regex problem with sed

ta0kira · 06-16-2005, 01:13 AM

I am working on a configuration file parser, written in a bash script. I am parsing a file where each line is in the form of

[comment][option title][comment]: [option value]

where [comment] is either " *[anything]* " or "". The problem is that egrep will find the lines I need, but sed will not remove the option tag for me.

Here is a simplified version of my script (sorry, no Linux here so I can't test it...):

Code:

#!/bin/bash

beg="^"
end=":"
mandspc="[ 	]*"
cmt="($mandspc\*.*\*$mandspc|)"
error="config_error"

function extract()
{
  tag="$beg$cmt$1$cmt$end$mandspc"
  if [ `egrep -c "$tag" "$file"` -ne 1 ]; then
    echo "$error"
  else
    egrep -h "$tag" "$file" | sed "s/$tag//"
  fi
}

echo "`extract \"extport_fs\"`"

(In case you can't tell, that is a space and a tab in mandspc.)

Here is the part of my config file applicable (file named ext-config):

Code:

* file system for extport *	extport_fs:     ext2

egrep can find the line just fine, however sed will not remove the "file system for extport * extport_fs: " from the line. Can you see anything I'm doing wrong as far as the regex? Is there something besides sed that I can use? It's still in "working" status, therefore I am going to change the style of the lines once I get it working. Thanks.
ta0kira

[EDIT] Made it a little more easy to read...

murugesan · 06-16-2005, 02:05 AM

Are you looking for something like this ?

Code:

$> cat > data
* file system for extport * extport_fs * file system for extport * :     ext2
* file system for extport *     extport_fs:     ext2
* file system for extport * extport_fs * file system for extport * :     ext2

* file system for extport *     extport_fs:     ext2

$> sed 's/\*[a-zA-Z0-9 ]*\*//g' data
 extport_fs  :     ext2
     extport_fs:     ext2
 extport_fs  :     ext2

     extport_fs:     ext2

ta0kira · 06-16-2005, 02:58 AM

Close, but the problem seems to be with either the .* or the spaces. [a-zA-Z0-9 ]* doesn't do the entire job because I want it to be a true comment; it can contain anything except the comment end delimeter. The big question is is there a difference in how I need to denote space, tab, and "anything besides nothing" between egrep and sed? I tried using /t in place of tab, but that did not work (actually that was with grep, so I'll try with egrep). Maybe I could use [\w\W]* instead of .*? Thanks.
ta0kira

murugesan · 06-16-2005, 03:16 AM

How about this ?

Code:

sed 's/\*.[^\*]*\*//g' data

Hope this solves your problem.

murugesan · 06-16-2005, 03:22 AM

How about this ?

Code:

sed 's/\*.[^\*]*\*//g' data

If you want to remove the tabs or spaces you can use something like

Code:

sed -e 's/\*.[^\*]*\*//g' -e 's/[        ]//g' data

Here -e 's/[ ]//g' the quare bracket [] consists of space followed by a tab.

Hope this solves your problem.

ta0kira · 06-16-2005, 04:20 AM

What does .[^\*]* do? Looks like [single char][anything besides * repeated indefinitely]. That will not work for a delimeter larger than one char though. Can I use [:space:] in place of space tab?
ta0kira

murugesan · 06-17-2005, 01:19 AM

Here we are replacing any string which starts with star [ \* ] and followed by any character [ . ] which is not a star [ ^\* ] and also ends with star [ \* ]. You can use space (key) for replacing.

ta0kira · 06-20-2005, 12:33 AM

Got it figured out; I needed the -r option with sed. Turns out my expression was fine for what I needed. Thanks for your help.
ta0kira