Date: Mon, 14 May 2001 23:52:34 -0700 From: Dima Dorfman <dima@unixfreak.org> To: Peter Pentchev <roam@orbitel.bg> Cc: Garance A Drosihn <drosih@rpi.edu>, Cyrille Lefevre <clefevre@poboxes.com>, Brian Somers <brian@Awfulhak.org>, hackers@FreeBSD.ORG Subject: Re: xargs(1) "replstr" patch Message-ID: <20010515065234.8412C3E0B@bazooka.unixfreak.org> In-Reply-To: <20010515093121.A94119@ringworld.oblivion.bg>; from roam@orbitel.bg on "Tue, 15 May 2001 09:31:22 %2B0300"
next in thread | previous in thread | raw e-mail | index | archive | help
Peter Pentchev <roam@orbitel.bg> writes:
> On Mon, May 14, 2001 at 11:31:53PM -0400, Garance A Drosihn wrote:
> > Still, let me say that I do hope to get back to 'xargs', and add
> > the -I option. I must admit my enthusiasm for doing -I wore off
> > after seeing the current code to 'xargs'. Not only is -J more
> > useful, but it is much less work to implement (given the current
> > code as a starting point) than -I would be. Still, it *would* be
> > nice to say we have '-I', just to match what the various standards
> > list for options to 'xargs'.
> >
> > While I do like the idea of adding it, I'll admit that it isn't a
> > particularly high priority on my list of things to do...
>
> Just as a side thought: one we have -J, I guess -I could be done
> as simply as emulating -n 1 and falling through to -J :)
> Then there would need to be another couple of checks, like -n and -I
> or -J and -I not being used simultaneously, but yes, I think it would
> really be *that* easy.
Sorry, no. -I can be used more than once, and the replstr does not
have to be given as a separate pointer in argv. For example, given
the input,
one\ntwo\nthree\n
this command
xargs -I [] mv srcdir/[] destdir/[]
should generate:
mv srcdir/one destdir/one
mv srcdir/two destdir/two
mv srcdir/three destdir/three
This is a simplistic example that can be done in many other ways
(including using -J), but it shows what -I is supposed to be able to
do. -J doesn't work with the above since it only looks for the
replstr once, and will not find it unless it's given as a separate
argument. For example, this works:
xargs -J [] echo this is [] a test
but this does not:
xargs -J [] echo this is[] a test
That said, -I isn't all that hard to implement. It's just hard to
implement without rewriting xargs as it is. However, that may not be
a bad thing. -I would pose a performance problem, anyway; it would
have to iterate over the command line arguments and standard input
more than once, possibly having to cache what it finds. If we
special-case -I with something like,
if (flag_I_found) {
alternate_main_function(ac, av, ...);
/*NOTREACHED*/
}
..code as it is now goes here..
then we don't have the problem of destroying the current code, and the
performance impact will only be seen if -I is used (which is
reasonable).
Of course, "not that hard" is relative. Adding -J was rather trivial;
the delta is +30 -6, and most of those +30 are simple things like
defining or initializing variables. I guess that adding -I would mean
adding on the order of 100 or 150 lines; again, not hard, but not
something one can do in 10 minutes.
Just food for thought, I guess. I'm sure Garance wouldn't mind if
someone did this for him :-).
Regards,
Dima Dorfman
dima@unixfreak.org
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-hackers" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20010515065234.8412C3E0B>