django saas多租户 分库分表 一个租户一个数据库

  • 企业(租户)应用,一个企业算一个租户
|----其他应用
|----organize
     |----management    自定义迁移命令
          |----commands
                |-----__inint__.py
                |----clearmigrations.py       清除所有迁移
                |----organize_migrate.py    企业数据库迁移
     |----patch
           |----__init__.py
           |----connection.py      添加企业数据库连接setting.DATABASES 
           |----contenttype.py     重构contenttype
           |----permission.py      重构权限
           |----request.py           请求接口根据用户企业id设置数据库连接
     |----app.py     应用初始化
     |----signal.py    信号,创建企业的时候触发创建数据库
     |----其他文件

多数据流程:
-、以企业id小写创建数据库
-、创建企业的时候通过信号触发create_database创建企业数据库
-、应用数据迁移由DATABASE_APPS_MAPPING决定哪个对应的应用
-、__connection_handler__getitem__会把不存在DATABASES的数据库添加的连接配置里
-、数据库读写由DBRouter路由
-、请求接口会根据request.user的organize id设置对应的数据库
-、设置数据set_current_db
-、获取数据库get_current_db,默认是default数据库
-、企业获取模型所有数据get_model_all_data
-、创建数据会根据新数据带的organize  id指定数据库
-、获取数据api_view.get_queryset(自定义视图)会查询对应的数据库
-、数据库迁移命令:
   迁移所有企业(不包括default):python manage.py organize_migrations --all
   迁移xx企业(xx=>企业id小写):python manage.py organize_migrations --database=xxx
  • setting.py 配置
INSTALLED_APPS = [
    # 共用应用
    ‘organize’,
     ...
    # 租户应用
    'organizeapp'
    ...
]

"""
定制化应用,给租户专有的应用指定数据库,租户共有的应用不需要指定,
不属于租户的应用都要指定到默认default数据库
"""
DATABASE_APPS_MAPPING = {
  'organize': 'default'
}

# 数据库
DATABASES = {
    'default': {
        'ENGINE': 'django.db.backends.mysql',
        'NAME': 'databasename',
        'USER': 'user',
        'PASSWORD': 'password',
        'HOST': '127.0.0.1',
        'PORT': '3306'
    }
}

# 数据库路由
DATABASE_ROUTERS = ['utils.api_db.DBRouter']
  • api_db.py 数据库处理
import copy
from django.conf import settings
from django.db.models import Model, Q
from threading import local, Thread
from django.db.utils import load_backend
from django.apps import apps as django_apps

_thread_local = local()


def create_database(org_id: str, org_name: str):
    """创建企业数据库"""
    # 创建连接
    db_config = get_default_db_config(None)
    backend = load_backend(db_config["ENGINE"])
    connection = backend.DatabaseWrapper(db_config, org_id)
    # 获取游标,执行sql语句,用企业id(自定义生成id)作为数据库名称
    sql = f"CREATE DATABASE IF NOT EXISTS {org_id} character set utf8;"
    with connection.cursor() as cursor:
        cursor.execute(sql)

    # 数据迁移
    thread = Thread(target=migrate, args=[org_id, org_name])
    thread.start()
    # execute_from_command_line(['manage.py', 'migrate', f'--database={self.id}'])
    # logger.info('migrate successfuly!')

    return True


def migrate(org_id: str, org_name: str):
    """数据迁移"""
    try:
        from django.core.management import execute_from_command_line
    except ImportError as exc:
        raise ImportError(
            "Couldn't import Django. Are you sure it's installed and "
            "available on your PYTHONPATH environment variable? Did you "
            "forget to activate a virtual environment?"
        ) from exc
    execute_from_command_line(['manage.py', 'migrate', f'--database={org_id}'])


def get_model_all_data(queryset=None, filter_conditions=None, view=None, organize=''):
    """
    获取模型的所有数据
    分两种情况:
    1:视图(接口)获取数据 传入view
    2: 非视图获取数据 传入queryset,和过滤条件filter_conditions
    所有数据按创建时间和id降序后返回
    如果指定了企业(id),则直接查询该企业的数据
    """

    if organize != '':
        res_data = get_db_data(queryset=queryset, filter_conditions=filter_conditions, view=view, organize=organize)
    else:
        Organize = get_model('organize.Organize')
        org_ids = Organize.objects.values_list('id', flat=True)
        res_data = list()
        for pk in org_ids:
            org_data = get_db_data(queryset=queryset, filter_conditions=filter_conditions, view=view, organize=pk)
            res_data.extend(org_data)

