Shell Programming and Scripting

Hello,

I have huge xml files and I need to replace only part of the data within a particular xml tag. This doesnt seem to be as simple as it sounds. I have searched everywhere and couldnt find any solution.

Ex:
In the below case I would like "def" to be replaced by "xyz" only when found in tag2 and not anywhere else.

sample line in xml file:
<tag1>def ghi</tag1><tag2>abc def ghi</tag2><tag3>def jkl</tag3>

desired output:
<tag1>def ghi</tag1><tag2>abc xyz ghi</tag2><tag3>def jkl</tag3>

Using pointer,any programming language can do this.
But it's dangerous if you do the job incorrectly,it's possible to damage data.

First to find out the line you want to replace,then build new line with same length,last write new line to file in corresponding position.

You can googleython file pointer.Beacause i used python.

Code:

perl -pe 's:(<tag2>.*?)def(.*?</tag2>):$1xyz$2:g' input.xml

Thank you Balaji! I was actually looking for a solution using sed. is that possible? However, the above perl option is working as expected. Can you please explain the command by breaking it?

Code:

sed 's:\(<tag2>.*\)def\(.*<\/tag2>\):\1xyz\2:g' input.xml

Part to be substituted:
\(<tag2>.*\) => Searches for <tag2> and characters after it and stores it in \1. (note that this greedily matches for all characters after <tag2>)
def => Tries to look for 'def'. But since .* would have already matched all the characters after <tag2>, it will give up characters one by one till it matches 'def'.
\(.*<\/tag2>\) => Searches for characters that end with </tag2> and hold it in \2.

Part that will be replaced:
\1xyz\2 => Print value in \1 followed by xyz followed by \2.

To summarise: Search for string starting with '<tag2>' followed by some characters followed by def followed by some more characters, followed by </tag2>.. Now replace only 'def' in searched pattern by 'xyz'.

Thank you. A nice trick using back references however, if there are two "def" words within the same tag then the substitution to xyz would happen only for the second def. The first def in the tag will be ignored and wont be replaced with xyz. Any solution to this?

Code:

$ cat input
<tag1>def ghi</tag1><tag2>abc def def</tag2><tag3>def jkl</tag3>
<tag1>def ghi</tag1><tag2>abc def ghi</tag2><tag3>def jkl</tag3>
$
$ perl -pe 's/(<tag2>.*?<\/tag2>)//g;$x=$1;$x=~s/def/xyz/g;s/(<\/tag1>)/$1$x/' input
<tag1>def ghi</tag1><tag2>abc xyz xyz</tag2><tag3>def jkl</tag3>
<tag1>def ghi</tag1><tag2>abc xyz ghi</tag2><tag3>def jkl</tag3>