Bash: Split a string exactly like Readline would split it

Question

Given a string representation of an arbitrary Bash "simple command", how can I split it into an array containing its individual "parts", i.e. the command name and individual parameters, just like the the shell itself (i.e. Readline) would split it when parsing it and deciding which executable/function to run and which parameters to pass it?

---

My specific use-case is needing to parse user-defined alias definitions. E.g. an alias might be defined as:

alias c2="cut -d' ' -f2"  # just an example... arbitrary commands should be handled!

And this is how my bash script would try to parse it:

alias_name="c2"
alias_definition=$(alias -p | grep "^alias $alias_name=") # "alias c2='cut -d'\'' '\'' -f2'"
alias_command=${alias_definition##alias $alias_name=}     # "'cut -d'\'' '\'' -f2'"
alias_command=$(eval "echo $alias_command")               # "cut -d' ' -f2"

alias_parts=($alias_command) # WRONG - SPLITS AT EVERY WHITESPACE!

echo "command name: ${alias_parts[0]}"

for (( i=1; i <= ${#alias_parts}; i++ )); do
  echo "parameter $i : ${alias_parts[$i]}"
done

Output:

command name: cut
parameter 1 : -d'
parameter 2 : '
parameter 3 : -f2

Desired output:

command name: cut
argument 1  : -d' '
argument 2  : -f2

What would I need to replace the alias_parts=($alias_command) line with, to achieve this?

Answer 1

as l0b0 said, it's not readline. It's the shell itself doing the splitting. So use the shell itself to do the parsing.

alias c2="cut -d' ' -f2"

split_parts() {
    alias_parts=("$@")
}

alias_defn=$(alias c2)
# 2 evals needed to get rid of quotes
eval eval split_parts ${alias_defn#alias c2=}

for (( i=0; i < ${#alias_parts}; i++ )); do
  echo "parameter $i : \"${alias_parts[$i]}\""
done

outputs

parameter 0 : "cut"
parameter 1 : "-d "
parameter 2 : "-f2"

Note that the -d includes the trailing space that the shell actually sees.

Answer 2

To minimalize "evil otto's" solution:

alias c2="cut -d' ' -f2"
alias_definition=$(alias c2)
eval eval alias_parts=( "${alias_definition#alias c2=}" )

You can use `declare -p' to do a quick array print:

$ declare -p alias_parts
declare -a alias_parts='([0]="cut" [1]="-d " [2]="-f2")'

Also useful may be `printf %q' to quote an argument "in a way that can be reused as shell input" (from: help printf):

$ printf %q ${alias_parts[1]}
-d\

Freddy Vulto
http://fvue.nl/wiki/Bash

Answer 3

That's not readline splitting, it's getopt or getopts. For example:

params="$(getopt -o d:h -l directory:,help --name "$0" -- "$@")"

eval set -- "$params"
unset params

while true
do
    case "${1-}" in
        -d|--directory)
            directory="$2"
            shift 2
            ;;
        -h|--help)
            usage
            exit
            ;;
        --)
            shift
            if [ "${1+defined}" = defined ]
            then
                usage
            fi
            break
            ;;
        *)
            usage
            ;;
    esac
done

Answer 4

The set built-in can be used for splitting strings.

bash$ set -- cut -d ' ' -f2

bash$ echo "'$3'"
' '

Edit: If the string you want to split is already in a variable, that's a lot trickier. You might play around with eval but in this case I'd say that complicates things, rather than simplifies them.

bash$ a="cut -d ' ' -f2"

bash$ eval set -- $a  # No quoting!

bash$ echo "'$3'"
' '

Answer 5

If we put each of alias_command's arguments on its own line, and then (locally) set IFS=\n, we're done:

parsealias ()
{
   alias_command_spaces=$(eval "echo $(alias $1)" | sed -e "s/alias $1=//") # "cut -d' ' -f2"
   alias_command_nl=$(eval each_arg_on_new_line $alias_command_spaces)      # "cut\n-d' '\n-f2"
   local IFS=$'\n' # split on newlines, not on spaces
   alias_parts=($alias_command_nl) # each line becomes an array element, just what we need
   # now do useful things with alias_parts ....
}

Now we only need to write the command each_arg_on_new_line used above, e.g:

#!/usr/bin/env perl

foreach (@ARGV) {
  s/(\s+)/'$1'/g; # put spaces whithin quotes
  print "$_\n";
}