Apache HttpComponents中的cookie匹配策略

1 简介

在clojure中使用clj-http抓取网页时，需要提交自定义cookie,总是不能成功发送，研究一下HttpClient中Cookie的工作方式。

clj-http包装了HttpClient库，对于请求页面时发回的状态可以自动处理，但是需要自己往请求中添加cookie时总是失败，折腾了很久，了解了HttpClient处理Cookie的细节，关于HttpClient中HTTP的状态管理，可以参考HttpClient的官方指南

2 示例

从bing获取搜索结果，代码如下：

deps.edn文件如下:

{
 :deps {
        org.clojure/clojure {:mvn/version "1.10.0"},
        clj-http {:mvn/version "3.9.1"}, ; http
        reaver {:mvn/version "0.1.2"}  ; jsoup, html parser
        }
}

实际代码：

(require '[clj-http.client :as client])
(require '[clj-http.cookies :as cookies])
(require '[reaver :as html])

(def ua "Mozilla/5.0 (Windows NT 6.1; WOW64; rv:33.0) Gecko/20120101 Firefox/33.0")
(def cs (cookies/cookie-store))

(def http-header {:headers {"user-agent" ua
                            "accept-charset" "utf-8"}
                  :proxy-host "localhost" ;本地代理,用于测试
                  :proxy-port 8080
                  :cookie-store cs
                  :insecure? true})

(defn search
  "搜索关键字kw，返回[{:url :title :desc} ...]
  kw 为搜索关键字
  opt为附加选项:
  :option http请求额外参数
  :max-page 查询的最大页数,默认为3"
  [kw & [opt]]
  (let [base-url "https://www.bing.com"
        option (get opt :option {})
        max-page (get opt :max-page 3)]
    (loop [page 1
           url (str base-url "/search?q=" kw) ;这里没有用url编码，仅为演示用
           result []]
      (if (> page max-page)
        result
        (let [doc (-> (client/get url
                                  (merge http-header option))
                      :body
                      html/parse)
              entrys (html/extract-from doc "li.b_algo"
                                        [:url :title :desc]
                                        "h2 > a" (html/attr "href")
                                        "h2 > a" html/text
                                        "div > p" html/text)
              r (apply conj result entrys)]
          (if-let [next-path (-> (html/select doc "a.sb_pagn" )
                                 (html/attr "href"))]
            (recur (inc page)
                   (str base-url next-path)
                   r)
            r))))))

(def googles (search "google" {:max-page 3}))
(count googles)
;; => 30
(first googles)
;; => {:url "http://www.google.cn/", :title "Google", :desc "2016-12-8 · google.com.hk 请收藏我们的网址"}

可以看到一页返回10个结果，如果要返回50(bing设置里最大结果数量)，则要设置cookie项： "SRCHHPGUSR" 的值为： "NRSLT=50"

下面添加cookie：

(cookies/clear-cookies cs) ;清除之前请求产生的cookies

(def usr-cookie (cookies/to-basic-client-cookie
                 ["SRCHHPGUSR" {
                                :discard false
                                :domain ".bing.com",
                                :path "/",
                                :value "NRSLT=50"
                                :expires (java.util.Date. 9000 1 1)
                                }]))
(cookies/add-cookie cs usr-cookie)

(cookies/get-cookies cs)
;; => {"SRCHHPGUSR" {:discard false, :domain ".bing.com", :expires #inst "10900-01-31T16:00:00.000-00:00", :path "/", :secure false, :value "NRSLT=50", :version 0}}

(def googles (search "google" {:max-page 3}))
(count googles)
;; => 30

可以看到cookie并没有生效，在代理中也可以看到第一次请求时并没有带上添加的cookie。原因是HttpClient默认的Cookie specifications不会把.bing.com匹配到www.bing.com。

下面Api用法可以参考CookieOrigin API文档和cookie package文档:

(import org.apache.http.cookie.CookieOrigin)
(import (org.apache.http.impl.cookie DefaultCookieSpec
                                     RFC6265LaxSpec
                                     RFC6265StrictSpec
                                     RFC2965Spec
                                     RFC2109Spec
                                     NetscapeDraftSpec
                                     IgnoreSpec
                                     BasicClientCookie2))

(def bing-co (CookieOrigin. "www.bing.com" 80 "/" false))

(def match-spec #(.match %1 usr-cookie bing-co))
(def default-spec (DefaultCookieSpec.))
(match-spec default-spec)
;; => false
(def rfc6265-lax-spec (RFC6265LaxSpec.))
(match-spec rfc6265-lax-spec)
;; => false
(def rfc6265-strict-spec (RFC6265StrictSpec.))
(match-spec rfc6265-strict-spec)
;; => false
(def rfc2965-spec (RFC2965Spec.))
(match-spec rfc2965-spec)
;; => true
(def rfc2109-spec (RFC2109Spec.))
(match-spec rfc2109-spec)
;; => true
(def netscape-spec (NetscapeDraftSpec.))
(match-spec netscape-spec)
;; => true

;; 通过测试的几个spec,可以看到，默认只有rfc2*和netscape可以匹配.bing.com
;; 下面设置cookie的attr
(str usr-cookie)
;; => "[version: 0][name: SRCHHPGUSR][value: NRSLT=50][domain: .bing.com][path: /][expiry: Mon Feb 01 00:00:00 CST 10900]"
(.setAttribute usr-cookie BasicClientCookie2/DOMAIN_ATTR "true")
(str usr-cookie)
;; => "[version: 0][name: SRCHHPGUSR][value: NRSLT=50][domain: .bing.com][path: /][expiry: Mon Feb 01 00:00:00 CST 10900]"
;; 从表面上看不出设置了attr的区别，只有匹配时不同：
(.getAttribute usr-cookie "domain") ; DOMAIN_ATTR的值是"domain"
;; => "true"

(def match-spec #(.match %1 usr-cookie bing-co))
(match-spec default-spec)
;; => true
(match-spec rfc6265-lax-spec)
;; => true
(match-spec rfc6265-strict-spec)
;; => true
(match-spec rfc2965-spec)
;; => true
(match-spec rfc2109-spec)
;; => true
(match-spec netscape-spec)
;; => true

;;再看搜索结果
(cookies/clear-cookies cs)
(cookies/add-cookie cs usr-cookie)
(def googles (search "google" {:max-page 3}))
(count googles)
;; => 78
;; 不是每页都有50条,不过重要的是通过.setAttribute,cookie起作用了

具体原因是rfc6265的规定,参考overflow的回答

经过上面的测试，也可以在clj-http中使用netscape的cookie-policy来达到目的,因为standard属于rfc6265-lax,默认也不会匹配:

(def usr-cookie (cookies/to-basic-client-cookie
                 ["SRCHHPGUSR" {
                                :discard false
                                :domain ".bing.com",
                                :path "/",
                                :value "NRSLT=50"
                                :expires (java.util.Date. 9000 1 1)
                                }]))

(cookies/clear-cookies cs)
(cookies/add-cookie cs usr-cookie)
(def googles (search "google" {:max-page 3
                               :option {:cookie-policy :netscape}}))
(count googles)
;; => 78

但是这种方法是不推荐的，还是用.setAttribute比较好。

作者: ntestoc

Created: 2019-01-25 五 20:18