extract part of a string and reinsert back into original

tlin · 02-15-2012, 03:42 AM

Hi,

I'm new to LQ and bash programming. Currently I'm faced with this task:

File name is abc.xml
inside this file look like this
...
...
...
<some text1 value="WWXXRRR222BBBCC" some1 moretext2/>
<some text2 value="YYXXRRR444KKKCC" some4 moretext7/>
<some text1 value="WWXXYYY222BBBTT" some6 moretext5/>
<some text2 value="GGG66RRIIIBBBCC" some8 moretext0/>
<some text1 value="WWXXRRR222VVVEE" some9 moretext3/>
...
...
...

The value (WWXXRRR222BBBCC) can be totally random but have a fixed length.
The some# moretest# following the value (WWXXRRR222BBBCC) are also random, but always starts with (" some).
The string before the value always starts with (value=")
I need to extract the value, double the value (WWXXRRR222BBBCC) into (WWXXRRR222BBBCCWWXXRRR222BBBCC) and insert it back into the exact place.

I'm able to get to the point where I can get just a list of values but I don't know how to process it one line at a time and reinsert it back into the original file.
I used code:

grep "<Key parity" abc.xml | sed -e 's/.*value=\"//' -e 's/\" from.*//'

Thanks in advance!

druuna · 02-15-2012, 03:54 AM

Hi,

Give this a try:

Code:

sed 's/"\(.*\)"/"\1\1"/' infile

Example run:

Code:

$ sed 's/"\(.*\)"/\1\1/' infile 
<some text1 value="WWXXRRR222BBBCCWWXXRRR222BBBCC" some1 moretext2/>
<some text2 value="YYXXRRR444KKKCCYYXXRRR444KKKCC" some4 moretext7/>
<some text1 value="WWXXYYY222BBBTTWWXXYYY222BBBTT" some6 moretext5/>
<some text2 value="GGG66RRIIIBBBCCGGG66RRIIIBBBCC" some8 moretext0/>
<some text1 value="WWXXRRR222VVVEEWWXXRRR222VVVEE" some9 moretext3/>

The above uses back referencing. All between \( and \) in the search part can be represented by \1 in the replace part.

PS: If this needs to be done in-place (changes are made in the original file), change the command to

Code:

sed -i 's/"\(.*\)"/"\1\1"/' infile

Hope this helps.

tlin · 02-15-2012, 05:42 AM

Thanks druuna for the quick reply!

Your solutions works for my example but when I try to run it on my actual file I'm not getting the result I'm looking for.

I guess it's because my actual file have a lot of "" in them.

It looks more like this:
<some text="1" value="WWXXRRR222BBBCC" some="ABC" moretext="1"/>
<some text="2" value="YYXXRRR444KKKCC" some="3" moretext="4"/>
<some text="1" value="WWXXYYY222BBBTT" some="7 moretext="5"/>
<some text="2" value="GGG66RRIIIBBBCC" some="8" moretext="0"/>
<some text="1" value="WWXXRRR222VVVEE" some="6" moretext="3"/>

There is also some text before this and after this that I don't want to change at all.

Thanks!

druuna · 02-15-2012, 06:00 AM

Hi,

Assuming you did post a relevant example this time:

Code:

sed 's/value="\(.*\)" some/value="\1\1" some/' infile

Quote:

There is also some text before this and after this that I don't want to change at all.

You need to provide a relevant example, this doesn't help. What is before/after?

Hope this helps.

tlin · 02-15-2012, 08:43 AM

Sorry about give a bad example to start with.
I figured out how to do everything using your code as a start.

Here is what I end up writing which works for me:

Code:

sed '/^\s*<some text=/s/"\([^"]*\)"\([^"]*\)"\([^"]*\)"/"\1"\2"\3\3"/' abc.xml

I want to change only lines starting with <some text=

so it end up looking like this:

Code:

sed '/^\s*<some text=/s/"\([^"]*\)"\([^"]*\)"\([^"]*\)"/"\1"\2"\3\3"/' abc.xml

<some other text="6" something else="AAYURYE" more text="YYUU" number="234332">  #Something before (No change)
<some text="1" value="WWXXRRR222BBBCCWWXXRRR222BBBCC" some="ABC" moretext="1"/>  #Changed because it start with <some text=
<some text="2" value="YYXXRRR444KKKCCYYXXRRR444KKKCC" some="3" moretext="4"/>    #Changed because it start with <some text=
<some text="1" value="WWXXYYY222BBBTTWWXXYYY222BBBTT" some="7 moretext="5"/>     #Changed because it start with <some text=
<some text="2" value="GGG66RRIIIBBBCCGGG66RRIIIBBBCC" some="8" moretext="0"/>    #Changed because it start with <some text=
<some text="1" value="WWXXRRR222VVVEEWWXXRRR222VVVEE" some="6" moretext="3"/>    #Changed because it start with <some text=
<some other text="8" something else="AAYTTBB" more text="IIEE" number="278462">  #Something after (No change)

Hopefully this will help someone else.
Note: I was told this sed usage may not work for Unix but should be just fine with Linux.

Thanks druuna!

ntubski · 02-15-2012, 01:38 PM

Alternate solution with xmlstarlet:

Code:

xml ed --inplace -u '//Key[@parity]/value' -x 'concat(.,.)' abc.xml