Temporarily change current working directory in bash to run a command

Question

I know I can use cd command to change my working directory in bash.

But if I do this command:

cd SOME_PATH && run_some_command

Then the working directory will be changed permanently. Is there some way to change the working directory just temporarily like this?

PWD=SOME_PATH run_some_command

Solution 1

You can run the cd and the executable in a subshell by enclosing the command line in a pair of parentheses:

(cd SOME_PATH && exec_some_command)

Demo:

$ pwd
/home/abhijit
$ (cd /tmp && pwd)  # directory changed in the subshell
/tmp 
$ pwd               # parent shell's pwd is still the same
/home/abhijit

Solution 2

bash has a builtin

pushd SOME_PATH
run_stuff
...
...
popd 

Solution 3

Something like this should work:

sh -c 'cd /tmp && exec pwd'

Comments

[Sahil]

why not keep it simple cd SOME_PATH && run_some_command && cd - the last command will take you back to the last pwd directory.

[Eyal]

@Sahil then it can't be run in parallel

[tripleee]

That sort of invalidates the point of using exec, don't you think?

[codaddict]

@tripleee: I guess OP meant to execute any executable and not the exec.

[Fr0sT]

+1, pushd/popd is ideal for this. Just don't forget to popd before you exit.

[ron rothman]

Not necessarily a good solution if run_stuff can fail (and the script exits). You'd be stuck in SOME_PATH.

[Allan Ruin]

not working in shell file

[galois]

@ron.rothmanℝℝ couldn't you just do something like pushd PATH; (run_stuff); [[ "$?" != 0 ]] && popd; ...; popd

[jez]

@BeC use function rather than alias. Better two years late than never.

[octoquad]

In my case, if the script execution failed e.g. git pull, I wanted to investigate in the directory with git log.

[Jason R Stevens CFA]

What an excellent way to quickly update targets under another project while testing. Thanks!

[mjs]

best answer. works with multiline wrappers as well and premature exits

[Damian]

best answer for me!

[Hexworks]

I won't see the output of the command if it runs in a subshell.

[Hexworks]

This should be the accepted answer! A subshell breaks the log output!

[Pierre]

worth noting: if your subshell to run the commands (in parentheses now) and that code does an "exit" (you want to stop execution), control goes back to the parent shell (the parent shell does NOT exit). This might be a difference if you refactor your code (execution doesn't stop anymore if something bad happens in the parentheses)