Bash – how to locally redefine ‘command_not_found_handle’

bashcommand-not-founderror handlingfunction

I'd like to make a particular bash script failfast when it cannot find a command, while retaining globally the usual friendly command_not_found behavior. E.g., if I save the following to /tmp/foo.sh, …

# ----------------------------start foo.sh----------------------------
THIS_FP="$0"
THIS_FN="$(basename ${THIS_FP})"
THIS_DIR="$(dirname ${THIS_FP})"

function setup {
  for CMD in \
    'foo' \
  ; do
    echo -e "\n$ ${THIS_FN}::${FUNCNAME[0]}::${CMD}"
    eval "${CMD}"
  done
} # end function setup

function teardown {
  for CMD in \
    "ls -alt ${THIS_DIR} | head" \
  ; do
    echo -e "\n$ ${THIS_FN}::${FUNCNAME[0]}::${CMD}"
    eval "${CMD}"
  done
} # end function teardown

for CMD in \
  'setup' \
  'teardown' \
; do
  echo -e "\n$ ${THIS_FN}::main loop::${CMD}"
  eval "${CMD}"
done
# ------------------------------end foo.sh----------------------------

… and make it executable, and run it, I get

# ----------------------------not what I want-------------------------
me@it:~$ /tmp/foo.sh 

$ foo.sh::main loop::setup

$ foo.sh::setup::foo
/tmp/foo.sh: line 10: foo: command not found

$ foo.sh::main loop::teardown

$ foo.sh::teardown::ls -alt /tmp | head
total 68840
drwxrwxrwt 22 root  root           1600 May  6 18:49 .
-rwxr-xr-x  1 me    me         527 May  6 18:49 foo.sh
drwx------  2 me    me          40 May  6 17:54 plugtmp-1
srw-------  1 me    me           0 May  6 17:28 tramp.13004aZf
drwx------  2 me    me        4200 May  6 17:22 matecorba-me
drwx------  2 me    me          40 May  6 02:20 plugtmp
-rw-------  1 me    me     2034335 May  4 14:21 s_TTRuhW.mp3.part
-rw-------  1 me    me     1658381 May  2 12:21 +rM4IttD.mp3.part
-rw-r--r--  1 me    me       86420 May  1 19:22 duplicity_20130501_1917.txt
# ----------------------------not what I want-------------------------

Instead, I want to make only this script foo.sh fail, and to fail as soon as it encounters the 'command not found': e.g.,

# ------------------------------what I want---------------------------
me@it:~$ /tmp/foo.sh 

$ foo.sh::main loop::setup

$ foo.sh::setup::foo
/tmp/foo.sh: ERROR: foo: command not found

# Throwing me back to the shell from which I invoked foo.sh,
# *without* changing the usual friendly behavior outside it, e.g.:

me@it:~$ foo
No command 'foo' found, did you mean:
 Command 'fio' from package 'fio' (main)
 Command 'xoo' from package 'xoo' (main)
 Command 'fop' from package 'fop' (main)
 Command 'fox' from package 'objcryst-fox' (main)
 Command 'zoo' from package 'zoo' (main)
 Command 'goo' from package 'goo' (main)
foo: command not found
# ------------------------------what I want---------------------------

Can I provide this behavior by editing only foo.sh, without touching .bashrc or any other currently-installed file ? If so, how? If not, why not?

Best Answer

To have your script exit immediately after error run your script with bash -e /tmp/foo.sh or put this line at the beginning of the script: set -e.

-e or errexit tells bash to exit the script immediately if an "error" occurs. Note that errexit is a bit more complicated than that and it does not always do what you might expect. Read this for more information: http://mywiki.wooledge.org/BashFAQ/105 ("Why doesn't set -e (or set -o errexit, or trap ERR) do what I expected?")

The command not found handling feature is (normally) not enabled inside a bash script. That is because a bash script does not read bashrc and the command_not_found_handle is (usually) defined in some bashrc (in debian it is /etc/bash.bashrc).

If you really want to be really sure that the command not found handling feature is disabled in your script you can undefine the function like this:

if [ x"$(type -t command_not_found_handle)" = x"function" ]; then
    unset -f command_not_found_handle
fi

If you do this inside a bash script and you execute the script, instead of sourcing, then command_not_found_handle will only be undefined inside the bash script. The state of command_not_found_handle in your interactive shell will stay in whatever state it was before executing the script.

Read this for more information: https://superuser.com/questions/176783/what-is-the-difference-between-executing-a-bash-script-vs-sourcing-it

And since you are using eval please also read: http://mywiki.wooledge.org/BashFAQ/048 ("Eval command and security issues")

Related Question