fasterxml - 最快的XML解析器

1.概述
2.编译安装
3.基本使用
4.性能
5.最后

------------------------------------------------------------

1.概述

fasterxml是一个快速SAX模式的XML解析器。它直接解析XML文本，调用注册的事件函数，快速访问XML中感兴趣的内容。
fasterxml主要用于只读方式挖掘XML数据到应用对象中，如XML配置文件解析、大型XML报文解包。因为它不像DOM模式需要先构造一个完整的文档树，所以速度非常快，几乎接近strlen的性能。
fasterxml非常小巧，整个源代码由一对fasterxml.h,fasterxml.c组成，500行，13KB，不依赖于任何其它库，容易嵌入到您的项目中。

2.编译安装

以Linux操作系统为例，每个目录里的构造脚本均是makefile.Linux，外层目录makefile.Linux递归下级子目录
[code]
$ cd src
$ make -f makefile.Linux install
gcc -g -fPIC -O2 -Wall -Werror -fno-strict-aliasing -I.  -c fasterxml.c
gcc -g -fPIC -O2 -Wall -Werror -fno-strict-aliasing -o libfasterxml.so fasterxml.o -shared -L.
cp -f libfasterxml.so /usr/lib/exlib/
cp -f fasterxml.h /usr/include/fasterxml/
[/code]

3.基本使用

fasterxml使用接口函数只有三个，下面依次说明之

[code]
typedef int funcCallbackOnXmlNode( int type , char *xpath , int xpath_len , int xpath_size , char *node , int node_len , char *properties , int properties_len , char *content , int content_len , void *p );
int TravelXmlBuffer( char *xml_buffer , char *xpath , int xpath_size
		 , funcCallbackOnXmlNode *pfuncCallbackOnXmlNode
		 , void *p );
int TravelXmlBuffer4( char *xml_buffer , char *xpath , int xpath_size
		 , funcCallbackOnXmlNode *pfuncCallbackOnXmlNode
		 , funcCallbackOnXmlNode *pfuncCallbackOnEnterXmlNode
		 , funcCallbackOnXmlNode *pfuncCallbackOnLeaveXmlNode
		 , funcCallbackOnXmlNode *pfuncCallbackOnXmlLeaf
		 , void *p );
[/code]
函数TravelXmlBuffer读入XML文本并快速解析之，当遇到XML树枝或XML树叶时调用pfuncCallbackOnXmlNode呼叫应用访问，应用可根据type确定读到了树枝事件或树叶事件等，以及xpath，筛选访问感兴趣的XML节点。
函数TravelXmlBuffer4是TravelXmlBuffer事件细分版本。
还可根据需要，快速解析访问当前节点的属性集合，函数原型如下
[code]
typedef int funcCallbackOnXmlProperty( char *xpath , int xpath_len , int xpath_size , char *propname , int propname_len , char *propvalue , int propvalue_len , void *p );
_WINDLL_FUNC int TravelXmlPropertiesBuffer( char *properties , int properties_len , char *xpath , int xpath_len , int xpath_size
					   , funcCallbackOnXmlProperty *pfuncCallbackOnXmlProperty
					   , void *p );
[/code]

下面是一个简单的示例
[code]
funcCallbackOnXmlProperty CallbackOnXmlProperty ;
int CallbackOnXmlProperty( char *xpath , int xpath_len , int xpath_size , char *propname , int propname_len , char *propvalue , int propvalue_len , void *p )
{
	printf( "    PROPERTY p[%s] xpath[%s] propname[%.*s] propvalue[%.*s]\n" , (char*)p , xpath , propname_len , propname , propvalue_len , propvalue );
	
	return 0;
}

funcCallbackOnXmlNode CallbackOnXmlNode ;
int CallbackOnXmlNode( int type , char *xpath , int xpath_len , int xpath_size , char *nodename , int nodename_len , char *properties , int properties_len , char *content , int content_len , void *p )
{
	int		nret = 0 ;
	
	if( type & FASTXML_NODE_BRANCH )
	{
		if( type & FASTXML_NODE_ENTER )
		{
			printf( "ENTER-BRANCH p[%s] xpath[%s] nodename[%.*s] properties[%.*s]\n" , (char*)p , xpath , nodename_len , nodename , properties_len , properties );
		}
		else if( type & FASTXML_NODE_LEAVE )
		{
			printf( "LEAVE-BRANCH p[%s] xpath[%s] nodename[%.*s] properties[%.*s]\n" , (char*)p , xpath , nodename_len , nodename , properties_len , properties );
		}
		else
		{
			printf( "BRANCH       p[%s] xpath[%s] nodename[%.*s] properties[%.*s]\n" , (char*)p , xpath , nodename_len , nodename , properties_len , properties );
		}
	}
	else if( type & FASTXML_NODE_LEAF )
	{
		printf( "LEAF         p[%s] xpath[%s] nodename[%.*s] properties[%.*s] content[%.*s]\n" , (char*)p , xpath , nodename_len , nodename , properties_len , properties , content_len , content );
	}

	if( properties && properties[0] )
	{
		nret = TravelXmlPropertiesBuffer( properties , properties_len , xpath , xpath_len , xpath_size , & CallbackOnXmlProperty , p ) ;
		if ( nret )
			return nret;
	}
	
	return 0;
}

...
char	xpath[ 1024 + 1 ] ;
char	*xml_buffer = NULL ;
char	*p = "hello world" ;

memset( xpath , 0x00 , sizeof(xpath) );
nret = TravelXmlBuffer( xml_buffer , xpath , sizeof(xpath) , & CallbackOnXmlNode , p ) ;
free( xml_buffer );
if( nret && nret != FASTXML_INFO_END_OF_BUFFER )
{
	printf( "TravelXmlTree failed[%d]\n" , nret );
	return nret;
}
...
[/code]

