Openstack : 14、Openstack-开发基础 Pecan学习

来源:互联网 发布:业务网站源码 编辑:程序博客网 时间:2024/06/05 03:49

参考文档:

http://pecan.readthedocs.io/en/latest/quick_start.html 

http://blog.csdn.net/allison_ywt/article/details/54140730

    Pecan是一个路由对象分发的oython web框架。本质上可以将url通过分割为每一部分,然后对每一部分查找对应处理该URL部分的处理类,处理后,继续交给后面部分的URL处理,直到所有URL部分都被处理后,调用最后分割的URL对应的处理函数处理。

1 基础知识介绍

安装: sudo pip install pecan
创建项目:pecan create test_project
部署: python setup.py develop
开启服务: 进入含有config.py的目录下,执行: pecan serve config.py
测试:curl http://localhost:8080/catalog/books/bestsellers
解析过程应该是将每个路径例如: catalog加上controller拼接为CatalogController,转到该类处理

RootController:可以认为是应用入口,根url路径,例如: 
http://localhost:8080/
index():是公共可访问的,通过expose()使用index.html,返回的是映射,映射渲染特定模板
    @expose(generic=True, template='index.html')
    def index(self):
        return dict()

index_post():方法接收post的参数请求q,所有使用post的请求都会转到index_post方法
pecan采用对象分发来映射来将http请求映射到controller,你可以想想你的应用的
控制器是一颗树,对象树的分支直接映射到url路径
如果方法不提供expose()装饰器,Pecan将不会将请求路由给它
@expose():
@expose('html_template.mako'):可以指定模板,然后将方法中的映射传入到模板中做渲染
@expose('json') : 将模板命名空间渲染为json文本
@pecan.expose(route='some-path'):指定明确的路径
pecan.route('some-path', RootController.some_path)
分发到子控制器
pecan.route(RootController, 'child-path', ChildController())
路由算法: _lookup(), _default(), and _route()
_lookup():可以对部分url处理,返回一个新的controller对象来继续处理剩余部分
_default():该方法在没有其他控制器可以匹配url的时候调用
 _route():该方法允许控制器controller完全重写Pecan的路由机制
pecan.Request 和 pecan.Response:可以与Pecan控制器交互
abort() :抛出http异常,例如abort(404)


2 下面内容来自于: http://blog.csdn.net/allison_ywt/article/details/54140730

使用Pecan框架时,OpenStack项目一般会把API服务的实现都放在一个api目录下,比如magnum项目是这样的:

➜ ~/openstack/env/p/magnum git:(master) $ tree magnum/apimagnum/api├── app.py├── auth.py├── config.py├── controllers│   ├── base.py│   ├── __init__.py│   ├── link.py│   ├── root.py│   └── v1│       ├── base.py│       ├── baymodel.py│       ├── bay.py│       ├── certificate.py│       ├── collection.py│       ├── container.py│       ├── __init__.py│       ├── magnum_services.py│       ├── node.py│       ├── pod.py│       ├── replicationcontroller.py│       ├── service.py│       ├── types.py│       ├── utils.py│       └── x509keypair.py├── expose.py├── hooks.py├── __init__.py├── middleware│   ├── auth_token.py│   ├── __init__.py│   └── parsable_error.py├── servicegroup.py└── validation.py

你也可以在Ceilometer项目中看到类似的结构。介绍一下几个主要的文件,这样你以后看到一个使用Pecan的OpenStack项目时就会比较容易找到入口。

  • app.py 一般包含了Pecan应用的入口,包含应用初始化代码

  • config.py 包含Pecan的应用配置,会被app.py使用

  • controllers/ 这个目录会包含所有的控制器,也就是API具体逻辑的地方

  • controllers/root.py 这个包含根路径对应的控制器

  • controllers/v1/ 这个目录对应v1版本的API的控制器。如果有多个版本的API,你一般能看到v2等目录。

代码变少了:application的配置

Pecan的配置很容易,通过一个Python源码式的配置文件就可以完成基本的配置。这个配置的主要目的是指定应用程序的root,然后用于生成WSGI application。我们来看Magnum项目的例子。Magnum项目有个API服务是用Pecan实现的,在magnum/api/config.py文件中可以找到这个文件,主要内容如下:

app = {    'root': 'magnum.api.controllers.root.RootController',    'modules': ['magnum.api'],    'debug': False,    'hooks': [        hooks.ContextHook(),        hooks.RPCHook(),        hooks.NoExceptionTracebackHook(),    ],    'acl_public_routes': [        '/'    ],}

