【Python笔记】python第三方库lxml源码安装&&使用笔记

python第三方库lxml使用笔记

1. 源码安装
        由于想折腾一下，故选择源码安装。
        从官网下载lxml最新版本(lxml-3.3.3)源码，linux机器可用curl工具拉取lxml源码，参考命令如下

curl -v --cacert ./https-ca/cacert.pem https://pypi.python.org/packages/source/l/lxml/lxml-3.3.3.tar.gz#md5=f2675837b4358a5ecab5fd9a783fd0e5 -o lxml-3.3.3.tar.gz


        其中：&＃8211;cacert ./https-ca/cacert.pem指定我机器上https认证文件的全路径，其它机器则该路径需要修改。关于认证文件的获取方法，可参考我之前写的这篇笔记。

        下载完成后解压，由INSTALL.txt可知，这货安装前需要预先装好libxml2和libxslt两个包，于是继续折腾：从xmlsoft.org官网下载libxml2和libxslt的最新版源码后分别解压安装。

        安装时的注意事项：

        a. 先装libxml2，后装libxslt，否则安装libxslt时报错

        b. 执行./configure时，可通过参数&＃8211;with-python=[DIR]指定自己想bind的python全路径(在多人共用的机器上很有必要指定)

        c. make libxslt源码前，需通过./configure中指定python全路径及libxml2的3个参数，参考命令如下：

/configure --prefix=/home/slvher/tools/3rd-libs-for-py/libxslt-1.1.28/ --with-python=/home/slvher/tools/python-2.7.5/bin/python --with-libxml-prefix=/home/slvher/tools/3rd-libs-for-py/libxml2-2.9.0 --with-libxml-include-prefix=/home/slvher/tools/3rd-libs-for-py/libxml2-2.9.0/include/ --with-libxml-libs-prefix=/home/slvher/tools/3rd-libs-for-py/libxml2-2.9.0/lib/

        libxml2和libxslt安装成功后，进入lxml源码目录，在某临时目录(假设为/home/slvher/tools/lxml-package)下安装lxml，参考命令如下：

python ./setup.py install root=/home/slvher/tools/lxml-package --with-xml2-cOnfig=/home/slvher/tools/3rd-libs-for-py/libxml2-2.9.0/bin/xml2-config --with-xslt-cOnfig=/home/slvher/tools/3rd-libs-for-py/libxslt-1.1.28/bin/xslt-config


        将该临时安装目录下lxml和lxml-3.3.3-py2.7.egg-info两个文件夹cp至python安装目录的site-packages路径下，至此，lxml安装完成

        个人体会：如果机器系统有现成的包管理工具或easy_install之类的工具支持一键安装，并且自己不想折腾，建议不要通过源码安装，对于老手来说，这个过程实在是有点浪费时间 -_-

2. 测试是否可用
        lxml及其egg-info拷贝到python的site-packages目录后，可以在python中试用lxml：

>>> from lxml import etree
Traceback (most recent call last):
File "", line 1, in
ImportError: /home/slvher/tools/python-2.7.5/lib/python2.7/site-packages/lxml/etree.so: undefined symbol: exsltStrXpathCtxtRegister

        符号找不到，难道是安装有问题?

        进入python的site-packages目录，终端输入命令ldd lxml/etree.so，输出结果如下：

lxml/etree.so: /usr/lib64/libxslt.so.1: no version information available (required by lxml/etree.so)
lxml/etree.so: /usr/lib64/libxslt.so.1: no version information available (required by lxml/etree.so)
lxml/etree.so: /usr/lib64/libxslt.so.1: no version information available (required by lxml/etree.so)
lxml/etree.so: /usr/lib64/libxslt.so.1: no version information available (required by lxml/etree.so)
lxml/etree.so: /usr/lib64/libxslt.so.1: no version information available (required by lxml/etree.so)
lxml/etree.so: /usr/lib64/libxslt.so.1: no version information available (required by lxml/etree.so)
lxml/etree.so: /usr/lib64/libxslt.so.1: no version information available (required by lxml/etree.so)
libxslt.so.1 => /usr/lib64/libxslt.so.1 (0x00007fd95d51f000)
libexslt.so.0 => /usr/lib64/libexslt.so.0 (0x00007fd95d40e000)
libxml2.so.2 => /home/slvher/tools/3rd-libs-for-py/libxml2-2.9.0/lib/libxml2.so.2 (0x00007fd95d1c2000)
libz.so.1 => /usr/lib64/libz.so.1 (0x00007fd95d0ae000)
libm.so.6 => /lib64/tls/libm.so.6 (0x00007fd95cf28000)
libpthread.so.0 => /lib64/tls/libpthread.so.0 (0x00007fd95ce13000)
libc.so.6 => /lib64/tls/libc.so.6 (0x00007fd95cbde000)
libgcrypt.so.11 => /usr/lib64/libgcrypt.so.11 (0x00007fd95ca94000)
libgpg-error.so.0 => /usr/lib64/libgpg-error.so.0 (0x00007fd95c991000)
libdl.so.2 => /lib64/libdl.so.2 (0x00007fd95c88d000)
/lib64/ld-linux-x86-64.so.2 (0x000000302ad00000)
libnsl.so.1 => /lib64/libnsl.so.1 (0x00007fd95c776000)

        可见，etree.so中依赖的libxslt.so.1和libexslt.so.0指向了机器/usr/lib64/路径下老版本的so库

       
解决办法如下：

export LD_LIBRARY_PATH=/home/slvher/tools/3rd-libs-for-py/libxml2-2.9.0/lib/:/home/slvher/tools/3rd-libs-for-py/libxslt-1.1.28/lib/:$LD_LIBRARY_PATH

        然后在python终端执行from lxml import etree

        一切正常

       
提示：

        若想在其它机器上使用lxml又不想安装，则可以将已装有lxml包的python的site-packages目录中与lxml有关的2个文件夹拷贝至目标机器python的site-packages目录下，然后将etree依赖的3个so库拷贝至目标机器的某指定路径下，最后通过LD_LIBRARY_PATH将该路径包含至python包对so的搜索路径中即可。

【参考资料】
1. lxml官网文档

========================== EOF ==========================