如何管理应用工厂模式中的额外模块？

Question

我正在使用应用工厂模式的 Flask。我知道应用工厂模式只在工厂函数中管理配置对象。（如下代码）

def create_app(config):
    app.config.from_object(config)
    sentry.init(app)
    ...

    return app

但是如何管理需要该配置但无法在应用程序创建时初始化的额外模块？

所以我想做一些类似的事情

def create_app(config):
    some_module_obj = Module(host=config.host, port=config.port)

    app.config.from_object(config)
    sentry.init(app)

    return some_module_obj, app

而不是

# I don't want to use `config` outside of the `create_app` function!
some_module_obj = Module(host=config.host, port=config.port)

def create_app(config):
    app.config.from_object(config)
    sentry.init(app)

    return app

Answer 1

决定将自定义 class 用于初始化对象作为工厂模式。

这是示例：

class CustomFactory(metaclass=ABCMeta):
    @abstractmethod
    def init_factory(self, config):
        pass

    @property
    @abstractmethod
    def app(self):
        pass

    def __getattr__(self, item):
        return getattr(self.app, item)


class RQSchedulerFactory(CustomFactory):
    def __init__(self):
        self._app = None

    def init_factory(self, config):
        self._app = Scheduler(connection=Redis(host=config.REDIS_HOST, port=config.REDIS_PORT))

    @property
    def app(self):
        return self._app


class FireDBFactory(CustomFactory):
    @property
    def app(self):
        return self._app

    def __init__(self):
        self._app = None

    def init_factory(self, config):
        cred = credentials.Certificate(config.FIREBASE_KEY_FILE)
        firebase_admin.initialize_app(cred)

        self._app = firestore.client()

并且在 __init__.py 中（具有 create_app 函数）

scheduler = RQSchedulerFactory()
fire_db = FireDBFactory()

然后在create_app函数中，初始化如下：

def create_app(config):
    app.config.from_object(config)

    # Scheduler initialization
    scheduler.init_factory(config)

    # Fire store initialization
    fire_db.init_factory(config)

Answer 2

不确定这是否是您所需要的，但您在评论中要求提供一个带有 inject + Flask 的小示例。据我了解，主要问题与 Flask + configuration + initialization 有关。这只是它如何工作的一个例子。

app.py:

from flask import Flask

from api import bp
from configurator import configure


def create_app():
    app = Flask(__name__)
    # configure Flask app config as you wish... (app.config.from_object(config))
    # just some settings for demonstration
    app.config.update(dict(
        MODULE1_TIMER=1,
        MODULE2_LIMIT=2,
    ))
    # configure inject using app context and Flask config
    with app.app_context():
        configure()
    # demo blueprint
    app.register_blueprint(bp)

    return app


if __name__ == '__main__':
    create_app().run(debug=True)

假设我们有一些模块：

# mod1.py
class Module1:
    def __init__(self, timer: int) -> None:
        self._timer = timer

# mod2.py
class Module2:
    def __init__(self, limit: int) -> None:
        self._limit = limit

    def get_limit(self):
        return self._limit

# mod3.py - works with mod1 and mod2
class Module3:
    def __init__(self, module1, module2) -> None:
        self._module1 = module1
        self._module2 = module2

    def get_limit(self):
        return self._module2.get_limit()

configurator.py:

import inject
from flask import current_app

from mod1 import Module1
from mod2 import Module2
from mod3 import Module3


@inject.params(
    module1=Module1,
    module2=Module2,
)
def _init_module3(module1, module2):
    # module1 and module2 are injected instances
    return Module3(module1, module2)


def _injector_config(binder):
    # initialization of Module1 and Module2 using Flask config
    binder.bind(Module1, Module1(current_app.config['MODULE1_TIMER']))
    binder.bind(Module2, Module2(current_app.config['MODULE2_LIMIT']))
    # initialization of Module3 using injected Module1 + Module2
    # you can use bind_to_constructor + any function
    binder.bind_to_constructor(Module3, _init_module3)


def configure():
    def config(binder):
        binder.install(_injector_config)
        # one more binder.install... etc...

    inject.clear_and_configure(config)

api.py:

import inject
from flask import Blueprint, jsonify

from mod1 import Module1
from mod2 import Module2
from mod3 import Module3

bp = Blueprint('api', __name__)


@bp.route('/test')
def test():
    # get instances which was created using inject
    return jsonify(dict(
        module1=str(type(inject.instance(Module1))),
        module2=str(type(inject.instance(Module2))),
        module3=str(type(inject.instance(Module3))),
    ))


# you can inject something as arg
@bp.route('/test2')
@inject.params(module3=Module3)
def test2(module3: Module3):
    return jsonify(dict(module3=str(type(module3))))


@bp.route('/test3')
def test3():
    # you can inject something into anything
    class Example:
        module3 = inject.attr(Module3)

        @inject.params(module2=Module2)
        def __init__(self, module2: Module2) -> None:
            self.module2 = module2

    return jsonify({
        'MODULE2_LIMIT': Example.module3.get_limit(),
        'example': dir(Example()),
    })

运行服务器，开启/test,test2,/test3.

关于好处的几句话：

初始化和配置一分
降低对current_app、flask 配置/上下文等的依赖
减少递归导入的问题
易于编写测试

希望对您有所帮助。

如何管理应用工厂模式中的额外模块？

How can I manage extra modules in app factory pattern?

factory-pattern

flask