上面这个app对象就是Pecan的配置,每个Pecan应用都需要有这么一个名为app的配置。app配置中最主要的就是root的值,这个值表示了应用程序的入口,也就是从哪个地方开始解析HTTP的根path:/hooks对应的配置是一些Pecan的hook,作用类似于WSGI Middleware。

有了app配置后,就可以让Pecan生成一个WSGI application。在Magnum项目中,magnum/api/app.py文件就是生成WSGI application的地方,我们来看一下这个的主要内容:

def get_pecan_config():    # Set up the pecan configuration    filename = api_config.__file__.replace('.pyc', '.py')    return pecan.configuration.conf_from_file(filename)def setup_app(config=None):    if not config:        config = get_pecan_config()    app_conf = dict(config.app)    app = pecan.make_app(        app_conf.pop('root'),        logging=getattr(config, 'logging', {}),        wrap_app=middleware.ParsableErrorMiddleware,        **app_conf    )    return auth.install(app, CONF, config.app.acl_public_routes)

get_pecan_config()方法读取我们上面提到的config.py文件,然后返回一个pecan.configuration.Config对象。setup_app()函数首先调用get_pecan_config()函数获取application的配置,然后调用pecan.make_app()函数创建了一个WSGI application,最后调用了auth.install()函数(也就是magnum.api.auth.install()函数)为刚刚生成的WSGI application加上Keystone的认证中间件(确保所有的请求都会通过Keystone认证)。

到这边为止,一个Pecan的WSGI application就已经准备好了,只要调用这个setup_app()函数就能获得。至于如何部署这个WSGI application,请参考WSGI简介这篇文章。

从Magnum这个实际的例子可以看出,使用了Pecan之后,我们不再需要自己写那些冗余的WSGI application代码了,直接调用Pecan的make_app()函数就能完成这些工作。另外,对于之前使用PasteDeploy时用到的很多WSGI中间件,可以选择使用Pecan的hooks机制来实现,也选择使用WSGI中间件的方式来实现。在Magnum的API服务就同时使用了这两种方式。其实,Pecan还可以和PasteDeploy一起使用,Ceilometer项目就是这么做的,大家可以看看。

确定路由变得容易了:对象分发式的路由

Pecan不仅缩减了生成WSGI application的代码,而且也让开发人员更容易的指定一个application的路由。Pecan采用了一种对象分发风格(object-dispatch style)的路由模式。我们直接通过例子来解释这种路由模式,还是以Magnum项目为例。

上面提到了,Magnum的API服务的root是magnum.api.controllers.root.RootController。这里的RootController的是一个类,我们来看它的代码:

class RootController(rest.RestController):    _versions = ['v1']    """All supported API versions"""    _default_version = 'v1'    """The default API version"""    v1 = v1.Controller()    @expose.expose(Root)    def get(self):        # NOTE: The reason why convert() it's being called for every        #       request is because we need to get the host url from        #       the request object to make the links.        return Root.convert()    @pecan.expose()    def _route(self, args):        """Overrides the default routing behavior.        It redirects the request to the default version of the magnum API        if the version number is not specified in the url.        """        if args[0] and args[0] not in self._versions:            args = [self._default_version] + args        return super(RootController, self)._route(args)

别看这个类这么长,我来解释一下你就懂了。首先,你可以先忽略掉_route()函数,这个函数是用来覆盖Pecan的默认路由实现的,在这里去掉它不妨碍我们理解Pecan(这里的_route()函数的作用把所有请求重定向到默认的API版本去)。去掉_route()和其他的东西后,整个类就变成这么短:

class RootController(rest.RestController):    v1 = v1.Controller()    @expose.expose(Root)    def get(self):        return Root.convert()
  • 首先,你要记住,这个RootController对应的是URL中根路径,也就是path中最左边的/

  • RootController继承自rest.RestController,是Pecan实现的RESTful控制器。这里的get()函数表示,当访问的是GET /时,由该函数处理。get()函数会返回一个WSME对象,表示一个形式化的HTTP Response,这个下面再讲。get()函数上面的expose装饰器是Pecan实现路由控制的一个方式,被expose的函数才会被路由处理。

  • 这里的v1 = v1.Controller()表示,当访问的是GET /v1或者GET /v1/...时,请求由一个v1.Controller实例来处理。

为了加深大家的理解,我们再来看下v1.Controller的实现:

class Controller(rest.RestController):    """Version 1 API controller root."""    bays = bay.BaysController()    baymodels = baymodel.BayModelsController()    containers = container.ContainersController()    nodes = node.NodesController()    pods = pod.PodsController()    rcs = rc.ReplicationControllersController()    services = service.ServicesController()    x509keypairs = x509keypair.X509KeyPairController()    certificates = certificate.CertificateController()    @expose.expose(V1)    def get(self):        return V1.convert()    ...

