#!/usr/bin/perl -w

# Multispell is a program that lets your editor talk to two spellers
# simultaneously. It launches two speller programs, one Hebrew (hspell),
# one English (ispell), and merges their results.
#
# Public Domain
#
# Please report bugs to: Mooffie <mooffie@typo.co.il>
#
# TODO
# - UTF-8 support (+cmdline opt to select offsets type: byte-based
#   or char-based)
# - intl support. e.g.:
#	multispell -T utf8 --ispell-encoding=koi8-r
#	or:
#	multispell -T utf8 --ispell-cmd="multispell --ispell-cmd=..." ...
# - handle aspell's '$$' ?

use strict;
use Getopt::Long;
use IPC::Open2;
use IO::Handle;

# options to pass to both hspell and ispell:
my @common_opts	    = qw/ --dummy /;
# options to pass to hspell only:
my @hspell_opts	    = qw/ -n --notes --verbose /;
my $hspell_cmd	    = "hspell";
my $ispell_cmd	    = "ispell";
my $remove_niqqud   = 0;

my $debug	    = 0;
my $debug_file	    = "$ENV{HOME}/multispell-dbg.txt";
my $rc_file	    = "$ENV{HOME}/.multispellrc";
my $rc_file_global  = "/etc/multispellrc";

my ($hsplrd, $hsplwr); # pipes to hspell
my ($isplrd, $isplwr); # pipes to ispell

my $VERSION = "0.4a";

sub DBG {
    if ($debug) {
	my $msg = shift;
	print DBGFL $msg;
    }
}

sub help {
    print <<EOM;
SYNOPSIS:
    multispell [multispell-options] [hspell-options] [ispell-options]

DESCRIPTION:
    Multispell is a program that lets your editor talk to two spellers
    simultaneously. It launches two speller programs, one Hebrew (hspell),
    one English (ispell), and merges their results.

    Currently the editor can talk with multispell only in ISO-8859-8
    or CP1255.
    
VALID MULTISPELL-OPTIONS:
    --debug
	    Write debug info to ~/multispell-dbg.txt
    
    --hspell-cmd PROG
	    Use PROG instead of "$hspell_cmd"
    
    --ispell-cmd PROG
	    Use PROG instead of "$ispell_cmd"

    --remove-niqqud
	    Erase words containing CP1255's niqqud (Hebrew points),
	    and convert maqaf (0xCE) to ASCII dash.

    --drop OPTNAME
	    Don't pass option OPTNAME to ispell or to hspell. It's
	    useful in ~/.multispellrc when you want to drop an option
	    the editor is determined to pass. You can specify multiple
	    options, separated by commas. If the option requires an
	    argument, append ':' to its name.

    --help -h
	    Print this message

FILES:
    /etc/multispellrc
	    Additional command-line options to use. These _override_
	    the ones specified on the command-line.

    ~/.multispellrc
	    If found, used _instead_ of /etc/multispellrc

EOM
    exit;
}

sub load_spellers
{
    sub load_speller {
	my ($cmd) = @_;
	local (*RD, *WR);
	DBG("Loading: $cmd\n");
	open2(*RD, *WR, $cmd);
	my $signature = <RD>;
	if (!$signature || $signature !~ /^@\(#\)/) {
	    DBG("Loading failed (please check stderr)\n");
	    die "Can't load $cmd";
	} else {
	    DBG("Loaded OK: $signature");
	}
	autoflush WR 1;
	return (*RD, *WR);
    }

    ($hsplrd, $hsplwr) = load_speller($hspell_cmd);
    ($isplrd, $isplwr) = load_speller($ispell_cmd);
}

sub print_signature {
    print "@(#) International Ispell Version 3.1.20 ".
	    "(but really multispell $VERSION)\n";
}

# do_loop() is the crux of the program. it reads ispell-a requests
# from stdin, sends them to ispell and/or hspell, and writes the merged
# replies to stdout.

