Path: utzoo!utgpu!news-server.csri.toronto.edu!rutgers!cs.utexas.edu!convex!usenet
From: tchrist@convex.COM (Tom Christiansen)
Newsgroups: comp.unix.questions
Subject: Re: Convert newlines to something else.
Message-ID: <1991Feb10.143018.14664@convex.com>
Date: 10 Feb 91 14:30:18 GMT
References: <6011@idunno.Princeton.EDU> <1991Feb10.030329.2094@jpradley.jpr.com>
Sender: usenet@convex.com (news access account)
Reply-To: tchrist@convex.COM (Tom Christiansen)
Distribution: usa
Organization: CONVEX Software Development, Richardson, TX
Lines: 49
Nntp-Posting-Host: pixel.convex.com

From the keyboard of jpr@jpradley.UUCP (Jean-Pierre Radley):
:In article <6011@idunno.Princeton.EDU> tvz@phoenix.Princeton.EDU (Timothy Van Zandt) writes:
:>How does one convert newlines to something else in a text file? I cannot
:>figure out how to do it with sed, if it is possible at all.
:This script uses sed to either add or delete CR's, depending on how it's called
:(i.e, store it as addcr, linked to delcr).
:	# addcr : adds CRs
:	# delcr : removes CRs
:	[ $# -ne 2 ] && echo Usage\: $0 infile outfile && exit
:	case $0 in
:		*addcr) sed s+$+^M+ <$1 >$2;;
:		*delcr) sed s-^M--g <$1 >$2;;
:	esac

That doesn't work.  The main problem is that carriage-returns (\r, 015)
are not the same as newlines (\n, 012), and under UNIX, it's newlines
that you have in your textfiles.  You will find that replacing the ^M's
with ^J's doesn't work in sed, nor are \n's going to help you without
a lot more monkeying around.

Minor nits: those sed expressions need to be quoted -- my sh blew up on
the addcr sed, although my ksh was content with them; so much for bug
compatibility.  Another one is that were this to actually work, it could
have quite easily been written as a filter, rather than a program that
requires known (and singular) input and output files.  For example, the
delcr line could have been written like this (assuming that this expr
worked, which it doesn't):

		*delcr) sed 's+$+^M+' $*;;


One perl version of addcr is:

    perl -pe 's/\n/\n\n/' 

and one for delcr is:

    perl -pe 's/\n//' 

Now, it's actually faster to do this for delcr:

    perl -pe chop

But that might scare aware folks who want it to look more like how 
sed really ought to work. :-)

--tom
--
 "All things are possible, but not all expedient."  (in life, UNIX, and perl)