a

Trying running perl -e 'while(1) { 1 }' sometime and see what that does to your CPU. :-)

As for how long to sleep, it depends on what you're doing: sometimes sleep(1) is enough, othertimes something a little (or a lot) longer is preferable.

Definitely something for you to experiment with, but the question you need to ask is how long can you afford to postpone another iteration of the loop?

Also, if you're feeling particularly nutty, you can use...

  while ( sleep(10) ) {
      ...
  }
[download]

... instead of the more traditional (?) ...

  while ( 1 ) {
      ...;
      sleep(10);
  }
[download]

Although the former ends up being as if sleep(10) was the first thing called inside the loop.

    --k.

It can. Remember you're checking if that 1 is still 1 on every loop. How long does a file usually stick around before being deleted?

Update: And of course the file test itself. Depending on how your disk caching is happening your results may vary.

My first big project had files being written across Samba mounts and then -e tests as well. CPU time wasn't much of factor, but the network could get clogged.
Getting the various perl programs talking directly to each other was much faster. So my experience might have nothing to do with your stuff

The file is there all the time - when it is deleted, it is a trigger for action. It is deleted by a user to trigger that action.

Is there a better way to idle than while(1)? It only uses about .05% cpu right now, so its not that bad...

The point is that while(1){} isn't idle, it's spinning madly behind the scenes, testing 1 as often as possible. In your case you probably have lots of other processes that push your script down to .05%, but I'm just guessing on that. On this old workstation the while(1){} goes up above 90% when not much is happening. Now I'm certainly no expert on process scheduling, but then neither is Win32 ;-).

I've been taught (though it may be deprecated like lots of other things I've learned) that it's just bad manners for a program to basically waste time spinning madly, as you don't know what other apps might want to do with those cycles. If you sleep() you're letting the system decide what to do with those cycles you'd otherwise be wasting. Of course as others have already said, it depends on just how urgent your task is - do you need to act the same millisecond the file is deleted, or is 1/10 of a second fast enough (and don't forget your network latency is probably killing way more response time than the sleep() will).

Maybe if you explain your application a bit more fully the answer will become clearer.

--
I'd like to be able to assign to an luser

If you're using Win32, there's always Win32::ChangeNotify. This will let you wait until your specified file is changed, then take action. I use it in a script on my NT box, and it doesn't take much CPU (especially compared to while(1)!).

Guildenstern
Negaterd character class uber alles!

However, if your delay is big, a better answer might be to use a scheduler to run you check program instead of idle in the background. On a *nix box use cron, not sure on Win32 box. This isn't as worthwhile with a small delay because of the overhead of loading Perl.

While this may not use much CPU resources, since you are doing I/O, it is also using disk resources. I'd set the delay as long as you can stand.

=Blue
...you might be eaten by a grue...

until (! $opt{time}) { # loop every specified seconds

{ check_disk_space($opt{dir} ) ; # run my df subroutine on specified d
+irectory

sleep ($opt{time}) # wait specified time, then begin again

}
[download]

if ($mr_leisure) { bow; }
[download]