Find newest file

capella · 03-04-2004, 09:10 AM

I have a directory where a new file gets created every 2 hours, I need a way to find the newest file in that directory regardless of the filename, can someone please help me out.

homey · 03-04-2004, 09:32 AM

something like this?
find /mnt/backup -type f -name '*' -cmin -120 {} \;

This example takes an action on the files it finds, deletes them in this case.
find /mnt/backup -type f -name '*.gz' -mtime +90 -exec rm {} \;

capella · 03-04-2004, 09:44 AM

Thanks but I really need to find the newest file regardless of time created as there will be other backups moving in that directory, so the solution cannot be base on cmin etc..

itsjustme · 03-04-2004, 09:52 AM

How about doing an 'ls -lt' and piping that to a file and then parse the first line with a small script to get the filename.

capella · 03-04-2004, 09:58 AM

I read somewhere that i can do that and combine it with tail-1, but i have no idea what that means im a database guy and was hoping someone could spoonfeed me the solution

libCognition · 04-27-2012, 04:44 PM

I just came up with a solution to this using recursion:

Code:

latest_file() {
  if [[ ! -z ${2} ]]
  then
    ls -l $2
    latest_file $1 $(find $1 -type f -newer "$2" | head -1)
  fi
}

Then run it with first argument being the folder to search, and the second argument being any arbitrary file in that folder. e.g. "latest_file /etc/ /etc/fstab | tail -n 1"

Limitations: probably doesn't work if a filename has spaces.

catkin · 04-27-2012, 10:25 PM

Code:

ls -1rt --group-directories-first | tail -1

(that's a number 1 in -1rt)

libCognition · 04-28-2012, 02:41 AM

Quote:

Originally Posted by catkin

Code:

ls -1rt --group-directories-first | tail -1

(that's a number 1 in -1rt)

That misses all the files that have a depth greater than zero.

This version of that approach would get some penetration:

Code:

ls -1rt --group-directories-first * */* */*/* | tail -1

but then you need to know how deep the structure goes. Or substitute with:

Code:

ls -1rt --group-directories-first $(find . -type d | sed 's/[^/]/*/g' | tr -s '*' | sort | uniq) | tail -1

What's interesting is the 'ls' version (above) got a different answer than 'find -newer' (from post 6), but both files were created the same minute. So one of the two approaches looks at the seconds.

catkin · 04-28-2012, 04:15 AM

Quote:

Originally Posted by libCognition

That misses all the files that have a depth greater than zero.

True but capella wrote in the OP "I have a directory" so its not a problem.

catkin · 04-28-2012, 04:26 AM

Quote:

Originally Posted by libCognition

What's interesting is the 'ls' version (above) got a different answer than 'find -newer' (from post 6), but both files were created the same minute. So one of the two approaches looks at the seconds.

That is interesting but experimentation showed that both find and ls are very fine grained regards time:

Code:

c@CW8:/tmp/tmp$  for i in $( seq 1 100 ); do touch $i; touch a$i; done
ls -lrt
[files listed in created order]
c@CW8:/tmp/tmp$ find . -newer 99
./100
./a100
./a99