sed高级用法

sed 高级用法
##############the best of all
[liguo@localhost ttcn]$ echo "apricot (bird) cake (diamond) eland" |sed  's/([^)]*)/QQ/g'
apricot QQ cake QQ eland
+++minimum match
[liguo@localhost ttcn]$ echo "apricot (bird) cake (diamond) eland" |sed  's/(.*)/QQ/g'
apricot QQ eland
+++maximum match
[liguo@localhost ttcn]$ echo "apricot (bird) cake (diamond) eland" |sed  's/.*(.*)/QQ/g'
QQ eland
+++first part before the () match
[liguo@localhost ttcn]$ echo "apricot (bird) cake (diamond) eland" |sed  's/(.*).*/QQ/g'
apricot QQ
+++last patr after the ()match




################
匹配两个距离最大之间的所有字符，第一个Z匹配第一个，第二个匹配最后一个
zxx@gll-bac:~/sedex$ echo "1Z2Z3Z4Z5Z" |sed 's/Z.*Z/Q/'
1Q

zxx@gll-bac:~/sedex$ echo "1Z2Z3Z4Z5Z" |sed 's/Z.\+Z/Q/'
1Q

匹配两个距离最小之间的所有字符，第一个Z匹配第一个，第二个匹配第二一个
zxx@gll-bac:~/sedex$ echo "1Z2Z3Z4Z5Z" |sed 's/Z.\?Z/Q/'
1Q3Z4Z5Z
zxx@gll-bac:~/sedex$ echo "1Z2Z3Z4Z5Z" |sed 's/Z[^Z]*Z/Q/'
1Q3Z4Z5Z

for lazy match
need the oposite of the pattern
[liguo@localhost ~]$ echo "t  ////test" |sed   "s/^\([^/]*\)\/\/.*$/\1/"
t
[liguo@localhost ~]$ echo "t  ////test" |sed   "s/^\(.*\)\/\/.*$/\1/"
t  //


 sed -n '1~2{h;n;x;H;x;p} ' flist
交换奇偶行的顺序
 sed -n '1~2{h;n;x;H;x;};p ' flist

sed -e '1{h;d}' -e'2{g}' flist
交换第一行和第二行的顺序
sed '1{h;d};2{g}' flist

sed 's/ /\\&/g'
在文件中把空格前都加入反斜杠

sed -n '1{h;d};1!{x;H};${x;H;p}'  ta.txt
把文件的行逆序输出


zxx@gll-bac:~/sedex$ for((i=1;i<=4;i++)); do case $i in 1) sed -n 4p ta.txt;; 4)sed -n 1p ta.txt;; *)sed -n ${i}p ta.txt;; esac; done
4
2
3
1
//exchang 2 lines,kown line number

no line numbers; read all lines between pater1 and pater2,两行不相邻
sed -n '/pater1/{:a; N; /pater2/! b a; p};'


zxx@gll-bac:~/sedex$ sed  '/2/{:a; N; /5/! b a; s/\([^\n]*\)\n\(.*\)\n\(.*\)/\3\n\2\n\1/; }' ta.txt
1ZZ
5
3ZZ
4ZZ
2ZZ
6
7
8

zxx@gll-bac:~/sedex$ sed -n '/2/{:a; N; /5/! b a; s/\([^\n]*\)\n\(.*\)\n\(.*\)/\3\n\2\n\1/; p;}' ta.txt
5
3ZZ
4ZZ
2ZZ

相邻行之间
sed -n '/2/{:a; N; /3/! b a; s/\([^\n]*\)\n\(.*\)/\2\n\1/; p;}' ta.txt
3ZZ
2ZZ

综合以上，不管相邻与否都可以：
sed  '/patern2/{:a; N; /patern3/! b a; /^[^\n]*\n[^\n]*$/{s/\([^\n]*\)\n\(.*\)/\2\n\1/; }; s/\([^\n]*\)\n\(.*\)\n\(.*\)/\3\n\2\n\1/; }  }' ta.txt

