通过 Python 装饰器实现DRY(不重复代码)原则
jopen
8年前
Python 装饰器 是一个消除冗余的强大工具。随着将功能模块化为大小合适的方法，即使是最复杂的工作流，装饰器也能使它变成简洁的功能。
例如让我们看看 Django web框架 ，该框架处理请求的方法接收一个方法对象，返回一个响应对象：
def handle_request(request): return HttpResponse("Hello, World")defhandle_request(request): return HttpResponse("Hello, World")我最近遇到一个案例，需要编写几个满足下述条件的api方法：
- 返回json响应
- 如果是GET请求，那么返回错误码
做为一个注册api端点例子，我将会像这样编写：
def register(request): result = None # check for post only if request.method != 'POST': result = {"error": "this method only accepts posts!"} else: try: user = User.objects.create_user(request.POST['username'], request.POST['email'], request.POST['password']) # optional fields for field in ['first_name', 'last_name']: if field in request.POST: setattr(user, field, request.POST[field]) user.save() result = {"success": True} except KeyError as e: result = {"error": str(e) } response = HttpResponse(json.dumps(result)) if "error" in result: response.status_code = 500 return responsedefregister(request): result = None # check for post only if request.method != 'POST': result = {"error": "this method only accepts posts!"} else: try: user = User.objects.create_user(request.POST['username'], request.POST['email'], request.POST['password']) # optional fields for fieldin ['first_name', 'last_name']: if fieldin request.POST: setattr(user, field, request.POST[field]) user.save() result = {"success": True} exceptKeyErroras e: result = {"error": str(e) } response = HttpResponse(json.dumps(result)) if "error" in result: response.status_code = 500 return response然而这样我将会在每个api方法中编写json响应和错误返回的代码。这将会导致大量的逻辑重复。所以让我们尝试用装饰器实现DRY原则吧。
装饰器简介
如果你不熟悉装饰器，我可以简单解释一下，实际上装饰器就是有效的函数包装器，python解释器加载函数的时候就会执行包装器，包装器可以修改函数的接收参数和返回值。举例来说，如果我想要总是返回比实际返回值大一的整数结果，我可以这样写装饰器：
# a decorator receives the method it's wrapping as a variable 'f' def increment(f): # we use arbitrary args and keywords to # ensure we grab all the input arguments. def wrapped_f(*args, **kw): # note we call f against the variables passed into the wrapper, # and cast the result to an int and increment . return int(f(*args, **kw)) + 1 return wrapped_f # the wrapped function gets returned.
# a decorator receives the method it's wrapping as a variable 'f' defincrement(f): # we use arbitrary args and keywords to # ensure we grab all the input arguments. defwrapped_f(*args, **kw): # note we call f against the variables passed into the wrapper, # and cast the result to an int and increment . return int(f(*args, **kw)) + 1 return wrapped_f # the wrapped function gets returned.
现在我们就可以用@符号和这个装饰器去装饰另外一个函数了：
@increment def plus(a, b): return a + b result = plus(4, 6) assert(result == 11, "We wrote our decorator wrong!")
@increment defplus(a, b): return a + b result = plus(4, 6) assert(result == 11, "We wrote our decorator wrong!")
装饰器修改了存在的函数，将装饰器返回的结果赋值给了变量。在这个例子中，’plus’的结果实际指向increment(plus)的结果。
对于非post请求返回错误
现在让我们在一些更有用的场景下应用装饰器。如果在django中接收的不是POST请求，我们用装饰器返回一个错误响应。
def post_only(f): """ Ensures a method is post only """ def wrapped_f(request): if request.method != "POST": response = HttpResponse(json.dumps( {"error": "this method only accepts posts!"})) response.status_code = 500 return response return f(request) return wrapped_fdefpost_only(f): """ Ensures a method is post only """ defwrapped_f(request): if request.method != "POST": response = HttpResponse(json.dumps( {"error": "this method only accepts posts!"})) response.status_code = 500 return response return f(request) return wrapped_f现在我们可以在上述注册api中应用这个装饰器：
@post_only def register(request): result = None try: user = User.objects.create_user(request.POST['username'], request.POST['email'], request.POST['password']) # optional fields for field in ['first_name', 'last_name']: if field in request.POST: setattr(user, field, request.POST[field]) user.save() result = {"success": True} except KeyError as e: result = {"error": str(e) } response = HttpResponse(json.dumps(result)) if "error" in result: response.status_code = 500 return response@post_only defregister(request): result = None try: user = User.objects.create_user(request.POST['username'], request.POST['email'], request.POST['password']) # optional fields for fieldin ['first_name', 'last_name']: if fieldin request.POST: setattr(user, field, request.POST[field]) user.save() result = {"success": True} exceptKeyErroras e: result = {"error": str(e) } response = HttpResponse(json.dumps(result)) if "error" in result: response.status_code = 500 return response现在我们就有了一个可以在每个api方法中重用的装饰器。
发送json响应
为了发送json响应（同时处理500状态码），我们可以新建另外一个装饰器：
def json_response(f): """ Return the response as json, and return a 500 error code if an error exists """ def wrapped(*args, **kwargs): result = f(*args, **kwargs) response = HttpResponse(json.dumps(result)) if type(result) == dict and 'error' in result: response.status_code = 500 return response
defjson_response(f): """ Return the response as json, and return a 500 error code if an error exists """ defwrapped(*args, **kwargs): result = f(*args, **kwargs) response = HttpResponse(json.dumps(result)) if type(result) == dictand 'error' in result: response.status_code = 500 return response
现在我们就可以在原方法中去除json相关的代码，添加一个装饰器做为代替：