sub do_loop {
    my $heb_chars	 = '\xE0-\xFA';
    my $niqqud_chars	 = '\xC0-\xCD\xCF';
    my $niqqud_heb_chars = $heb_chars . $niqqud_chars;

    print_signature();

    while (my $line = <STDIN>) {
	DBG(">$line");
	my @replies = ();

	# if the last line of the input does not end in newline, add it ourselves
	# so that hspell doesn't hang waiting for it.
	# e.g.: echo -n word | multispell -a
	$line .= "\n" if $line !~ /\n/;

	# a workaround for some versions of Aspell that ignore empty input lines.
	# see http://ivrix.org.il/bugzilla/show_bug.cgi?id=3
	$line = " \n" if $line eq "\n";

	if ($remove_niqqud) {
	    # erase words containing CP1255's niqqud (hebrew points)
	    $line =~ s/([$niqqud_heb_chars]+[$niqqud_chars][$niqqud_heb_chars]*)/
			    ' ' x length($1)/oeg;
	    # convert maqaf to ASCII dash
	    $line =~ tr/\xCE/-/;
	}
	
	if ($line =~ /^[#!~@%+&*-]/) {
	    if ($line =~ /^[@&*]/) {
		# "@word" - ignore word.
		# "*word" - add word to private dict.
		# "&word" - ditto.
		# We send hebrew words to hspell, otherwise to ispell
		if ($line =~ /[$heb_chars]/o
			|| $line eq "\@ActivateExtendedProtocol\n") {
		    DBG("HSPL>$line");
		    print $hsplwr $line;
		} else {
		    DBG("ISPL>$line");
		    print $isplwr $line;
		}
	    } else {
		DBG("IHSPL>$line");
		print $hsplwr $line;
		print $isplwr $line;
	    }
	    # this is a command, so we don't read replies.
	} else {
	    # words to spell-check.
	    DBG("HSPL>$line");
	    print $hsplwr $line;
	    while (<$hsplrd>) {
		DBG("HSPL<$_");
		last if ! /\S/;
		push @replies, $_;
	    }
	   
	    # delete all heb words before sending to ispell
	    $line =~ s/[$heb_chars]/ /og;
	    DBG("ISPL>$line");
	    print $isplwr $line;
	    while (<$isplrd>) {
		DBG("ISPL<$_");
		last if ! /\S/;
		push @replies, $_;
	    }

	    # We're about to sort the replies.
	    # But first we need to take care of non-standard replies (like
	    # spelling-hints). Such replies (so we believe) add information to
	    # the last standard reply (e.g. spelling-hints explain the last
	    # reported misspelled word), so we want to concatenate them with
	    # that reply instead of sorting them as if they were independent.
	    for (my $i = 1; $i <= $#replies; $i++) {
		if ($replies[$i] =~ /^[^*+&?#-]/) { # non-standard?
		    $replies[$i-1] .= $replies[$i];
		    splice @replies, $i, 1;
		    $i--;
		}
	    }

	    # sort the replies
	    sub getidx {
		$_ = shift;
		(/^[&?] \S+ \d+ (\d+)/ || /^# \S+ (\d+)/) ? $1 : 0;
	    }
	    @replies = sort { getidx($a) <=> getidx($b) } @replies;

	    push @replies, "\n";
	    DBG("<$_") foreach (@replies);
	    print @replies;
	}
    }
}

# extract_options() takes a list of option names and extracts them and
# their values from @ARGV.
# For example, if we pass ('-n', '-d:', '-S') to this function, and @ARGV
# is ('-i', '-nx', '-denglish', '-p'), it returns "-n -d english", and
# @ARGV becomes ('-i', '-x', '-p').

sub extract_options {
    my @options = @_;
    my %getopt_hash;
    my $args = "";

    foreach my $option (@options) {
	$option =~ tr/-//d;
	my $require_arg = ($option =~ s/:$/=s/);
	$getopt_hash{$option} = sub {
	    my ($optname, $optval) = @_;
	    $optname = (length($optname) > 1 ? "--" : "-") . $optname;
	    $args .= " $optname";
	    $args .= " '$optval'" if $require_arg;
	};
    }
    GetOptions(%getopt_hash);
    return $args;
}

