Passing named arguments to shell scripts

Question

Is there any easy way to pass (receive) named parameters to a shell script?

For example,

my_script -p_out '/some/path' -arg_1 '5'

And inside my_script.sh receive them as:

# I believe this notation does not work, but is there anything close to it?
p_out=$ARGUMENTS['p_out']
arg1=$ARGUMENTS['arg_1']

printf "The Argument p_out is %s" "$p_out"
printf "The Argument arg_1 is %s" "$arg1"

Is this possible in Bash or Zsh?

Answer 1

If you don't mind being limited to single-letter argument names i.e. my_script -p '/some/path' -a5, then in bash you could use the built-in getopts, e.g.

#!/bin/bash

while getopts ":a:p:" opt; do
  case $opt in
    a) arg_1="$OPTARG"
    ;;
    p) p_out="$OPTARG"
    ;;
    \?) echo "Invalid option -$OPTARG" >&2
    exit 1
    ;;
  esac

  case $OPTARG in
    -*) echo "Option $opt needs a valid argument"
    exit 1
    ;;
  esac
done

printf "Argument p_out is %s\n" "$p_out"
printf "Argument arg_1 is %s\n" "$arg_1"

Then you can do

$ ./my_script -p '/some/path' -a5
Argument p_out is /some/path
Argument arg_1 is 5

There is a helpful Small getopts tutorial or you can type help getopts at the shell prompt.

Edit: The second case statement in while loop triggers if the -p option has no arguments and is followed by another option, e.g. my_script -p -a5, and exits the program.

Answer 2

This is not a parser for positioned arguments, is for key=value aa=bb arguments;

for ARGUMENT in "$@"
do
   KEY=$(echo $ARGUMENT | cut -f1 -d=)

   KEY_LENGTH=${#KEY}
   VALUE="${ARGUMENT:$KEY_LENGTH+1}"

   export "$KEY"="$VALUE"
done

# use here your expected variables
echo "STEPS = $STEPS"
echo "REPOSITORY_NAME = $REPOSITORY_NAME"
echo "EXTRA_VALUES = $EXTRA_VALUES"

Usage

bash args_shell_parser.sh STEPS="ABC" REPOSITORY_NAME="stackexchange" \
           EXTRA_VALUES="KEY1=VALUE1 KEY2=VALUE2"

Console result:

STEPS = ABC
REPOSITORY_NAME = stackexchange
EXTRA_VALUES = KEY1=VALUE1 KEY2=VALUE2

Features:

• It does not matter what order the arguments are in
• Explicit declaration of all variables are not required
• values could have spaces.
• It handles complex cases, when the argument value contains "=" sign

Put this snippet at the start of your script.

Prior version of this script:

• v1

Answer 3

I stole this from drupal.org, but you could do something like this:

while [ $# -gt 0 ]; do
  case "$1" in
    --p_out=*)
      p_out="${1#*=}"
      ;;
    --arg_1=*)
      arg_1="${1#*=}"
      ;;
    *)
      printf "***************************\n"
      printf "* Error: Invalid argument.*\n"
      printf "***************************\n"
      exit 1
  esac
  shift
done

The only caveat is that you have to use the syntax my_script --p_out=/some/path --arg_1=5.

Answer 4

The probably closest syntax to that is:

p_out='/some/path' arg_1='5' my_script

Update 1

Nearly a decade later I realize that it gets much closer than that (not sure whether it already was back then).

man -P 'less -p "^\s+set \["' bash

set [-abefhkmnptuvxBCEHPT] [-o option-name] [--] [-] [arg ...]

-k      

All arguments in the form of assignment statements are placed in the environment for a command, not just those that precede the command name.

So you can do this, too:

set -k
my_script p_out='/some/path' arg_1='5'

A bit more complicated but safer:

if [[ $- =~ k ]]; then
    # set -k has already been set so no need to reverse it after the command
    my_script p_out='/some/path' arg_1='5'
else
    set -k
    my_script p_out='/some/path' arg_1='5'
    set +k
fi

You can test this with:

set -k
echo "\$foo: '$foo'"; bash -c 'echo $foo' foo=bar

Answer 5

I just came up with this script

while [ $# -gt 0 ]; do

   if [[ $1 == *"--"* ]]; then
        v="${1/--/}"
        declare $v="$2"
   fi

  shift
done

pass it like my_script --p_out /some/path --arg_1 5 and then in the script you can use $arg_1 and $p_out.

Answer 6

CAVEAT
The solution below does not check for incomplete / malformed command invocation! For example, if --p_out requires an argument, but the command is called as my-script.sh --p_out --arg_1 27 then it will happily assign the string "--arg_1" to the p_out variable.

This answer was initially an edit of @cdmo's answer (thanks to @Milkncookiez's comment also!), that got rejected as expected.

When using the variables, one can make sure that they have a default value set by using "${p_out:-"default value"}" for example.

From 3.5.3 Shell Parameter Expansion of the GNU Bash manual:

${parameter:-word}
If parameter is unset or null, the expansion of word is substituted. Otherwise, the value of parameter is substituted.

