使用sed查找并复制到新文件

Question

I have multiple lines in a file. Each line has a common start tag and end tag. I want to get the contents between the tag and put that in a new file separated by /r.

1) I tried the following .. but its copying the entire line and putting into the new file

#!/bin/sh

startline="<Mytag>"
endline="<Nexttag>"

echo $startline
echo $endline

sed "/$startline/,/$endline/!d" input.txtt > test.txt

2) Ideally the end tag should be </Mytag> but sed is not taking the '/' very well. How to overcome this? Should I use a '//' ?

Thanks

---

update

---

input.txt has the following lines

<?xml version="1.0" encoding="UTF-8" ?><InputRecord xmlns:xsi= "http://www.w3.org/2001/XMLSchema-instance" <tag1>blah</tag1><mytag>myinfo</mytag><tag2>blah</tag2></InputRecord>

<?xml version="1.0" encoding="UTF-8" ?><InputRecord xmlns:xsi= "http://www.w3.org/2001/XMLSchema-instance" <tag1>blah1</tag1><mytag>myinfo1</mytag><tag2>blah2</tag2></InputRecord>

expected output

myinfo
myinfo1

Answer 1

Answer for revised question

Given input:

 <?xml version="1.0" encoding="UTF-8" ?><InputRecord xmlns:xsi= "http://www.w3.org/2001/XMLSchema-instance" <tag1>blah</tag1><mytag>myinfo</mytag><tag2>blah</tag2></InputRecord> <?xml version="1.0" encoding="UTF-8" ?><InputRecord xmlns:xsi= "http://www.w3.org/2001/XMLSchema-instance" <tag1>blah1</tag1><mytag>myinfo1</mytag><tag2>blah2</tag2></InputRecord> 

the output should be:

 myinfo myinfo1 

Temporarily ignoring the fact that parsing XML with regular expressions is generally not sensible, this can be treated as a request to find the text between a start tag and an end tag on a single line. This translates to:

starttag="<mytag>"
endtag="</mytag>"
sed -n "\%.*$starttag\(.*\)$endtag.*% s//\1/p"

The \\% notation is required by POSIX sed to allow the use of something other than a slash as the delimiter for a regular expression. POSIX sed says:

... a context address (which consists of a BRE, as described in Regular Expressions in sed , preceded and followed by a delimiter, usually a <slash> )

and:

In a context address, the construction "\\cBREc" , where c is any character other than <backslash> or <newline> , shall be identical to "/BRE/" . If the character designated by c appears following a <backslash> , then it shall be considered to be that literal character, which shall not terminate the BRE. For example, in the context address "\\xabc\\xdefx" , the second x stands for itself, so that the BRE is "abcxdef" .

Answer for original version of question

Your script should work as written if you get the $endline value correct. However, IMNSHO, it is simpler to be positive about the range to print:

sed -n "/$startline/,/$endline/p" input.txtt > test.txt

The -n means 'do not print unless I tell you to' and the script ways 'print between the line matching the start line and the line matching the end line.

For the end tag with the slash in it, you need to escape the slash with a backslash:

endline="<\/Nexttag>"

Or you could use a . in place of the slash, which could in theory match the start of <XNexttag> but probably won't. The absence of the backslash would account for why you got everything from the start line to the end of file.

---

On the benefits of positivitity

Consider the data file:

line1
line2 start1
line3
line4 end1
line5
line6 start2
line7
line8 end2
line9

And consider the shell and sed commands:

echo Positive Single
sed -n -e '/start1/,/end1/p'  data
echo Negative Single
sed    -e '/start1/,/end1/!d' data

echo Positive Double
sed -n -e '/start1/,/end1/p'  -e '/start2/,/end2/p'  data
echo Negative Double
sed    -e '/start1/,/end1/!d' -e '/start2/,/end2/!d' data

The output from running that script is:

$ sh sed.scripts
Positive Single
line2 start1
line3
line4 end1
Negative Single
line2 start1
line3
line4 end1
Positive Double
line2 start1
line3
line4 end1
line6 start2
line7
line8 end2
Negative Double
$

For the case of a single pattern range to match, there's no problem with the !d formulation vs the -n plus p formulation.

However, the 'positive double' pattern works fine, producing the answer I'd expect, for 'print the lines between start1 and end1 and also the lines between start2 and end2 ', whereas the 'negative double' pattern does not work correctly any more. I'd rather use the extensible version than the version that has to be rewritten when the requirement changes.

Answer 2

To escape the slashes, precede them with a back slash, like this:

<\/Nexttag>

But you only need that because you've chosen to use a slash as your delimiter. You can use any character you want (slash is conventionally chosen because many other languages use it to delimit regexes). So chose a character that won't appear in tags, like a hash #:

sed "#$startline#,#$endline#!d" input.txtt > test.txt

Answer 3

This is probably not the most optimal solution, but it produces the expected output for your sample input:

#!/bin/sh

startline="<mytag>"
endline="<\/mytag>"

awk '{ gsub(">", "&\n"); gsub("<", "\n&"); print; }' | sed -e "/$startline/,/$endline/!d" -e "/$startline/d" -e "/$endline/d"

Redirect your sample input to this script for example like this:

sh script.sh < sample.txt

The awk in the middle is just to put a newline after all > and before < , because the sed script works only if the start and end tags are on their own line alone. (To be honest this is really not a great script.)