# tokenize_cmd_line() is used for parsing the RC files.
# it takes a command line string, e.g.:
#   -x --opt="one two" -i 'on"e" two'
# and returns an array:
#   ('-x', '--opt=one two', '-i', 'on"e" two')

sub tokenize_cmd_line {
    my $s = shift;
    return  map {
		s/['"]//;
		s/['"]$//;
		$_;
	    } grep {
		/\S/
	    } split( /(
		[^\s'"]*(?:'.*?'|".*?") |
		[^\s'"]+
	    )/x , $s);
}

main: {
    $| = 1;
    Getopt::Long::Configure("bundling_override", "pass_through");
    my $print_signature	= 0;
    my $ispella_mode	= 0;
    my $dictionary	= "";
    my @drop_opts	= ();

    # Applications use the '-d' option to select a dictionary.
    # We don't want aspell to consider the locale when '-d' is not specified,
    # because in most cases that's not what the user wants (e.g. he_IL), and,
    # anyway, we can't represent most languages other than English in
    # ISO-8859-8.
    $ENV{'LC_ALL'} = 'C';

    if (open(RC, $rc_file) || open(RC, $rc_file_global)) {
	while (<RC>) {
	    s/#.*//;
	    push @ARGV, tokenize_cmd_line($_);
	}
	close(RC);
    }

    my @saved_argv = @ARGV;
    GetOptions('debug'		=> \$debug,
	       'hspell-cmd=s'	=> \$hspell_cmd,
	       'ispell-cmd=s'	=> \$ispell_cmd,
	       'drop=s'		=> \@drop_opts,
	       'remove-niqqud'	=> \$remove_niqqud,
	       'remove-niqud'	=> \$remove_niqqud,
	       'remove-nikud'	=> \$remove_niqqud,
	       'd=s'		=> \$dictionary,
	       'v'		=> \$print_signature,
	       'a'		=> \$ispella_mode,
	       'i'		=> sub { ; }, # silently eat up hspell's '-i'
	       'help|h|?'	=> sub { help(); }
	      ) or help();

    if ($debug) {
	open(DBGFL, ">>$debug_file");
	autoflush DBGFL 1;
	DBG("\n\n" . "-" x 75 . "\n");
	DBG("I was invoked: $0");
	DBG(" '$_'") for @saved_argv;
	DBG("\n");
	DBG("On: " . localtime() . "\n");
	# According to SUSV, the following 'ps' command is portable,
	# but I don't want to take risks, so I run it only on linux.
	if ($^O =~ /linux/i) {
	    my $PPID = getppid();
	    DBG("By: " . `ps -p $PPID -o args=`);
	}
    }

    if ($print_signature) {
	print_signature();
	exit;
    }

    if (!$ispella_mode) {
	die "You can only use this speller in \"pipe-mode\" " .
	    "(using the '-a' option)\n";
    }
    
    # If $dictionary contains "hebrew", it's probably something
    # kspell or emacs passed. We ignore it since ispell knows
    # no such dictionary.
    if ($dictionary && $dictionary !~ /hebrew/i) {
	push @ARGV, '-d', $dictionary;
    }

    @drop_opts = split(/,/, join(",", @drop_opts));
    extract_options(@drop_opts) if @drop_opts;
    my $common_opts = extract_options(@common_opts);
    my $hspell_opts = extract_options(@hspell_opts) . $common_opts;
    my $ispell_opts = "@ARGV" . $common_opts;

    $hspell_cmd .= " -a $hspell_opts";
    $ispell_cmd .= " -a $ispell_opts";
    
    load_spellers();
    DBG("\n");
    do_loop();
}

# vim:ts=8:sts=4:sw=4: