flohmarkt_ynh/scripts/ynh_handle_getopts_args

#!/bin/bash

# Redisgn of ynh_handle_getopts_args for flohmarkt to be tested as `flohmarkt_ynh_handle_getopts_args`
# https://github.com/YunoHost/yunohost/pull/1856
# Internal helper design to allow helpers to use getopts to manage their arguments
#
# [internal]
#
# example: function my_helper()
# {
#     local -A args_array=( [a]=arg1= [b]=arg2= [c]=arg3 )
#     local arg1
#     local arg2
#     local arg3
#     ynh_handle_getopts_args "$@"
#
#     [...]
# }
# my_helper --arg1 "val1" -b val2 -c
#
# usage: ynh_handle_getopts_args "$@"
# | arg: $@    - Simply "$@" to tranfert all the positionnal arguments to the function
#
# This helper need an array, named "args_array" with all the arguments used by the helper
#   that want to use ynh_handle_getopts_args
# Be carreful, this array has to be an associative array, as the following example:
# local -A args_array=( [a]=arg1 [b]=arg2= [c]=arg3 )
# Let's explain this array:
# a, b and c are short options, -a, -b and -c
# arg1, arg2 and arg3 are the long options associated to the previous short ones. --arg1, --arg2 and --arg3
# For each option, a short and long version has to be defined.
# Let's see something more significant
# local -A args_array=( [u]=user [f]=finalpath= [d]=database )
#
# NB: Because we're using 'declare' without -g, the array will be declared as a local variable.
#
# Please keep in mind that the long option will be used as a variable to store the values for this option.
# For the previous example, that means that $finalpath will be fill with the value given as argument for this option.
#
# Also, in the previous example, finalpath has a '=' at the end. That means this option need a value.
# So, the helper has to be call with --finalpath /final/path, --finalpath=/final/path or -f /final/path, 
# the variable $finalpath will get the value /final/path
# If there's many values for an option, -f /final /path, the value will be separated by a ';' $finalpath=/final;/path
# For an option without value, like --user in the example, the helper can be called only with --user or -u. $user 
# will then get the value 1.
#
# To keep a retrocompatibility, a package can still call a helper, using getopts, with positional arguments.
# The "legacy mode" will manage the positional arguments and fill the variable in the same order than they are given 
# in $args_array. e.g. for `my_helper "val1" val2`, arg1 will be filled with val1, and arg2 with val2.

# Positional parameters (used to be the only way to use ynh_handle_getopts_args once upon a time) can be 
# used also:
# 
# '--'          start processing the rest of the arguments as positional parameters
# $legacy_args  The arguments positional parameters will be assign to
#               Needs to be composed of array keys of args_array. If a key for a predefined variable
#               is used multiple times the assigned values will be concatenated delimited by ';'.
#               If the long option variable to contain the data is predefined as an array (e.g. using
#               `local -a arg1` then multiple values will be assigned to its cells.
#               If the last positional parameter defined in legacy_args is defined as an array all 
#               the leftover positional parameters will be assigned to its cells.
#               (it is named legacy_args, because the use of positional parameters was about to be
#               deprecated before the last re-design of this sub)
#
# Requires YunoHost version 3.2.2 or higher.
# flohmarkt_ynh_handle_getopts_args() {
ynh_handle_getopts_args() {
    # Manage arguments only if there's some provided
    set +o xtrace # set +x
    if [ $# -eq 0 ]; then
      ynh_print_warn --message="ynh_handle_getopts_args called without arguments"
      return
    fi

