How to empty ("truncate") a file on linux that already exists and is protected in someway?

Question

I have a file called error.log on my server that I need to frequently truncate. I have rw permissions for the file. Opening the file in vi > deleting all content > saving works (obviously). But when I try the below

cat /dev/null > error.log

I get the message

File already exists.

Obviously there is some kind of configuration done on the server to prevent accidental overriding of files. Can anybody tell how do I "truncate" the file in a single command?

For a non-protected file truncate -s 0 file works. – Pramod Dec 30 '12 at 13:53 — Pramod, Dec 30 '12 at 13:53

score 161 · Accepted Answer · answered Mar 11 '10 at 07:45

161

You have the noclobber option set. The error looks like it's from csh, so you would do:

cat /dev/null >! file

If I'm wrong and you are using bash, you should do:

cat /dev/null >| file

in bash, you can also shorten that to:

>| file

answered Mar 11 '10 at 07:45

R Samuel Klatchko

74,869
16
134
187

you are right about the shell being csh. How did you know that? – Sumeet Pareek Mar 11 '10 at 08:02
6

@Wikidkaka - based on the error message. csh on my system gives the similar `File exists` error while bash gives the very different `cannot overwrite existing file` error. – R Samuel Klatchko Mar 11 '10 at 08:10
6

@Jarmund - because the user has noclobber set, `>file` does not work. – R Samuel Klatchko May 16 '13 at 00:03
What's the difference between `> file` and `>| file`? – Will Jul 28 '14 at 02:27
2

@Will - if you have the `noclobber` option set, trying to redirect to an existing file fails. To override for the specific redirection while leaving noclobber on, you can use `>|` – R Samuel Klatchko Jul 28 '14 at 16:37

score 75 · Answer 2 · answered Aug 26 '13 at 06:12

75

You can also use function truncate

$truncate -s0 yourfile

if permission denied, use sudo

$sudo truncate -s0 yourfile

Help/Manual: man truncate

tested on ubuntu Linux

answered Aug 26 '13 at 06:12

risnandar

5,513
1
26
17

1

truncate also touches the file's modification time - while ">file" does not touch if if the file is unchanged. – dsteinkopf Mar 17 '17 at 21:11
It doesn't release the disk space that file holds. At first it shows the 0 bytes. But when you write something to that file, the file size is = before truncate + new data size. – Aniket Kulkarni Dec 02 '19 at 11:44
@AniketKulkarni that's because your file has not been opened with APPEND flag, so it keeps the write buffer, and when it updates that buffer it flushes the whole buffer to the file, and that's how you see it with the same + increased size. – Mladen B. Aug 13 '21 at 20:19

score 48 · Answer 3 · answered Mar 11 '13 at 22:50

48

This will be enough to set the file size to 0:

> error.log

answered Mar 11 '13 at 22:50

SIFE

5,567
7
32
46

6

actually, this **doesn't** work with noclobber (tested it myself), so it's NOT a valid solution in this case ("-bash: error.log: cannot overwrite existing file"); see commend above ("@Jarmund - because the user has noclobber set, >file does not work."). I'm just wondering what kind of "intelligent people" upvote obviously wrong answers like this... – Apr 12 '14 at 16:14

score 13 · Answer 4 · answered Apr 09 '15 at 09:07

13

the credit goes for my senior colleague for this:

:> filename

This will not break log files, so you can even use it on syslog, for example.

answered Apr 09 '15 at 09:07

petermolnar

1,626
14
23

9

Please explain. How does it work? What is that notation doing? – user31986 Jan 15 '16 at 22:00
3

@user31986 `:` is a no-effect command (almost a comment), then `>` is redirection as usual, thus this just redirects the lack of output from a command that does nothing into the file. If noclobber is enabled, you need `:>| filename` (bash) (and I presume `:>! filename` in csh assuming csh has `:`). – Rhubbarb Aug 31 '17 at 12:47

score 4 · Answer 5 · answered Oct 31 '12 at 13:42

4

false|tee fileToTruncate

may work as well

answered Oct 31 '12 at 13:42

Dodger Web

114
1
5

That's just silly. Or if this is useful, obscurely getting `dd` to output nothing would be even better than the plain readable `false`? – tripleee Nov 16 '16 at 06:29

score 4 · Answer 6 · edited May 23 '17 at 12:18

4

Since sudo will not work with redirection >, I like the tee command for this purpose

echo "" | sudo tee fileName

edited May 23 '17 at 12:18

Community

1
1

answered Jan 09 '14 at 07:09

sakhunzai

13,900
23
98
159

well, actually it does, as described in both the question you linked and here above. "sudo sh -c '> error.log'"... – Apr 12 '14 at 16:11
it depends on the permissions on the target file `error.log`, if that is already writable there is not need to sudo. I think the other issue is you are passing command as string argument to `sh` , I don't know how that is handled internally but that is another layer of complexity/redirection added – sakhunzai Apr 14 '14 at 05:36
2

This won't actually truncate the file, because `echo ""` will add a newline. You can use `echo -n ""` to suppress the newline – Seb Feb 19 '16 at 10:43

score 1 · Answer 7 · edited Jul 13 '18 at 13:37

1

Any one can try this command to truncate any file in linux system

This will surely work in any format :

truncate -s 0 file.txt

edited Jul 13 '18 at 13:37

Jouby

2,196
2
21
33

answered Jul 13 '18 at 12:34

user3301460

11
4

score 1 · Answer 8 · answered Sep 04 '19 at 08:22

1

I do like this: cp /dev/null file

answered Sep 04 '19 at 08:22

Den129

11
3

score -4 · Answer 9 · answered May 27 '13 at 08:22

-4

You can try also:

echo -n > /my/file

answered May 27 '13 at 08:22

Oden

756
8
21

8

See the answer with a ton of votes? Try reading it, and look up `noclobber` and then ask yourself if your answer makes any sense at all. – May 27 '13 at 08:26

How to empty ("truncate") a file on linux that already exists and is protected in someway?

9 Answers9

Linked