[LONG] UNIX shell script for word lookup

[Robin Lee Powell <rlpowell@csclub.uwaterloo.ca>]

Here's a little script I wrote that looks up lojban words in the cmavo
and gismu files, and makes the output nice and pretty and readable.  I'm
quite proud of it; feedback welcomed.  :-)

Note that it assumes that the wordlists are gzip'd: the first call to
grep is actually to zgrep.  If you have the space available to keep them
in plain text, remove the 'z' at the beginning of both lines that start
with 'z$GREP'.

Edit the defines at the top as appropriate.  If you don't have any of
the programs used, fmt in particular, you're on your own.  I call it
'll', but you can call it whatever you want.  Just don't call me
Shirley.  It's a ksh script, but is probably sh compatible; most of my
scripts are.

-Robin


#!/usr/bin/ksh
##############
# ll = Lojban Lookup
##############

GISMU=~/lojban/gismu
CMAVO=~/lojban/cmavo
TEMP=/tmp/ll.1
WORD="[] 	.,'\"()[]"	#Word boundary pattern
GREP="grep"

USAGE="
$0 [-a] [-n] [-i] [-g] [-c] <word>

    $0 looks through the gismu and cmavo word files for the given word,
in that order, as described below.  The gismu and cmavo entries can be
told apart by the (always uppercase) selma'o name in the rafsi columns
(in effect, column 2),

    By default, $0 looks for <word> as either the exact word in the
first column (i.e. the exact lojbanic word) or an exact word in the
rafsi, gloss or mnemonic columns.  IOW, $0 by default only looks in the
first line of the paragraphs it outputs (this was made a reasonable
proposition by the discover that neither file contains the character
'$'.)

    -a:	Find the word anywhere, but still an exact word match. 

    -n: No space checking, find anywhere at all.  'mil' finds 'milti'.
Note that, keeping shell quoting issues in mind, an arbitrary grep
expression can be passed to -n with the expected results.
    
    -i: Case insensitive.

    -c: cmavo only.
    
    -g: gismu only.

    -n overrides -a.  Using both -g and -c is a NOP. The difference
between the arguments is well exemplified by running '$0 coi', '$0 -a
coi', '$0 -ai coi', '$0 -n coi'.
"

while getopts anigc c
do
    case $c in
	a)	ANY="ANY";;
	n)	NS="NS";;
	i)	GREP="$GREP -i";;
	g)	GONLY="g";;
	c)	CONLY="c";;
	\?)     echo "$USAGE"
		exit 2;;
    esac
done
shift `expr $OPTIND - 1`

if [ $# != 1 ]
then
    echo "$USAGE"
    exit 2
fi

pat=$1

touch $TEMP

if [ -n "$GONLY" -o -z "$CONLY" ]
then
    z$GREP "$pat" $GISMU >> $TEMP
fi
if [ -n "$CONLY" -o -z "$GONLY" ]
then
    z$GREP "$pat" $CMAVO >> $TEMP
fi

cat $TEMP | $GREP -v '^ [0-9]* Lojban' | \
nawk '{ printf("$%s$%20s%s$$%20s%s\n", \
substr($0, 1, 62), "", substr($0, 63, 96), "", substr( $0, 170)); }' | \
cat >$TEMP.2

if [ "$NS" ]
then
    ANY=""
    cat $TEMP.2 | \
    tr '$' '\n' | fmt > $TEMP
fi

if [ "$ANY" ]
then
    cat $TEMP.2 | \
    $GREP "$WORD$pat$WORD" | \
    tr '$' '\n' | fmt > $TEMP
fi

if [ -z "$ANY" -a -z "$NS" ]
then
    cat $TEMP.2 | \
    $GREP "^\\$[^\$]*$WORD$pat$WORD[^\$]*\\$" | \
    tr '$' '\n' | fmt > $TEMP
fi

num=`cat $TEMP | wc -l`
if [ $num -gt 10 ]
then
    less $TEMP
else
    cat $TEMP
fi
rm $TEMP