how to "execute command on file in multiple subdirectories within containing folder" i.e cd into each subdirectory containing file and execute command

Question

I have a file named filename.sh in different directories. I am trying to simultaneously submit this file to run on a number of cluster nodes. I need to find this file wherever they may be, cd into their containing folder and execute the command "sbatch run_code.sh" This is what I have tried.

find . -mindepth 1 -type f -name filename.sh -exec sbatch run_code.sh {} \;

And it works except it runs command from current folder which does not contain dependencies needed to run the code.

alternatively I am leaning towards a for loop like this

for d in dirname; do 
  cd "$(dirname "$(find . -type f -name filename.sh)")" && sbatch run_code.sh)
done 

which does not work. I look forward to getting some help here. Please.

Answer 1

Sounds like you want:

find . -name filename.sh -type f -execdir sbatch run_code.sh {} ';'

-mindepth 1 is superflous. It excludes ., but . is excluded anyway by both -type f and -name filename.sh.

-execdir (from BSD), is not a standard find predicate, but chances are that if your find supports -mindepth (also non-standard though from GNU), it will also support -execdir.

-execdir is exactly for that. It's like -exec except it runs the command in the parent directory of the selected file. Here, depending on the find implementation, the command being run will be either sbatch run_code.sh ./filename.sh or sbatch run_code.sh filename.sh. Remove the {} if you don't want the filename.sh to be passed as an argument to the command.

With find implementations that don't support -execdir, you could do:

find . -name filename.sh -type f -exec sh -c '
  cd "${1%/*}" && exec sbatch run_code.sh "${1##*/}"' sh {} ';'

${1%/*} is $1 but with the shortest tail matching /* removed, so it will act like "$(dirname -- "$1")" but is more efficient and more reliable¹. ${1##*/} removes the longest head matching */ giving you a basename-like result.

Or to avoid running one sh per file:

find . -name filename.sh -type f -exec sh -c '
  for file do
    (cd "${file%/*}" && exec sbatch run_code.sh "${file##*/}"
  done' sh {} +

With the zsh shell, you could also do:

for f (./**/filename.sh(ND.)) (cd $f:h && sbatch run_code.sh $f:t)

(the exec (to save a process) is not necessary as zsh does it implicitly already for the last command run in the subshell).

$f:h expands to $f's head (dirname) and $f:t to its tail (basename).

---

^{¹ would still work for instance if the dir name ended in newline characters; there are cases where using dirname (or zsh's $var:h) gives better results though such when the path to get the dirname of doesn't have any / or ends in / characters.}