For example,if I have an index “{a:1, b:1, c:-1}”,it works well for find({a:1,b:2}).sort({c:-1}), but  very slowly for find( {a:{$in:[1,2,3]}, b:4}).sort({c:-1}),here’s the output of the explain plan:

{
	"Cursor:BtreeCursor" : a_1_b_1_c_-1,
	"nscanned" : 13567
       	...
	"scanAndOrder" : true,
	...
}

There are many documents said that the sort column should be the last column used in the index,but it’s not always true,especailly if you are doing a range query, such as $lt,$in and etc.it  does complicate the query processing and make it hard to use the index to satisfy the sort conditions.indeed,in this case,you should put the sort field before range query field in your index,for example, “{c:-1,a:1,b:1}”  in this case.

rememberer = request.environ['repoze.who.plugins']['cookie']
identity = {'repoze.who.userid': user.username}
response.headerlist = response.headerlist + \
	rememberer.remember(request.environ, identity) 

Happy birthday, My sweet little EQ.

Wu-Manber算法是一个不错的解决方案，尤其是针对包含有大量关键词的搜索。Wu-Manber算法的C语言实现非常丰富，为了方便如C#，VB.net等编程语言使用该算法，我用C/C++写了一个基于改进的Wu-Manber算法的多关键词搜索组件：wu-manber-com。

Download wu-manber-com

使用方法

以vb.net/c#为例，使用前，先运行regsvr32 stringsearch.dll注册该组件，然后将stringsearch.dll添加到项目引用中。调用示例代码如下：

StringSearchLib.WuManber s = new StringSearchLib.WuManber();
s.AddPatterns("keyword1,keyword2,keyword3", ",", false);
int index = s.Search("... some text for searching...");
Console.WriteLine(index);

该控件不会搜寻所有存在的关键词，碰到第一个匹配的关键词，即返回该匹配的关键词在原文中的位置索引,没有找到则返回-1。

The /MP option can reduce the total time to compile the source files on the command line. The /MP option causes the compiler to create one or more copies of itself, each in a separate process. Then these copies simultaneously compile the source files. Consequently, the total time to build the source files can be significantly reduced.

and here is the screenshot when compiled with /MP flag on an 8 core processor system:

that is because the custom build step generated by QT for each header files that contain the Q_OBJECT macro are dependent on mocinclude.tmp,which being created every time the project is built,the mocinclude.tmp’s date is later than the header files,so the custom build rule will be run every time.

to solve this problem,open %QTDIR%/makespecs/features/moc.prf,comment out  these two lines(add ‘#’’ at the start of line):

!isEmpty(INCLUDETEMP):moc_source.depends += $$INCLUDETEMP
......
!isEmpty(INCLUDETEMP):moc_header.depends += $$INCLUDETEMP

and run qmake –tp vc again to regenerate the project files.

with WriteFIleGather, the disk is always keep busy without entering and leaving from kernel mode. log data can be written to disk at a rate that approaches the limits of the physical characteristics of the hardware device.that’s really boost the I/O performance.

because the CPU is mostly idle during write operation,to maximum the performance,I have compressed the data before writing to the disk to reduce the amount of data to be written,the compress is always faster than the I/O operation,so that I get the compression for “free”.

after these changes,the performance for logging system has increased by over 43%.

if you have any questions/issues,please submit an issue at http://code.google.com/p/openssl-for-windows/issues/list

Download roptipng

To optimization of multiple directories recursively with rOptiPng,simply by using the ** wildcard in file path.
for examples:
roptipng **/*.png
roptipng c:\photos\**\*.png
roptipng c:\photos\**\*.png d:\website\**\*.png
roptipng c:\photos\**\image8.png

if you find it useful, a time saver,or even just mildly interesting,then my effort was worth it.
if you have any suggestions on how to improve it, I’d love to hear it.

Enjoy:)

Feel free to use this tool and give me feedback if you have any.

1>inffas32.asm(649) : error A2070: invalid instruction operands
1>inffas32.asm(663) : error A2070: invalid instruction operands
1>inffas32.asm(720) : error A2070: invalid instruction operands

these errors are due to an issue with Microsoft Macro Assembler ,included with Visual C++ 2005.it refuses to assemble a MOVD instruction with a memory operand with an implied size, and requires that "dword ptr" prefix the memory operand.

to fix these problems,simply adding”DWORD PTR’ before the operands:

movd mm5,dword ptr [esp+4](649)
movd mm7,dword ptr [esi](663)
movd mm7,dword ptr [esi](720)