[SOLVED] Using sed to remove text in fgrep string

manwithaplan · 10-10-2009, 12:16 AM

I would like to use sed to remove text between "( )" in a string pattern... Question is after reading the man pages and looking at examples, I am still slightly confused on how to implement.

Here is a example of what I wanna do:

Code:

  
cd $pkg_edit
    
      for edt in acl attr cloop gcc; 
      do
        cd $edt
        cat PKGBUILD | fgrep depends | sed ..? # This is where I need an example
        cd ..
      done

An example of the fgrep string:

Code:

$  cat PKGBUILD | fgrep depends
depends=('attr>=2.4.41')

I need to remove the text in between "( )"

I could remove the entire line in the file... But I am keeping the variable, just in case of a manual edit. I am not completely helpless ... So, if you show me how i can match a pattern with sed in the string, then remove then text...

rikijpn · 10-10-2009, 12:42 AM

So, according to your example, you want this:

Code:

depends=('attr>=2.4.41')

to be printed like this:

Code:

depends=()

right?

Then this should do it:

Code:

 'sed s/(.*)/()/g'

This tells sed to substitute every character including "(" till ")" with a "()". You don't really need the /g part, but in case you put this same command with sed on a file (not piping) you should use it.

manwithaplan · 10-10-2009, 01:37 AM

Quote:

Originally Posted by rikijpn

So, according to your example, you want this:

Code:

depends=('attr>=2.4.41')

to be printed like this:

Code:

depends=()

right?

Then this should do it:

Code:

 'sed s/(.*)/()/g'

This tells sed to substitute every character including "(" till ")" with a "()". You don't really need the /g part, but in case you put this same command with sed on a file (not piping) you should use it.

Yes that's right, I want to remove the text between "( )".

Thanks for the example... Though it seems the sed isn't matching my fgrep.

So I done away w/ the cat and fgrep... This sed seemed to work with the example you gave me.

After reading the man page. The -i extension was needed to edit the file in place.

Code:

sed -i 's/depends=(.*)/depends=()/g' PKGBUILD >/dev/null

I was having trouble with sed replacing all of my "( )" with empty strings, so with your help on the /s regexp, I was able to fix it.

Though still don't understand the "/g" Here is what the man page says:

Quote:

g G Copy/append hold space to pattern space.

I assume it holds the appended text in the same line..?

Thanks for the help.

lutusp · 10-10-2009, 02:04 AM

Quote:

Originally Posted by manwithaplan

I would like to use sed to remove text between "( )" in a string pattern... Question is after reading the man pages and looking at examples, I am still slightly confused on how to implement.

Here is a example of what I wanna do:

Code:

  
cd $pkg_edit
    
      for edt in acl attr cloop gcc; 
      do
        cd $edt
        cat PKGBUILD | fgrep depends | sed ..? # This is where I need an example
        cd ..
      done

An example of the fgrep string:

Code:

$  cat PKGBUILD | fgrep depends
depends=('attr>=2.4.41')

I need to remove the text in between "( )"

I could remove the entire line in the file... But I am keeping the variable, just in case of a manual edit. I am not completely helpless ... So, if you show me how i can match a pattern with sed in the string, then remove then text...

The solutions in this thread are way too complicated. Have you considered the easy way?:

Code:

$ x="depends=('attr>=2.4.41')"
$ echo "${x//\'*\'}"
depends=()

manwithaplan · 10-10-2009, 02:09 AM

Quote:

Originally Posted by lutusp

The solutions in this thread are way too complicated. Have you considered the easy way?:

Code:

$ x="depends=('attr>=2.4.41')"
$ echo "${x//\'*\'}"
depends=()

Yes... the thing is in your example is, that the variable $x is static. I need to run this in a for loop, that goes through multiple directories to append that one line on the same filename.

Code:

 cd $abs_bd 
   
	    for pkg in acl attr fakeroot pacman libfetch libarchive openssl wget rsync;  
	    do
	     cd $pkg 
		sed -i 's/depends=(.*)/depends=()/g' PKGBUILD >/dev/null  # Removes all dependencies from the build
		makepkg -s $pkg || pkgfail  # Rebuild function
	     cd ..
	    done

lutusp · 10-10-2009, 02:25 AM

Quote:

Originally Posted by manwithaplan

Yes... the thing is in your example is, that the variable $x is static. I need to run this in a for loop, that goes through multiple directories to append that one line on the same filename.

So read the file line by line, which you are already doing, and use the simple string method as a filter. Put this inside your present file scanning loop:

Code:

while read line
do
    echo "${line//\'*\'}"
done < filename > tempfile

mv tempfile filename

Be sure to create a temp file as shown above -- don't try to read and write to the same file at the same time. Do this regardless of which method you end up using -- it is vital that you not try to read and write to the same file at once.

rikijpn · 10-11-2009, 01:00 AM

Quote:

Originally Posted by manwithaplan

Yes that's right, I want to remove the text between "( )".

Thanks for the example... Though it seems the sed isn't matching my fgrep.

So I done away w/ the cat and fgrep... This sed seemed to work with the example you gave me.

After reading the man page. The -i extension was needed to edit the file in place.

Code:

sed -i 's/depends=(.*)/depends=()/g' PKGBUILD >/dev/null

I was having trouble with sed replacing all of my "( )" with empty strings, so with your help on the /s regexp, I was able to fix it.

Though still don't understand the "/g" Here is what the man page says:
I assume it holds the appended text in the same line..?

Thanks for the help.

Well, good thing you got it. You sure can edit the a file with the shell or sed, although sed is probably just shorter and easier to remember on most times.

The /g options is for edit the file globally, meaning replacement in all lines. Otherwise it would replace only the first string match it finds, and then end the program (ignoring the rest of the file). If you would like to read more about sed probably this page is the best you can find: http://www.grymoire.com/Unix/Sed.html#uh-6 (link to the part about the /g option).