执行之，输出为
[code]
$ cat test_basic.xml
<?xml version="1.0" encoding="GBK"?>
<root>
	<leaf>content</leaf>
	<sub_branch>
		<sub_leaf>sub_content</sub_leaf>
	</sub_branch>
	<self_branch />
	<self_branch2/>
	<branch_has_property propname1=propvalue1>
		<leaf_has_property propname1="prop value1" propname2="prop value2" >content</leaf_has_property>
	</branch_has_property>
</root>
$ test_fasterxml test_basic.xml
BRANCH       p[hello world] xpath[/xml] nodename[xml] properties[ version="1.0" encoding="GBK"?]
    PROPERTY p[hello world] xpath[/xml.version] propname[version] propvalue[1.0]
    PROPERTY p[hello world] xpath[/xml.encoding] propname[encoding] propvalue[GBK]
ENTER-BRANCH p[hello world] xpath[/root] nodename[root] properties[]
LEAF         p[hello world] xpath[/root/leaf] nodename[leaf] properties[] content[content]
ENTER-BRANCH p[hello world] xpath[/root/sub_branch] nodename[sub_branch] properties[]
LEAF         p[hello world] xpath[/root/sub_branch/sub_leaf] nodename[sub_leaf] properties[] content[sub_content]
LEAVE-BRANCH p[hello world] xpath[/root/sub_branch] nodename[sub_branch] properties[]
BRANCH       p[hello world] xpath[/root/self_branch] nodename[self_branch] properties[ /]
BRANCH       p[hello world] xpath[/root/self_branch2] nodename[self_branch2] properties[/]
ENTER-BRANCH p[hello world] xpath[/root/branch_has_property] nodename[branch_has_property] properties[ propname1=propvalue1]
    PROPERTY p[hello world] xpath[/root/branch_has_property.propname1] propname[propname1] propvalue[propvalue1]
LEAF         p[hello world] xpath[/root/branch_has_property/leaf_has_property] nodename[leaf_has_property] properties[ propname1="prop value1" propname2="prop value2" ] content[content]
    PROPERTY p[hello world] xpath[/root/branch_has_property/leaf_has_property.propname1] propname[propname1] propvalue[prop value1]
    PROPERTY p[hello world] xpath[/root/branch_has_property/leaf_has_property.propname2] propname[propname2] propvalue[prop value2]
LEAVE-BRANCH p[hello world] xpath[/root/branch_has_property] nodename[branch_has_property] properties[]
LEAVE-BRANCH p[hello world] xpath[/root] nodename[root] properties[]
[/code]

客户可以改造事件回调函数，实现快速遍历访问XML，比如根据XML某些XPATH数据填充一个C结构体相关成员。

4.性能

fasterxml的性能非常高，下面是fasterxml与目前宣称最快XML解析器rapidxml的压测比较
压测用XML文件test_big_and_lang.xml约622KB(17870行)，里面包含大量节点、重复节点、各种编码以及少量注释。
[code]
~/exsrc/fasterxml-1.0.0/test $ ls -l test_big_and_lang.xml
-rw-r--r-- 1 calvin calvin 622511 Sep 21 10:51 test_big_and_lang.xml
~/exsrc/fasterxml-1.0.0/test $ wc test_big_and_lang.xml
17870  26332 622511 test_big_and_lang.xml
[/code]

首先是fasterxml开跑10次，取最好成绩
[code]
~/exsrc/fasterxml-1.0.0/test $ time ./press_fasterxml test_big_and_lang.xml 1000
Elapse 5 seconds
real    0m5.130s
user    0m3.402s
sys     0m0.007s
[/code]
遍历测试XML文件所有树枝、树叶、属性1000次，耗时约5秒。

然后是rapidxml开跑10次，取最好成绩
[code]
~/exsrc/fasterxml-1.0.0/test_rapidxml $ time ./press_rapidxml ../test/test_big_and_lang.xml 1000
Elapse 19 seconds

real    0m19.169s
user    0m0.923s
sys     0m18.265s
[/code]
遍历测试XML文件所有树枝、树叶、属性1000次，耗时约19秒。

可见fasterxml比rapidxml快了约3倍，fasterxml夺取了XML(SAX)性能桂冠。

5.最后

欢迎使用fasterxml，如果你使用中碰到了问题请告诉我，谢谢 ^_^
首页传送门 : [url]http://git.oschina.net/calvinwilliams/fasterxml[/url]
作者邮箱   : calvinwilliams.c@gmail.com