sed学习笔记（不断更新补充）

在字符串处理中，sed非常强大，从前就了解皮毛，而在笔者某次网络电面中就曾遇到，只能现查man文档解决问题，非常窘迫。现进行系统学习，笔记于此。

参考

sed新手使用进阶全功略！

http://www.linuxsir.org/bbs/showthread.php?t=189620

在学习前，首先要保证知道正则表达式的基本知识，此可参见我的另一学习笔记

http://blog.sina.com.cn/s/blog_4d1bbec70100qmxg.html

当然，学习sed的最好办法，是man sed后，阅读其文档。当然，比较舒服的办法还是跟随大牛，理论和实际相结合，学习过程如下。

sed主要用来查找替换，对字符串进行处理，功能强大，其基本格式如下：

sed [-n][-e] 'command' file(s) sed [-n] -f scriptfile file(s)

不用去理会以上格式如何使用，因为我自己在学习时，将其说明阅读若干边依然一头雾水。我是在诸多实例中，明白了其参数和格式的用法和含义。

sed从一个文件的一个文本行或从标准输入的几种格式中读取数据，将之拷贝到一个编辑缓冲区，然后读命令行或脚本的第一条命令，并使用这些命令查找模式或定位行号编辑它。重复此过程直到命令结束。但sed不和输入的文件打交道，它操作的只是文件的一个拷贝（除非用-i和输出重定向改动原文件），然后所有的改动如果没有重定向到一个文件，将输出到屏幕。

sed选项（其具体用法将会示例中展开）：

-n 不打印编辑行到标准输出（sed缺省为打印所有行，包括原始行和编译行）；

p命令可用来打印编辑行；

-e 允许同一行中执行多条命令

-f 

-c 编译命令

-i 编译原文件

-d 删除命令

sed基础用法

sed在文本中查询时，可用两种定位文本的方式：使用行号；使用正则表达式。方法如下：

x                    此处x为一行号，如1
x,y                 表示行号范围从x到y，如2，5表示从第2行到第5行
/pattern/       查询包含模式的行。例如/disk/或/[a-z]/
/pattern/pattern/   查询包含两个模式的行。例如/disk/disks/
/pattern/,x   在给定行号上查询包含模式的行。如/ribbon/,3
x,/pattern/    通过行号和模式查询匹配行。3,/vdu/
x,y!                查询不包含指定行号x和y的行。1,2!

基本sed编辑命令：
sed编辑命令
p 打印匹配行
= 显示文件行号
a\ 在定位行号后附加新文本信息
i\ 在定位行号后插入新文本信息
d 删除定位行
c\ 用新文本替换定位文本
s 使用替换模式替换相应模式
r 从另一个文件中读文本
w 写文本到一个文件
q 第一个模式匹配完成后推出或立即推出
l 显示与八进制A S C I I代码等价的控制字符
{ } 在定位行执行的命令组
n 从另一个文件中读文本下一行，并附加在下一行
g 将模式2粘贴到/pattern n/
y 传送字符
n 延续到下一输入行；允许跨行的模式匹配语句