    # 按创建时间和id降序,如果创建时间相同,则按id降序
    res_data.sort(key=lambda x: (x['create_time'], x['id']), reverse=True)

    return res_data

def get_db_data(queryset=None, filter_conditions=None, view=None, organize=None):
    """
    获取数据库数据
    分两种情况:
    1:视图(接口)获取数据 传入view
    2: 非视图获取数据 传入queryset,和过滤条件filter_conditions
    所有数据按创建时间和id降序后返回
    """
    # 设置数据库
    db_name = organize.lower()
    set_current_db(db_name)
    # 过滤条件
    conditions = []
    if filter_conditions:
        conditions = [Q(**item) for item in filter_conditions]

    if view is not None:
        # 视图的数据需要过滤和序列化
        org_queryset = view.filter_queryset(view.get_queryset()) \
            .order_by('-create_time', '-id')
        serializer = view.get_serializer(instance=org_queryset, many=True)
        org_data = serializer.data
    else:
        # 非视图数据按条件过滤
        org_data = queryset.filter(*conditions) \
            .order_by('-create_time', '-id').values()
        org_data = list(org_data)

    return org_data


def get_current_db():
    return getattr(_thread_local, 'db_name', 'default')


def set_current_db(db_name):
    setattr(_thread_local, 'db_name', db_name.lower())


def get_default_db_config(db=None):
    """获取默认数据库配置"""
    db_config = copy.deepcopy(settings.DATABASES.get('default', None))
    if db_config is None:
        raise KeyError('获取不到数据库配置')
    db_config['NAME'] = None if db is None else db.lower()

    return db_config


class DBRouter:
    """数据库读写"""

    def db_for_read(self, model: Model, **hints) -> str:
        if model._meta.app_label in settings.DATABASE_APPS_MAPPING:
            return settings.DATABASE_APPS_MAPPING[model._meta.app_label]

        return get_current_db()

    def db_for_write(self, model: Model, **hints):
        if model._meta.app_label in settings.DATABASE_APPS_MAPPING:
            return settings.DATABASE_APPS_MAPPING[model._meta.app_label]

        return get_current_db()

    def allow_migrate(self, db: str, app_label: str, **hints) -> bool:
        """平台只迁移default数据库的app,企业只迁移租户(非default)的app"""
        if app_label == 'contenttypes':
            return True
        app_db = settings.DATABASE_APPS_MAPPING.get(app_label, None)
        if app_db == 'default' and db == 'default':
            return True
        elif app_db != 'default' and db != 'default':
            return True
        else:
            return False

def get_custom_id(model, pre_str, num):
    """
    获取自定义id 先从redis获取最后一个id,如果是None则从数据库获取
    查询结果会去重,必须保证所有数据的id都不一样
    :param model: 模型
    :param pre_str: id的前缀
    :param num: id预设长度
    :return: str 自定义id
    """
    redis_con = get_redis_connection("default")
    last_id = redis_con.hget('E:ADMIN:LASTID', model._meta.app_label)

    if last_id is None:
        try:
            last_data = model.objects.using('default').order_by('create_time', 'id').last()
            last_id = last_data.id
        except:
            last_data = get_model_all_data(model.objects.all())
            last_id = last_data[0]['id']

        if last_id is None or not last_id.startswith(pre_str):
            # 第一个以EID开头的id
            last_id = 0
        else:
            # 在已有的id上增加
            last_id = int(re.sub(r'\D+0*', '', last_id))
    else:
        last_id = int(last_id)

    new_id = str(last_id + 1)
    redis_con.hset('E:ADMIN:LASTID', model._meta.app_label, new_id)
    # 补零个数
    zero_num = 0 if len(new_id) > num else num - len(new_id)
    return pre_str + '0' * zero_num + new_id
  • clearmigrations.py 清除所有迁移
import os
import shutil
from django.core.management.base import BaseCommand
from django.apps import apps


class Command(BaseCommand):
    help = "Clear all migration files in project."

