Shell Scripts: Safe Short/Long Option Parsing

Updated: 2025-10 | Published: 2025-08

Stop manually parsing $1, $2, etc. in your shell scripts! Use getopt to define and parse arguments like a command-line pro.

Inside every good shell script is a pearl of wisdom. For command-line options, that pearl is getopt.

The Problem I Want to Solve

I want a clean and robust shell script that can handle:

• Optional arguments with default: -a, --arga[=VAL]
• Flags: -b, --argb
• Required arguments: -c, --argc VAL
• Help: --help.
• Non-option arguments: -- this and that.

The Core Idea: The "Option String" Menu

The only way to tell getopt what to expect is by giving it an "option string", one for short options (-o) and one for long options (--long).

You can look at the option string as the menu you hand to getopt. It tells getopt what "meals" (options) are available and which ones go with a "side" (an argument) 😋

Short Option String (-o)

• Each character is an option.
  • abc means you accept -a, -b, and -c.
• A colon : after a character means it requires an argument.
  • c: means -c must be followed by a value (e.g., -c my-value).
• Two colons :: after a character means it has an optional argument.
  • a:: means -a can be used alone or with a value (e.g., -a or -a my-value).

Example:

Consider -o f:gh::i::

• -f requires an argument.
• -g is a simple flag (no argument).
• -h has an optional argument.
• -i requires an argument.

Long Option String (--long)

The rules are the same, but the options are full words separated by commas.

Example:

Consider --long file:,verbose,user::,help.

• --file requires an argument (--file /path/to/it).
• --verbose is a simple flag.
• --user has an optional argument (--user or --user=bahman).
• --help is a simple flag.

Rich Pattern: GNU getopt (Long + Short Options)

This is a safe, complete script you can use in any scenario - obviously after renaming the options 😁

#!/usr/bin/env bash

# Strict mode on
set -Eeuo pipefail

script_name="${0##*/}"

usage() {
  cat <<EOF
Usage: $script_name [OPTIONS] [--] [ARGS...]

Options:
  -a, --arga[=VAL]   Optional argument (default: "some default value")
  -b, --argb         Flag (off by default)
  -c, --argc VAL     Required argument
  -h, --help         Show this help and exit
  -v, --version      Show version and exit

Examples:
  $script_name -b -c hello
  $script_name --arga=foo --argc "hi there" -- file1 file2
  $script_name -a -cval -- some-positional
EOF
}

version() { echo "$script_name 1.0.0"; }

# Defaults
ARG_A="some default value"
ARG_B=0
ARG_C=""

# Resolve getopt binary first, then test it. GNU getopt (util-linux) returns exit status 4 for -T/--test.
GETOPT_BIN="${GETOPT:-getopt}"
"$GETOPT_BIN" -T >/dev/null 2>&1 || true
getopt_status=$?
if [[ $getopt_status -ne 4 ]]; then
  echo "Error: This script requires GNU 'getopt' (util-linux)." >&2
  echo "On macOS: brew install gnu-getopt && brew link --force gnu-getopt" >&2
  echo "Or export GETOPT=/<path>/getopt to point at the GNU binary." >&2
  echo "Homebrew path (incl. Apple Silicon): export GETOPT=\"$(brew --prefix gnu-getopt)/bin/getopt\"" >&2
  exit 1
fi

# Define short and long options:
# - Short: a:: (optional), b (flag), c: (required), h (help), v (version)
# - Long:  arga::,  argb,  argc:,  help,  version
PARSED="$("$GETOPT_BIN" --options=a::bc:hv \
  --long=arga::,argb,argc:,help,version --name "$script_name" -- "$@")" || {
  # getopt already printed an error; exit non-zero.
  exit 2
}

# Why 'eval set -- "$PARSED"'?  GNU getopt returns a shell-escaped string
# representing the normalised argv. This safely re-tokenises it so "$1", "$2"...
# are correct. It's safe because $PARSED is generated by getopt, not user input.
eval set -- "$PARSED"

while true; do
  case "$1" in
    -a|--arga)
      # Optional argument placeholder: util-linux inserts an empty string if omitted.
      if [[ -n "${2:-}" ]]; then
        ARG_A="$2"
        shift 2
      else
        # No value provided: leave default
        shift 2
      fi
      ;;
    -b|--argb)
      ARG_B=1
      shift
      ;;
    -c|--argc)
      ARG_C="$2"
      shift 2
      ;;
    -h|--help)
      usage
      exit 0
      ;;
    -v|--version)
      version
      exit 0
      ;;
    --)
      shift
      break
      ;;
    *)
      echo "Internal error: unexpected option '$1'" >&2
      exit 3
      ;;
  esac
done

