cwb-make-safe

#! /bin/sh

# cwb-make-safe
#
# A wrapper script for cwb-make provided in the CWB Perl package
# working around a bug in cwb-huffcode that causes an error if a
# positional attribute has only a single value in the corpus.
#
# Run "cwb-make-safe -h" to get a usage message.

# TODO:
# - Use cwb-make also for single-value attributes if available and if
#   CWB is recent enough to have been fixed in that respect (version
#   3.4.11 or later).


progname=`basename $0`
progdir=`dirname $0`
cmdline="$@"

shortopts="r:M:Vg:p:Dh"
longopts="registry:,memory:,validate,group:,permissions:,debug,help"

. $progdir/korp-lib.sh

registry_opt=
cwb_make_opts=
cwb_makeall_opts=
group=
perms=
corpus=
attrs=


find_prog_cwb () {
    if ! find_prog cwb-$1 $CWB_BINDIR $default_cwb_bindirs; then
	$2 "Program cwb-$1 not found on PATH or in directories $(echo $CWB_BINDIR $default_cwb_bindirs | sed -e 's/ \+/, /g')"
    fi
}

# Use set -e to propagate the possible error exit from $(find_prog_cwb ...)
set -e
cwb_describe_corpus=$(find_prog_cwb describe-corpus error)
cwb_makeall=$(find_prog_cwb makeall error)
cwb_huffcode=$(find_prog_cwb huffcode error)
cwb_compress_rdx=$(find_prog_cwb compress-rdx error)
set +e
cwb_make=$(find_prog_cwb make warn)

# Try to find the library directory for CWB Perl based on the location
# of cwb-make (which needs it): replace trailing /bin with
# /lib/site_perl.
# FIXME: The library might not always be in that default location.
# What if it is not found? Should we use find to try to find CWB.pm?
cwb_perl_libdir=${cwb_make%/bin/*}/lib/site_perl
if [ -d $cwb_perl_libdir ]; then
    if [ "x$PERL5LIB" != x ]; then
	PERL5LIB=$PERL5LIB:$cwb_perl_libdir
    else
	PERL5LIB=$cwb_perl_libdir
    fi
fi
export PERL5LIB


usage () {
    cat <<EOF
Usage: $progname [options] corpus [attributes]

Index and compress the attributes of an encoded CWB corpus.

This is a wrapper script for cwb-make provided in the CWB Perl package
working around a bug in cwb-huffcode that causes an error if a
positional attribute has only a single value in the corpus.

The options are the same as for cwb-make:
EOF
    if [ "x$cwb_make" = x ]; then
	echo "cwb-make not found; cannot list options"
    else
	$cwb_make -h 2>&1 |
	awk '/^Options:/ {p = 1; next} p && /./ {print}'
    fi
cat <<EOF

However, please note that the registry directory name may not contain
whitespace characters.
EOF
    exit 0
}

# Process options
while [ "x$1" != "x" ] ; do
    case "$1" in
	-h | --help )
	    usage
	    ;;
	-r | --registry )
	    shift
	    cwb_makeall_opts="$cwb_makeall_opts -r $1"
	    registry_opt="-r $1"
	    ;;
	-M | --memory )
	    shift
	    cwb_makeall_opts="$cwb_makeall_opts -M $1"
	    ;;
	-V | --validate )
	    cwb_makeall_opts="$cwb_makeall_opts -V"
	    ;;
	-g | --group )
	    shift
	    group=$1
	    cwb_make_opts="$cwb_make_opts -g $1"
	    ;;
	-p | --permissions )
	    shift
	    perms=$1
	    cwb_make_opts="$cwb_make_opts -p $1"
	    ;;
	-D | --debug )
	    cwb_makeall_opts="$cwb_makeall_opts -D"
	    ;;
	-- )
	    shift
	    break
	    ;;
	--* )
	    warn "Unrecognized option: $1"
	    ;;
	* )
	    break
	    ;;
    esac
    shift
done

if [ "x$1" = x ]; then
    error "No corpus specified"
fi
corpus=$1
shift
attrs=$@

cwb_make_opts="$cwb_makeall_opts $cwb_make_opts"


get_single_valued_pos_attrs () {
    opt=
    case $1 in
	-* )
	    opt=$1
	    shift
	    ;;
    esac
    $cwb_describe_corpus $registry_opt -s $corpus |
    grep '^p-ATT' |
    grep $opt '.*\b1 types' |
    awk '{print $2}' |
    if [ "x$1" != x ]; then
	grep -F "$(echo $@ | tr ' ' '\n')"
    else
	cat
    fi
}

get_corpus_data_dir () {
    $cwb_describe_corpus $registry_opt $1 |
    grep -E '^home directory: *' |
    sed -e 's/home directory: *//'
}

cwb_make_manual () {
    datadir=$(get_corpus_data_dir $corpus)
    datafiles_patt=
    make_attrs=
    for attr in "$@"; do
	if [ ! -e $datadir/$attr.corpus.cnt ] ||
	    [ ! -e $datadir/$attr.lexicon.srt ] ||
	    [ ! -e $datadir/$attr.huf ] ||
	    [ ! -e $datadir/$attr.crx ] ||
	    [ $datadir/$attr.corpus -nt $datadir/$attr.corpus.cnt ] ||
	    [ $datadir/$attr.lexicon -nt $datadir/$attr.lexicon.srt ] ||
	    [ $datadir/$attr.corpus -nt $datadir/$attr.huf ] ||
	    [ $datadir/$attr.corpus.rev -nt $datadir/$attr.crx ];
	then
	    make_attrs="$make_attrs $attr"
	    rm_files=
	    for suf in corpus.cnt corpus.rdx corpus.rev lexicon.srt \
		hcd huf syn crc crx;
	    do
		rm_files="$rm_files $datadir/$attr.$suf"
	    done
	    rm -f $rm_files
	    datafiles_patt="$datafiles $datadir/$attr.*"
	fi
    done
    if [ "x$make_attrs" != x ]; then
	$cwb_makeall $cwb_makeall_opts $corpus $make_attrs
	# echo to convert newlines to spaces
	single_valued=$(echo $single_valued_attrs)
	for attr in $make_attrs; do
	    if ! word_in $attr $single_valued_attrs; then
		$cwb_huffcode -P $attr $corpus
		if [ $? = 0 ]; then
		    rm $datadir/$attr.corpus
		fi
		$cwb_compress_rdx -P $attr $corpus
		if [ $? = 0 ]; then
		    rm $datadir/$attr.corpus.rev $datadir/$attr.corpus.rdx
		fi
	    fi
	done
	datafiles=$(ls $datafiles_patt)
	if [ "x$perms" != x ]; then
	    chmod $perms $datafiles
	fi
	if [ "x$group" != x ]; then
	    chgrp $group $datafiles
	fi
    fi
}


single_valued_attrs=$(get_single_valued_pos_attrs $attrs)

if [ "x$cwb_make" != x ]; then
    if [ "x$single_valued_attrs" = x ]; then
	$cwb_make $cwb_make_opts $corpus $attrs
    else
	non_single_valued_attrs=$(get_single_valued_pos_attrs -v $attrs)
	if [ "x$non_single_valued_attrs" != x ]; then
	    $cwb_make $cwb_make_opts $corpus $non_single_valued_attrs
	fi
	cwb_make_manual $single_valued_attrs
    fi
else
    warn "cwb-make not found: using cwb-makeall for all attributes"
    if [ "x$attrs" = x ]; then
	attrs=$(corpus_list_attrs $corpus pos)
    fi
    cwb_make_manual $attrs
fi