Shell Programming and Scripting

I don't know how large your log files will be, but greping or seding them will always read them from the start, which may become serious overhead.
If you just want to know if there's a DONE occurred in your log file ever since your last visit, why don't you keep the recent file size in a variable:

Code:

$ skip=$(stat -c%s file)

, and then, n seconds later, check the lines added ever since with

Code:

$ dd if=file ibs=1  skip=$skip | grep DONE

. You'll always have exactly the last chunk of data added with no overhead!
See the time differences for a 200MB log file::

Code:

$ time grep -n DONE file |tail -1
7040081:DONE

real    0m0.653s
user    0m0.516s
sys    0m0.132s
$ time dd if=file ibs=1 skip=$skip |grep DONE
DONE

real    0m0.005s
user    0m0.000s
sys    0m0.004s

dd is standard for every *nix implementation while stat may not be; then ls and awk will be your friends...

Hi RudiC-

Wow, I was really excited to see this idea too since I'm a big fan of efficiency.

Unfortunately, as you mentioned may be the case, "stat" isn't available in my environment (Solaris).

I'm not as familiar with dd (it appears I need "stat" beforehand to make it use of "dd" ? ).

If that's true, perhaps I can combine your method with Hanson's way of getting the (original) last line with done:

Code:

LAST_LINE_WITH_DONE=`grep -n DONE log | tail -n 1 | cut -d : -f 1`

....then, perhaps there's some way of modifying the below line:

Code:

NEWLAST_LINE_WITH_DONE=`grep -n DONE $LOGFILE | tail -1 | cut -d : -f 1`

...so that it can start searching the file after line number $LAST_LINE_WITH_DONE ?

In the worst case, this is more of a fun challenge to make things more efficient, since Hanson's idea (ps: thanks again!) has been working very well too.

Thanks again everyone!
CG

As I stated in my post, you don't NEED stat; it's just more efficient than this method to get the files sizes in bytes:

Code:

$ ls -l file newaa
-rw-rw-r-- 1 xxxxx xxxxx 1300 Apr 14 22:26 file
-rw-rw-r-- 1 xxxxx xxxxx  111 Apr  7 22:37 newaa
$ ls -l file newaa | awk '{print $5}'
1300
111
$ stat -c%s file newaa
1300
111

The result ist the same, as you can see. dd just has to skip last time's file sizes' count of bytes to get at the new lines added.
I could imagine that Solaris has a stat equivalent command, btw, and you just need to search a bit...

And, the two methods would be difficult to combine, as dd needs a number of bytes or blocks to skip, and grep identifies lines that can be of varying length. Would sort of work if one could assume an average line length, but this were far from being exact.