示例：有文件test.txt

cat test.txt 

1 The honeysuckle band played all night long for only $90.

2 It was an evening of splendid music and company.

3 Too bad the disco floor fell through at 23:00.

4 The local nurse Miss P.Neave was in attendance.

命令：

$ sed -e '1p' test.txt

1 The honeysuckle band played all night long for only $90.

1 The honeysuckle band played all night long for only $90.

2 It was an evening of splendid music and company.

3 Too bad the disco floor fell through at 23:00.

4 The local nurse Miss P.Neave was in attendance.

$ sed -n -e '1p' test.txt

1 The honeysuckle band played all night long for only $90.

从上面的例子中可以看出，命令-e在此处可有可无，而-n则限制是否将未编辑的初始行打印，1为第一行，p则是指出需要将匹配的行打印出。同样，

$ sed -n -e '2p' test.txt

2 It was an evening of splendid music and company.

打印第二和第三行

$ sed -n -e '2,3p' test.txt

2 It was an evening of splendid music and company.

3 Too bad the disco floor fell through at 23:00.

打印含有company的匹配行

$ sed -n -e '/company/p' test.txt

2 It was an evening of splendid music and company.

打印第2行和含有23：00的匹配行（注意，是分别打印，并不是逻辑与的关系）

$ sed -n -e '2,/23:00/p' test.txt

2 It was an evening of splendid music and company.

3 Too bad the disco floor fell through at 23:00.

打印非第2和第3行（注意其间的逻辑关系）

$ sed -n -e '2,3!p' test.txt

1 The honeysuckle band played all night long for only $90.

4 The local nurse Miss P.Neave was in attendance.

用=打印行号，此处再次注意-n的重要作用

$ sed -e '=' test.txt

1

1 The honeysuckle band played all night long for only $90.

2

2 It was an evening of splendid music and company.

3

3 Too bad the disco floor fell through at 23:00.

4

4 The local nurse Miss P.Neave was in attendance.

和（-n需要放在-e前）

$ sed -n -e '=' test.txt

1

2

3

4

下面还有一些例子：

这个不解释

$ sed -n -e '/music/p' test.txt

2 It was an evening of splendid music and company.

yyang@Yanux:~/shell$ sed -n -e '/music/=' test.txt

2

此处要执行多个匹配，-e的作用亦在此，倘若不加-e，则会报错，只能执行第一个匹配。

$ sed -n -e '/music/p' -e '/music/=' test.txt

2 It was an evening of splendid music and company.

2

还有，

$ sed -e '=;p' test.txt 

1

1 The honeysuckle band played all night long for only $90.

1 The honeysuckle band played all night long for only $90.

2

2 It was an evening of splendid music and company.

2 It was an evening of splendid music and company.

3

3 Too bad the disco floor fell through at 23:00.

3 Too bad the disco floor fell through at 23:00.

4

4 The local nurse Miss P.Neave was in attendance.

4 The local nurse Miss P.Neave was in attendance.

yyang@Yanux:~/shell$ sed -e 'p' test.txt 

1 The honeysuckle band played all night long for only $90.

1 The honeysuckle band played all night long for only $90.

2 It was an evening of splendid music and company.

2 It was an evening of splendid music and company.

3 Too bad the disco floor fell through at 23:00.

3 Too bad the disco floor fell through at 23:00.

4 The local nurse Miss P.Neave was in attendance.

4 The local nurse Miss P.Neave was in attendance.

yyang@Yanux:~/shell$ sed -n -e '=;p' test.txt 

1

1 The honeysuckle band played all night long for only $90.

2

2 It was an evening of splendid music and company.

3

3 Too bad the disco floor fell through at 23:00.

4

4 The local nurse Miss P.Neave was in attendance.

通过上面的三个例子，可以看出；用来分隔两个匹配模式。另外，

$ sed -n -e '=' -e 'p' test.txt

1

1 The honeysuckle band played all night long for only $90.

2

2 It was an evening of splendid music and company.

3

3 Too bad the disco floor fell through at 23:00.

4

4 The local nurse Miss P.Neave was in attendance.

这个例子也说明了上述问题。

a为附加命令，将附加内容附加至匹配行后

$ sed -e 'a\this line will be added to the end of each line!' test.txt

1 The honeysuckle band played all night long for only $90.

this line will be added to the end of each line!

2 It was an evening of splendid music and company.

this line will be added to the end of each line!

3 Too bad the disco floor fell through at 23:00.

this line will be added to the end of each line!

4 The local nurse Miss P.Neave was in attendance.

this line will be added to the end of each line!

$ sed -e '/music/a\this line will be added to the end of each line!' test.txt

1 The honeysuckle band played all night long for only $90.

2 It was an evening of splendid music and company.

this line will be added to the end of each line!

3 Too bad the disco floor fell through at 23:00.

4 The local nurse Miss P.Neave was in attendance.

i为插入命令，将需要插入的内容插入至匹配行前

$ sed -e 'i\this line will be inserted to the begin of each line!' test.txt

this line will be inserted to the begin of each line!

1 The honeysuckle band played all night long for only $90.

this line will be inserted to the begin of each line!

2 It was an evening of splendid music and company.

this line will be inserted to the begin of each line!

3 Too bad the disco floor fell through at 23:00.

this line will be inserted to the begin of each line!

4 The local nurse Miss P.Neave was in attendance.

$ sed -e '/music/i\this line will be inserted to the begin of each line!' test.txt

1 The honeysuckle band played all night long for only $90.

this line will be inserted to the begin of each line!

2 It was an evening of splendid music and company.

3 Too bad the disco floor fell through at 23:00.

4 The local nurse Miss P.Neave was in attendance.

c为行更改命令，而d为行删除命令

$ sed -e '/music/c\this line will  be modified to the matching line!' test.txt

1 The honeysuckle band played all night long for only $90.

this line will  be modified to the matching line!