    # Store arguments in an array to keep each argument separated
    local arguments=("$@")

    # For each option in the array, reduce to short options for getopts (e.g. for [u]=user, --user will be -u)
    # And built parameters string for getopts
    # ${!args_array[@]} is the list of all option_flags in the array (An option_flag is 'u' in [u]=user, user is a value)
    local getopts_parameters=""
    local option_flag=""
    ## go through all possible options and replace arguments with short versions
    flohmarkt_print_debug "arguments = '${arguments[@]}"
    flohmarkt_print_debug "args_array = (${!args_array[@]})"
    for option_flag in "${!args_array[@]}"; do
        # TODO refactor: Now I'm not sure anymore this part belongs here. To make the
        # this all less hard to read and understand I'm thinking at the moment that it 
        # would be good to split the different things done here into their own loops:
        #
        # 1. build the option string $getopts_parameters
        # 2. go through the arguments and add empty arguments where needed to 
        #    allow for cases like '--arg= --value' where 'value' is a valid option, too
        # 3. replace long option names by short once
        # 4. (possibly add empty parameters for '-a -v' in cases where -a expects a value
        #    and -v is a valid option, too - but I dearly hope this will not be necessary)
        flohmarkt_print_debug "option_flag = $option_flag"
        # Concatenate each option_flags of the array to build the string of arguments for getopts
        # Will looks like 'abcd' for -a -b -c -d
        # If the value of an option_flag finish by =, it's an option with additionnal values. 
        # (e.g. --user bob or -u bob)
        # Check the last character of the value associate to the option_flag
        flohmarkt_print_debug "compare to '${args_array[$option_flag]: -1}'"
        if [ "${args_array[$option_flag]: -1}" = "=" ]; then
            # For an option with additionnal values, add a ':' after the letter for getopts.
            getopts_parameters="${getopts_parameters}${option_flag}:"
        else
            getopts_parameters="${getopts_parameters}${option_flag}"
        fi
        flohmarkt_print_debug "getopts_parameters = ${getopts_parameters}"
        # Check each argument given to the function
        local arg=""
        # ${#arguments[@]} is the size of the array
        ## for one possible option: look at each argument supplied:
        for arg in $(seq 0 $((${#arguments[@]} - 1))); do
            flohmarkt_print_debug "option_flag='$option_flag', arg = '$arg', argument = '${arguments[arg]}'"
            # the following cases need to be taken care of
            # '--arg=value'    → works
            # '--arg= value'   → works
            # '--arg=-value'   → works
            # '--arg= -v'      or
            # '--arg= --value' → works if not exists arg '[v]=value='
            #                  → $arg will be set to '-v' or '--value'
            #   but if exists '[v]=value=' this is not the expected behavior:
            #   → then $arg is expected to contain an empty value and '-v' or '--value'
            #     is expected to be interpreted as its own valid argument
            #   (found in use of ynh_replace_string called by ynh_add_config)
            #   solution:
            #   insert an empty arg into array arguments to be later interpreted by 
            #   getopts as the missing value to --arg=
            if [[ -v arguments[arg+1] ]] && [[ ${arguments[arg]: -1} == '=' ]]; then
                # arg ends with a '='
                local this_argument=${arguments[arg]}
                local next_argument=${arguments[arg+1]}
                # for looking up next_argument in args_array remove optionally trailing '='
                next_argument=$( printf '%s' "$next_argument" | cut -d'=' -f1 )
                flohmarkt_print_debug "MISSING PARAMETER: this_argument='$this_argument', next_argument='$next_argument'"

                # check if next_argument is a value in args_array
                # → starts with '--' and the rest of the argument excluding optional trailing '=' 
                #     of the string is a value in associative array args_array
                # → or starts with '-' and the rest of the argument is a valid key in args_array
                #   (long argument could already have been replaced by short version before)
                flohmarkt_print_debug "args_array values='${args_array[@]}'"
                flohmarkt_print_debug "args_array   keys='${!args_array[@]}'"
                flohmarkt_print_debug "{next_argument:2}='${next_argument:2}'"
                flohmarkt_print_debug "{next_argument:0:2}='${next_argument:0:2}'"
                if ( [[ "${next_argument:0:2}" == '--' ]] \
                    && printf '%s ' "${args_array[@]}" | fgrep -w "${next_argument:2}" > /dev/null ) \
                || ( [[ "${next_argument:0:1}" == '-' ]] \
                    && printf '%s ' "${!args_array[@]}" | fgrep -w "${next_argument:1:1}" > /dev/null )
                then
                    # insert an empty value to array arguments to be interpreted as the value
                    # for argument[arg]
                    arguments=( ${arguments[@]:0:arg+1} '' ${arguments[@]:arg+1})
                    flohmarkt_print_debug "now arguments='${arguments[@]}'"
                fi
            fi

