Re: Is there a way to get the output of a function without an extra bash process?

[Koichi Murase]

2020-05-11 6:20 Peng Yu <address@hidden>:
>
> Hi,
>
> $(fun) by default will always use a new bash process. This is wasteful
> in certain cases. In there a way to use the existing bash process
> (without using a temp file)? Thanks.

Another solution with a shell function.

  $ cat a.sh

  #!/bin/bash

  _ble_base_run=$(mktemp -d)
  trap -- 'rm -rf "$_ble_base_run"' EXIT

  _ble_util_assign_base=$_ble_base_run/$$.ble_util_assign.tmp
  _ble_util_assign_level=0
  function ble/util/assign {
    local _ble_local_tmp=$_ble_util_assign_base.$((_ble_util_assign_level++))
    builtin eval "$2" >| "$_ble_local_tmp"
    local _ble_local_ret=$? _ble_local_arr=
    ((_ble_util_assign_level--))
    mapfile -t _ble_local_arr < "$_ble_local_tmp"
    IFS=$'\n' eval "$1=\"\${_ble_local_arr[*]}\""
    return "$_ble_local_ret"
  }

  # sample
  function f { echo "$BASHPID"; }

  pid=$(f)
  echo "command subst  : $BASHPID,$pid"

  ble/util/assign pid f
  echo "ble/util/assign: $BASHPID,$pid"

  $ bash a.sh
  command subst  : 39173,39175
  ble/util/assign: 39173,39173

Actually, this function `ble/util/assign' is maintained as one of the
most important utilities in my script for more than five years and
tested well.

https://github.com/akinomyoga/ble.sh/blob/54769337/src/util.sh#L1414-L1446

Maybe you don't like it because it internally uses temporary files,
but you do not have to care about the handling of temporary files as
the shell function takes care of them.  If you do not want to cause
disk access, you can create temporary files/directories on in-memory
filesystems (tmpfs) such as /dev/shm or /tmp.  Even if the system does
not mount in-memory filesystems, filesystem access is much faster than
fork & exec.

--
Koichi