    def get_apps(self):
        for app in apps.get_app_configs():
            # path = os.path.join(
            #     settings.BASE_DIR, app.name.replace(".", "/"), "migrations"
            # )
            path = os.path.join(app.path, "migrations")
            if os.path.exists(path):
                yield app, path

    def handle(self, *args, **options):
        for app, path in self.get_apps():
            # 递归删除所有文件及其路径(删除整个文件夹)
            shutil.rmtree(path)
            # 创建文件夹,并生成__init__.py文件
            os.makedirs(path)
            with open(os.path.join(path, "__init__.py"), "w+") as file:
                file.write("")
            self.stdout.write(self.style.SUCCESS(f"Clear {path}"))

        self.stdout.write(self.style.SUCCESS("Successfully cleared!"))
  • organize_migrate.py 企业数据库迁移
import sys
from importlib import import_module

from django.apps import apps
from django.core.management.base import CommandError, no_translations
from django.core.management.sql import (
    emit_post_migrate_signal, emit_pre_migrate_signal,
)
from django.db import connections
from django.db.migrations.autodetector import MigrationAutodetector
from django.db.migrations.executor import MigrationExecutor
from django.db.migrations.loader import AmbiguityError
from django.db.migrations.state import ModelState, ProjectState
from django.utils.module_loading import module_has_submodule

from django.core.management.commands.migrate import Command as MigrateCommand


class Command(MigrateCommand):
    """
    迁移企业数据库
    api_db.DBRouter.allow_migrate会将模型迁移到对应的数据中去
    迁移xxx企业:python manage.py organize_migrate --database xxx
    迁移所有企业:python manage.py organize_migrate --all
    """
    help = "organize migrate database."
    requires_system_checks = []

    def add_arguments(self, parser):
        parser.add_argument(
            '--all', action='store_true',
            help='migrate all database.',
        )
        super().add_arguments(parser)

    @no_translations
    def handle(self, *args, **options):
        if options['all']:
            # 迁移所有企业
            from organize import get_all_organize_db
            dbs = list(get_all_organize_db().keys())
            for key in dbs:
                connections[key]
                self.execute_migrate(key, args, options)
        else:
            # 迁移xxx企业
            cur_database = options['database']
            if not cur_database in connections:
                connections[cur_database]
            self.execute_migrate(cur_database, args, options)

    def execute_migrate(self, cur_database, args, options):
        if not options['skip_checks']:
            self.check(databases=[cur_database])

        self.verbosity = options['verbosity']
        self.interactive = options['interactive']

        # Import the 'management' module within each installed app, to register
        # dispatcher events.
        for app_config in apps.get_app_configs():
            if module_has_submodule(app_config.module, "management"):
                import_module('.management', app_config.name)

        # Get the database we're operating from
        self.stdout.write(f'\nMigrate Database: {cur_database}')
        connection = connections[cur_database]
        # Hook for backends needing any database preparation
        connection.prepare_database()
        # Work out which apps have migrations and which do not
        executor = MigrationExecutor(connection, self.migration_progress_callback)

        # Raise an error if any migrations are applied before their dependencies.
        executor.loader.check_consistent_history(connection)

        # Before anything else, see if there's conflicting apps and drop out
        # hard if there are any
        conflicts = executor.loader.detect_conflicts()
        if conflicts:
            name_str = "; ".join(
                "%s in %s" % (", ".join(names), app)
                for app, names in conflicts.items()
            )
            raise CommandError(
                "Conflicting migrations detected; multiple leaf nodes in the "
                "migration graph: (%s).\nTo fix them run "
                "'python manage.py makemigrations --merge'" % name_str
            )

        # If they supplied command line arguments, work out what they mean.
        run_syncdb = options['run_syncdb']
        target_app_labels_only = True
        if options['app_label']:
            # Validate app_label.
            app_label = options['app_label']
            try:
                apps.get_app_config(app_label)
            except LookupError as err:
                raise CommandError(str(err))
            if run_syncdb:
                if app_label in executor.loader.migrated_apps:
                    raise CommandError("Can't use run_syncdb with app '%s' as it has migrations." % app_label)
            elif app_label not in executor.loader.migrated_apps:
                raise CommandError("App '%s' does not have migrations." % app_label)

