user_id(self):
        return self.corpus.user_id

    def delete(self):
        try:
            os.remove(self.path)
        except OSError as e:
            current_app.logger.error(e)
        db.session.delete(self)
        self.corpus.status = CorpusStatus.UNPREPARED

    def to_json(self, backrefs=False, relationships=False):
        _json = {
            'id': self.hashid,
            'url': self.url,
            'address': self.address,
            'author': self.author,
            'booktitle': self.booktitle,
            'chapter': self.chapter,
            'editor': self.editor,
            'institution': self.institution,
            'journal': self.journal,
            'pages': self.pages,
            'publisher': self.publisher,
            'publishing_year': self.publishing_year,
            'school': self.school,
            'title': self.title,
            **self.file_mixin_to_json(
                backrefs=backrefs,
                relationships=relationships
            )
        }
        if backrefs:
            _json['corpus'] = self.corpus.to_json(backrefs=True)
        return _json

class Corpus(HashidMixin, db.Model):
    '''
    Class to define a corpus.
    '''
    __tablename__ = 'corpora'
    # Primary key
    id = db.Column(db.Integer, primary_key=True)
    # Foreign keys
    user_id = db.Column(db.Integer, db.ForeignKey('users.id'))
    # Fields
    creation_date = db.Column(db.DateTime(), default=datetime.utcnow)
    description = db.Column(db.String(255))
    last_edited_date = db.Column(db.DateTime())
    status = db.Column(
        IntEnumColumn(CorpusStatus),
        default=CorpusStatus.UNPREPARED
    )
    title = db.Column(db.String(32))
    num_analysis_sessions = db.Column(db.Integer, default=0)
    num_tokens = db.Column(db.Integer, default=0)
    is_public = db.Column(db.Boolean, default=False)
    # Backrefs: user: User
    # Relationships
    files = db.relationship(
        'CorpusFile',
        backref='corpus',
        lazy='dynamic',
        cascade='all, delete-orphan'
    )
    # "static" attributes
    max_num_tokens = 2_147_483_647

    def __repr__(self):
        return f'<Corpus {self.title}>'

    @property
    def analysis_url(self):
        return url_for('corpora.analyse_corpus', corpus_id=self.id)

    @property
    def jsonpatch_path(self):
        return f'{self.user.jsonpatch_path}/corpora/{self.hashid}'

    @property
    def path(self):
        return os.path.join(self.user.path, 'corpora', str(self.id))

    @property
    def url(self):
        return url_for('corpora.corpus', corpus_id=self.id)

    @property
    def user_hashid(self):
        return self.user.hashid

    @staticmethod
    def create(**kwargs):
        corpus = Corpus(**kwargs)
        db.session.add(corpus)
        db.session.flush(objects=[corpus])
        db.session.refresh(corpus)
        try:
            os.mkdir(corpus.path)
            os.mkdir(os.path.join(corpus.path, 'files'))
            os.mkdir(os.path.join(corpus.path, 'cwb'))
            os.mkdir(os.path.join(corpus.path, 'cwb', 'data'))
            os.mkdir(os.path.join(corpus.path, 'cwb', 'registry'))
        except OSError as e:
            current_app.logger.error(e)
            db.session.rollback()
            raise e
        return corpus

    def build(self):
        corpus_element = ET.fromstring('<corpus>\n</corpus>')
        for corpus_file in self.files:
            normalized_vrt_path = os.path.join(self.path, 'cwb', f'{corpus_file.id}.norm.vrt')
            try:
                normalize_vrt_file(corpus_file.path, normalized_vrt_path)
            except:
                self.status = CorpusStatus.FAILED
                return
            element_tree = ET.parse(normalized_vrt_path)
            text_element = element_tree.getroot()
            text_element.set('author', corpus_file.author)
            text_element.set('title', corpus_file.title)
            text_element.set(
                'publishing_year',
                f'{corpus_file.publishing_year}'
            )
            text_element.set('address', corpus_file.address or 'NULL')
            text_element.set('booktitle', corpus_file.booktitle or 'NULL')
            text_element.set('chapter', corpus_file.chapter or 'NULL')
            text_element.set('editor', corpus_file.editor or 'NULL')
            text_element.set('institution', corpus_file.institution or 'NULL')
            text_element.set('journal', corpus_file.journal or 'NULL')
            text_element.set('pages', f'{corpus_file.pages}' or 'NULL')
            text_element.set('publisher', corpus_file.publisher or 'NULL')
            text_element.set('school', corpus_file.school or 'NULL')
            text_element.tail = '\n'
            # corpus_element.insert(1, text_element)
            corpus_element.append(text_element)
        ET.ElementTree(corpus_element).write(
            os.path.join(self.path, 'cwb', 'corpus.vrt'),
            encoding='utf-8'
        )
        self.last_edited_date = datetime.utcnow()
        self.status = CorpusStatus.SUBMITTED