@post_only @json_response def register(request): try: user = User.objects.create_user(request.POST['username'], request.POST['email'], request.POST['password']) # optional fields for field in ['first_name', 'last_name']: if field in request.POST: setattr(user, field, request.POST[field]) user.save() return {"success": True} except KeyError as e: return {"error": str(e) }@post_only @json_response defregister(request): try: user = User.objects.create_user(request.POST['username'], request.POST['email'], request.POST['password']) # optional fields for fieldin ['first_name', 'last_name']: if fieldin request.POST: setattr(user, field, request.POST[field]) user.save() return {"success": True} exceptKeyErroras e: return {"error": str(e) }现在，如果我需要编写新的方法，那么我就可以使用装饰器做冗余的工作。如果我要写登录方法，我只需要写真正相关的代码：
@post_only @json_response def login(request): if request.user is not None: return {"error": "User is already authenticated!"} user = auth.authenticate(request.POST['username'], request.POST['password']) if user is not None: if not user.is_active: return {"error": "User is inactive"} auth.login(request, user) return {"success": True, "id": user.pk} else: return {"error": "User does not exist with those credentials"}@post_only @json_response deflogin(request): if request.useris not None: return {"error": "User is already authenticated!"} user = auth.authenticate(request.POST['username'], request.POST['password']) if useris not None: if not user.is_active: return {"error": "User is inactive"} auth.login(request, user) return {"success": True, "id": user.pk} else: return {"error": "User does not exist with those credentials"}BONUS: 参数化你的请求方法
我曾经使用过 Tubogears 框架，其中请求参数直接解释转递给方法这一点我很喜欢。所以要怎样在Django中模仿这一特性呢？嗯，装饰器就是一种解决方案！
例如:
def parameterize_request(types=("POST",)): """ Parameterize the request instead of parsing the request directly. Only the types specified will be added to the query parameters. e.g. convert a=testdefparameterize_request(types=("POST",)): """ Parameterizetherequestinsteadofparsingtherequestdirectly. Onlythetypesspecifiedwillbeaddedto thequeryparameters. e.g. convert a=test注意这是一个参数化装饰器的例子。在这个例子中，函数的结果是实际的装饰器。
现在我就可以用参数化装饰器编写方法了！我甚至可以选择是否允许GET和POST，或者仅仅一种请求参数类型。
@post_only @json_response @parameterize_request(["POST"]) def register(request, username, email, password, first_name=None, last_name=None): user = User.objects.create_user(username, email, password) user.first_name=first_name user.last_name=last_name user.save() return {"success": True}@post_only @json_response @parameterize_request(["POST"]) defregister(request, username, email, password, first_name=None, last_name=None): user = User.objects.create_user(username, email, password) user.first_name=first_name user.last_name=last_name user.save() return {"success": True}现在我们有了一个简洁的、易于理解的api。
BONUS #2: 使用functools.wraps保存docstrings和函数名
很不幸，使用装饰器的一个副作用是没有保存方法名(__name__)和docstring(__doc__)值：
def increment(f): """ Increment a function result """ wrapped_f(a, b): return f(a, b) + 1 return wrapped_f @increment def plus(a, b) """ Add two things together """ return a + b plus.__name__ # this is now 'wrapped_f' instead of 'plus' plus.__doc__ # this now returns 'Increment a function result' instead of 'Add two things together'
defincrement(f): """ Increment a function result """ wrapped_f(a, b): return f(a, b) + 1 return wrapped_f @increment defplus(a, b) """ Add two things together """ return a + b plus.__name__ # this is now 'wrapped_f' instead of 'plus' plus.__doc__ # this now returns 'Increment a function result' instead of 'Add two things together'
这将对使用反射的应用造成麻烦，比如Sphinx，一个 自动生成文档的应用 。
为了解决这个问题，我们可以使用’wraps’装饰器附加上名字和docstring：
from functools import wraps def increment(f): """ Increment a function result """ @wraps(f) wrapped_f(a, b): return f(a, b) + 1 return wrapped_f @increment def plus(a, b) """ Add two things together """ return a + b plus.__name__ # this returns 'plus' plus.__doc__ # this returns 'Add two things together'
fromfunctoolsimportwraps defincrement(f): """ Increment a function result """ @wraps(f) wrapped_f(a, b): return f(a, b) + 1 return wrapped_f @increment defplus(a, b) """ Add two things together """ return a + b plus.__name__ # this returns 'plus' plus.__doc__ # this returns 'Add two things together'
BONUS #3: 使用’decorator’装饰器
如果仔细看看上述使用装饰器的方式，在包装器声明和返回的地方也有不少重复。
你可以安装python egg ‘decorator’，其中包含一个提供装饰器模板的’decorator’装饰器！
使用easy_install:
$ sudo easy_install decorator
$ sudoeasy_installdecorator
或者Pip:
$ pip install decorator
$ pipinstalldecorator
然后你可以简单的编写：
from decorator import decorator @decorator def post_only(f, request): """ Ensures a method is post only """ if request.method != "POST": response = HttpResponse(json.dumps( {"error": "this method only accepts posts!"})) response.status_code = 500 return response return f(request)fromdecoratorimportdecorator @decorator defpost_only(f, request): """ Ensures a method is post only """ if request.method != "POST": response = HttpResponse(json.dumps( {"error": "this method only accepts posts!"})) response.status_code = 500 return response return f(request)这个装饰器更牛逼的一点是保存了__name__和__doc__的返回值，也就是它封装了
functools.wraps的
功能！