Shell pipes and fifo

Most bash "blocks" accept a redirection parameter at the end of their declaration. This can replace where the output and errors are going

fn () {
  echo "error" >&3
} 3>&2 2>file
# Prints "error" to the standard error. The real stderr is stored in ./file

Functions, as above
Sub-shells: (echo "one"; echo "two") > file
Loops: while (1); do echo "loop"; done > file

Remember that redirection syntax looks like [fd][< | >][&fd | file_path], where fd is the index in the process' file descriptors. This works for < as well. |& is identical to using 2>&1 |

The order of IO redirection matters. Redirection will go to wherever the redirected destination currently goes. Redeclaring a redirection doesn't work

            ┌3 writes to stdout
            │    ┌stdout writes to stderr
            │    │    ┌stderr writes to where 3 writes, so stdout
            │    │    │
echo "one" 3>&1 1>&2 2>&3

To do this for the entire shell or script, use exec

exec 4>&1          # Holds onto reference to stdout
exec 1>/dev/null   # Silences stdout and loses track of it
exec 1>&4 4>&1     # Restore stdout through reference in 4 and close 4
# Everything back to normal

mkfifo creates a named pipe, which can be easier to interact with between scripts. Make sure it's being held open at use

mkfifo fifo
cat fifo >ofif &   # Holds open fifo pipe and copies anything in there to ofif
exec 3>fifo        # fd 3 reffers to pipe fifo
echo "double" >&3  # Writes to fifo. >fifo would close the pipe right after
exec 3>&-          # Manually close the open pipe

The example above works similar to tee in logging everything in the pipe

1.7 KiB Raw Blame History

Shell pipes and fifo

1.7 KiB

Raw Blame History