    def delete(self):
        shutil.rmtree(self.path, ignore_errors=True)
        db.session.delete(self)

    def to_json(self, backrefs=False, relationships=False):
        _json = {
            'id': self.hashid,
            'creation_date': f'{self.creation_date.isoformat()}Z',
            'description': self.description,
            'max_num_tokens': self.max_num_tokens,
            'num_analysis_sessions': self.num_analysis_sessions,
            'num_tokens': self.num_tokens,
            'status': self.status.name,
            'last_edited_date': (
                None if self.last_edited_date is None
                else f'{self.last_edited_date.isoformat()}Z'
            ),
            'title': self.title,
            'is_public': self.is_public
        }
        if backrefs:
            _json['user'] = self.user.to_json(backrefs=True)
        if relationships:
            _json['files'] = {
                x.hashid: x.to_json(relationships=True)
                for x in self.files
            }
        return _json
# endregion models


##############################################################################
# event_handlers                                                             #
##############################################################################
# region event_handlers


@db.event.listens_for(Corpus, 'after_delete')
@db.event.listens_for(CorpusFile, 'after_delete')
@db.event.listens_for(Job, 'after_delete')
@db.event.listens_for(JobInput, 'after_delete')
@db.event.listens_for(JobResult, 'after_delete')
def ressource_after_delete(mapper, connection, ressource):
    jsonpatch = [{'op': 'remove', 'path': ressource.jsonpatch_path}]
    room = f'users.{ressource.user_hashid}'
    socketio.emit('users.patch', jsonpatch, room=room)
    room = f'/users/{ressource.user_hashid}'
    socketio.emit('PATCH', jsonpatch, room=room)


@db.event.listens_for(Corpus, 'after_insert')
@db.event.listens_for(CorpusFile, 'after_insert')
@db.event.listens_for(Job, 'after_insert')
@db.event.listens_for(JobInput, 'after_insert')
@db.event.listens_for(JobResult, 'after_insert')
def ressource_after_insert_handler(mapper, connection, ressource):
    value = ressource.to_json()
    for attr in mapper.relationships:
        value[attr.key] = {}
    jsonpatch = [
        {'op': 'add', 'path': ressource.jsonpatch_path, 'value': value}
    ]
    room = f'/users/{ressource.user_hashid}'
    socketio.emit('PATCH', jsonpatch, room=room)


@db.event.listens_for(Corpus, 'after_update')
@db.event.listens_for(CorpusFile, 'after_update')
@db.event.listens_for(Job, 'after_update')
@db.event.listens_for(JobInput, 'after_update')
@db.event.listens_for(JobResult, 'after_update')
def ressource_after_update_handler(mapper, connection, ressource):
    jsonpatch = []
    for attr in db.inspect(ressource).attrs:
        if attr.key in mapper.relationships:
            continue
        if not attr.load_history().has_changes():
            continue
        if isinstance(attr.value, datetime):
            value = f'{attr.value.isoformat()}Z'
        elif isinstance(attr.value, Enum):
            value = attr.value.name
        else:
            value = attr.value
        jsonpatch.append(
            {
                'op': 'replace',
                'path': f'{ressource.jsonpatch_path}/{attr.key}',
                'value': value
            }
        )
    if jsonpatch:
        room = f'/users/{ressource.user_hashid}'
        socketio.emit('PATCH', jsonpatch, room=room)


@db.event.listens_for(Job, 'after_update')
def job_after_update_handler(mapper, connection, job):
    for attr in db.inspect(job).attrs:
        if attr.key != 'status':
            continue
        if not attr.load_history().has_changes():
            return
        if job.user.setting_job_status_mail_notification_level == UserSettingJobStatusMailNotificationLevel.NONE:
            return
        if job.user.setting_job_status_mail_notification_level == UserSettingJobStatusMailNotificationLevel.END:
            if job.status not in [JobStatus.COMPLETED, JobStatus.FAILED]:
                return
        msg = create_message(
            job.user.email,
            f'Status update for your Job "{job.title}"',
            'tasks/email/notification',
            job=job
        )
        mail.send(msg)
# endregion event_handlers


##############################################################################
# misc                                                                       #
##############################################################################
# region misc
@login.user_loader
def load_user(user_id):
    return User.query.get(int(user_id))
# endregion misc