bfilter/trunk/bfilter.pl

#!/usr/bin/perl -w
#

use strict;
use locale;

# maximum entries
my $max = 0;
# minimum letters to search by
my $min_len = shift @ARGV;
$min_len = 3 unless defined($min_len);
# if more than x elements, warn to increase min_len
my $increase_at = 500;

# name of generated index
my $headlines = 'headlines';

my $debug = 1;

sub print_file {
        my $f = shift || return;
        open(F, $f) || die "$f: $!";
        while(<F>) {
                print;
        }
        close(F);
}

print qq{
var $headlines = new Object();
};

my @part_arr;
my $last_part = '';
my $total = 0;

my $max_elements = 0;

sub escape_js {
        my $t = shift || return 'undef';
        # escape single quote and backspace
        $t =~ s/(['\\])/\\$1/g && print STDERR "ESCAPED '$t'\n";
        # quote string if not number
        $t = "'$t'" unless ($t =~ m/^\d+$/);
        return $t;
}

while(<STDIN>) {
        chomp;

        if (!m/\t/ || m/\t$/) {
                print STDERR "SKIP '$_': no tab\n";
                next;
        }

        my @data = split(/\t+/,$_);

        my $headline = shift @data || die "need at least headline!";

        if (length($headline) < $min_len) {
                print STDERR "SKIP '$_': too short\n";
                next;
        }


        # split into min_len part and rest
        my ($part,$rest) = ( substr($headline,0,$min_len), substr($headline,$min_len) );

        # make part lowercase
        $part = lc($part);

        $last_part = $part if (! $last_part);

        # new part?
        if ($part ne $last_part) {
                print STDERR $last_part,"\t",$#part_arr+1,"\n" if ($debug && $#part_arr > $increase_at);
                $max_elements = $#part_arr if ($#part_arr > $max_elements);
                print "${headlines}[",escape_js($last_part),"] = [\n ",join(",\n ",@part_arr),"];\n" if (@part_arr);
                $total += $#part_arr;
                @part_arr = ();
                $last_part = $part;
        }
        push @part_arr, "[".escape_js($headline).",".join(",",map { escape_js($_) } @data)."]";

        # break out?
        last if ($max && $total > $max);
}

print "${headlines}[",escape_js($last_part)."] = [\n ",join(",\n ",@part_arr),"];\n" if (@part_arr);
print qq{

${headlines}.min_len = $min_len;
${headlines}.length = $total;

};

print STDERR "You have more than $increase_at elements, so you should\nincrease min_len to ",$min_len+1," or higher for performance benefit.\n" if ($max_elements > $increase_at);
1	#!/usr/bin/perl -w
2	#
3
4	use strict;
5	use locale;
6
7	# maximum entries
8	my $max = 0;
9	# minimum letters to search by
10	my $min_len = shift @ARGV;
11	$min_len = 3 unless defined($min_len);
12	# if more than x elements, warn to increase min_len
13	my $increase_at = 500;
14
15	# name of generated index
16	my $headlines = 'headlines';
17
18	my $debug = 1;
19
20	sub print_file {
21	my $f = shift \|\| return;
22	open(F, $f) \|\| die "$f: $!";
23	while(<F>) {
24	print;
25	}
26	close(F);
27	}
28
29	print qq{
30	var $headlines = new Object();
31	};
32
33	my @part_arr;
34	my $last_part = '';
35	my $total = 0;
36
37	my $max_elements = 0;
38
39	sub escape_js {
40	my $t = shift \|\| return 'undef';
41	# escape single quote and backspace
42	$t =~ s/(['\\])/\\$1/g && print STDERR "ESCAPED '$t'\n";
43	# quote string if not number
44	$t = "'$t'" unless ($t =~ m/^\d+$/);
45	return $t;
46	}
47
48	while(<STDIN>) {
49	chomp;
50
51	if (!m/\t/ \|\| m/\t$/) {
52	print STDERR "SKIP '$_': no tab\n";
53	next;
54	}
55
56	my @data = split(/\t+/,$_);
57
58	my $headline = shift @data \|\| die "need at least headline!";
59
60	if (length($headline) < $min_len) {
61	print STDERR "SKIP '$_': too short\n";
62	next;
63	}
64
65
66	# split into min_len part and rest
67	my ($part,$rest) = ( substr($headline,0,$min_len), substr($headline,$min_len) );
68
69	# make part lowercase
70	$part = lc($part);
71
72	$last_part = $part if (! $last_part);
73
74	# new part?
75	if ($part ne $last_part) {
76	print STDERR $last_part,"\t",$#part_arr+1,"\n" if ($debug && $#part_arr > $increase_at);
77	$max_elements = $#part_arr if ($#part_arr > $max_elements);
78	print "${headlines}[",escape_js($last_part),"] = [\n ",join(",\n ",@part_arr),"];\n" if (@part_arr);
79	$total += $#part_arr;
80	@part_arr = ();
81	$last_part = $part;
82	}
83	push @part_arr, "[".escape_js($headline).",".join(",",map { escape_js($_) } @data)."]";
84
85	# break out?
86	last if ($max && $total > $max);
87	}
88
89	print "${headlines}[",escape_js($last_part)."] = [\n ",join(",\n ",@part_arr),"];\n" if (@part_arr);
90	print qq{
91
92	${headlines}.min_len = $min_len;
93	${headlines}.length = $total;
94
95	};
96
97	print STDERR "You have more than $increase_at elements, so you should\nincrease min_len to ",$min_len+1," or higher for performance benefit.\n" if ($max_elements > $increase_at);