Shell – difference between results and effectiveness after adding -f in rm and -print in find

shell-scriptxargs

I got a task to delete all files which are older than 90 days from /dir/prd/log recrusively i.e delete from subdirectries too. I made this command:

find /dir/prd/log* -mtime +90 | xargs rm

Then I was talking to a unix guy and he suggested

find /dir/prd/log* -mtime +90 -print | xargs rm -f

I just want to know what would have been his purpose to put "-print" and "-f" in command.

Best Answer

-f tells rm to never prompt (e.g. when it encounters read-only files) and to ignore missing files it’s asked to delete (instead of indicating an error). It will also not complain if not passed any file to remove.

Here, depending on the xargs implementation rm's stdin will be either the pipe from find or /dev/null, so without -f, find may end up reading the answer to those prompts from the output of find!

-print is the default action for find in Linux distributions and POSIX-compliant systems, but might need to be explicitly specified on very old Unix or Unix-like systems.

So his purpose was probably to make the command more robust and portable. You could do better with some variants of find (going beyond POSIX):

find /dir/prd/log -mtime +90 -print0 | xargs -0 rm -f

avoids problems with filenames containing “special” characters (including space, newline, tab, single quote, double quote, backslash), and if your find supports the -delete action,

find /dir/prd/log -mtime +90 -delete

avoids spawning other processes to perform the deletion (and also avoids some race condition issues). (Note that I’m specifying /dir/prd/log here to match your stated requirement from the first sentence of your question, as confirmed in the comments.)

With POSIX find, you can still avoid xargs and filename parsing by asking find to run rm itself:

find /dir/prd/log -mtime +90 -exec rm -f '{}' +

Beside being more portable and reliable, it also avoids the issue of rm reading prompt answers from find's output mentioned above (it also avoids running rm at all if no file is found).

If /dir/prod/log contains subdirectories, you’ll want to filter them to avoid error messages since rm can’t delete them:

find /dir/prd/log ! -type d -mtime +90 -exec rm -f '{}' +

Related Solutions

Shell – Creating a shebang pointing portably to an interpreter in the folder of a script

EDIT: (The original answer still responds to the stated question, but indeed makes file.js not a valid JS file). In order to allow the desired behavior, follows a shebang + 1st-line-combination that will feed the file.js through node skipping the two first lines, thus feeding it with just the JS code:

#!/bin/sh
sed '1,2d' $0 |node --harmony_arrow_functions; exit $?
/* Your JS code begins here */

--- Below the original answer ---

If you can count as certain your interpreter will be in the same directory of the interpreted file (which is the stated question), you can use $(dirname $0) to obtain its path.

Example:

#!/bin/sh
exec $(dirname $0)/nodeharmony.sh "$0" "$@"

In this case, when executing

$ cd ..
$ ./subfolder/file.js

The $(dirname $0) will resolve to ./subfolder, so exec will use ./subfolder/nodeharmony.sh as the interpreter.

Shell – Is it possible to write to the crontab from a multipurpose script

Per Ipor Sircer's answer about the usage of cron, i.e.

man crontab:

   crontab [ -u user ] file

   The  first  form  of this command is used to install a new crontab from
   some named file or standard  input  if  the  pseudo-filename  ``-''  is
   given.

this means that you send the lines you want in your crontab file, to the stdin of this command:

crontab -

crontab will recreate a new cron file containing those commands.

The script will first print your existing crontab using crontab -u $user -l 2>/dev/null.
_{you will need to assign the value of your user to $user or use $USER if its in your environment.}
It will print the new lines you want and capture the aggregated result into a pipe connected to stdin of crontab -.

Here's how it should look like in your general-purpose script:

#!/bin/bash

user=YOU_NEED_TO_ENTER_YOUR_USER_HERE

# use a subshell to capture both commands output into the pipe (    # prints the current value of crontab
    crontab -u $user -l 2>/dev/null

    # print your cron jobs to STDOUT    
    cat <<- 'EOF'
        0 8 * * *  tar -zcvf /home/USERNAME/backups/files/www-html-$(date +\%F-\%T-).tar.gz /var/www/html
        0 8 * * *  find /home/USERNAME/backups/files/* -mtime +30 -exec rm {} \;

        0 8 * * *  mysqldump -u benia -p121212 --all-databases > /home/USERNAME/backups/mysql/alldb_backup.sql
        1 8 * * *  tar -zcvf /home/USERNAME/backups/mysql/alldb_backup-$(date +\%F-\%T-).sql.tar.gz /home/USERNAME/backups/mysql/alldb_backup.sql
        2 8 * * *  rm /home/USERNAME/backups/mysql/alldb_backup.sql
        2 8 * * *  find /home/USERNAME/backups/mysql/* -mtime +30 -exec rm {} \;
    EOF

# Everything printed to stdout - inside the subshell will be connected
# to the pipe and feed to crontab on stdin - recreating a new crontab ) | crontab -

Best Answer

Related Solutions

Shell – Creating a shebang pointing portably to an interpreter in the folder of a script

Shell – Is it possible to write to the crontab from a multipurpose script

Related Question