转载一个文章 “Python vs. PHP”

Posted on 30th March 2005 by Nio in 程序人生

PHP Everywhere 贴出一篇《Python never had a chance against PHP》的文章，这并不是在对 Python 进行挑衅，而是较为客观地对 PHP 和 Python 进行分析，认为在 Web 应用方面，Python 比不上 PHP：

Python is not a template language, in the sense that you cannot mix code and html easily. PHP is a wonderfully flexible in this respect.

Python is a so-so string processing language. One reason being it treats strings as immutable. PHP has much better string processing facilities: embedded "$var in strings", mutable strings, auto-conversion of other data types to strings, output buffering, etc.

PHP's documentation is cleaner and much easier to understand than Python's. Probably because PHP is a much simpler language.

PHP has tighter integration of a lot of web related stuff. For example, HTTP and SERVER variables.

对于各种语言，我的观点从来都是：存在即是合理的，每个语言都有其应用的空间。就对于 Python 而言，在 Web 方面不如 PHP，但其在通用编程方面功能却要比 PHP 强大得多。而现在流行 Web 编程，所以 PHP 这方面的优势也就决定了增长势头要比 Python 猛得多。

关于此文章的评论也很值得一看。I'll certainly concede that PHP has some important advantages over Python, especially in terms of deployment, isolation, and the ease of starting projects. But I can't agree with much of these particular reasons.

* Templating language: this is true, though you can template perfectly well in Python ― there exist PHP-like systems, as well as templating languages implemented on top of Python. Python itself isn't a templating language so there's no single convention. Of course, even PHP has things like Smarty.

* Python is way, way, WAY better at handling strings compared to PHP. This is the thing that has always driven me crazy in PHP. The $'s are nice ― except for the quoting rules so that $'s aren't interpreted in single quoted strings. That drives me nuts. Python does very good auto-conversion in interpolated strings (%s), and explicit conversion is easy enough. Python also has lots of techniques for buffering ― the specific pattern that PHP uses isn't common, but it's easy to do anyway (StringIO). Python has a concise and fast set of string methods, it has decent regular expression support (more convenient than PHP's at least), and it has really good unicode support. Oh, and NO MAGIC QUOTING. Quoting bugs don't come up that often in Python.

* I don't know what to say about the documentation. I like php.net's comments. I find the original documentation quite poor, and obviously the structure of the language is non-existant, lacking namespaces, modules, classes for most (all?) core types, and other basic organization. Also, most of the Python standard library is implemented in Python, and the source is readable. But it's not an uncommon complaint, so apparently it's something that bothers people.

* Don't good PHP programmers avoid the HTTP and SERVER variables? (register globals or whatever it is) If you just mean a global variable with those values from the request ($_POST?), that's easy in Python, and some frameworks use that. Python works with markup, like HTML and XML, better than PHP ― that's the kind of thing I'd think of if I was thinking about web integration.

� Ian BickingPlease don't forget:

* Smarty is just a rework of PHP itself - its a templating language written in a language already capable of templating. A pointless project from the start, in my opinion, which increases the workload, development time, execution time, and adds an aditional layer where issues can occur.

* Single quotes are there for a reason - so that we don't have to escape special characters all the time. I use single quotes and concattenation for everything, which can be a major speed increase when working with strings in PHP. We can, and its recommended to, turn Magic Quoting off.

* PHP has released a version of its original documentation which incorporates the comments from the website. I find this very useful. However, it only works on windows.

* As for the structure of PHP… IMHO it makes it easy for people to get started, and then (if they wish, as I did) to move to more structured development techniques, then on to other lanugages. I've recently started to look at Python as my next development language, after PHP, JavaScript (Client-side), XSL, and XQuery. Without PHP's quick learning curve, I don't think I'd have ever gotted started.

* No, we don't avoid the global variables, we just avoid the depricated ones. $HTTP_SERVER_VARS and the like. Register globals is the method that allows any parameter passed to a script in GET or POST format to be instantly accessible within the script. This is a serious security threat, and therefore we access them now through the $_POST and $_GET arrays. Variables such as $_SERVER allow us access to information which is passed around via http headers from client to server. Very handy.

* With the advent of PHP5, working with HTML and XML has become very easy. We can now quickly populate objects with a DOM, and modify elements/nodes and attributes with ease. We can also perform XPath queries easily on them. XSLT has been around for a while too, so a lot on the time things can be collected and translated without PHP doing much work.

PHP has its place, and in the future I hope that it will grow to incorporate some of the benefits of other languages, such as persistance. But for the moment while developing for the web, I'll first look at whether PHP can provide a solution before considering other platforms.