上面这个Controller也是继承自rest.RestController。所以它的get函数表示,当访问的是GET /v1的时候,要做的处理。然后,它还有很多类属性,这些属性分别表示不同URL路径的控制器:

  • /v1/bays 由bays处理

  • /v1/baymodels 由baymodels处理

  • /v1/containers 由containers处理

其他的都是类似的。我们再继续看bay.BaysController的代码:

class BaysController(rest.RestController):    """REST controller for Bays."""    def __init__(self):        super(BaysController, self).__init__()    _custom_actions = {        'detail': ['GET'],    }    def get_all(...):        def detail(...):        def get_one(...):        def post(...):        def patch(...):    def delete(...):

这个controller中只有函数,没有任何类属性,而且没有实现任何特殊方法,所以/v1/bays开头的URL处理都在这个controller中终结。这个类会处理如下请求:

  • GET /v1/bays

  • GET /v1/bays/{UUID}

  • POST /v1/bays

  • PATCH /v1/bays/{UUID}

  • DELETE /v1/bays/{UUID}

  • GET /v1/bays/detail/{UUID}

看了上面的3个controller之后,你应该能大概明白Pecan是如何对URL进行路由的。这种路由方式就是对象分发:根据类属性,包括数据属性和方法属性来决定如何路由一个HTTP请求。Pecan的文档中对请求的路由有专门的描述,要想掌握Pecan的路由还是要完整的看一下官方文档。

内置RESTful支持

我们上面举例的controller都是继承自pecan.rest.RestController,这种controller称为RESTful controller,专门用于实现RESTful API的,因此在OpenStack中使用特别多。Pecan还支持普通的controller,称为Generic controller。Generic controller继承自object对象,默认没有实现对RESTful请求的方法。简单的说,RESTful controller帮我们规定好了get_one()get_all()get(),post()等方法对应的HTTP请求,而Generic controller则没有。关于这两种controller的区别,可以看官方文档Writing RESTful Web Services with Generic Controllers,有很清楚的示例。

对于RestController中没有预先定义好的方法,我们可以通过控制器的_custom_actions属性来指定其能处理的方法。

class RootController(rest.RestController):    _custom_actions = {        'test': ['GET'],    }    @expose()    def test(self):        return 'hello'

上面这个控制器是一个根控制器,指定了/test路径支持GET方法,效果如下:

 $ curl http://localhost:8080/testhello% 

那么HTTP请求和HTTP响应呢?

上面讲了这么多,我们都没有说明在Pecan中如何处理请求和如何返回响应。这个将在下一章中说明,同时我们会引入一个新的库WSME

WSME

Pecan对请求和响应的处理

在开始提到WSME之前,我们先来看下Pecan自己对HTTP请求和响应的处理。这样你能更好的理解为什么会再引入一个WSME库。

Pecan框架为每个线程维护了单独的请求和响应对象,你可以直接在请求处理函数中访问。pecan.requestpecan.response分别代表当前需要处理的请求和响应对象。你可以直接操作这两个对象,比如指定响应的状态码,就像下面这个例子一样(例子来自官方文档):

@pecan.expose()def login(self):    assert pecan.request.path == '/login'    username = pecan.request.POST.get('username')    password = pecan.request.POST.get('password')    pecan.response.status = 403    pecan.response.text = 'Bad Login!'

这个例子演示了访问POST请求的参数以及返回403。你也可以重新构造一个pecan.Response对象作为返回值(例子来自官方文档):

from pecan import expose, Responseclass RootController(object):    @expose()    def hello(self):        return Response('Hello, World!', 202)

另外,HTTP请求的参数也会可以作为控制器方法的参数,还是来看几个官方文档的例子:

class RootController(object):    @expose()    def index(self, arg):        return arg    @expose()    def kwargs(self, **kwargs):        return str(kwargs)

这个控制器中的方法直接返回了参数,演示了对GET请求参数的处理,效果是这样的:

$ curl http://localhost:8080/?arg=foofoo$ curl http://localhost:8080/kwargs?a=1&b=2&c=3{u'a': u'1', u'c': u'3', u'b': u'2'}

有时候,参数也可能是URL的一部分,比如最后的一段path作为参数,就像下面这样:

class RootController(object):    @expose()    def args(self, *args):        return ','.join(args)

效果是这样的:

$ curl http://localhost:8080/args/one/two/threeone,two,three

另外,我们还要看一下POST方法的参数如何处理(例子来自官方文档):

class RootController(object):    @expose()    def index(self, arg):        return arg

效果如下,就是把HTTP body解析成了控制器方法的参数:

$ curl -X POST "http://localhost:8080/" -H "Content-Type: application/x-www-form-urlencoded" -d "arg=foo"foo

返回JSON还是HTML?

