找到最新的修改文件并执行

Question

I want to find out the latest built rpm in a directory and then exec something on it. Something similar to below.

/bin/ls -1t srcdir/*.rpm | head -1 

But with the find command

find srcdir/*.rpm <get-only-the-most-recently-modified-file> -exec "<do-something>"

Answer 1

If you just wish to see the latest rpm :

ls -1tr /pat/to/directory/*.rpm | tail -n 1

is good. But since [ ls output shouldn't be parsed ] , you need to look to find for a remedy if you plan to use the file for some purpose as is in your case.

My Solution would be :

#!/bin/bash
declare -i lts # latest time stamp
declare -i cts # current time stamp
declare latestrpmfile="No file available"
lts=0
while read -r -d '' fname
do
  cts="$( stat --printf=%Y "$fname" )"
if [ "$cts" -gt "$lts" ]
then
  lts="$cts"
  latestrpmfile="$fname"
fi
done < <( find /directory/to/rpm/files/ -type f -iname "*.rpm" -print0 )

#At this point $latestrpmfile will contain the full path to the 
#latest rpm file
#Now, say, to copy the latest rpm file to a directory, do
if [ -e "$latestrpmfile" ]
then
  cp "$latestrpmfile" /path/to/where/to/copy/
fi

---

Problem specific solution

lbt=0 #latest build time
lb="nosuchfile" #latest build
for i in "$srcdir"/*.rpm # source dir could be an argument to script
do
bt=$(rpm -q --queryformat "%{BUILDTIME}" -p "$i" 2>/dev/null)
if [ "$bt" -gt "$lbt" ]
then
  lbt=$bt
  lb="$i"
fi
done
#Do something with "$lb"

Answer 2

Two approaches -- one portable to non-GNU platforms, the other fast with large directories (to the extent allowed by the filesystem):

---

Portable

What?

BashFAQ #3: How can I sort or compare files based on some metadata attribute (newest / oldest modification time, size, etc)? is relevant here. To summarize:

latest=
for f in "$srcdir"/*.rpm; do
  [ "$f" -nt "$latest" ] && latest=$f
done

How?

[ "$a" -nt "$b" ] checks whether the file named in variable a is newer than that named in variable b in ksh-derived shells.

---

Fast

...to be clear, this is faster with very large directories, as opposed to faster in all cases. That said, it's also easily adapted to find (for instance) the 5 or 10 newest files, or all files except those 5 or 10 newest, which the other approach could not do nearly as effectively.

What?

If you have GNU tools (GNU find, GNU sort), consider the following:

{ read -r -d ' ' mtime && IFS= read -r -d '' filename; } \
  < <(find /directory -type f -iname "*.rpm" -printf '%T@ %p\0' | sort -z -r -n)

This will put your latest file's time (in seconds-since-epoch) in the shell variable mtime and the name of that file in filename . Thus, you can then operate on that file:

if [[ -e $filename ]]; then
  # do whatever arbitrary operations you're looking for on that resulting filename
  cp -- "$filename" /path/to/where/to/copy
fi

---

How?

To explain how this works:

find ... -printf '%T@ %p\0'

...emits contents in the format <epoch_mtime> <filename><NUL> , where epoch_mtime is the number of seconds since January 1st, 1970.

sort -z -r -n

...then sorts that output, expecting it to be NUL-delimited, on the numbers at the beginning.

{ read -r -d ' ' mtime && IFS= read -r -d '' filename; }

...reads content into the mtime variable up to the first space in the first line, and then reads forward to the first NUL into the filename variable.