# Remaining args after '--' are positional arguments:
POSITIONAL=("$@")

# Validate required options
if [[ -z "$ARG_C" ]]; then
  echo "Error: --argc is required." >&2
  echo >&2
  usage
  exit 2
fi

# Do something useful with parsed values
printf 'ARG_A = %s\n' "$ARG_A"
printf 'ARG_B = %s\n' "$ARG_B"
printf 'ARG_C = %s\n' "$ARG_C"
printf 'POSITIONAL (%d): %s\n' "${#POSITIONAL[@]}" "${POSITIONAL[*]-}"

Niceties to Note About The Script

• Uses strict mode (set -Eeuo pipefail) for fail‑fast behaviour.
• Explicitly identifies GNU getopt using -T (exit status 4), avoiding silent portability traps.
• Handles optional arguments safely and predictably.
• Use -- to separate options from positional arguments.
• Safely uses eval set -- "$PARSED". That is the correct pattern for GNU getopt.

Portable Alternative: Pure POSIX getopts (Short Options Only)

Occasionally, when I don’t need long options, I reach for getopts.

The good thing is that it’s built‑in and portable with no external dependencies, but that comes at the cost of long options and optional arguments. Ugh!

#!/usr/bin/env bash

# Strict mode on.
set -Eeuo pipefail

script_name="${0##*/}"

usage() {
  cat <<EOF
Usage: $script_name [-a VAL] [-b] -c VAL [--] [ARGS...]

Options:
  -a VAL    Optional argument (simulate optional by defaulting if omitted)
  -b        Flag (off by default)
  -c VAL    Required argument
  -h        Show help

Note: getopts does not support long options or true optional arguments.
EOF
}

ARG_A="some default value"  # Simulate optional by default
ARG_B=0
ARG_C=""

# Leading colon (:) -> silent error reporting; we handle cases in *)
while getopts ":a:bc:h" opt; do
  case "$opt" in
    a) ARG_A="$OPTARG" ;;
    b) ARG_B=1 ;;
    c) ARG_C="$OPTARG" ;;
    h) usage; exit 0 ;;
    :)
      # Missing required argument for option
      echo "Error: Option -$OPTARG requires an argument." >&2
      usage
      exit 2
      ;;
    \?)
      echo "Error: Invalid option: -$OPTARG" >&2
      usage
      exit 2
      ;;
  esac
done
shift $((OPTIND - 1))

POSITIONAL=("$@")

if [[ -z "$ARG_C" ]]; then
  echo "Error: -c is required." >&2
  usage
  exit 2
fi

printf 'ARG_A = %s\n' "$ARG_A"
printf 'ARG_B = %s\n' "$ARG_B"
printf 'ARG_C = %s\n' "$ARG_C"
printf 'POSITIONAL (%d): %s\n' "${#POSITIONAL[@]}" "${POSITIONAL[*]-}"

Optional Arguments: Handle the Ambiguity

Optional arguments are tricky.

When you write --user file.txt, does file.txt belong to --user or is it a positional argument?

GNU getopt solves this by inserting empty strings when values are omitted:

# Input:  --user file.txt
# Output: --user '' -- file.txt

Best practice: Use = for long options, stick values to short options:

script --user=bahman     # Clear
script -ubahman          # Clear
script --user file.txt   # Ambiguous - avoid

Common Gotchas

• Always try to validate required flags early in your script and fail fast with friendly usage messages.

• macOS ships a BSD getopt that’s not compatible with GNU getopt 🤦

  • Install GNU getopt via Homebrew: brew install gnu-getopt && brew link --force gnu-getopt.
  • Then point at it explicitly: export GETOPT="$(brew --prefix gnu-getopt)/bin/getopt". On Linux, you almost always have GNU already.

• Optional arguments are a nasty creature:

  • Prefer --opt=value for optional arguments; it avoids ambiguity.
  • For short options, -oVALUE is okay; avoid -o VALUE when optional.

• Don’t drop the -- before positional arguments:

  • script -- -not-an-option file.txt ensures -not-an-option isn’t parsed.

• Do not eval user input. I repeat: Do. Not.

  • Only eval the string produced by GNU getopt. That output is already shell‑escaped and safe to re‑tokenise.

• set -e interactions:

  • Using getopt -T in a conditional is safe; set -e doesn’t exit on failures in if tests.

• getopts limitations:

  • No long options; no true optional arguments. Simulate optional by setting defaults.

---

Like the good old mailing lists: one e-mail per post, sent as soon as it’s published.

Get new posts by e-mail