Say, I have a script that gets called with this line:
./myscript -vfd ./foo/bar/someFile -o /fizz/someOtherFile
or this one:
./myscript -v -f -d -o /fizz/someOtherFile ./foo/bar/someFile
What's the accepted way of parsing this such that in each case (or some combination of the two) $v, $f, and $d will all be set to true and $outFile will be equal to /fizz/someOtherFile?
getopt()/getopts() is a good option. Copied from here:
The simple use of "getopt" is shown in this mini-script:
#!/bin/bash
echo "Before getopt"
for i
do
echo $i
done
args=`getopt abc:d $*`
set -- $args
echo "After getopt"
for i
do
echo "-->$i"
done
What we have said is that any of -a, -b, -c or -d will be allowed, but that -c is followed by an argument (the "c:" says that).
If we call this "g" and try it out:
bash-2.05a$ ./g -abc foo
Before getopt
-abc
foo
After getopt
-->-a
-->-b
-->-c
-->foo
-->--
We start with two arguments, and "getopt" breaks apart the options and puts each in its own argument. It also added "--".
Expanding on @bruno-bronosky's answer, I added a "preprocessor" to handle some common formatting:
--longopt=val into --longopt val-xyz into -x -y -z-- to indicate the end of flags#!/bin/bash
# Report usage
usage() {
echo "Usage:"
echo "$(basename "$0") [options] [--] [file1, ...]"
}
invalid() {
echo "ERROR: Unrecognized argument: $1" >&2
usage
exit 1
}
# Pre-process options to:
# - expand -xyz into -x -y -z
# - expand --longopt=arg into --longopt arg
ARGV=()
END_OF_OPT=
while [[ $# -gt 0 ]]; do
arg="$1"; shift
case "${END_OF_OPT}${arg}" in
--) ARGV+=("$arg"); END_OF_OPT=1 ;;
--*=*)ARGV+=("${arg%%=*}" "${arg#*=}") ;;
--*) ARGV+=("$arg") ;;
-*) for i in $(seq 2 ${#arg}); do ARGV+=("-${arg:i-1:1}"); done ;;
*) ARGV+=("$arg") ;;
esac
done
# Apply pre-processed options
set -- "${ARGV[@]}"
# Parse options
END_OF_OPT=
POSITIONAL=()
while [[ $# -gt 0 ]]; do
case "${END_OF_OPT}${1}" in
-h|--help) usage; exit 0 ;;
-p|--password) shift; PASSWORD="$1" ;;
-u|--username) shift; USERNAME="$1" ;;
-n|--name) shift; names+=("$1") ;;
-q|--quiet) QUIET=1 ;;
-C|--copy) COPY=1 ;;
-N|--notify) NOTIFY=1 ;;
--stdin) READ_STDIN=1 ;;
--) END_OF_OPT=1 ;;
-*) invalid "$1" ;;
*) POSITIONAL+=("$1") ;;
esac
shift
done
# Restore positional parameters
set -- "${POSITIONAL[@]}"
getopt(s)I was inspired by the relatively simple answer by @bronson and tempted to try to improve it (without adding too much complexity). Here's the result:
-n [arg], -abn [arg], --name [arg] and --name=arg styles of options;$@ after the loop;-- to force remaining arguments to be treated as positional;getopt(s) or external tools (one feature uses a simple sed command);# Convenience functions.
usage_error () { echo >&2 "$(basename $0): $1"; exit 2; }
assert_argument () { test "$1" != "$EOL" || usage_error "$2 requires an argument"; }
# One loop, nothing more.
if [ "$#" != 0 ]; then
EOL=$(printf '\1\3\3\7')
set -- "$@" "$EOL"
while [ "$1" != "$EOL" ]; do
opt="$1"; shift
case "$opt" in
# Your options go here.
-f|--flag) flag='true';;
-n|--name) assert_argument "$1" "$opt"; name="$1"; shift;;
# Arguments processing. You may remove any unneeded line after the 1st.
-|''|[!-]*) set -- "$@" "$opt";; # positional argument, rotate to the end
--*=*) set -- "${opt%%=*}" "${opt#*=}" "$@";; # convert '--name=arg' to '--name' 'arg'
-[!-]?*) set -- $(echo "${opt#-}" | sed 's/\(.\)/ -\1/g') "$@";; # convert '-abc' to '-a' '-b' '-c'
--) while [ "$1" != "$EOL" ]; do set -- "$@" "$1"; shift; done;; # process remaining arguments as positional
-*) usage_error "unknown option: '$opt'";; # catch misspelled options
*) usage_error "this should NEVER happen ($opt)";; # sanity test for previous patterns
esac
done
shift # $EOL
fi
# Do something cool with "$@"... \o/
Note: I know... An argument with the binary pattern 0x01030307 could break the logic. But, if anyone passes such an argument in a command-line, they deserve it.