来源：http://www.nioxiao.com/archives/250

转载：Facebook 架构学习

在 QCon 2008 (旧金山站) 上Facebook 做的这个技术分享有不少值得借鉴的东西。所以，暂停对 QCon 北京的回顾，临时插播一贴。

设计原则

• 尽可能的使用开源软件，并且在需要优化的时候进行优化
• Unix 哲学。包括，模块化原则；整合化原则；清晰化原则等
• 任何组件具备扩展性
• 最小化故障影响
• 简化，简化，简化！

架构概览

Facebook 是 LAMP 的坚定支持者，也差不多是用 LAMP (或许用 LAM²P 更适合) 实现的最大的动态站点。

基础组件加上服务，中间用自己实现的一些工具进行粘合。其中关于运维细节的事情基本不会说出来的，这是很多公司的软实力所在。

PHP 经验

参见《Facebook 的 PHP 性能与扩展性》

MySQL 经验

• 主要用于做 Key-Value 类型的存储操作，数据随机分布在多台逻辑实例上，访问多数基于全局 ID 。
• 逻辑实例分散在多台物理主机上(超过1800台)，负载均衡在物理层进行。
• 不做读复制。
• 尽量不做逻辑数据迁移(成本太高)。
• 不做 JOIN 操作 (豆瓣在 QCon 上也阐述了这一点)。数据是随机分布的，关联操作反而带来了极大的复杂度。
• 对于数据访问，主要的操作集中在最新的数据上，针对这部分做优化，旧的数据进行归档。
• 在中心 DB 绝不存储非静态数据。
• 使用服务或者 Memcached 进行全局查询。

Memcached 经验

参见我以前的笔记：Facebook 的 Memcached 扩展经验。Facebook 对 Memcached 做了不小的改进。另外，顺便说一下，前两天 Memcached 刚在 1.2.7 发布几天之后又发布了新版本 1.2.8，修正了一些问题。

一个比较有价值的是关于个人页面数据的获取的描述。这个就完全是需要做单页面 Benchmark 的细致活儿了，可能还需要产品经理能够理解工程师的"抵抗"。

• 获取个人信息数据：通过Cache，隐性通过用户所在的 DB 获取(基于 User-ID 获知 DB)
• 获取朋友连接信息：通过Cache，否则的话通过DB(基于 User-ID 获知 DB)
• 并行抓取每个朋友的 10个照片相册 ID ，从Cache抓取，如果失效，再从 DB 抓取(基于相册 ID)
• 并行抓取最近相册中的照片数据
• 运行PHP 把整个业务逻辑跑出来
• 返回数据给用户

然后是对 Facebook 非 LAMP 体系的东西做了一番介绍，基本上也开源了。最后参考两个架构图。

Facebook NewsFeed 的架构示意图

Facebook 搜索功能的架构示意图

管中窥豹，盲人摸象而已。

--EOF--

By Fenng on Apr 12, 2009 4:10 PM | Twitter | Del.icio.us | Comments (5) | Arch | Edit

来源：http://passing.tk/index.php?q=YUhSMGNEb3ZMM2QzZHk1a1ltRnViM1JsY3k1dVpYUXZZWEpqYUM5bVlXTmxZbTl2YTE5aGNtTm9YMjV2ZEdVdWFIUnRiQT09（唉！理解一下吧！国内上网的限制！）

Gentoo下安装Oracle 11g

折腾了一个下午，终于在gentoo下安装成功。安装过程都是记忆中仅存的，系统环境为去年某时装的gentoo 2007.0 amd64，由于gentoo下没有安装桌面环境，所以在一台windows的机器上装了xming，设置xming的启动参数增加-ac，否则gentoo在设置了DISPLAY后会提示connect refuse的信息。

安装前的准备工作：

1： 增加用户和组

groupadd -g 10000 dba

groupadd -g 10001 oinstall

useradd -u 10000 -g dba -G dba,oinstall -s /sbin/bash -m -d /oracle oracle

2：安装一些需要（也许需要，因为除了错误就找解决的方法，安装了以下的）软件包

emerge sun-jdk libaio logger libstdc++-v3 app-emulation/emul-linux-x86-baselibs app-emulation/emul-linux-x86-xlibs

