#!/bin/bash

PROGNAME=$0
usage() {
    echo "
gebruik:

    $PROGNAME [opties] -i
    $PROGNAME [opties] corpusfile(s) en/of corpusdirectory(s)

    -i     : kies interactief

overige opties:

    -n int : max aantal resultaten
    -p     : gebruik pager
    -s     : tel hits één keer per bericht
    -v     : verbose
    "
    exit
}

SINGLE=0
SELECT=0
USEPAGER=0
LIMIT=0
VERBOSE=0
while getopts 'sin:pv' opt
do
    case "$opt" in
        i)
            SELECT=1
            ;;
        n)
            LIMIT="$OPTARG"
            ;;
        p)
            USEPAGER=1
            ;;
        s)
            SINGLE=1
            ;;
	v)
	    VERBOSE=1
	    ;;
        *)
            usage
            ;;
    esac
done
shift "$(($OPTIND -1))"

TAIL=''
if [ $LIMIT -gt 0 ]
then
    TAIL=" | head -n $LIMIT"
fi
if [ $USEPAGER = 1 ]
then
    TAIL="$TAIL | ${PAGER:-less}"
fi

EXPR='fp://node[@cat="mwu" and node[@pt="spec"] and not(@his="normal") and not(@his_1="decap")]'
if [ $SINGLE = 1 ]
then
    SORT="sed -e 's/\.[0-9][0-9]*$//' | sort | uniq | sed -e 's/\t.*//'"
    TEMPLATE='tt:%w\t%I'
else
    SORT=sort
    TEMPLATE='tt:%w'
fi

search () {
    if [ $VERBOSE = 1 ]
    then
        echo "alto $@ '$EXPR' '$TEMPLATE' | $SORT | uniq -c | sort -nr 2> /dev/null$TAIL"
    fi
    eval "alto $@ '$EXPR' '$TEMPLATE' | $SORT | uniq -c | sort -nr 2> /dev/null$TAIL"
}

for i in "$@"
do
    case "$i" in
        *.xml|*.dact|*.dbxml|*.data.dz|*.index|*.zip)
            ;;
        *)
            if [ ! -d "$i" ]
            then
                usage
            fi
            ;;
    esac
done

if [ $# == 0 -a $SELECT = 1 ]
then
    cd /net/corpora/nlnieuws
    select i in `find . -name '*data.dz' | sort`
    do
        search $i
    done
elif [ $# -gt 0 -a $SELECT = 0 ]
then
    search "$@"
else
    usage
fi