From mboxrd@z Thu Jan  1 00:00:00 1970
Received: from oldp.astro.wisc.edu ([128.104.39.15]) by hawkwind.utcs.toronto.edu with SMTP id <2223>; Sun, 16 May 1993 02:49:29 -0400
Received: by oldp.astro.wisc.edu (5.65/DEC-Ultrix/4.3)
	id AA09549; Sun, 16 May 1993 01:49:14 -0500
Message-Id: <9305160649.AA09549@oldp.astro.wisc.edu>
To:	rc@hawkwind.utcs.toronto.edu
Subject: Re: read
Date:	Sun, 16 May 1993 02:49:13 -0400
From:	Alan Watson <alan@oldp.astro.wisc.edu>
X-Mts: smtp

I have just realized that my implementation of read returns the
wrong status if you type:

   ; read x
   foo<ctrl-d>

More seriously, and strengthening my suggestion that read should be a
builtin, I cannot find a way to get `awk' to simulate `line'.  In
response to the above input, line would print "foo\n" and set status to
1.  The obvious awk replacement, i.e.,

   ; fn line { awk '{print $0; exit 0}END{print $0; exit 1}' }

does not work in this way, rather I have to press <ctrl-d> twice, and
then it prints "foo\n0\nfoo\n1\n".

Your mission, should you choose to accept it, is to come up with a
portable (i.e., you can't use line) read which implements v7 semantics
on EOF (i.e., if anything was typed before EOF then set the variable to
this, otherwise unset it, and return 1 if an EOF is seen).  Actually, a
working implementation of line will be sufficient (although, obviously,
the C code for line will not do).

I'd also very much appreciate it if someone could restore my faith in
awk by explaining its output!

Alan.