Execute only if it is a bash function

Question

I'm looking for something similar to Bash's built-in command that will only run something if it is a function. So currently I have an insecure way of doing:

# Go through arguments in order
for i in $*; do
    if [ -z `which $i` ]; then
        # Run function
        $i && echo -n ' '
    fi
done

This if statement doesn't work properly. Anyway, even if I could check if it's a command and not a function, I can't run explicitly run a function, which is bad because if anyone has any programs in $PATH that are the same name as my function, they will be run. If I nullify PATH or set it to anything else, anyone could still use $i to run a program explicit, so that's also not a solution.

Any way I can "secure" my shell script?

Answer 1

With bash, you can do something like this:

for f do
  if declare -F -- "$f" >/dev/null 2>&1; then
    : "$f" is a function, do something with it
  fi
done

declare -F -- "$f" >/dev/null 2>&1 will return success code if $f is a bash function, output nothing.

You might also want to disable some special builtin commands when bash run in POSIX mode by adding builtin enable -n -- "$f".

Answer 2

The formal way Bash provides to do it is using the type builtin as of type -t:

function fun() {
  : foo bar
}

if [ function = $(type -t fun) ]; then
  echo fun is a function.
fi

If you are suspicious about the function name from your parameter list, just double-quote it: $(type -t "$arg")

Answer 3

While you're probably well set in a bash shell, to do similar portably you might employ something like the following...

eval "  for c in $(unalias -a
        for c do case ${c#function}  in
                (*[!_[:alnum:]]*) ;; (?*)
                PATH=   command -v "$c" >&2 &&
                PATH=   command -V "$c"     &&
                        printf '\n'"'$c'"'\\\0'
        esac 2>/dev/null; done  |
        tr  '\t\n\0' '  \n'     |
        sed -ne's/.* function .* / /p')"'
;do     "$c"; printf %02s
done'

It works in stages:

1. First weed out all args which contain any characters that might disqualify them as shell names in the first place.
2. If the current arg is a valid name and a valid shell builtin or function print its name and type to stdout. Because of the way shell's can vary their output for command -V its output cannot be counted upon to be a single line. So follow the output with our verified name and and a null-byte.
3. translate all newlines and tabs to spaces, and all null-bytes to newlines.
4. with sed, s///ubstitute away all of any line up the last space character thereon and print the results if the line also matches the string function.
5. All of sed's output is rounded up as iterators for the outer for loop and each of these is executed in turn, and after each is executed printf writes out two spaces.