        if options['app_label'] and options['migration_name']:
            migration_name = options['migration_name']
            if migration_name == "zero":
                targets = [(app_label, None)]
            else:
                try:
                    migration = executor.loader.get_migration_by_prefix(app_label, migration_name)
                except AmbiguityError:
                    raise CommandError(
                        "More than one migration matches '%s' in app '%s'. "
                        "Please be more specific." %
                        (migration_name, app_label)
                    )
                except KeyError:
                    raise CommandError("Cannot find a migration matching '%s' from app '%s'." % (
                        migration_name, app_label))
                targets = [(app_label, migration.name)]
            target_app_labels_only = False
        elif options['app_label']:
            targets = [key for key in executor.loader.graph.leaf_nodes() if key[0] == app_label]
        else:
            targets = executor.loader.graph.leaf_nodes()

        plan = executor.migration_plan(targets)
        exit_dry = plan and options['check_unapplied']

        if options['plan']:
            self.stdout.write('Planned operations:', self.style.MIGRATE_LABEL)
            if not plan:
                self.stdout.write('  No planned migration operations.')
            for migration, backwards in plan:
                self.stdout.write(str(migration), self.style.MIGRATE_HEADING)
                for operation in migration.operations:
                    message, is_error = self.describe_operation(operation, backwards)
                    style = self.style.WARNING if is_error else None
                    self.stdout.write('    ' + message, style)
            if exit_dry:
                sys.exit(1)
            return
        if exit_dry:
            sys.exit(1)

        # At this point, ignore run_syncdb if there aren't any apps to sync.
        run_syncdb = options['run_syncdb'] and executor.loader.unmigrated_apps
        # Print some useful info
        if self.verbosity >= 1:
            self.stdout.write(self.style.MIGRATE_HEADING("Operations to perform:"))
            if run_syncdb:
                if options['app_label']:
                    self.stdout.write(
                        self.style.MIGRATE_LABEL("  Synchronize unmigrated app: %s" % app_label)
                    )
                else:
                    self.stdout.write(
                        self.style.MIGRATE_LABEL("  Synchronize unmigrated apps: ") +
                        (", ".join(sorted(executor.loader.unmigrated_apps)))
                    )
            if target_app_labels_only:
                self.stdout.write(
                    self.style.MIGRATE_LABEL("  Apply all migrations: ") +
                    (", ".join(sorted({a for a, n in targets})) or "(none)")
                )
            else:
                if targets[0][1] is None:
                    self.stdout.write(
                        self.style.MIGRATE_LABEL('  Unapply all migrations: ') +
                        str(targets[0][0])
                    )
                else:
                    self.stdout.write(self.style.MIGRATE_LABEL(
                        "  Target specific migration: ") + "%s, from %s"
                                      % (targets[0][1], targets[0][0])
                                      )

        pre_migrate_state = executor._create_project_state(with_applied_migrations=True)
        pre_migrate_apps = pre_migrate_state.apps
        emit_pre_migrate_signal(
            self.verbosity, self.interactive, connection.alias, apps=pre_migrate_apps, plan=plan,
        )

        # Run the syncdb phase.
        if run_syncdb:
            if self.verbosity >= 1:
                self.stdout.write(self.style.MIGRATE_HEADING("Synchronizing apps without migrations:"))
            if options['app_label']:
                self.sync_apps(connection, [app_label])
            else:
                self.sync_apps(connection, executor.loader.unmigrated_apps)

        # Migrate!
        if self.verbosity >= 1:
            self.stdout.write(self.style.MIGRATE_HEADING("Running migrations:"))
        if not plan:
            if self.verbosity >= 1:
                self.stdout.write("  No migrations to apply.")
                # If there's changes that aren't in migrations yet, tell them how to fix it.
                autodetector = MigrationAutodetector(
                    executor.loader.project_state(),
                    ProjectState.from_apps(apps),
                )
                changes = autodetector.changes(graph=executor.loader.graph)
                if changes:
                    self.stdout.write(self.style.NOTICE(
                        "  Your models in app(s): %s have changes that are not "
                        "yet reflected in a migration, and so won't be "
                        "applied." % ", ".join(repr(app) for app in sorted(changes))
                    ))
                    self.stdout.write(self.style.NOTICE(
                        "  Run 'manage.py makemigrations' to make new "
                        "migrations, and then re-run 'manage.py migrate' to "
                        "apply them."
                    ))
            fake = False
            fake_initial = False
        else:
            fake = options['fake']
            fake_initial = options['fake_initial']
        post_migrate_state = executor.migrate(
            targets, plan=plan, state=pre_migrate_state.clone(), fake=fake,
            fake_initial=fake_initial,
        )
        # post_migrate signals have access to all models. Ensure that all models
        # are reloaded in case any are delayed.
        post_migrate_state.clear_delayed_apps_cache()
        post_migrate_apps = post_migrate_state.apps

