I'm working on a shell script that does certain changes on a txt file only if it does exist, however this test loop doesn't work, I wonder why? Thank you!
while [ ! -f /tmp/list.txt ] ;
do
sleep 2
done
-
4I can't say that I'm surprised; that loop doesn't attempt to change anything.Ignacio Vazquez-Abrams– Ignacio Vazquez-Abrams2010-03-04 14:12:04 +00:00Commented Mar 4, 2010 at 14:12
-
5Works for me - the loop terminates when the file is created outside of the script.anon– anon2010-03-04 14:14:15 +00:00Commented Mar 4, 2010 at 14:14
-
1in fact, this loop only serves to wait until the file is there, the rest of my script does the changes... :pZenet– Zenet2010-03-04 14:14:25 +00:00Commented Mar 4, 2010 at 14:14
-
2Then the while loop works, it's just me... sorry.Zenet– Zenet2010-03-04 14:15:53 +00:00Commented Mar 4, 2010 at 14:15
-
1Given the "the while loop works; it's just me" comment, presumably this should be closed as 'no longer relevant', except that reason for closing is no longer available...?Jonathan Leffler– Jonathan Leffler2010-03-04 14:31:35 +00:00Commented Mar 4, 2010 at 14:31
|
Show 1 more comment
8 Answers
When you say "doesn't work", how do you know it doesn't work?
You might try to figure out if the file actually exists by adding:
while [ ! -f /tmp/list.txt ]
do
sleep 2 # or less like 0.2
done
ls -l /tmp/list.txt
You might also make sure that you're using a Bash (or related) shell by typing 'echo $SHELL'. I think that CSH and TCSH use a slightly different semantic for this loop.
3 Comments
valentt
Why are you using inverted file check? Shouldn't while [ -f /tmp/list.txt ] be used instead?
Kenyakorn Ketsombut
@valentt No hew loop say literally "while NOT file exists do sleep" .. if you would remove the 'NOT' the loop would break instantly
jaques-sam
in 1 line:
while [ ! -f /tmp/list.txt ]; do sleep 2; done; ls -l /tmp/list.txtIf you are on linux and have inotify-tools installed, you can do this:
file=/tmp/list.txt
while [ ! -f "$file" ]
do
inotifywait -qqt 2 -e create -e moved_to "$(dirname $file)"
done
This reduces the delay introduced by sleep while still polling every "x" seconds. You can add more events if you anticipate that they are needed.
6 Comments
Michał Šrajer
+1 for efficiency. Pooling with sleep is ugly. For folks not knowing inotifywait - it's in package inotify-tools.
Craig Ringer
That's an exceedingly handy tool. For anyone wondering why the loop, it's to deal with possible race conditions between creation and waiting and because inotifywait has
--exclude to filter out filenames, but not --include to ignore everything except the filename. The above command should use the -qq argument instead of >&/dev/null though.
Alex Dean
t is the
--timeout, not the frequency of checking, no? The point of inotifywait is that there is no polling
yingted
@AlexDean The timeout is to prevent a race condition. Polling with sleep is slow because the loop won't exit during the sleep, but inotifywait will exit before the timeout if it sees an event.
yingted
@AlexDean Yes, but it's necessary to prevent a TOCTTOU race condition. Otherwise,
inotifywait could hang indefinitely if a file is created just before it starts listening for events. |
I had the same problem, put the ! outside the brackets;
while ! [ -f /tmp/list.txt ];
do
echo "#"
sleep 1
done
Also, if you add an echo inside the loop it will tell you if you are getting into the loop or not.
Comments
Here is a version with a timeout so that after an amount of time the loop ends with an error:
# After 60 seconds the loop will exit
timeout=60
while [ ! -f /tmp/list.txt ];
do
# When the timeout is equal to zero, show an error and leave the loop.
if [ "$timeout" == 0 ]; then
echo "ERROR: Timeout while waiting for the file /tmp/list.txt."
exit 1
fi
sleep 1
# Decrease the timeout of one
((timeout--))
done
Comments
I ran into a similar issue and it lead me here so I just wanted to leave my solution for anyone who experiences the same.
I found that if I ran cat /tmp/list.txt the file would be empty, even though I was certain that there were contents being placed immediately in the file. Turns out if I put a sleep 1; just before the cat /tmp/list.txt it worked as expected. There must have been a delay between the time the file was created and the time it was written, or something along those lines.
My final code:
while [ ! -f /tmp/list.txt ];
do
sleep 1;
done;
sleep 1;
cat /tmp/list.txt;
Hope this helps save someone a frustrating half hour!
1 Comment
BryanC
In your specific use case, I would opt for
while ! [ -s /tmp/list.txt]; do sleep 1; done; cat /tmp/list.txt<br/> -s in while checks is the file is not 0 bytes, so this negates it to keep looping until it has data written. Alternatively, until is a good option too: until [ -s /tmp/list.txt ]; do sleep 1; done; cat /tmp/list.txtLike @zane-hooper, I've had a similar problem on NFS. On parallel / distributed filesystems the lag between you creating a file on one machine and the other machine "seeing" it can be very large, so I could wait up to a full minute after the creation of the file before the while loop exits (and there also is an aftereffect of it "seeing" an already deleted file).
This creates the illusion that the script "doesn't work", while in fact it is the filesystem that is dropping the ball.
This took me a while to figure out, hope it saves somebody some time.
PS This also causes an annoying number of "Stale file handler" errors.
Comments
works with bash and sh both:
touch /tmp/testfile
sleep 10 && rm /tmp/testfile &
until ! [ -f /tmp/testfile ]
do
echo "testfile still exist..."
sleep 1
done
echo "now testfile is deleted.."
Comments
do it like this
while true
do
[ -f /tmp/list.txt ] && break
sleep 2
done
ls -l /tmp/list.txt
