用 .NET 进行编程就像是在麦当劳做烹饪。那里有全套神奇的设备能够自动的造出所有的东西。只有你按一下正确的按钮，跟着提示灯的指示，你就可以批量的制造出完美无缺的1.6盎司的汉堡，比地球上的任何其他人都要快。

然而，如果你想做出一个1.7盎司的汉堡，很抱歉，你做不到。很简单，没有这样的按钮。 馅饼是预先做成了那样的尺寸。它们做出来之后就被冷冻，所有你无法把它们揉到一起做成其它尺寸，解冻机和烹饪机是一体的，你没有办法在它们之间进行拦截。 麦当劳的厨房产出的食品精确的和麦当劳的菜单保持一致的——这种模式使你的大脑不需要任何的思考。可是，它不能偏离菜单，对烹饪机器的任何你妄想的压挤变 形都会导致它停止工作，而被送回返厂维修。

sudo apt-get install tomcat7 apache2 libapache2-mod-jk

2. 修改workers.properties

sudo nano /etc/libapache2-mod-jk/workers.properties

把其中一行：

workers.tomcat_home=/usr/share/tomcat6

替换为：

workers.tomcat_home=/usr/share/tomcat7

3. 配置tomcat7

sudo nano /etc/tomcat7/server.xml

把port 8009这行的默认注释给去掉：


<Connector port="8009" protocol="AJP/1.3" redirectPort="8443" />

并添加host中添加自己的应用：

<Context path="/geonetwork"
docBase="/var/www/xining/geonetwork/web/geonetwork/"
crossContext="false"
debug="0"
reloadable="false" />

4. 修改apache2 site配置

sudo nano /etc/apache2/sites-enabled/xining

在其中，添加：

JkMount /geonetwork/* ajp13_worker
JkMount /geonetwork ajp13_worker

以让mod_jk生效。

5. 重启相关服务

sudo service apache2 restart
sudo service tomcat7 restart

6. 安装sun-java6-jdk
我们应用的GEONETWORK软件在使用默认的OPENJDK时会有错误产生，如不能登录、不能查询，因此需要安装SUN JDK。

sudo apt-get install python-software-properties
sudo add-apt-repository ppa:ferramroberto/java
sudo apt-get update
sudo apt-get install sun-java6-jdk
sudo update-alternatives –config java

安装完成后，可以通过java -version命令确认：

java version “1.6.0_26″
Java(TM) SE Runtime Environment (build 1.6.0_26-b03)
Java HotSpot(TM) 64-Bit Server VM (build 20.1-b02, mixed mode)

update metadata m set source=s.uuid from source s right join datasource d on s.id=d.sourceid where m.uuid=d.uuid

bing api默认的请求参数是上面这样的。后面经发现’Options’=>’EnableHighlighting’,有问题。如果指定
Bing Search API时如果指定Options=EnableHighlighting（如下），则返回的Jason字符串中在关键字两边会有全角空格字符，而全角空格字符在IE下显示为很长的空格，在Firefox上显示为乱码。

最终解决方方案：在查询請求中去除Options=EnableHighlighting的参数设置后正常。

After download the 1.0.1 version, and try some tests in its code. There is no problem in the text replacing, but if I insert into the odt file an image, the OpenOffice (What I use is LibreOffice) always tell me this file need repair, although the repair result nice. Dig into another search, find this problem and solution in the author’s blog.

Use the latest version (1.3) in the author’s blog, this bug is disappeared.

There is another requirement in my site yet, maybe it is not a problem: The image file which is dynamic produced has not a extension in the url, So the code in odtphp: setImage must change its behavior here.

So I patched the code, and released it here (it’s GPL license):

wlx@wlxpc:~$ diff odf.php odtphp-1.0.1_modified_files/odf.php
111a112,113
> $filename = strtok(strrchr($value, '/'), '/.');
> $file = substr(strrchr($value, '/'), 1);
116,118d117
< $filename=md5($value);;
< $imgext=array('.gif','.jpg','.png','.swf','.psd','.bmp','.tif','.tif','.jpc','.jp2','.jpx','.jb2','.swc','.iff','.wbmp','.xbm','.ico');
< $file=$filename.$imgext[$size[2]];
287,289c286
< //$this->file->addFile($imageKey, 'Pictures/' . $imageValue);
< $this->file->addFromString('Pictures/' . $imageValue,file_get_contents($imageKey));
<
---
> $this->file->addFile($imageKey, 'Pictures/' . $imageValue);

In short, what I do: change the filenmae to a hash string (md5(source)), and save it with file_get_contents function (which could parse the url).

