Search for a pattern and if it exists change some other text on the line
 

I am very new to sed and cannot seem to get my command to work. I am trying to search each line for the value 'onnection' and if it exists anywhere in the line change Abc to Xyz and write the line to a new file which is a parameter passed to the shell. My code currently looks like this
file="first_file.exp"
sed -n "/onnection/p" $file|sed -e "s/_Abc/_Xyz/g" $file > "$file_test1_parsed"

What am I doing wrong? It is changing all occurrences of Abc not just the lines that have connect in them.

TIA,
Kathy

You don't need 2 sed parts, have a look at this:The blue part selects lines with onnection in them and if so Abc is turned into Xyz

You can tell sed to use line or pattern ranges as well:These links might help:

Sed resources:

• GNU Sed, a stream editor
• Sed - An Introduction and Tutorial
• The Sed FAQ
• Sed - one-liners
• Sed One-Liners Explained, Part I
• O'Reilly: Sed/Awk

What's the actual error?

Try this version:
Edit: I think you don't have to pass the file on the second sed command.

@konsolebox: Why 2 piped sed commands in the first place?

@druuna I missed your post as I had my focus on the variables and jumped quickly on it sorry. After making a quick notice I just made a quick edit of my answer.

Anyway I think you should adjust the command to this:
I don't think the OP wants the other lines as described.

And I focussed too much on the range stuff.

If only the lines need to be printed that contain onnection, it can be done like this (simpler?) as well:

I see. I don't really try to remember that s accepts p subcommand, so I just chose the 'g; p;', but I actually had the idea before I modified it. Just to be safe.

Thanks all. Great suggestions! I do want all the lines in the file not just the ones I have changed.

What would I change in sed -n '/onnection/s/Abc/Xyz/gp' $file to get all the lines from $file to newfile with my changes?

Sed has a -i for in place changes (changes the original file):If you want to put the output in a different file:

I tried -i but I get sed: illegal option -- i
When I use the syntax suggested above sed -n '/onnection/s/Abc/Xyz/gp' $file > $newfile
all I get in newfile are the changed lines and I want newfile to contain the whole file containing the changed and unchanged lines.

My suggestion was actually to just print the lines that contain the pattern. If you want to include other lines as well, you could just use druuna's first post:

Got it! Thanks all. I am all set.

Welcome. Glad I was able to help, miss :)

By the way, as for the actual error in the OP:

The first sed is reading from the file, and then the second sed is also reading from the same file. The piped data between them is being ignored. Remove the "$file" argument from the second command and that should fix it.

(Although as shown you you only really need a single command.)


PS: Don't forget to quote all of your variables.

http://mywiki.wooledge.org/Arguments
http://mywiki.wooledge.org/WordSplitting
[url]http://mywiki.wooledge.org/Quotes[url]

Edit:

Only the gnu version of sed included in most Linux distros has the -i option. You're apparently using a different implementation that doesn't provide it. Check your man page.