On Thu, Apr 13, 2006 at 09:09:47AM +1000, Peter Rundle wrote:
> Matthew Hannigan wrote:
> >Yeah, not a bad call - it follows the 'correct first, fast later
> >/ premature optimization is the root of all evil' principle.
>
> Obviously xargs is a "better" solution based on the unix concept of only
> doing one thing and doing it well. xargs adds the capability to any data
> stream where as -exec is tied to find. I just reckon that people are
> just a bit quick to diss -exec. Consider these two cmds;
>
> find . -exec echo {} \;
>
> find . -print0 | xargs -0 echo
>
> Don't produce the same result.
>
> >Still, it can be a lot slower to use -exec.
>
> Maybe but.... given that rm is the command in this case, the time to
> remove the files from disk is orders of magnitude greater than the time
> to load and execute the program.
Lots of good points.
> Also what happens if find returns a million file names? xargs can't put
> them all on a single command line else the dreaded "too many args" error
> will occur. So xargs must have some smarts in it to call the command
> multiple times, passing it 1,2,...N args at a time? Speculation on my
> part, I don't know this for a fact but man xargs says;
>
> If any invocation of the command exits with a status of 255,
> xargs will stop immediately without reading any further input.
>
> So it obviously breaks the args up into chunks and invokes the command
> multiple times in any case.
Indeed it does.
> The stopping on error could be a "good thing" or a "bad thing" depending
> on what you are trying to do. In the original example the find statement
> will pick up directory names, hence passing it to "xargs rm -f" should
> cause an error straight up as the first name found will be a directory.
> Hence the find with xargs method won't clear the smtpd logs which was
> Terry's intention. (Or will it as what error code does a failed rm
> return?) On the other hand, -exec will continue on blindly executing the
> given command for each filename passed to it, which may or may not be "a
> good thing" depending on circumstance.
Yep, and another thing. Since xargs might run command with just 1
args, the command may behave differently. E.g. grep when given one file
will not print the filename as part of the match, but will when given
many files.
So
find . -print0 | xargs -0 grep | dosomething
might misbehave very occasionally, depending on what exactly dosomething does.
The fix is to use -H with grep (non-portable) or just
add a certain to be not matched file after grep, so xargs runs
grep with at least 2 files each time:
find . -print0 | xargs -0 grep /dev/null | dosomething
Of course using -exec avoids the problem entirely.
Matt
--
SLUG - Sydney Linux User's Group Mailing List - http://slug.org.au/
Subscription info and FAQs: http://slug.org.au/faq/mailinglists.html