An elegant option parser for shell scripts (full support for all POSIX shells) https://github.com/ko1nksm/getoptions (Update: v3.3.0 released on 2021-05-02)
getoptions is a new option parser (generator) written in POSIX-compliant shell script and released in august 2020. It is for those who want to support the POSIX / GNU style option syntax in your shell scripts.
The supported syntaxes are -a, +a, -abc, -vvv, -p VALUE, -pVALUE, --flag, --no-flag, --with-flag, --without-flag, --param VALUE, --param=VALUE, --option[=VALUE], --no-option --.
It supports subcommands, validation, abbreviated options, and automatic help generation. And works with all POSIX shells (dash 0.5.4+, bash 2.03+, ksh88+, mksh R28+, zsh 3.1.9+, yash 2.29+, busybox ash 1.1.3+, etc).
#!/bin/sh
VERSION="0.1"
parser_definition() {
setup REST help:usage -- "Usage: example.sh [options]... [arguments]..." ''
msg -- 'Options:'
flag FLAG -f --flag -- "takes no arguments"
param PARAM -p --param -- "takes one argument"
option OPTION -o --option on:"default" -- "takes one optional argument"
disp :usage -h --help
disp VERSION --version
}
eval "$(getoptions parser_definition) exit 1"
echo "FLAG: $FLAG, PARAM: $PARAM, OPTION: $OPTION"
printf '%s\n' "$@" # rest arguments
It's parses the following arguments:
example.sh -f --flag -p VALUE --param VALUE -o --option -oVALUE --option=VALUE 1 2 3
And automatic help generation.
$ example.sh --help
Usage: example.sh [options]... [arguments]...
Options:
-f, --flag takes no arguments
-p, --param PARAM takes one argument
-o, --option[=OPTION] takes one optional argument
-h, --help
--version
It is also an option parser generator, generates the following simple option parsing code. If you use the generated code, you won't need getoptions. Achieve true portability and zero dependency.
FLAG=''
PARAM=''
OPTION=''
REST=''
getoptions_parse() {
OPTIND=$(($#+1))
while OPTARG= && [ $# -gt 0 ]; do
case $1 in
--?*=*) OPTARG=$1; shift
eval 'set -- "${OPTARG%%\=*}" "${OPTARG#*\=}"' ${1+'"$@"'}
;;
--no-*|--without-*) unset OPTARG ;;
-[po]?*) OPTARG=$1; shift
eval 'set -- "${OPTARG%"${OPTARG#??}"}" "${OPTARG#??}"' ${1+'"$@"'}
;;
-[fh]?*) OPTARG=$1; shift
eval 'set -- "${OPTARG%"${OPTARG#??}"}" -"${OPTARG#??}"' ${1+'"$@"'}
OPTARG= ;;
esac
case $1 in
'-f'|'--flag')
[ "${OPTARG:-}" ] && OPTARG=${OPTARG#*\=} && set "noarg" "$1" && break
eval '[ ${OPTARG+x} ] &&:' && OPTARG='1' || OPTARG=''
FLAG="$OPTARG"
;;
'-p'|'--param')
[ $# -le 1 ] && set "required" "$1" && break
OPTARG=$2
PARAM="$OPTARG"
shift ;;
'-o'|'--option')
set -- "$1" "$@"
[ ${OPTARG+x} ] && {
case $1 in --no-*|--without-*) set "noarg" "${1%%\=*}"; break; esac
[ "${OPTARG:-}" ] && { shift; OPTARG=$2; } || OPTARG='default'
} || OPTARG=''
OPTION="$OPTARG"
shift ;;
'-h'|'--help')
usage
exit 0 ;;
'--version')
echo "${VERSION}"
exit 0 ;;
--)
shift
while [ $# -gt 0 ]; do
REST="${REST} \"\${$(($OPTIND-$#))}\""
shift
done
break ;;
[-]?*) set "unknown" "$1"; break ;;
*)
REST="${REST} \"\${$(($OPTIND-$#))}\""
esac
shift
done
[ $# -eq 0 ] && { OPTIND=1; unset OPTARG; return 0; }
case $1 in
unknown) set "Unrecognized option: $2" "$@" ;;
noarg) set "Does not allow an argument: $2" "$@" ;;
required) set "Requires an argument: $2" "$@" ;;
pattern:*) set "Does not match the pattern (${1#*:}): $2" "$@" ;;
notcmd) set "Not a command: $2" "$@" ;;
*) set "Validation error ($1): $2" "$@"
esac
echo "$1" >&2
exit 1
}
usage() {
cat<<'GETOPTIONSHERE'
Usage: example.sh [options]... [arguments]...