如果你不是明确的返回一个Response对象,那么Pecan中方法的返回内容类型就是由expose()装饰器决定的。默认情况下,控制器的方法返回的content-type是HTML。

class RootController(rest.RestController):    _custom_actions = {        'test': ['GET'],    }    @expose()    def test(self):        return 'hello'

效果如下:

 $ curl -v http://localhost:8080/test* Hostname was NOT found in DNS cache*   Trying 127.0.0.1...* Connected to localhost (127.0.0.1) port 8080 (#0)> GET /test HTTP/1.1> User-Agent: curl/7.38.0> Host: localhost:8080> Accept: */*>* HTTP 1.0, assume close after body< HTTP/1.0 200 OK< Date: Tue, 15 Sep 2015 14:31:28 GMT< Server: WSGIServer/0.1 Python/2.7.9< Content-Length: 5< Content-Type: text/html; charset=UTF-8<* Closing connection 0hello% 

也可以让它返回JSON:

class RootController(rest.RestController):    _custom_actions = {        'test': ['GET'],    }    @expose('json')    def test(self):        return 'hello'

效果如下:

 curl -v http://localhost:8080/test* Hostname was NOT found in DNS cache*   Trying 127.0.0.1...* Connected to localhost (127.0.0.1) port 8080 (#0)> GET /test HTTP/1.1> User-Agent: curl/7.38.0> Host: localhost:8080> Accept: */*>* HTTP 1.0, assume close after body< HTTP/1.0 200 OK< Date: Tue, 15 Sep 2015 14:33:27 GMT< Server: WSGIServer/0.1 Python/2.7.9< Content-Length: 18< Content-Type: application/json; charset=UTF-8<* Closing connection 0{"hello": "world"}% 

甚至,你还可以让一个控制器方法根据URL path的来决定是返回HTML还是JSON:

class RootController(rest.RestController):    _custom_actions = {        'test': ['GET'],    }    @expose()    @expose('json')    def test(self):        return json.dumps({'hello': 'world'})

返回JSON:

 $ curl -v http://localhost:8080/test.json* Hostname was NOT found in DNS cache*   Trying 127.0.0.1...* Connected to localhost (127.0.0.1) port 8080 (#0)> GET /test.json HTTP/1.1> User-Agent: curl/7.38.0> Host: localhost:8080> Accept: */*>* HTTP 1.0, assume close after body< HTTP/1.0 200 OK< Date: Wed, 16 Sep 2015 14:26:27 GMT< Server: WSGIServer/0.1 Python/2.7.9< Content-Length: 24< Content-Type: application/json; charset=UTF-8<* Closing connection 0"{\"hello\": \"world\"}"% 

返回HTML:

 $ curl -v http://localhost:8080/test.html* Hostname was NOT found in DNS cache*   Trying 127.0.0.1...* Connected to localhost (127.0.0.1) port 8080 (#0)> GET /test.html HTTP/1.1> User-Agent: curl/7.38.0> Host: localhost:8080> Accept: */*>* HTTP 1.0, assume close after body< HTTP/1.0 200 OK< Date: Wed, 16 Sep 2015 14:26:24 GMT< Server: WSGIServer/0.1 Python/2.7.9< Content-Length: 18< Content-Type: text/html; charset=UTF-8<* Closing connection 0{"hello": "world"}% 

这里要注意一下:

  1. 同一个字符串作为JSON返回和作为HTML返回是不一样的,仔细看一下HTTP响应的内容。

  2. 我们的例子中在URL的最后加上了.html后缀或者.json后缀,请尝试一下不加后缀的化是返回什么?然后,调换一下两个expose()的顺序再试一下。

从上面的例子可以看出,决定响应类型的主要是传递给expose()函数的参数,我们看下expose()函数的完整声明:

pecan.decorators.expose(template=None,                        content_type='text/html',                        generic=False)
  • template参数用来指定返回值的模板,如果是'json'就会返回JSON内容,这里可以指定一个HTML文件,或者指定一个mako模板。

  • content_type指定响应的content-type,默认值是'text/html'。

  • generic参数表明该方法是一个“泛型”方法,可以指定多个不同的函数对应同一个路径的不同的HTTP方法。

看过参数的解释后,你应该能大概了解expose()函数是如何控制HTTP响应的内容和类型的。

用WSME来做什么?