And, a big Thanks to you and your odtphp project: Vikas Mahajan.
Download patch file: odf.php.patch

1. 实际测试平台
UBUNTU 11.04，DELL T7400 INTEL XEON X5450@3.0GHz（双CPU，各四核，测试时只用1核）。
g++ 4.5.2
it++ 4.0.7 （系统自带）
armadillo 1.2.0 (手工安装）
2. 编译方法
armadillo采用的是如下编译方法：

g++ -o test -larmadillo -O2 -fwhole-program xxx.cpp

IT++采用的是如下编译方法：

g++ -o test -litpp -O2 -fwhole-program xxx.cpp

我们的矩阵类用的是如下的编译方法：

g++ -o test -larmadillo -O2 -fwhole-program -ldas/include/ ldas/src/util/exception.cpp cas1.cpp

3. 测试方法及代码
主要测试矩阵的加法运算和乘法运算，构建了ABZ三个1000*1000矩阵，分别测试了Z=A+B、Z=A*B以及Z=A+B+C、Z=A*B*C各100次，比较其计算时间。
armadillo代码：

#include <iostream>
#include <armadillo>
int main() {
int size = 1000; int N = 100;
//Armadillo
// size and N are specified by the user on the command line
arma::mat A = arma::randu(size,size);
arma::mat B = arma::randu(size,size);
arma::mat C = arma::randu(size,size);
arma::mat Z = arma::zeros(size,size); int i;
arma::wall_clock timer; timer.tic();
for(i=0; i<N; ++i) Z = A+B+C; // or Z = A+B+C ... etc
std::cout << "armadillo: time taken for addition = " << timer.toc() / double(N) << std::endl;
timer.tic();
for(i=0; i<N; ++i) Z = A*B*C; // or Z = A+B+C ... etc
std::cout << "armadillo: time taken for multiplication = " << timer.toc() / double(N) << std::endl;
}

IT++测试代码：

#include <itpp/itbase.h>
#include <itpp/base/random.h>
#include <itpp/base/timing.h>
int main()
{ int size = 1000; int N = 100;
//IT++
itpp::mat A2 = itpp::randu(size,size);
itpp::mat B2 = itpp::randu(size,size);
itpp::mat C = itpp::randu(size,size);
itpp::mat Z2 = itpp::zeros(size,size); int i;
itpp::Real_Timer timer; timer.tic();
for(i=0; i<N; ++i) Z2 = A2+B2+C;
std::cout << "IT++: time taken for addition = " << timer.toc() / double(N) << std::endl;
timer.tic();
for(i=0; i<N; ++i) Z2 = A2*B2*C; // or Z = A+B+C ... etc
std::cout << "IT++: time taken for multiplication = " << timer.toc() / double(N) << std::endl;
}

自己实现的矩阵类：

#include <iostream>
#include <armadillo>
#include "matrix.hpp"
#include "exception.h"
int main() {
int size = 1000; int N = 100;
//Armadillo
// size and N are specified by the user on the command line
ldas::Matrix<double> A(size,size);
for(int i=0;i<size;i++)
for(int j=0;j<size;j++) A.set(i+1,j+1,rand()/RAND_MAX);
ldas::Matrix<double> B(size,size);
for(int i=0;i<size;i++) for(int j=0;j<size;j++) B.set(i+1,j+1,rand()/RAND_MAX);
ldas::Matrix<double> C(size,size);
for(int i=0;i<size;i++) for(int j=0;j<size;j++) C.set(i+1,j+1,rand()/RAND_MAX);
ldas::Matrix<double> Z(size,size);
arma::wall_clock timer; timer.tic();
for(int i=0; i<N; ++i) Z = A+B; // or Z = A+B+C ... etc
std::cout << "our Matrix: time taken for addition = " << timer.toc() / double(N) << std::endl;
timer.tic();
for(int i=0; i<N; ++i) Z = A*B; // or Z = A+B+C ... etc
std::cout << "our Matrix: time taken for multiplication = " << timer.toc() / double(N) << std::endl;
}

4. 测试结果
A+B&A*B
armadillo: time taken for addition = 0.00803794
armadillo: time taken for multiplication = 0.117536
IT++: time taken for addition = 0.0119361
IT++: time taken for multiplication = 0.119165
our Matrix: time taken for addition = 0.029522
our Matrix: time taken for multiplication = 17.4216

A+B+C&A*B*C
armadillo: time taken for addition = 0.0107693
armadillo: time taken for multiplication = 0.230288
IT++: time taken for addition = 0.0254268
IT++: time taken for multiplication = 0.237969
our Matrix: time taken for addition = 0.05127
our Matrix: time taken for multiplication = 34.428

5. 测试结果分析
如armadillo作者自己所言，armadillo的计算速度最快，矩阵的加法，特别有长算式的计算，armadillo的优势非常明显。我们自己的矩阵类由于乘法是采用常规方法实现的，没有使用BLAS，因此效率差距很明显。

6. 参考文献
it++ vs armadillo

{
"translatorID":"d5770df3-b41a-45cb-bb3a-261948c9af49",
"translatorType":4,
"label":"SCAR",
"creator":"Liangxu Wang<wangliangxu＠gmail.com>",
"target":"http://www.scar.ac.cn/hhkxen/ch/reader",
"minVersion":"1.0",
"maxVersion":"",
"priority":100,
"inRepository":true,
"lastUpdated":"2009-5-27 22:33:00"
}

if(detectWeb(doc, url) == "multiple") {
Zotero.debug("Enter multiple~");
// TODO: implement the multiple function.
} else {
var urls = [url];
}
Zotero.debug(urls);
// 下面对每条url进行解析
Zotero.Utilities.processDocuments(urls, scrape, function() { Zotero.done(); });
Zotero.wait();
}
也可以直接下载：scar.js

int main()
{
float t=0.000000277;
cout<<t<<endl;
return 0;
}
在LINUX上：

2.77e-07

而在WINDOWS上结果如下：

2.77e-007