How to use `find -exec` to execute command in directory of found file (not current directory)?

Question

Let's say I'm in a directory which has some subdirectories, dir1, dir2, and dir3. Each of these directories have a file foo and I would like to execute the same command on each foo in each of the directories and have that command executed from inside that directory.

If I were to do it "by hand", it would look something like this:

cd dir1
(execute on foo)
cd ../dir2
(execute on foo)
cd ../dir3
(execute on foo)

It is imperative that the command be executed from each of the directories. foo is a batch scheduling script (for HTCondor, if you care to know) and must be executed from each subdirectory so that the output of the runs started by the scheduling scripts will end up in each subdirectory.

The question "Find a file and execute a command in the file's directory", as far as I can tell, does not answer my question. The first answer to that question is more of a workaround which will not work in my case, and the second answer does not have enough of an explanation for me to know how to use it.

Answer 1

Assuming your find supports it, use the -execdir option instead of -exec

find . -type f -name 'foo' -execdir pwd \;

This will change into each directory in turn, and if there's a foo file present there it will run pwd.

NB You cannot repurpose this solution for running commands inside directories named foo. For that you'd need to revert to a simple -exec sh -c '...' approach, iterating across the args supplied to sh and using cd for each in turn. This example runs pwd inside each directory:

find . -type d -name 'foodir' -exec sh -c '
    for d in "$@"; do ( cd "$d" || exit; pwd ); done
    ' _ {} +

Answer 2

Assuming you know exactly what directories to run your command in,

for dir in dir1 dir2 dir3; do
    ( cd "$dir" && somecommand foo )
done

or, looping over the foo files,

for foo in ./*/foo; do
  ( cd "${foo%/foo}" && somecommand foo )
done

The subshell prevents the cd to affect the working directory of the rest of the script. The ${foo%/foo} parameter substitution expands to the name of the directory where the foo file lives (by removing the knows suffix string /foo from the value of $foo).

Using standard find to do the same thing (in the case where the foo files may live deep down in a directory structure):

find . -type f -name foo -exec sh -c '
    for foo do
        ( cd "${foo%/foo}" && somecommand foo )
    done' sh {} +

In words: Find all regular files called "foo" anywhere in the current directory or below and give their pathnames to this shell script in batches. The shell script shall iterate over the given pathnames, change directory to the directory portion of each pathname and if that succeeds, execute somecommand foo.

For a solution that uses the nonstandard -execdir, see roaima's answer. His answer adapted to my example:

find . -type f -name foo -execdir somecommand foo ';'

Answer 3

Maybe not beautiful, but I think it does the job

find . -name 'foo' -type f -exec sh -c 'cd $(echo {}|cut -d"/" -f2) && ./foo' \;

Answer 4

Tested in WSL and GitBash over Win 10.

execdir from previous solutions won't work because the command will execute using the current directory as an argument, not into it.
You can just cd into the directory and do what you want.
E.g. this is what I use to update all my repos before starting to work

find $HOME/workspace/ -maxdepth 1 -type d -name "repo*" -exec bash -c "cd \"{}\" && echo ____Pulling from \"{}\" && git pull" \;

EDIT:
If you're not on your personal PC/laptop and need more robust code, you should write a script not injectable with external commands, like:

#!/bin/sh
find "$HOME/workspace/" -maxdepth 1 -type d -exec sh -c '
  for dir in "$@"
  do
    echo "____Pulling from $dir"
    cd "$dir"
    git pull
  done' sh {} +