3 Too bad the disco floor fell through at 23:00.

4 The local nurse Miss P.Neave was in attendance.

$ sed -e '/music/,3d' test.txt 

1 The honeysuckle band played all night long for only $90.

4 The local nurse Miss P.Neave was in attendance.

sed替换

替换命令为s

$ sed -e 's/h/ooooo/' test.txt 

1 Toooooe honeysuckle band played all night long for only $90.

2 It was an evening of splendid music and company.

3 Too bad toooooe disco floor fell through at 23:00.

4 Toooooe local nurse Miss P.Neave was in attendance.

上面的例子中，只用ooooo替换了每行的第一个h；而使用参数g，则可以进行全局替换

$ sed -e 's/h/ooooo/g' test.txt 

1 Toooooe ooooooneysuckle band played all nigooooot long for only $90.

2 It was an evening of splendid music and company.

3 Too bad toooooe disco floor fell tooooorougooooo at 23:00.

4 Toooooe local nurse Miss P.Neave was in attendance.

下面的例子需要理解

$ sed -e '1,2s/h/ooooo/g' test.txt

1 Toooooe ooooooneysuckle band played all nigooooot long for only $90.

2 It was an evening of splendid music and company.

3 Too bad the disco floor fell through at 23:00.

4 The local nurse Miss P.Neave was in attendance.

yyang@Yanux:~/shell$ sed -e '1,2s/h/ooooo/' test.txt

1 Toooooe honeysuckle band played all night long for only $90.

2 It was an evening of splendid music and company.

3 Too bad the disco floor fell through at 23:00.

4 The local nurse Miss P.Neave was in attendance.

在下面的例子中，\$为转义符转义$，命令为只为含有$的行进行全局替换

sed -e '/\$/s/h/ooooo/g' test.txt

1 Toooooe ooooooneysuckle band played all nigooooot long for only $90.

2 It was an evening of splendid music and company.

3 Too bad the disco floor fell through at 23:00.

4 The local nurse Miss P.Neave was in attendance.

而下面的命令中，含有It和23:00的行都将进行替换

$ sed -e '/It/,/23:00/s/l/ooooo/g' test.txt

1 The honeysuckle band played all night long for only $90.

2 It was an evening of spoooooendid music and company.

3 Too bad the disco fooooooor feoooooooooo through at 23:00.

4 The local nurse Miss P.Neave was in attendance.

替换修改字符串&

yyang@Yanux:~/shell$ sed -e 's/$90/&230/g' test.txt 

1 The honeysuckle band played all night long for only $90230.

2 It was an evening of splendid music and company.

3 Too bad the disco floor fell through at 23:00.

4 The local nurse Miss P.Neave was in attendance.

$ sed -e 's/90/120,&/g' test.txt 

1 The honeysuckle band played all night long for only $120,90.

2 It was an evening of splendid music and company.

3 Too bad the disco floor fell through at 23:00.

4 The local nurse Miss P.Neave was in attendance.

不用解释，明白人一看就懂。

sed的基本应用就这样，但更好的掌握则需要不断地使用，在实际应用中扩展对它的了解，熟练应用。今后所得，将会补充于此。

补：

1、语法间的分隔符，不仅仅可以使用/，其他也行。比如下面例子：

$ echo $PWD | sed -e 's/\//@/g'

@home@yyang@shell

我将输出的/转换为@，如果采用上面的方法，各种表达方式错综复杂，不方便阅读。因此可以

yyang@Yanux:~/shell$ echo $PWD | sed -e 's:/:@:g'

@home@yyang@shell

yyang@Yanux:~/shell$ echo $PWD | sed -e 's;/;@;g'

@home@yyang@shell

结论：可以采用其他易于辨认的符号作为语法成分间的分隔符

2、引号的问题

此处对引号的讨论，源于参考中的一个例子，关于更加详细的引号用法，可以参见

说说sed中引号的用法（抛砖引玉！）http://www.chinaunix.net/jh/24/129474.html

参考中的例子如下：

要用sed把字符“\”转化成“'”该怎么写？比如：

$ echo "hello\hh\haha"

hello\hh\haha

按照常理，我们都会如此来书写，但无法得到预期结果

$ echo "hello\hh\haha" | sed -e 's:\:':g' 

> ^C

估计原因出在单引号的匹配上，由于在sed的匹配式中出现了三个单引号'，于是系统认为命令还未完结，因此在第二行出现了>，等待继续输入。参考中的正确答案如下，同事也得到了正确的结果。

$ echo "hello\hh\haha" | sed -e 's/\\/'"'"'/g'

hello'hh'haha

经思考，认为s/\\/后的第一个单引号'是用来匹配s前的第一个单引号，而最后的'则是用来匹配g后的单引号，中间的成对双引号" "将一个单引号'引出，表示用于替换的'，为了验证想法是否正确，实验如下（双引号中的'被；号替换）：

$ echo "hello\hh\haha" | sed -e 's/\\/'";"'/g'

hello;hh;haha

实验成功。

3、分域的问题

4、创建sed脚本的问题