        # Re-render models of real apps to include relationships now that
        # we've got a final state. This wouldn't be necessary if real apps
        # models were rendered with relationships in the first place.
        with post_migrate_apps.bulk_update():
            model_keys = []
            for model_state in post_migrate_apps.real_models:
                model_key = model_state.app_label, model_state.name_lower
                model_keys.append(model_key)
                post_migrate_apps.unregister_model(*model_key)
        post_migrate_apps.render_multiple([
            ModelState.from_model(apps.get_model(*model)) for model in model_keys
        ])

        # Send the post_migrate signal, so individual apps can do whatever they need
        # to do at this point.
        emit_post_migrate_signal(
            self.verbosity, self.interactive, connection.alias, apps=post_migrate_apps, plan=plan,
        )

  • connection.py 添加企业数据库连接setting.DATABASES
import logging
from django.db.utils import ConnectionHandler

from utils.api_db import get_default_db_config

logger = logging.getLogger('django')


def __connection_handler__getitem__(self, alias: str) -> ConnectionHandler:
    if isinstance(alias, str):
        try:
            return getattr(self._connections, alias)
        except AttributeError:
            if alias not in self.settings:
                organize_db = get_default_db_config(alias)
                if organize_db:
                    # 添加没有配置的企业数据库
                    self.settings[alias] = organize_db
                else:
                    logger.error(f"The connection '{alias}' doesn't exist.")
                    raise self.exception_class(f"The connection '{alias}' doesn't exist.")
        conn = self.create_connection(alias)
        setattr(self._connections, alias, conn)
        return conn

    else:
        logger.error(f'The  connection alias [{alias}] must be string')
        raise Exception(f'The  connection alias [{alias}] must be string')


ConnectionHandler.__getitem__ = __connection_handler__getitem__

  • contenttype.py 重构contenttype
from django.apps import apps as global_apps
from django.contrib.contenttypes import management
from django.contrib.contenttypes.management import get_contenttypes_and_models
from django.db import DEFAULT_DB_ALIAS

from organize import get_common_apps


def create_contenttypes(
    app_config,
    verbosity=2,
    interactive=True,
    using=DEFAULT_DB_ALIAS,
    apps=global_apps,
    **kwargs
):
    """
    Create content types for models in the given app.
    """
    if not app_config.models_module:
        return

    app_label = app_config.label
    try:
        app_config = apps.get_app_config(app_label)
        ContentType = apps.get_model('contenttypes', 'ContentType')
    except LookupError:
        return

    common_applist = get_common_apps()
    if app_config.name in common_applist:
        return None

    content_types, app_models = get_contenttypes_and_models(app_config, using, ContentType)
    if not app_models:
        return

    cts = [
        ContentType(
            app_label=app_label,
            model=model_name,
        )
        for (model_name, model) in app_models.items()
        if model_name not in content_types
    ]
    ContentType.objects.using(using).bulk_create(cts)
    if verbosity >= 2:
        for ct in cts:
            print("Adding content type '%s | %s'" % (ct.app_label, ct.model))


management.create_contenttypes = create_contenttypes

  • permission.py 重构权限
"""
Creates permissions for all installed apps that need permissions.
"""

from django.apps import apps as global_apps
from django.contrib.contenttypes.management import create_contenttypes
from django.db import DEFAULT_DB_ALIAS, router
from django.contrib.auth.models import Permission
from django.contrib.auth.management import _get_all_permissions
from django.contrib.auth import management
from django.contrib.auth import backends

from organize import get_common_apps


def create_permissions(app_config, verbosity=2, interactive=True, using=DEFAULT_DB_ALIAS, apps=global_apps, **kwargs):
    if not app_config.models_module:
        return

    create_contenttypes(app_config, verbosity=verbosity, interactive=interactive, using=using, apps=apps, **kwargs)

    app_label = app_config.label
    try:
        app_config = apps.get_app_config(app_label)
        ContentType = apps.get_model('contenttypes', 'ContentType')
        Permission = apps.get_model('auth', 'Permission')
    except LookupError:
        return

    common_applist = get_common_apps()
    if app_config.name in common_applist:
        return

    if not router.allow_migrate_model(using, Permission):
        return

    # This will hold the permissions we're looking for as
    # (content_type, (codename, name))
    searched_perms = []
    # The codenames and ctypes that should exist.
    ctypes = set()

    for klass in app_config.get_models():
        # Force looking up the content types in the current database
        # before creating foreign keys to them.

