home.public.bin/bin/some

#! /bin/bash
# Script to print "some" lines of a file, like head or tail but, random start
#
# Usage: some [[-#] FILE]

set -e; set -u; set -o pipefail # be safe out there

# Helper functions
PROG=`basename "$0" | tr -d '\n'`

function info()  { echo `date +%c` ${PROG}\: info: "$@" 1>&2; }
function warn()  { echo `date +%c` ${PROG}\: warning: "$@" 1>&2; }
function error() { echo `date +%c` ${PROG}\: error: "$@" 1>&2; }
function debug() { [[ -v DEBUG ]] && echo `date +%c` ${PROG}\: debug: "$@" 1>&2 || true ; }
function die()   { echo `date +%c` ${PROG}\: fatal: "$@" 1>&2 && exit 1; }


function truncate() {
    # trunchate a FILE to MAXLINES

    local FILE=$1
    local MAXLINES=$2

    local TEMP=`mktemp`

    cp ${FILE} ${TEMP} && \
        tail -${MAXLINES} ${TEMP} > \
             ${FILE} && \
        rm ${TEMP}
}



function someLines {
    # Functon to print "some" lines of a file, like head or tail but, random start
    #
    # ARGS:
    #
    #   $1 - "-#" or FILENAME if only one arg
    #   $2 - FILENAME if present

    local FILE=$1
    local HOW_MANY=$2


    # Count the lines to bound display
    LINES=`wc -l $FILE | sed -e 's/ .*//'`
    # TODO Do something sensible for stdin here
    #      Either
    #         - Save /dev/stdin to a tmp file (up to a max?), count THAT and use as FILE
    #         - revert to head-like behavior (with a timeout?)  (since we have no idea how long stdin is)

    # pick a random starting line at least HOW_MANY back from the end
    FIRST=$((1 + RANDOM % (LINES - HOW_MANY + 1)))
    FIRST=$((FIRST>LINES ? LINES : FIRST))

    LAST=$((FIRST + HOW_MANY - 1))
    LAST=$((LAST>LINES ? LINES : LAST))

    # log line numbers to allow re-extraction of randomly chosen lines
    info lines $FIRST to $LAST \(of $LINES\) of `readlink -f $FILE` |& cat  >> ${SOMELOG}

    # only keep MAXLOG lines of SOMELOG
    truncate $SOMELOG $MAXLOG

    # Lets see some lines !
    awk "NR >= $FIRST && NR <= $LAST" $FILE
}

function usage() {
    if [ "$1" != "" ]; then
        # TODO allow for undefined $1.  Now, have to set to "" or get undefined error.
        warn "${1}"
    fi

    USAGE=$(cat <<-END


Usage: $PROG [options] [FILENAME]

    Print some random lines from a file

options:
    -#            The number of lines to print.   Default 10.
    -h            Print usage.

arguments:
    FILENAME      The file to sample.  Default /dev/stdin
END
)
    info "$USAGE"

}


function main() {
    # main - parse args, set defaults, call someLines

    # Bash Version 3 required (it also works with ksh)
    [[ ${BASH_VERSINFO[0]} -lt 3 ]] && exit 1

    # Defaults
    local HOW_MANY=10 # Number of lines to print.  Default.
    local SOMELOG="${HOME}/.somelog" # log of queries in case you want to find that chunk again
    local MAXLOG=10

    # Put all arguments in a new array (because BASH_ARGV is read only)
    ARGS=( "$@" )

    for i in "${!ARGS[@]}"; do
        case "${ARGS[i]}" in
            '')             # Skip if element is empty (happens when its unsetted before)
                continue
                ;;
            -[0-9]*)     # Use +1 to access next array element and unset it
                HOW_MANY="${ARGS[i]}"
                HOW_MANY=`echo $HOW_MANY | sed 's/^-//'`
                shift
                ;;

            -h)     # Use +1 to access next array element and unset it
                usage ""
                exit 0
                ;;

            --)             # End of arguments
                unset 'ARGS[i]'
                break
                ;;
            *)              # Skip unset if our argument has not been matched
                continue
                ;;
        esac
        unset 'ARGS[i]'
    done
    # TODO add -n argumet to print line numbers

    FILE=${1:-"/dev/stdin"}

    if [ "$#" -gt 1 ]; then
        usage "Too many arguments"
        exit 1
    fi

    someLines $FILE $HOW_MANY

}



main $*