Options:
-f, --flag takes no arguments
-p, --param PARAM takes one argument
-o, --option[=OPTION] takes one optional argument
-h, --help
--version
GETOPTIONSHERE
}
I wanna submit my project : https://github.com/flyingangel/argparser
source argparser.sh
parse_args "$@"
Simple as that. The environment will be populated with variables with the same name as the arguments
There are several ways to parse cmdline args (e.g. GNU getopt (not portable) vs BSD (MacOS) getopt vs getopts) - all problematic. This solution
= separator between option and argument-vxf--color vs --color=always),-- to signal end of options, andExamples: Any of
# flag
-f
--foo
# option with required argument
-b"Hello World"
-b "Hello World"
--bar "Hello World"
--bar="Hello World"
# option with optional argument
--baz
--baz="Optional Hello"
#!/usr/bin/env bash
usage() {
cat - >&2 <<EOF
NAME
program-name.sh - Brief description
SYNOPSIS
program-name.sh [-h|--help]
program-name.sh [-f|--foo]
[-b|--bar <arg>]
[--baz[=<arg>]]
[--]
FILE ...
REQUIRED ARGUMENTS
FILE ...
input files
OPTIONS
-h, --help
Prints this and exits
-f, --foo
A flag option
-b, --bar <arg>
Option requiring an argument <arg>
--baz[=<arg>]
Option that has an optional argument <arg>. If <arg>
is not specified, defaults to 'DEFAULT'
--
Specify end of options; useful if the first non option
argument starts with a hyphen
EOF
}
fatal() {
for i; do
echo -e "${i}" >&2
done
exit 1
}
# For long option processing
next_arg() {
if [[ $OPTARG == *=* ]]; then
# for cases like '--opt=arg'
OPTARG="${OPTARG#*=}"
else
# for cases like '--opt arg'
OPTARG="${args[$OPTIND]}"
OPTIND=$((OPTIND + 1))
fi
}
# ':' means preceding option character expects one argument, except
# first ':' which make getopts run in silent mode. We handle errors with
# wildcard case catch. Long options are considered as the '-' character
optspec=":hfb:-:"
args=("" "$@") # dummy first element so $1 and $args[1] are aligned
while getopts "$optspec" optchar; do
case "$optchar" in
h) usage; exit 0 ;;
f) foo=1 ;;
b) bar="$OPTARG" ;;
-) # long option processing
case "$OPTARG" in
help)
usage; exit 0 ;;
foo)
foo=1 ;;
bar|bar=*) next_arg
bar="$OPTARG" ;;
baz)
baz=DEFAULT ;;
baz=*) next_arg
baz="$OPTARG" ;;
-) break ;;
*) fatal "Unknown option '--${OPTARG}'" "see '${0} --help' for usage" ;;
esac
;;
*) fatal "Unknown option: '-${OPTARG}'" "See '${0} --help' for usage" ;;
esac
done
shift $((OPTIND-1))
if [ "$#" -eq 0 ]; then
fatal "Expected at least one required argument FILE" \
"See '${0} --help' for usage"
fi
echo "foo=$foo, bar=$bar, baz=$baz, files=${@}"
Based on other answers here, this my version:
#!/bin/bash
set -e
function parse() {
for arg in "$@"; do # transform long options to short ones
shift
case "$arg" in
"--name") set -- "$@" "-n" ;;
"--verbose") set -- "$@" "-v" ;;
*) set -- "$@" "$arg"
esac
done
while getopts "n:v" optname # left to ":" are flags that expect a value, right to the ":" are flags that expect nothing
do
case "$optname" in
"n") name=${OPTARG} ;;
"v") verbose=true ;;
esac
done
shift "$((OPTIND-1))" # shift out all the already processed options
}
parse "$@"
echo "hello $name"
if [ ! -z $verbose ]; then echo 'nice to meet you!'; fi
Usage:
$ ./parse.sh
hello
$ ./parse.sh -n YOUR_NAME
hello YOUR_NAME
$ ./parse.sh -n YOUR_NAME -v
hello YOUR_NAME
nice to meet you!
