#
# Split field()s s with optional in and out delimiters
#
#       usage: field r ({in,out}delim=whitespace, rangeset=r)
#       usage: field id r (indelim=id, outdelim=whitespace, rangeset=r)
#       usage: field id od r (indelim=id, outdelim=od, rangeset=r)
#
# XXX TODO specifying od but not id?
#
# Takes a range specifier and returns all the field numbers
# contained in the specification.
#
# Syntax: one of these three forms:
#
#       indelim outdelim range
#       indelim range
#       range
#
# Where unspecified, a delim defaults to a space or tab.
# This is currently the only way to use more than one
# character as a delimiter, since the code only handles
# specifying single character delimiters right now.
#
# Range specifier is a comma separated list of ranges, where
# ranges are either single numbers or first-last contiguous
# sequences, so for example:
#
#       1-3,4,8,7-10
#
# would return "1 2 3 4 8 7 8 9 10" to the caller
# NOTE: incomplete range ends will default to last entry
#
# XXX TODO BUG this function will eat whitespace, i.e.
# "dothis '1 2'" will be treated the same as "dothis '1   2'"
#
# XXX TODO does it eat backslashes? Much like my shell history
# stores "\symbol" as "symbol" but will store "'symbol'" as
# "'symbol'"
#
# XXX TODO: 10x faster!!!
#
#       awk '{
#               for (i = start; i <= end; i++)
#                       printf("%s ", $i);
#               printf("\n");
#       }'
#
# XXX TODO allow multiple delimiters
# XXX TODO allow regexes for delimiters
#

require range_to_list

field ()
{
        local \
                rangearg \
                indelim \
                outdelim

        local shellfalse=1

        case $# in
        (3) indelim="$1" outdelim="$2" rangearg="$3" ;;
        (2) indelim="$1" outdelim=' '  rangearg="$2" ;;
        (1) indelim=' '  outdelim=' '  rangearg="$1" ;;
        (*) return $shellfalse;;
        esac

        # we only handle single-character delimiters right now
        #
        ((${#indelim} == 1)) || return $shellfalse;
        ((${#outdelim} == 1)) || return $shellfalse;

        # ugly hack: we need to break on whitespace, not
        # just spaces, i.e. to include tabs (and newlines
        # too, but not yet), so by unsetting indelim, it
        # defaults later when we assign IFS
        #
        (($# == 1)) && unset indelim

        # expands the special-syntax list of ranges into a
        # single list of numbers occupying that range
        #
        # - split up the input into fields as delimited by
        #   the given delimiter, or just sameline whitespace
        #   if none was given
        #
        # - only echo the particular fields that were
        #   requested with the range argument
        #
        # - separate output fields with the specified output
        #   delimiter, which defaults to a space
        #
        (
                local -i fieldnum iterations
                local -a fields
                local -a outarray

                IFS="${indelim:-$' \t'}"
                while read -a line
                do (
                        IFS=' '
                        outarray=()
                        fields=($(range_to_list $rangearg ${#line[@]}))
                        iterations=${#fields[@]}

                        for ((i = 0; i < iterations; i++))
                        do
                                fieldnum=$((${fields[i]} - 1))
                                outarray+=("${line[fieldnum]}")
                        done

                        (IFS="$outdelim"; echo "${outarray[*]}")
                ); done
        )
}
# vim:syn=sh:ft=sh