从最简单的开始：(sed 的工作原理)
lily@willow:~/notessever$ cat tnlist
bc
de
fg
hi
ye
fq
（这里sed把命令循环起来，最后才跳出命令行，这时已把所有的行都读入模式空间了）
lily@willow:~/notessever$ sed  ':a;N; s/\n/ZZ/g;ba' tnlist
bcZZdeZZfgZZhiZZyeZZfq
（sed 每次读下一行进入模式空间后在对引号里的命令顺序执行，执行完所有命令后模式空间将被清空，
来读取下一行，这里从第一行开始，N；命令读入第二行，处理后模式空间被输出后清空，再读第三行）
lily@willow:~/notessever$ sed  'N; s/\n/ZZ/g;' tnlist
bcZZde
fgZZhi
yeZZfq


for adress range:
cat tlist
-----------------------
AA
1
BB
AA
2
BB
------------------------------
sed '/AA/,/BB/ {:a; /BB/! N; s/\n/ZZ/; /BB/n; ba }; ' tlist
AAZZ1BB
AAZZ2BB
but   sed  '/AA/,/BB/ {:a;  N; s/\n/ZZ/;  ba  }; ' st
AAZZ1BBAAZZ2BB



lily@willow:~/notessever$ sed  'N; 4! b a;s/\n/ZZ/g;; :a;' tnlist
bc
de
fgZZhi
ye
fq

###################
#sed can only locate consecutive two character in greedy  mode, when in
#the maximum pattern match, it will fail out.
* How to debug sed
suing sedsed to debug sed
http://aurelio.net/projects/sedsed/
** two blank lines should /^\n$/, but only blank line could make this pattern
--------------
guolili@cougar archive]$ cat -n cd
     1  ab
     2
     3
     4  cd
[guolili@cougar archive]$ cat a.sed
/^$/{
N
/^\n$/d
}
[guolili@cougar archive]$ ~/test/sedsed-1.0 -d  -f a.sed cd
PATT:ab$
HOLD:$
COMM:/^$/ {
PATT:ab$
HOLD:$
ab
PATT:$              ###it's a blank line
HOLD:$
COMM:/^$/ {         #/^$/ match PATT:$
COMM:N
PATT:\n$            #so \n of line 2 will be added into PATT
HOLD:$
COMM:/^\n$/ d       #/^\n$/ math pattern \n$, so this is two blank lines N pattern result
PATT:cd$
HOLD:$
COMM:/^$/ {
PATT:cd$
HOLD:$
cd
----------------------------------------

** process a pair of something
text need to be processed
----------------------
{
....
123
}
=====
othercc
{
...
456
}
dfdfdfa
------------------------
/{/ {   #### /{/,/}/ {  this is not correct usage, don't use it.
:a
N
/}/!ba   ###get the end pair from here
w ss
}
---------------------------



the file ss would be 
---------------------
{
...
123
}
{
...
456456456456
}
---------------------------

*** here if the delimiter is BEG and blank line like this
-----------
BEG
123

dddddd
BEG
456

BEG
789
---------------

 ------------------------
/BEG/ {   
:a
N
//!ba  #here blank line is not a symbol to match  
w ss
}
---------------------------
So here should replace blank lines with some real thing like END
-----------
BEG
123
END
dddddd
BEG
456
END
BEG
789
END
---------------

------------------------
/BEG/ {   
:a
N
/END/!ba  #here END  a symbol to match  
w ss
}
---------------------------



*** sed could be used to process a pair of things, but not for netsted one
for  ex:
--------------
{
  {
    i
  }
}
---------------
this could only get the {{} ont {{}}


** append line number after a group
[guolili@cougar archive]$ cat -n ab
     1  {
     2  ab
     3  }
     4  iiiiiiii
     5  END
     6
     7  {
     8  dfdf
     9  dfd
    10  }
[guolili@cougar archive]$ sed -n  -f b.sed ab
{ab}
3
{dfdfdfd}
10
[guolili@cougar archive]$ sed -n  -f b.sed ab |sed 'N; s/\n//g'
{ab}3
{dfdfdfd}10
[guolili@cougar archive]$ cat b.sed
/{/ {
:a
N
/}/!ba
s/\n//g
p    # print all the pattern space 
=    # print the linenumber
}
[guolili@cougar archive]$


** reverse everline of file
[guolili@cougar archive]$ sed -n  -f ef.sed ef
3
2
1
[guolili@cougar archive]$ sed -n  '1{h;d}; 1!{x;  H;  }; ${x; H; p} ' ef
[guolili@cougar archive]$ cat ef.sed
1{h;d}
1!{x;H}
${x;H;p}
[guolili@cougar archive]$ cat ef
1
2
3
[guolili@cougar archive]$