From f3e5a3158547ed32073f104b1bb37ff77e872376 Mon Sep 17 00:00:00 2001
From: Sebastian Nagel <snagel@apache.org>
Date: Wed, 10 Oct 2012 21:54:37 +0000
Subject: [PATCH] NUTCH-706 (applied correct patch)

git-svn-id: https://svn.apache.org/repos/asf/nutch/trunk@1396817 13f79535-47bb-0310-9956-ffa450edef68
---
 conf/regex-normalize.xml.template                          | 2 +-
 .../sample/regex-normalize-default.test                    | 7 ++++++-
 .../urlnormalizer-regex/sample/regex-normalize-default.xml | 2 +-
 3 files changed, 8 insertions(+), 3 deletions(-)

diff --git a/conf/regex-normalize.xml.template b/conf/regex-normalize.xml.template
index 4bc1611a6a..d14174eaac 100644
--- a/conf/regex-normalize.xml.template
+++ b/conf/regex-normalize.xml.template
@@ -29,7 +29,7 @@
 
 <!-- removes session ids from urls (such as jsessionid and PHPSESSID) -->
 <regex>
-  <pattern>([;_]?\b((?i)l|j|bv_)?((?i)sid|phpsessid|sessionid)=.*?)(\?|&amp;|#|$)</pattern>
+  <pattern>(?i)(;?\b_?(l|j|bv_)?(sid|phpsessid|sessionid)=.*?)(\?|&amp;|#|$)</pattern>
   <substitution>$4</substitution>
 </regex>
 
diff --git a/src/plugin/urlnormalizer-regex/sample/regex-normalize-default.test b/src/plugin/urlnormalizer-regex/sample/regex-normalize-default.test
index 09fb9188c7..dbb08b5680 100644
--- a/src/plugin/urlnormalizer-regex/sample/regex-normalize-default.test
+++ b/src/plugin/urlnormalizer-regex/sample/regex-normalize-default.test
@@ -11,8 +11,13 @@ http://www.foo.com/foo.html;jsessionid=1E6FEC0D14D044541DD84D2D013D29ED http://w
 http://www.foo.com/foo.html?param=1&another=2;jsessionid=1E6FEC0D14D044541DD84D2D013D29ED http://www.foo.com/foo.html?param=1&another=2
 http://www.foo.com/foo.html;jsessionid=1E6FEC0D14D044541DD84D2D013D29ED?param=1&another=2 http://www.foo.com/foo.html?param=1&another=2
 http://www.foo.com/foo.php?&x=1&sid=xyz&something=1 http://www.foo.com/foo.php?x=1&something=1
-# but NewsId is not a session id (NUTCH-706, NUTCH-1328)
+http://www.foo.com/foo.html?_sessionID=824A6C0A13a7e11205wxN28F44E3 http://www.foo.com/foo.html
+http://www.foo.com/foo.php?_sessionid=qmyrcedt&outputformat=html&path=/3_images/foo http://www.foo.com/foo.php?outputformat=html&path=/3_images/foo
+http://www.foo.com/foo.php?_pid=2&_spid=0&lang=en&_sessionid=e36902d5bb2d0d922fc24b43 http://www.foo.com/foo.php?_pid=2&_spid=0&lang=en
+http://www.foo.com/foo.php?app=content&content=overview&lang=en&_sid=587fba8f825b05844526519fdb7d75c8&b=35&m=47 http://www.foo.com/foo.php?app=content&content=overview&lang=en&b=35&m=47
+# but NewsId (and similar) is not a session id (NUTCH-706, NUTCH-1328)
 http://www.foo.com/fa/newsdetail.aspx?NewsID=1567539 http://www.foo.com/fa/newsdetail.aspx?NewsID=1567539
+http://www.foo.com/home.cfm?language=en&country=uk&addressid=250646&pagingpos=0 http://www.foo.com/home.cfm?language=en&country=uk&addressid=250646&pagingpos=0
 
 # test removal default pages
 http://www.foo.com/home/index.html http://www.foo.com/home/
diff --git a/src/plugin/urlnormalizer-regex/sample/regex-normalize-default.xml b/src/plugin/urlnormalizer-regex/sample/regex-normalize-default.xml
index 14e2b4fb8b..69a8eba26a 100644
--- a/src/plugin/urlnormalizer-regex/sample/regex-normalize-default.xml
+++ b/src/plugin/urlnormalizer-regex/sample/regex-normalize-default.xml
@@ -13,7 +13,7 @@
 
 <!-- removes session ids from urls (such as jsessionid and PHPSESSID) -->
 <regex>
-  <pattern>([;_]?\b((?i)l|j|bv_)?((?i)sid|phpsessid|sessionid)=.*?)(\?|&amp;|#|$)</pattern>
+  <pattern>(?i)(;?\b_?(l|j|bv_)?(sid|phpsessid|sessionid)=.*?)(\?|&amp;|#|$)</pattern>
   <substitution>$4</substitution>
 </regex>