$ ./parse.sh -v -n YOUR_NAME
hello YOUR_NAME
nice to meet you!
$ ./parse.sh -v
hello
nice to meet you!
Assume we create a shell script named test_args.sh as follow
#!/bin/sh
until [ $# -eq 0 ]
do
name=${1:1}; shift;
if [[ -z "$1" || $1 == -* ]] ; then eval "export $name=true"; else eval "export $name=$1"; shift; fi
done
echo "year=$year month=$month day=$day flag=$flag"
After we run the following command:
sh test_args.sh -year 2017 -flag -month 12 -day 22
The output would be:
year=2017 month=12 day=22 flag=true
I wanted to share what I made for parsing options. Some of my needs were not fulfilled by the answers here so I had to come up with this: https://github.com/MihirLuthra/bash_option_parser
This supports:
Let's say we have a command named fruit with usage as follows:
fruit <fruit-name> ...
[-e|—-eat|—-chew]
[-c|--cut <how> <why>]
<command> [<args>]
-e takes no args
-c takes two args i.e. how to cut and why to cut
fruit itself takes at least one argument.
<command> is for suboptions like apple, orange etc. (similar to git which has suboptions commit, push etc. )
So to parse it:
parse_options \
'fruit' '1 ...' \
'-e' , '--eat' , '--chew' '0' \
'-c' , '--cut' '1 1' \
'apple' 'S' \
'orange' 'S' \
';' \
"$@"
Now if there was any usage error, it can be printed using option_parser_error_msg as follows:
retval=$?
if [ $retval -ne 0 ]; then
# this will manage error messages if
# insufficient or extra args are supplied
option_parser_error_msg "$retval"
# This will print the usage
print_usage 'fruit'
exit 1
fi
To check now if some options was passed,
if [ -n "${OPTIONS[-c]}" ]
then
echo "-c was passed"
# args can be accessed in a 2D-array-like format
echo "Arg1 to -c = ${ARGS[-c,0]}"
echo "Arg2 to -c = ${ARGS[-c,1]}"
fi
Suboption parsing can also be done by passing $shift_count to parse_options_detailed which makes it start parsing after shifting args to reach args of suboption. It is demonstrated in this example.
A detailed description is provided in the readme and examples in the repository.
Here is a getopts that achieves the parsing with minimal code and allows you to define what you wish to extract in one case using eval with substring.
Basically eval "local key='val'"
function myrsync() {
local backup=("${@}") args=(); while [[ $# -gt 0 ]]; do k="$1";
case "$k" in
---sourceuser|---sourceurl|---targetuser|---targeturl|---file|---exclude|---include)
eval "local ${k:3}='${2}'"; shift; shift # Past two arguments
;;
*) # Unknown option
args+=("$1"); shift; # Past argument only
;;
esac
done; set -- "${backup[@]}" # Restore $@
echo "${sourceurl}"
}
Declares the variables as locals instead of globals as most answers here.
Called as:
myrsync ---sourceurl http://abc.def.g ---sourceuser myuser ...
The ${k:3} is basically a substring to remove the first --- from the key.
I wrote down a script that can assist with parsing command-line arguments easily - https://github.com/unfor19/bargs
$ bash example.sh -n Willy --gender male -a 99
Name: Willy
Age: 99
Gender: male
Location: chocolate-factory
$ bash example.sh -n Meir --gender male
[ERROR] Required argument: age
Usage: bash example.sh -n Willy --gender male -a 99
--person_name | -n [Willy] What is your name?
