Find a directory which contains all the files in my list

Question

I have moved a file, which makes a reference to four other files with the same parent directory. And now I am stock trying to figure out which parent directory that is since there a number of files with the same name in my build environment. But I don't want to take any chances and copy the wrong file. So I want to know if there a way to tell Linux find command or some other command to give me the results matching two or more strings. So I only want to see matches for locations which contain files:

foo & bar & buz

I can always pipe this into grep for the desired common directory, but if there is an easier way then I would love it. Something like:

common_direcoty/foo & common_directory/bar & common_directory/buz 

In other words, I am looking for the copy of these files with a common directory. I am not looking for the file that matches the name alone.

Here is the output of the command which gave me what I was after:

sansari@ubuntu:~/WORKING_DIRECTORY$ find . -name io.h -exec sh -c '[ -e "${0%/*}/kernel.h" ] && [ -e "${0%/*}/init.h" ]' {} \; -print
./include/linux/io.h
sansari@ubuntu:~/WORKING_DIRECTORY/include/linux$ ls module.h
module.h
sansari@ubuntu:~/WORKING_DIRECTORY/include/linux$ ls kernel.h 
kernel.h
sansari@ubuntu:~/WORKING_DIRECTORY/include/linux$ ls init.h
init.h
sansari@ubuntu:~/WORKING_DIRECTORY/include/linux$ ls io.h
io.h

Answer 1

One way is to look for one of the file names (pick the rarest one if you know which one it is) then filter the matches to retain only those where the other file exists.

find . -name foo -exec sh -c '[ -e "${0%/*}/bar" ] && [ -e "${0%/*}/buz" ] && echo "${0%/*}"' {} \;

You can use the shell snippet's return code if you want to use a find action, but that would apply to the first file, not to the directory. For example the following snippet prints output of the form ./path/to/foo:

find . -name foo -exec sh -c '[ -e "${0%/*}/bar" ] && [ -e "${0%/*}/buz" ]' {} \; -print

If you need to pass the names as arguments:

find_multiple_files () {
  root=$1; shift
  name1=$2; shift
  find "$root" -name "$name1" -exec sh -c 'dir=${0%/*}; for x; do [ -e "$dir/$x" ] || exit 1; done; echo "$dir"' {} "$@" \;
}
find_multiple_files . foo bar buz

An alternative approach is to generate lists of names and filter them. This is a little more cumbersome and less efficient, but it has the advantage that you can switch from find to locate if you have an index of file names. I assume that the file names don't contain newlines.

locate foo |              # or: find -name foo
sed 's!/[^/]*$!!' |
xargs -I {} sh -c 'for x; do [ -e "$0/$x" ] || exit 1; done; echo "$0"' {} bar buz

Answer 2

I am unaware of any find option that can do it all by itself. Using greps you can achieve that - put all the file names that you want to find in a separate file , say /tmp/files_to_find, with the following pattern

In your case files_to_find would be

\/foo$
\/bar$
\/buz$

Then issue the following command

find root_dir_to_search_in | grep -f /tmp/files_to_find | xargs -I{} dirname {} | sort | uniq -c | grep `wc -l /tmp/files_to_find | awk '{print $1}'`

Answer 3

You should be able to group find patterns with -or and -and to select different patterns. I don't use such options but its

$ find . \( -name foo -or -name bar -or -name buz \) -and -path '*/G/*'|9 mc
./G/buz ./G/bar ./G/foo

where G = common_directory. I tested this against a file structure

$ find . -name foo -or -name bar -or -name buz|9 mc
./G/buz ./G/bar ./G/foo ./F/buz ./F/bar ./F/foo

where I also have foo,bar and baz in the non common_directory.

The 9 mc command is the multi column formatter from p9p, just to tune the output.