#!/usr/bin/env bash

while [ $# -gt 0 ]; do
  case "$1" in
    -p|-p_out|--p_out)
      p_out="$2"
      ;;
    -a|-arg_1|--arg_1)
      arg_1="$2"
      ;;
    *)
      printf "***************************\n"
      printf "* Error: Invalid argument.*\n"
      printf "***************************\n"
      exit 1
  esac
  shift
  shift
done

# Example of using the saved command line arguments.
# WARNING: Any of these may be an empty string, if the corresponding
#          option was not supplied on the command line!

echo "Without default values:"
echo "p_out: ${p_out}"
echo "arg_1: ${arg_1}"; echo

# Example of using parsed command line arguments with DEFAULT VALUES.
# (E.g., to avoid later commands being blown up by "empty" variables.)

echo "With default values:"
echo "p_out: ${p_out:-\"27\"}"
echo "arg_1: ${arg_1:-\"smarties cereal\"}"

If the above script is saved into my-script.sh, then these invocations are all equivalent:

$ . my-script.sh -a "lofa" -p "miez"
$ . my-script.sh -arg_1 "lofa" --p_out "miez"
$ . my-script.sh --arg_1 "lofa" -p "miez"

(where . is Bash's source command.)

To demonstrate the default values part, here's the output without any arguments given:

$ . my-script.sh
Without default values:
p_out:
arg_1:

With default values:
p_out: "27"
arg_1: "smarties cereal" 

Answer 7

I found the solution from cdmo the best, because it is not only limited to a single letter. With a slight adjustment it will consume either whitespace separated arguments like --url www.example.com and those with an assignment like, for example the synonym parameter -u=www.example.com. Even flag arguments can be parsed:

while [ $# -gt 0 ]; do
  case "$1" in
    --url*|-u*)
      if [[ "$1" != *=* ]]; then shift; fi # Value is next arg if no `=`
      URL="${1#*=}"
      ;;
    --file*|-f*)
      if [[ "$1" != *=* ]]; then shift; fi
      FILE="${1#*=}"
      ;;
    --help|-h)
      printf "Meaningful help message" # Flag argument
      exit 0
      ;;
    *)
      >&2 printf "Error: Invalid argument\n"
      exit 1
      ;;
  esac
  shift
done

Answer 8

With zsh, you'd use zparseopts:

#! /bin/zsh -
zmodload zsh/zutil
zparseopts -A ARGUMENTS -p_out: -arg_1:

p_out=$ARGUMENTS[--p_out]
arg1=$ARGUMENTS[--arg_1]

printf 'Argument p_out is "%s"\n' "$p_out"
printf 'Argument arg_1 is "%s"\n' "$arg_1"

But you'd call the script with myscript --p_out foo.

Note that zparseopts doesn't support abbreviating long options or the --p_out=foo syntax like GNU getopt(3) does.

Answer 9

I have just made this one. It doesn't require "=" and supports non-valued parameters. Mind the need of additional shift commands for valued parameters.

#!/bin/bash

if [ $# -eq 0 ]; then
    printf "Utilizacao:\n"
    printf "$0 [[--imagem_inicial|-i] <tag_inicial>] [[--imagem_final|-f] <tag_final>] [--verbose|-v]\n"
    exit 1
fi

while [ $# -gt 0 ]; do
  case "$1" in
    --imagem_inicial|-i)
      export tag_imagem_inicial="${2}"
      shift
      ;;
    --imagem_final|-f)
      export tag_imagem_final="${2}"
      shift
      ;;
    --verbose|-v)
      export verbose_option=1
      ;;
    *)
      printf "ERRO: Parametros invalidos\n"
      printf "Execute sem parametros para a sintaxe.\n"
      exit 1
  esac
  shift
done

echo tag_imagem_inicial=${tag_imagem_inicial}
echo tag_imagem_final=${tag_imagem_final}

Answer 10

I made a combination of some approaches recommended here. In my case, I need a script that could handle arguments in this format: --arg1=test So I used the following:

args=()

for argument in "$@"
do

    key=$(echo $argument | cut -f1 -d=)
    value=$(echo $argument | cut -f2 -d=)   

    if [[ $key == *"--"* ]]; then
        v="${key/--/}"
        declare $v="${value}" 
   fi
done

args+=( '--local_dir' ${local_dir})
args+=( '--format' ${format})

python3 ./test.py "${args[@]}"

You can see that I added the arguments together to call a python script with those arguments: test_script.sh --local_dir=test --format=other

Thanks to Shahzad Malik and JRichardsz, for the guide to join the two approaches.

Answer 11

I want the script to function consistently with other utilities, providing the least surprise to the user. Here is what I came up with, combining several ideas here (especially the answer from cdmo) and some additional research. It has the following features:

• Only accepts known named parameters.
• Accepts positional parameters as well.
• Named parameters can be in any order.
• Named (so-called "long parameters") can be in the form foo=bar or --foo bar.
• --help is recognized.
• -- is recognized to signal remaining parameters are all positional parameters.

