[SOLVED] BASH: how to substitute just one occurency of a pattern in a text file

carolflb · 11-19-2009, 02:14 AM

I want to know if it's possible to use 'sed' to substitute just the first occurency of a searched pattern in the file, so if I have:

text
text PATTERN
text
PATTERN
text PATTERN text
text

is there a way of substituting the first occurency of PATTERN in the file? So that at the end I have:

text
text SUBSTITUTED PATTERN
text
PATTERN
text PATTERN text
text

catkin · 11-19-2009, 02:37 AM

If you want to do it using sed then it is a sed question not a bash question (so the thread title is misleading and I cannot help you).

If you are happy to do it in bash script then it is easy but will take longer to execute than when done by sed. That may or may not be a problem, depending on the size of the file.

David the H. · 11-19-2009, 02:47 AM

How do I match only the first occurrence of a pattern?

From the sed faq.

ghostdog74 · 11-19-2009, 03:09 AM

@OP : awk,sed etc are just tools to help your work. What's more important is to get your algorithm and steps to solve the problem right. You only want to substitute the first occurrence. That means you can set a toggle flag to turn substitution "off" when the first pattern is found

Code:

#!/bin/bash
toggle=0
while read -r line
do
    [ "$toggle" -eq 1 ] && echo $line && continue
    if [ "$toggle" -eq 0 ] ; then
        case $line in
            *PATTERN*) 
                    line=${line/PATTERN/SUBSTITUTED}
                    toggle=1
                    ;;
        esac
    fi
    echo $line
done <"file"

output

Code:

# ./shell.sh
text
text SUBSTITUTED
text
PATTERN
text PATTERN text
text

i think the code is self explanatory. read up more on shell scripting (see my sig).

carolflb · 11-19-2009, 03:32 AM

Hi guys,

sorry for the confusion. The script I'm writing is a .sh file and I'm open to any solution. I just wrote 'with sed' because it is the only command I was using until now.

@David: Thanks for the link! I just tried to use:

sed '0,/RE/s//to_that/' file to change only the first match, but it didn't work.

I wrote in my script:

sed '0,/<declaration>/s//<declaration>\nint p_covg['$access_array_size'];/' file.xml

The pattern I want to find is <declaration> and it should be substituted by <declaration>\nint p_covg['$access_array_size'];, where access_array_size is an integer.

And the text in the file.xml is the following:

<?xml version='1.0' encoding='utf-8'?>
<!DOCTYPE nta PUBLIC '-//Uppaal Team//DTD Flat System 1.1//EN' 'http://www.it.uu.se/research/group/darts/uppaal/flat-1_1.dtd'>
<nta>
<declaration>
clock global;
broadcast chan dummybchannel;
...

Would you know why it doesn't work?

carolflb · 11-19-2009, 04:01 AM

Hi Ghostdog,

nice that you are answering again, thanks. I used your idea but it doesn't save the changes in the original file. What do I need to do, so that the changes are saved?

anon237 · 11-19-2009, 04:05 AM

Hi,

It looks like this: sed '0,/<declaration>/s//<declaration>\nint p_covg['$access_array_size'];/' file.xml
is not correct, you are missing the part between the red slashes.

The <declaration> part should be there:

sed '0,/<declaration>/s/<declaration>/<declaration>\nint p_covg['$access_array_size'];/' file.xml

Hope this helps.

carolflb · 11-19-2009, 04:16 AM

You were right druuna. It works now

Thanks a lot!

ghostdog74 · 11-19-2009, 04:31 AM

Quote:

Originally Posted by carolflb

doesn't save the changes in the original file. What do I need to do, so that the changes are saved?

redirect to a file and rename it back.