            # Replace long option with = (match the beginning of the argument)
            arguments[arg]="$(printf '%s\n' "${arguments[arg]}" \
                | sed "s/^--${args_array[$option_flag]}/-${option_flag}/")"
            # And long option without = (match the whole line)
            arguments[arg]="$(printf '%s\n' "${arguments[arg]}" \
                | sed "s/^--${args_array[$option_flag]%=}$/-${option_flag}/")"
            flohmarkt_print_debug "                            arg = '$arg', argument = '${arguments[arg]}'"
        done
        flohmarkt_print_debug "====> end loop: arguments = '${arguments[@]}'"
    done
    flohmarkt_print_debug '================= end first loop ================='
    
    # Parse the first argument, return the number of arguments to be shifted off the arguments array
    # The function call is necessary here to allow `getopts` to use $@
    parse_arg() {
        flohmarkt_print_debug "========= parse_arg started ======== , arguments='$@', getopts_parameters: '$getopts_parameters'"
        # Initialize the index of getopts
        OPTIND=1
        # getopts will fill $parameter with the letter of the option it has read.
        local parameter=""
        getopts ":$getopts_parameters" parameter || true
        flohmarkt_print_debug "after getopts - parameter='$parameter', OPTIND='${OPTIND:-none}', OPTARG='${OPTARG:-none}'"

        if [ "$parameter" = "?" ]; then
            ynh_die --message="Invalid argument: -${OPTARG:-}"
            flohmarkt_print_debug "Invalid argument: -${OPTARG:-}"
            exit 255
        elif [ "$parameter" = ":" ]; then
            ynh_die --message="-$OPTARG parameter requires an argument."
            echo "-$OPTARG parameter requires an argument."
            exit 255
        else
            # Use the long option, corresponding to the short option read by getopts, as a variable
            # (e.g. for [u]=user, 'user' will be used as a variable)
            # Also, remove '=' at the end of the long option
            # The variable name will be stored in 'option_var' as a nameref
            option_var="${args_array[$parameter]%=}"
            flohmarkt_print_debug "option_var='$option_var'"
            # if there's a '=' at the end of the long option name, this option takes values
            if [ "${args_array[$parameter]: -1}" != "=" ]; then
                # no argument expected for option - set option variable to '1'
                option_value=1
            else
                # remove leading and trailing spaces from OPTARG
                OPTARG="$( printf '%s' "${OPTARG}" | sed -e 's/^[[:space:]]*//' -e 's/[[:space:]]*$//')"
                option_value="${OPTARG}"
            fi
            flohmarkt_print_debug "option_value='$option_value'"
            # set shift_value according to the number of options interpreted by getopts
            shift_value=$(( $OPTIND - 1 ))
            flohmarkt_print_debug "shift_value='$shift_value'"
        fi
    }

    # iterate over the arguments: if first argument starts with a '-' feed arguments to getopts
    # if first argument doesn't start with a '-' enter mode to read positional parameters
    local argument
    local positional_mode=0 # state is getopts mode at the beginning, not positional parameters
    local positional_count=0 # counter for positional parameters
    local option_var=''   # the variable name to be filled
    # Try to use legacy_args as a list of option_flag of the array args_array
    # Otherwise, fill it with getopts_parameters to get the option_flag. 
    # (But an associative arrays isn't always sorted in the correct order...)
    # Remove all ':' in getopts_parameters, if used.
    legacy_args=${legacy_args:-${getopts_parameters//:/}}
    while [[ -v 'arguments' ]] && [[ ${#arguments} -ne 0 ]]; do
        flohmarkt_print_debug '======= start while loop ======='
        local shift_value=0
        local option_value='' # the value to be filled into ${!option_var}
        argument=${arguments[0]}
        flohmarkt_print_debug "argument='$argument'"
        # if state once changed to positional parameter mode, all the rest of the arguments will
        # be interpreted in positional parameter mode even if they start with a '-'
        if [ $positional_mode == 0 ] && [ "${argument}" == '--' ];then
            flohmarkt_print_debug "found '--', start positional parameter mode"
            positional_mode=1
            shift_value=1
        elif [ $positional_mode == 0 ] && [ "${argument:0:1}" == '-' ]; then
            flohmarkt_print_debug "getopts, arguments='${arguments[@]}', starting parse_arg"
            parse_arg "${arguments[@]}"
        else
            positional_mode=1 # set state to positional parameter mode
            flohmarkt_print_debug "positional parameter, argument='$argument'"

