File Coverage

blib/lib/KinoSearch1/Search/Similarity.pm

Criterion	Covered	Total	%
statement	45	45	100.0
branch	7	10	70.0
condition			n/a
subroutine	14	14	100.0
pod	0	3	0.0
total	66	72	91.6

package KinoSearch1::Search::Similarity;
use strict;
use warnings;
use KinoSearch1::Util::ToolSet;
use base qw( KinoSearch1::Util::CClass );

BEGIN { __PACKAGE__->init_instance_vars(); }

# See _float_to_byte.
*encode_norm = *_float_to_byte;
*decode_norm = *_byte_to_float;

# Calculate the Inverse Document Frequecy for one or more Term in a given
# collection (the Searcher represents the collection).
#
# If multiple Terms are supplied, their idfs are summed.
sub idf {
    my ( $self, $term_or_terms, $searcher ) = @_;
    my $max_doc = $searcher->max_doc;
    my $terms
        = ref $term_or_terms eq 'ARRAY' ? $term_or_terms : [$term_or_terms];

    return 1 unless $max_doc;    # guard against log of zero error

    # accumulate IDF
    my $idf = 0;
    for my $term (@$terms) {
        my $doc_freq = $searcher->doc_freq($term);
        $idf += 1 + log( $max_doc / ( 1 + $searcher->doc_freq($term) ) );
    }
    return $idf;
}

# Normalize a Query's weight so that it is comparable to other Queries.
sub query_norm {
    my ( $self, $sum_of_squared_weights ) = @_;
    return 0 if ( $sum_of_squared_weights == 0 );  # guard against div by zero
    return ( 1 / sqrt($sum_of_squared_weights) );
}

# KLUDGE -- see comment at STORABLE_thaw.
sub STORABLE_freeze {
    my ( $self, $cloning ) = @_;
    return if $cloning;
    return "1";
}

package KinoSearch1::Search::TitleSimilarity;
use strict;
use warnings;
use KinoSearch1::Util::ToolSet;
use base qw( KinoSearch1::Search::Similarity );

sub new {
    my $self = shift->SUPER::new(@_);
    $self->_use_title_tf;
    return $self;
}

sub lengthnorm {
    return 0 unless $_[1];
    return 1 / sqrt( $_[1] );
}

1;

__END__