3：设置oracle用户的.bash_profile

export ORACLE_BASE=/oracle
export ORACLE_HOME=$ORACLE_BASE/11g
export ORA_CRS_HOME=$ORACLE_BASE/crs
export ORACLE_PATH=$ORACLE_BASE/common/oracle/sql:.:$ORACLE_HOME/rdbms/admin
export ORACLE_SID=orcl1
export PATH=$ORACLE_HOME/bin:$ORA_CRS_HOME/bin:${PATH}:$HOME/bin
export PATH=${PATH}:/usr/bin:/bin:/usr/bin/X11:/usr/local/bin
export PATH=${PATH}:$ORACLE_BASE/common/oracle/bin
export ORACLE_TERM=xterm
export TNS_ADMIN=$ORACLE_HOME/network/admin
export ORA_NLS10=$ORACLE_HOME/nls/data
export LD_LIBRARY_PATH=${LD_LIBRARY_PATH}:/lib:/usr/lib:/usr/local/lib:$ORACLE_HOME/oracm/lib:$ORACLE_HOME/lib
export LIBPATH=$LIBPATH:$ORA_CRS_HOME/lib:$ORACLE_HOME/lib
export CLASSPATH=${CLASSPATH}:$ORACLE_HOME/rdbms/jlib:$ORACLE_HOME/jlib:$ORACLE_HOME/network/jlib:$ORACLE_HOME/JRE
export THREADS_FLAG=native
export TEMP=/tmp
export TMPDIR=/tmp
export NLS_LANG=american_america.WE8ISO8859P1
export DISPLAY=10.60.4.137:0
export EDITOR=vi

4：sysctl

  　　kernel.shmall = 2097152 　　kernel.shmmax = 2147483648   　　kernel.shmmni = 4096 　　kernel.sem = 250 32000 100 128   　　fs.file-max = 65536 　　net.ipv4.ip_local_port_range = 1024 65000   　　net.core.rmem_default=262144 　　net.core.wmem_default=262144   　　net.core.rmem_max=262144 　　net.core.wmem_max=262144

5：安装oracle

su – oracle

cd database

./runInstaller

安装过程中碰到的问题：

  Error 1:     undefined reference to __pthread_unwind@GLIBC_PRIVATE 解决方法:     编辑$ORACLE_HOME/lib/sysliblist,在-ldl -lm -lpthread -lnsl -lirc -lipgo后加-lrt  Error 2:     cannot find -lagtsh 解决方法:     将database/stage/Components/oracle.rdbms.util/11.1.0.6.0/1/DataFiles/filegroup14.jar文件解压缩,copy 32位的libagtsh.so.1.0文件到$ORACLE_HOME/lib32目录中  Error 3:     skipping incompatible /usr/lib64/libpthread_nonshared.a when searching for /usr/lib64/libpthread_nonshared.a 解决方法:     修改$ORACLE_HOME/bin/genclntsh:        [ "$1" = "lib32" ] && ULIB="lib32" && LOOP="DONE" && CF=-m32 && USRLIB32=-L/usr/lib32        LD="gcc ${CF} -shared -Wl,-relax ${STUBS} ${USRLIB32} -L${OLIB}"     修改$ORACLE_HOME/bin/genagtsh:        if [ $1 != "-32" ]; then        LIB_NAME=$1 # Library name        LIB_VER=$2 # Library version number        LIB=lib        NON64_LDOPT=        USRLIB32=        else        LIB_NAME=$2 # Library name        LIB_VER=$3 # Library version number        LIB=lib32        LOOP="done"        NON64_LDOPT="-m32"        USRLIB32=-L/usr/lib32        fi         LD="gcc ${NON64_LDOPT} -shared ${USRLIB32} -L${ORACLE_HOME}/${LIB} -L${ORACLE_HOME}/${LIB}/stubs"    修改$ORACLE_HOME/ctx/lib/env_ctx.mk:        LDFLAGS32=$(AMD32FLAGS) -o $@ -L/usr/lib32 $(LDPATHFLAG)$(PRODLIBHOME32) $(LDPATHFLAG)$(LIBHOME32) $(LDPATHFLAG)$(LIBHOME32)stubs/    修改$ORACLE_HOME/rdbms/lib/env_rdbms.mk:        REDEFINES32=LIBDIR=lib32 LDFLAGS='-m32 -o $$@ -L/usr/lib32 $$(LDPATHFLAG)$$(PRODLIBHOME) $$(LDPATHFLAG)$$(LIBHOME) $$(LDPATHFLAG)$$(LIBHOME)stubs/'  Error 4:     cannot find -lclntsh 解决方法:     下载oracle client 32bit的zip包,解压缩,然后将client/stage/Components/oracle.rdbms.ic/11.1.0.6.0/1/DataFiles/filegroup4.jar文件解压缩,copy 32位的libclntsh.so.11.1文件到$ORACLE_HOME/lib32目录中  Error 5:     gcc: /usr/lib64/libstdc++.so.6: No such file or directory 解决方法:     cd /usr/lib64/ ; ln -s libstdc++.so.5 libstdc++.so.6

参考:

http://forums.gentoo.org/viewtopic-p-4883219.html