Let's assume we have the following command, accepting three positional parameters and a named foo parameter:

munge [--foo=bar] x y z

#!/bin/bash
set -eu

usage() {
  echo 'Usage: munge [--foo=bar] <x> <y> <z>' >&2
}

args=()
foo="default value"
while [ $# -gt 0 ]; do
  case "$1" in
### start custom parameter checks
    --foo=*)
      foo="${1#*=}"
      shift
      ;;
    --foo*)
      foo="$2"
      shift
      shift
      ;;
### end custom parameter checks
    --help)
      usage
      exit 0
      ;;
    --)
      shift
      while [ $# -gt 0 ]; do args+=("$1"); shift; done;
      break;
      ;;
    --*)
      echo "**Bad parameter: $1**" >&2
      usage
      exit 1
      ;;
    *)
      args+=("$1");
      shift
      ;;
  esac
done

if [[ ${#args[@]} -ne 3 ]]; then
  echo "**Incorrect number of arguments: ${#args[@]}**" >&2
  usage
  exit 1
fi

# test
echo foo: $foo
echo arg 1: ${args[0]}
echo arg 2: ${args[1]}
echo arg 3: ${args[2]}

This accepts the following:

• munge blah blah blah
• munge --foo=bar blah blah blah
• munge --foo bar blah blah blah
• munge --foo=bar -- blah blah blah
• munge --foo bar -- blah blah blah
• munge blah --foo=bar blah blah
• munge blah --foo bar blah blah
• …
• munge blah blah blah --foo=bar
• munge blah blah blah --foo bar

Caveats:

• The program must henceforth use the args array (e.g. {$args[0]}) rather than $* (e.g. $1) and the like.
• Usage is always sent to stderr, when for --help it should more correctly be sent to stdout.
• If a known long parameter comes at the end without the value, the error message will say $2: unbound variable instead of something useful. (The additional error handling would just be too verbose for the benefit.)

Yeah, it's verbose and clumsy, but it is most consistent with other CLI utilities. Bash really needs something better. It this area Bash is really pitiful in comparison to PowerShell (which of course has its own problems).

Answer 12

My solution is based on that of JRichardsz, but fixing a few issues and adding more transparency about what's actually going on − I'm not a fan of "copy this random code that you don't understand, and run it on your personal computer. I swear it's safe" :P

for argument in "$@"
do
  key=$(echo $argument | cut --fields 1 --delimiter='=')
  value=$(echo $argument | cut --fields 2 --delimiter='=')

  case "$key" in
    "wallet")        wallet="$value" ;;
    "network")       network="$value" ;;
    *)
  esac
done

echo "We'll be using wallet $wallet and network $network"

Invoking this script can be done as:

myScript.sh wallet=myWallet.json network=https://granadanet.smartpy.io

Answer 13

If a function or an application has more than zero arguments, it always has a last argument.

If you want to read option flag and value pairs, as in: $ ./t.sh -o output -i input -l last

And you want to accept a variable number of option/value pairs,

And do not want a huge "if .. then .. else .. fi" tree,

Then after checking for an argument count of non-zero and even,

Write a while loop with these four eval statements as the body, followed by a case statement using the two values determined in each pass through the loop.

The tricky part of the scripting is demonstrated here:

#!/bin/sh    

# For each pair - this chunk is hard coded for the last pair.
eval TMP="'$'$#"
eval "PICK=$TMP"
eval TMP="'$'$(($#-1))"
eval "OPT=$TMP"

# process as required - usually a case statement on $OPT
echo "$OPT \n $PICK"

# Then decrement the indices (as in third eval statement) 

:<< EoF_test
$ ./t.sh -o output -i input -l last
-l 
last
$ ./t.sh -o output -l last
-l 
last
$ ./t.sh  -l last
-l 
last
EoF_test

Answer 14

function satunnaisluku () { let $1=$RANDOM ;}; satunnaisluku apu; satunnaisluku kipu; echo $apu; echo $kipu

• there are several other methods to use nameparameters. Forexample method to find maximum from matrix:

function haemaksimi () { maksimi=$(echo $(sort -n <(declare -p $1 | tr ' ' '\n' | tr -dc '\n'[0-9])) | sed 's/.* //') ;} ; apu=""; for n in {1..10000}; do apu=$apu" "$RANDOM$RANDOM; done; unset maksimi; time haemaksimi matrix; echo $maksimi

---

I have made IDE for BASH-scripting to myself - some few thousand lines - and I have used a function like those for ten years - for reading items in different tasks in the same session - nowadays my days are full of scripting because a busted leg.

If you use something very heavily for ten years and it always operates impeccably without doing any harm to other programs isn't it quite the same how it is called?

I stumble constantly to warnings: BASH cannot do that ... Sour grapes I say.

Answer 15

mitsos@redhat24$ my_script "a=1;b=mitsos;c=karamitsos"
#!/bin/sh
eval "$1"

you've just injected command line parameters inside script scope !!