LinuxQuestions.org - Read and write to a co-process in bash

- Linux - Software (https://www.linuxquestions.org/questions/linux-software-2/)

- - Read and write to a co-process in bash (https://www.linuxquestions.org/questions/linux-software-2/read-and-write-to-a-co-process-in-bash-685772/)

Read and write to a co-process in bash

Hello all,
Ksh provides the option to read from a co-process by means of:

Code:

#!/bin/ksh

find . -name "*.log" |&

while read -p; do

# whatever

 :

done

How can this be done in bash?
Using a pipeline is no option for me as I set some shell variables inside the while loop that I need to access later on.

Thanks,
Opher.

It depends. The evident way is to use a file to hold the data, and read it back from the loop.

You can take the whole list into the loop with <<<, like this

Code:

IFS=' '

n=1

while read line

do

  echo "${line}"

  n=$((n+1))

done <<< $(find .)

echo $n

IFS is the internal field separator marker. <<< will import the whole list into the loop. $n is there so you can check that the var is accessible after while is over.

Thanks for the reply,

Quote:

Originally Posted by i92guboj (Post 3353030)

It depends. The evident way is to use a file to hold the data, and read it back from the loop.

Indeed, but then I have to clean up temporary files.
Also, I'm not sure how this behaves if I background (&) the first command and in my script read faster than the command generates output ...?

Quote:

You can take the whole list into the loop with <<<, like this

Code:

IFS=' '

n=1

while read line

do

  echo "${line}"

  n=$((n+1))

done <<< $(find .)

echo $n

IFS is the internal field separator marker. <<< will import the whole list into the loop. $n is there so you can check that the var is accessible after while is over.

I'm afraid I was not specific enough. In my script I have several loops and commands reading from the co-process output. So I would need to brace ({ ... }) the bulk of my script. Also the natural order of things is lost

Code:

{

while read; do

# first loop

 :

done

read

...

while read; do

# second loop

 :

done

} < <(find .)

Some more research (google -> bash FAQ) revealed that co-processes are not supported directly.
Instead, one can use a named pipe pair (one for input and one for output).
There are two options: 1) place the extra code in your script.
2) source the functions/coproc.bash (from the 3.2 distribution)
ex.

Code:

  local fifo="/var/tmp/getoptions.$$.$RANDOM"

  mkfifo $fifo || {

    echo "$(basename $0): getoptions: Error creating named pipe. Exiting." >&2

    exit 2

  }

  ( find . >$fifo ; rm -f $fifo ) &

  trap 'echo Received SIGPIPE >&2' SIGPIPE

  globalvar=

  while read; do

  # first loop

    :

  done <$fifo

  # do something with globalvar

  while read; do

  #second loop

    :

  done <$fifo

  trap - SIGPIPE