Search code examples
bash

How to get process output into variable without executing subshell

I have a script reading an identifier

ID=$(GPU.py --get)
#Do something
GPU.py --release $ID

Now from GPU.py I want to test what was the PID of the --get calling process so I call os.getppid(). On release I can then test if the given ID was released by the same process.

~~In case of the Bash script above, it does not work, while it calls subshell.~~

Actually it does! I was confused but as accepted answer points out there is some optimization not running subshell! The only problem was I have rewritten it as below and it stopped working.

I was thinking I will be lucky doing

read ID < <(GPU.py --get)

But it seems also here subshell is created and os.getppid() does not return PID of the script. Could you help me to fix it? And the fix is not to call os.getpppid() while I can call the Python program not only from Bash.

Solution

  • When command substitution runs only a single command, which is your case, then bash optimizes by running directly the command without creating a subshell.

    You can verify with the following :

    echo $$ $(python -c 'import os; print(os.getppid())')