Hi all : - Is there anyway to delete nnn number of lines from the top of a file without opening it? Can it be done using head or tail? I tried tail and piped the output to another file and renamed it to original (after renaming the original to something else). It worked, but the problem is the renamed original is still being written by the application instead of the new file with the original name. This is a 10g AS log file.

Any other ideas on how to shrink that logfile size?

Thanks.

Use sed with the -i option. For example, if you want to remove the top 1000 lines from your file:

What is the problem with it being a new file? So long as it contains the correct information, etc, what difference does having it created by a script from the output of another file make?

Maybe ways around this issue can then be suggested.

James

EDIT: Beaten to it by a proper solution!

I agree with jimbo. If it's just a log file you can just rotate it out. It will still contain the info that you need. Check out logrotate. It should get the job done.

Thank you for the response all. Here is what I found somewhere else on the internet: This is using Perl.

perl -i -ne 'print unless 1..500' <file_name> --> this deletes the top 500 lines of that file.

However I am interested in the 'sed' solution suggested by Matthew Gates. I tried but no success. Matthew if you don't mind, can you please give an example:

Regards.

Sure. Say you have a file named "bobdobbs", containing:
All you need to do to remove the top, say 3 lines is this:
After this command completed, the file bobdobbs contains only:
It's well worth doing a sed tutorial. If you're like me you'll leave it too late, and when you finally get round to it you'll think "****! If I'd known this a few years ago I would have saved myself a massive amount of work".

Ho hum.

Oh, perhaps a little explanation of the sed command might help. As a reminder, here's the command:
-n means "don't print lines to the output file unless specifically requested (with the p command)".

-i means "edit input files in place - don't write output to standard output - save any output to the input files".

The bobdobbs is the name of input file.

The rest is the command. The command is in single quotes to prevent the shell from interpretting meta characters and passing sed something else - sed needs to see this as a literal string.

OK. The command is pretty simple:
Sed commands can be prefixed with an address. This is to say some expression which tells sed which lines of the input files to execute the rest of the command on. In this case the address is ...which means from line 4 to the end of the file. If we said
...it would mean from line 4 to line 12 (inclusive). $ is a special address meaning "the last line of the input file".

The second part is the actual operation to be performed on lines which match the address. In this case the command is simply p - print the line. Usually sed will print the line after any operation you specify, but we told it (with the -n option, as described above) not to print output unless we explicitly request it.

So, in summary, 4,$ p means "if the line number is between 4 and the end of the file (inclusive), print the line".

Hope that helps.

Learn sed. It's time well spent.