--age | -a [Required]
--gender | -g [Required]
--location | -l [chocolate-factory] insert your location
$ bash example.sh -h
Usage: bash example.sh -n Willy --gender male -a 99
--person_name | -n [Willy] What is your name?
--age | -a [Required]
--gender | -g [Required]
--location | -l [chocolate-factory] insert your location
I ended up implementing the dash (or /bin/sh) version of the accepted answer, basically, without array usage:
while [[ $# -gt 0 ]]; do
case "$1" in
-v|--verbose) verbose=1; shift;;
-o|--output) if [[ $# -gt 1 && "$2" != -* ]]; then
file=$2; shift 2
else
echo "-o requires file-path" 1>&2; exit 1
fi ;;
--)
while [[ $# -gt 0 ]]; do BACKUP="$BACKUP;$1"; shift; done
break;;
*)
BACKUP="$BACKUP;$1"
shift
;;
esac
done
# Restore unused arguments.
while [ -n "$BACKUP" ] ; do
[ ! -z "${BACKUP%%;*}" ] && set -- "$@" "${BACKUP%%;*}"
[ "$BACKUP" = "${BACKUP/;/}" ] && break
BACKUP="${BACKUP#*;}"
done
Simple and easy to modify, parameters can be in any order. this can be modified to take parameters in any form (-a, --a, a, etc).
for arg in "$@"
do
key=$(echo $arg | cut -f1 -d=)`
value=$(echo $arg | cut -f2 -d=)`
case "$key" in
name|-name) read_name=$value;;
id|-id) read_id=$value;;
*) echo "I dont know what to do with this"
ease
done
I use it to iterate over key => value from the end. A first optional argument is caught after the loop.
Usage is ./script.sh optional-first-arg -key value -key2 value2
#!/bin/sh
a=$(($#-1))
b=$(($#))
while [ $a -gt 0 ]; do
eval 'key="$'$a'"; value="$'$b'"'
echo "$key => $value"
b=$(($b-2))
a=$(($a-2))
done
unset a b key value
[ $(($#%2)) -ne 0 ] && echo "first_arg = $1"
Sure you can do it from the left to the right with a few changes.
This snippet code shows the key => value pairs and the first argument if it exists.
#!/bin/sh
a=$((1+$#%2))
b=$((1+$a))
[ $(($#%2)) -ne 0 ] && echo "first_arg = $1"
while [ $a -lt $# ]; do
eval 'key="$'$a'"; value="$'$b'"'
echo "$key => $value"
b=$(($b+2))
a=$(($a+2))
done
unset a b key value
Tested with 100,000 arguments, fast.
You can also iterate key => value and first optional arg from the left to the right without eval :
#!/bin/sh
a=$(($#%2))
b=0
[ $a -eq 1 ] && echo "first_arg = $1"
for value; do
if [ $b -gt $a -a $(($b%2)) -ne $a ]; then
echo "$key => $value"
fi
key="$value"
b=$((1+$b))
done
unset a b key value
Another solution which calls getopt separately for each option instead of using a switch case statement:
cat >/tmp/demo.sh <<'EOF'
#!/bin/bash
# set to the value of "-o", "--outfile" or the default "."
outfile="$(getopt -q -u -l "outfile:" "o:" "$@" | grep -oP "(?<= --outfile | -o ).*(?= --.*)" || printf ".")"
# set to "1" if option "-V" or "--verbose" is present
verbose="$(getopt -q -u -l "verbose" "V" "$@" | grep -oP "(?<= --verbose | -V ).*(?=--.*)" && printf "1")"
# set to last positional argument
infile="${*: -1:1}"
echo "infile:$infile"
echo "outfile:$outfile"
echo "verbose:$verbose"
EOF
chmod +x /tmp/demo.sh
Example 1:
# /tmp/demo.sh -Vo /tmp/outfile /tmp/infile
infile:/tmp/infile
outfile:/tmp/outfile
verbose:1
Example 2:
# /tmp/demo.sh /tmp/infile
infile:/tmp/infile
outfile:.
verbose:
Downside:
If the user makes the mistake and uses long and short option names at the same time, the variable is filled with the option name itself:
# /tmp/demo.sh -o /tmp/outfile --outfile=/tmp/outfile /tmp/infile
infile:/tmp/infile
outfile:/tmp/outfile --outfile /tmp/outfile
verbose:
But this is something which can be easily catched with further checks.