Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.2 9/18/84; site brl-tgr.ARPA
Path: utzoo!watmath!clyde!burl!ulysses!allegra!mit-eddie!think!harvard!seismo!brl-tgr!tgr!bilbo.niket@locus.ucla.edu
From: bilbo.niket@locus.ucla.edu (Niket K. Patwardhan)
Newsgroups: net.unix
Subject: Re: INFO-UNIX Digest  V2#006
Message-ID: <3492@brl-tgr.ARPA>
Date: Fri, 22-Nov-85 12:49:58 EST
Article-I.D.: brl-tgr.3492
Posted: Fri Nov 22 12:49:58 1985
Date-Received: Sun, 24-Nov-85 06:47:59 EST
Sender: news@brl-tgr.ARPA
Lines: 29

The pattern matching allows you to break up the pattern into parts and replace
specific bits of it in the substitute command. Try this (you may have to escape
stuff if this is on the sed command line)

s/\(#{garbage}\)#/\1/

What this does is bracket the first part of the pattern so you can refer to it
later: (the \(\) does this). Then when you are doing the replacement you ask
for only the first part rather than the whole thing: ( \1 rather than &). If
you have multiple pieces you bracket the pieces you want to refer to, and you
refer to the pieces by counting the number of \( starting from the left. Eg.

s/ \(\(xyz\)abc\(pqr\)\) / \1\2\3/

on
 xyzabcpqrxxx

results in
 xyzabcpqrxyzpqrxxx

\1 is xyzabcpqr
\2 is xyz
\3 is pqr

If the trailing # is at the end of the line just use

s/#$//

Good luck!