        ctype = ContentType.objects.db_manager(using).get_for_model(klass, for_concrete_model=False)

        ctypes.add(ctype)
        for perm in _get_all_permissions(klass._meta):
            searched_perms.append((ctype, perm))

    # Find all the Permissions that have a content_type for a model we're
    # looking for.  We don't need to check for codenames since we already have
    # a list of the ones we're going to create.
    all_perms = set(Permission.objects.using(using).filter(
        content_type__in=ctypes,
    ).values_list(
        "content_type", "codename"
    ))

    perms = [
        Permission(codename=codename, name=name, content_type_id=ct.id)
        for ct, (codename, name) in searched_perms
        if (ct.pk, codename) not in all_perms
    ]
    Permission.objects.using(using).bulk_create(perms)
    if verbosity >= 2:
        for perm in perms:
            print("Adding permission '%s'" % perm)


def _get_group_permissions(self, user_obj):
    user_groups_field = user_obj._meta.get_field('groups')
    user_groups_query = 'group__%s' % user_groups_field.related_query_name()
    return Permission.objects.filter(**{user_groups_query: user_obj})


backends.ModelBackend._get_group_permissions = _get_group_permissions

management.create_permissions = create_permissions

  • request.py 请求接口根据用户企业id设置数据库连接
from rest_framework.request import Request
from rest_framework import exceptions

from utils.api_db import set_current_db


def __request_authenticate(self):
    """
    Attempt to authenticate the request using each authentication instance
    in turn.
    """
    for authenticator in self.authenticators:
        try:
            user_auth_tuple = authenticator.authenticate(self)
        except exceptions.APIException:
            self._not_authenticated()
            raise

        if user_auth_tuple is not None:
            self._authenticator = authenticator
            self.user, self.auth = user_auth_tuple
            if self.user and self.user.organize:
                # 设置对应的企业数据库
                set_current_db(self.user.organize.id.lower())
            return

    self._not_authenticated()


Request._authenticate = __request_authenticate

  • app.py 应用初始化
from django.apps import AppConfig
from django.db import connections

from organize import get_all_organize_db


class OrganizeConfig(AppConfig):
    default_auto_field = 'django.db.models.BigAutoField'
    name = 'organize'

    def ready(self) -> None:
        from .signal import create_data_handler  # 创建企业的时候创建数据库
        from .patch.connection import ConnectionHandler
        from .patch import request  # 给请求接口设置数据库
        from .patch.contenttype import management
        from .patch.permission import management

        dbs = list(get_all_organize_db().keys())

        for db in dbs:
            # 连接数据库
            connections[db]
        return super().ready()

  • signal.py 信号,创建企业的时候触发创建数据库
from django.dispatch import Signal  # 自定义信号
from django.dispatch import receiver
from utils.api_db import create_database
from utils.logger import logger

create_organize = Signal()  # 创建企业


@receiver(create_organize)
def create_data_handler(sender, org_id, org_name, **kwargs):
    try:
        # 给企业创建数据库
        create_database(org_id, org_name)
        logger.info(f'create database : [{org_id}] successfuly for {org_name}')
    except Exception as e:
        logger.error(e)
  • 其他
# 触发创建企业数据库信号
create_organize.send(sender=Organize, org_id=result['uid'].lower(), org_name=result['value'])


# api_view.get_queryset
class APIView(GenericViewSet)
    def get_queryset(self):
        db = get_current_db()
        organize = self.request.query_params.get('organize', '') or self.request.data.get('organize', '')
        if db == 'default' and organize is not None:
            # 平台调用企业数据(单条)
            set_current_db(organize.lower())
            return super().get_queryset()
        elif db != 'default' and self.basename in get_common_apps():
            # 企业查询平台数据
            set_current_db('default')
            return super().get_queryset()
        else:
            return super().get_queryset()


class AllDataModelMixin(object):
    """平台获取所有企业数据"""