上面两节已经说明了Pecan可以比较好的处理HTTP请求中的参数以及控制HTTP返回值。那么为什么我们还需要WSME呢?因为Pecan在做下面这个事情的时候比较麻烦:请求参数和响应内容的类型检查(英文简称就是typing)。当然,做是可以做的,不过你需要自己访问pecan.request和pecan.response,然后检查指定的值的类型。WSME就是为解决这个问题而生的,而且适用场景就是RESTful API。

WSME简介

WSME的全称是Web Service Made Easy,是专门用于实现REST服务的typing库,让你不需要直接操作请求和响应,而且刚好和Pecan结合得非常好,所以OpenStack的很多项目都使用了Pecan + WSME的组合来实现API(好吧,我看过的项目,用了Pecan的都用了WSME)。WSME的理念是:在大部分情况下,Web服务的输入和输出对数据类型的要求都是严格的。所以它就专门解决了这个事情,然后把其他事情都交给其他框架去实现。因此,一般WSME都是和其他框架配合使用的,支持Pecan、Flask等。WSME的文档地址是http://wsme.readthedocs.org/en/latest/index.html。

WSME的使用

用了WSME后的好处是什么呢?WSME会自动帮你检查HTTP请求和响应中的数据是否符合预先设定好的要求。WSME的主要方式是通过装饰器来控制controller方法的输入和输出。WSME中主要使用两个控制器:

  • @signature: 这个装饰器用来描述一个函数的输入和输出。

  • @wsexpose: 这个装饰器包含@signature的功能,同时会把函数的路由信息暴露给Web框架,效果就像Pecan的expose装饰器。

这里我们结合Pecan来讲解WSME的使用。先来看一个原始类型的例子:

from wsmeext.pecan import wsexposeclass RootController(rest.RestController):    _custom_actions = {        'test': ['GET'],    }    @wsexpose(int, int)    def test(self, number):        return number

如果不提供参数,访问会失败:

$ curl http://localhost:8080/test{"debuginfo": null, "faultcode": "Client", "faultstring": "Missing argument: \"number\""}% 

如果提供的参数不是整型,访问也会失败:

$ curl http://localhost:8080/test\?number\=a{"debuginfo": null, "faultcode": "Client", "faultstring": "Invalid input for field/attribute number. Value: 'a'. unable to convert to int"}% 

上面这些错误信息都是由WSME框架直接返回的,还没有执行到你写的方法。如果请求正确,那么会是这样的:

$ curl -v http://localhost:8080/test\?number\=1* Hostname was NOT found in DNS cache*   Trying 127.0.0.1...* Connected to localhost (127.0.0.1) port 8080 (#0)> GET /test?number=1 HTTP/1.1> User-Agent: curl/7.38.0> Host: localhost:8080> Accept: */*>* HTTP 1.0, assume close after body< HTTP/1.0 200 OK< Date: Wed, 16 Sep 2015 15:06:35 GMT< Server: WSGIServer/0.1 Python/2.7.9< Content-Length: 1< Content-Type: application/json; charset=UTF-8<* Closing connection 01% 

请注意返回的content-type,这里返回JSON是因为我们使用的wsexpose设置的返回类型是XML和JSON,并且JSON是默认值。上面这个例子就是WSME最简单的应用了。

那么现在有下面这些问题需要思考一下:

  • 如果想用POST的方式来传递参数,要怎么做呢?提示:要阅读WSME中@signature装饰器的文档。

  • 如果我希望使用/test/1这种方式来传递参数要怎么做呢?提示:要阅读Pecan文档中关于路由的部分。

  • WSME中支持对哪些类型的检查呢?WSME支持整型、浮点型、字符串、布尔型、日期时间等,甚至还支持用户自定义类型。提示:要阅读WSME文档中关于类型的部分。

  • WSME支持数组类型么?支持。

上面的问题其实也是很多人使用WSME的时候经常问的问题。我们将在下一篇文章中使用Pecan + WSME来继续开发我们的demo,并且用代码来回答上面所有的问题。


原创粉丝点击