            # Get the option_flag from getopts_parameters by using the option_flag according to the 
            # position of the argument.
            option_flag=${legacy_args:$positional_count:1}

            # increment counter for legacy_args if still args left. If no args left check if the 
            # last arg is a predefined array and let it cells be filled. Otherwise complain and 
            # return.
            flohmarkt_print_debug "positional_counter='$positional_count', max positional_counter='$(( ${#legacy_args} -1 ))'"
            if [[ $positional_count -le $((${#legacy_args} - 1)) ]]; then
                # set counter to for next option_flag to fill
                positional_count=$((positional_count+1))
                flohmarkt_print_debug "incremented positional_counter to '$positional_count'"

                # Use the long option, corresponding to the option_flag, as a variable
                # (e.g. for [u]=user, 'user' will be used as a variable)
                # Also, remove '=' at the end of the long option
                # The variable name will be stored in 'option_var'
                option_var="${args_array[$option_flag]%=}"
            elif [[ $positional_count -ge $((${#legacy_args} - 1)) ]] && 
                ! declare -p ${option_var} | grep '^declare -a'
            then
                # no more legacy_args to fill - legacy behaviour: complain and return
                ynh_print_warn --message="Too many arguments ! \"${arguments[$i]}\" will be ignored."
                return
            else
                flohmarkt_print_debug "array found - keep going"
            fi

            # value to be assigned to ${!option_var}
            option_value=$argument

            # shift off one positional parameter
            shift_value=1
        fi

        # fill option_var with value found
        # if ${!option_var} is an array, fill mutiple values as array cells
        # otherwise concatenate them seperated by ';'
        # nameref is used to access the variable that is named $option_var
        local -n option_ref=$option_var
        # this defines option_ref as a reference to the variable named "$option_var"
        # any operation on option_ref will be written to the variable named "$option_var"
        # 'option_ref="hello world"' will work like as if '${!option_var}="hello world"'
        # would be a valid syntax
        # see also: `man bash` part about commands 'declare' option '-n'
        flohmarkt_print_debug "option_ref declare: '$(declare -p option_ref)'"
        flohmarkt_print_debug "'$option_var' declare: '$(declare -p $option_var)'"
        flohmarkt_print_debug "option_value='$option_value'"
        if declare -p $option_var | grep '^declare -a ' > /dev/null; then
          # hurray it's an array
          flohmarkt_print_debug "hurray! '$option_var' is an array."
          option_ref+=( "${option_value}" )
        elif ! [[ -v "$option_var" ]] || [[ -z "$option_ref" ]]; then
          flohmarkt_print_debug "'$option_var' is unset or empty"
          option_ref=${option_value}
        else
          flohmarkt_print_debug "appending to string '$option_ref'"
          option_ref+=";${option_value}"
        fi
        flohmarkt_print_debug "now declared $option_var: '$(declare -p $option_var)'"

        # shift value off arguments array
        flohmarkt_print_debug "shifting '$shift_value' off arguments='${arguments[@]}'"
        arguments=("${arguments[@]:${shift_value}}")
    done

    # the former subroutine did this - no idea if it is expected somewhere
    unset legacy_args

    # re-enable trace
    set -o xtrace # set -x
}