    def all_data(self, request):
        organize = request.query_params.get('organize', '')
        res_data = get_model_all_data(view=self, organize=organize)
        # 返回分页
        page = self.paginator.paginate_queryset(res_data, request)
        return self.paginator.get_paginated_response(page,)
最后编辑于
©著作权归作者所有,转载或内容合作请联系作者
  • 序言:七十年代末,一起剥皮案震惊了整个滨河市,随后出现的几起案子,更是在滨河造成了极大的恐慌,老刑警刘岩,带你破解...
    沈念sama阅读 206,214评论 6 481
  • 序言:滨河连续发生了三起死亡事件,死亡现场离奇诡异,居然都是意外死亡,警方通过查阅死者的电脑和手机,发现死者居然都...
    沈念sama阅读 88,307评论 2 382
  • 文/潘晓璐 我一进店门,熙熙楼的掌柜王于贵愁眉苦脸地迎上来,“玉大人,你说我怎么就摊上这事。” “怎么了?”我有些...
    开封第一讲书人阅读 152,543评论 0 341
  • 文/不坏的土叔 我叫张陵,是天一观的道长。 经常有香客问我,道长,这世上最难降的妖魔是什么? 我笑而不...
    开封第一讲书人阅读 55,221评论 1 279
  • 正文 为了忘掉前任,我火速办了婚礼,结果婚礼上,老公的妹妹穿的比我还像新娘。我一直安慰自己,他们只是感情好,可当我...
    茶点故事阅读 64,224评论 5 371
  • 文/花漫 我一把揭开白布。 她就那样静静地躺着,像睡着了一般。 火红的嫁衣衬着肌肤如雪。 梳的纹丝不乱的头发上,一...
    开封第一讲书人阅读 49,007评论 1 284
  • 那天,我揣着相机与录音,去河边找鬼。 笑死,一个胖子当着我的面吹牛,可吹牛的内容都是我干的。 我是一名探鬼主播,决...
    沈念sama阅读 38,313评论 3 399
  • 文/苍兰香墨 我猛地睁开眼,长吁一口气:“原来是场噩梦啊……” “哼!你这毒妇竟也来了?” 一声冷哼从身侧响起,我...
    开封第一讲书人阅读 36,956评论 0 259
  • 序言:老挝万荣一对情侣失踪,失踪者是张志新(化名)和其女友刘颖,没想到半个月后,有当地人在树林里发现了一具尸体,经...
    沈念sama阅读 43,441评论 1 300
  • 正文 独居荒郊野岭守林人离奇死亡,尸身上长有42处带血的脓包…… 初始之章·张勋 以下内容为张勋视角 年9月15日...
    茶点故事阅读 35,925评论 2 323
  • 正文 我和宋清朗相恋三年,在试婚纱的时候发现自己被绿了。 大学时的朋友给我发了我未婚夫和他白月光在一起吃饭的照片。...
    茶点故事阅读 38,018评论 1 333
  • 序言:一个原本活蹦乱跳的男人离奇死亡,死状恐怖,灵堂内的尸体忽然破棺而出,到底是诈尸还是另有隐情,我是刑警宁泽,带...
    沈念sama阅读 33,685评论 4 322
  • 正文 年R本政府宣布,位于F岛的核电站,受9级特大地震影响,放射性物质发生泄漏。R本人自食恶果不足惜,却给世界环境...
    茶点故事阅读 39,234评论 3 307
  • 文/蒙蒙 一、第九天 我趴在偏房一处隐蔽的房顶上张望。 院中可真热闹,春花似锦、人声如沸。这庄子的主人今日做“春日...
    开封第一讲书人阅读 30,240评论 0 19
  • 文/苍兰香墨 我抬头看了看天上的太阳。三九已至,却和暖如春,着一层夹袄步出监牢的瞬间,已是汗流浃背。 一阵脚步声响...
    开封第一讲书人阅读 31,464评论 1 261
  • 我被黑心中介骗来泰国打工, 没想到刚下飞机就差点儿被人妖公主榨干…… 1. 我叫王不留,地道东北人。 一个月前我还...
    沈念sama阅读 45,467评论 2 352
  • 正文 我出身青楼,却偏偏与公主长得像,于是被迫代替她去往敌国和亲。 传闻我的和亲对象是个残疾皇子,可洞房花烛夜当晚...
    茶点故事阅读 42,762评论 2 345

推荐阅读更多精彩内容