#
# reads a tin filter file with wildmat filters on STDIN, converts it to
# regexp filters and returns it on STDOUT
#
# 2020-11-10 <[email protected]>
#
# NOTE: don't use w2r.pl on regexp filters
#
# for case optimization of your regexp filters use opt-case.pl, i.e.:
# w2r.pl < wildmat-filter-file | opt-case.pl > regexp-filter-file
#
# for joining regexp filters with the same group= and score= use
# joinf.pl (not written yet)

# perl 5 is needed for lookahead assertions and perl < 5.004 is know
# to be buggy
require 5.004;

# version Number
# $VERSION = "0.2.8";

while (defined($line = <>)) {
    chomp $line;

    # ignore comments etc.
    if ($line =~ m/^(?:[#\s]|$)/o) {
        print "$line\n";
        next;
    }

    # skip 'empty' patterns, they are nonsense
    next if ($line =~ m/^[^=]+=$/o);

    # lines which needs to be translated
    if ($line =~ m/^(subj|from|msgid(?:|_last|_only)|refs_only|xref|path)=(.*)$/o) {
        printf ("$1=%s\n", w2p($2));
        next;
    }

    # other lines don't need to be translated
    print "$line\n";
}


# turns a wildmat into a regexp
sub w2p {
    local ($wild)  = @_;    # input line
    my $cchar = "";        # current char
    my $lchar = "";        # last char
    my $reg = "";        # translated char
    $bmode = 0;        # inside [] ?
    $rval = "";        # output line

    # break line into chars
    while ($wild =~ s/(.)//) {
        $cchar = $1;

        # if char is a [, and we aren't allreay in []
        if ($lchar !~ m/\\/o && $cchar =~ m/\[/o) {
            $bmode++;
            $reg = $cchar;
        }

        # if char is a ], and we were in []
        if ($lchar !~ m/\\/o && $cchar =~ m/\]/o) {
            $bmode--;
            $reg = $cchar;
        }

        # usual cases
        if ($bmode == 0 && $lchar !~ m/\\/o) {
            $reg = $cchar;
            $reg =~ s/\t/\\t/o;    # translate tabs
            $reg =~ s/\./\\./o;    # quote .
            $reg =~ s/\)/\\)/o;    # quote )
            $reg =~ s/\(/\\(/o;    # quote (
            $reg =~ s/\*/\.*/o;    # translate *
            $reg =~ s/\?/\./o;    # translate ?
            $reg =~ s/\^/\\^/o;    # quote ^
            $reg =~ s/\$/\\\$/o;    # quote $
        }

        # if last char was a qute, current char can't be a meta
        if ($lchar =~ m/\\/o || $bmode != 0) {
            $reg = $cchar;
            $cchar =~ s/\\//o;    # skip 2nd \\ inside []
        }

        $lchar = $cchar;    # store last char
        $rval = $rval.$reg;    # build return string
    }

    # common abbreviations
    # TODO: make this global
    # replace [0-9] with [\d] in the first []
    # replace [a-zA-Z0-9_] with [\w] in the first []
    # replace [a-zA-Z0-9] with [^\W_] in the first []
    # replace [a-zA-Z] with [^\W\d_] in the first []
    $rval =~ s/^([^\[]*)\[0-9\]/$1\[\\d\]/o;
    $rval =~ s/([^\[]*)\[a-za-z0-9_\]/$1\[\\w\]/io;
    $rval =~ s/([^\[]*)\[a-za-z0-9\]/$1\[^\\W_\]/io;
    $rval =~ s/([^\[]*)\[a-za-z\]/$1\[^\\W\\d_\]/io;

    # optimizations
    #
    # add ^-anchor if needed
    $rval =~ s/^(?!\.\*)(.*)/\^$1/o;
    # add $-anchor if needed
    $rval =~ s/^((?:.*)(?:[^.][^*]))$/$1\$/o;
    # remove leading .* if allowed
    $rval =~ s/^\.\*(?!$)//o;
    # remove tailing .* if allowed
    $rval =~ s/(.+)\.\*$/$1/o;

    return $rval;
}

__END__

=head1 NAME

w2r.pl - Convert tin wildmat filters to tin regexp filters

=head1 SYNOPSIS

B<w2r.pl> E<lt> I<input> [E<gt> I<output>]

=head1 DESCRIPTION

B<w2r.pl> reads a L<tin(1)> filter file with wildmat filters on STDIN,
converts it to regexp filters and returns it on STDOUT.

=head1 NOTES

Don't use B<w2r.pl> on regexp filter files

=head1 AUTHOR

Urs Janssen E<lt>[email protected]<gt>

=head1 SEE ALSO

L<tin(1)>, L<tin(5)>

=cut