Koha/C4/Matcher.pm

package C4::Matcher;

# Copyright (C) 2007 LibLime
#
# This file is part of Koha.
#
# Koha is free software; you can redistribute it and/or modify it under the
# terms of the GNU General Public License as published by the Free Software
# Foundation; either version 2 of the License, or (at your option) any later
# version.
#
# Koha is distributed in the hope that it will be useful, but WITHOUT ANY
# WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR
# A PARTICULAR PURPOSE.  See the GNU General Public License for more details.
#
# You should have received a copy of the GNU General Public License along with
# Koha; if not, write to the Free Software Foundation, Inc., 59 Temple Place,
# Suite 330, Boston, MA  02111-1307 USA

use strict;
use C4::Context;
use MARC::Record;
use C4::Search;
use C4::Biblio;

use vars qw($VERSION);

BEGIN {
	# set the version for version checking
	$VERSION = 3.01;
}

=head1 NAME

C4::Matcher - find MARC records matching another one

=head1 SYNOPSIS

=over 4

my @matchers = C4::Matcher::GetMatcherList();

my $matcher = C4::Matcher->new($record_type);
$matcher->threshold($threshold);
$matcher->code($code);
$matcher->description($description);

$matcher->add_simple_matchpoint('isbn', 1000, '020', 'a', -1, 0, '');
$matcher->add_simple_matchpoint('Date', 1000, '008', '', 7, 4, '');
$matcher->add_matchpoint('isbn', 1000, [ { tag => '020', subfields => 'a', norms => [] } ]);

$matcher->add_simple_required_check('245', 'a', -1, 0, '', '245', 'a', -1, 0, '');
$matcher->add_required_check([ { tag => '245', subfields => 'a', norms => [] } ], 
                             [ { tag => '245', subfields => 'a', norms => [] } ]);

my @matches = $matcher->get_matches($marc_record, $max_matches);

foreach $match (@matches) {

    # matches already sorted in order of
    # decreasing score
    print "record ID: $match->{'record_id'};
    print "score:     $match->{'score'};

}

my $matcher_description = $matcher->dump();

=back

=head1 FUNCTIONS

=cut

=head2 GetMatcherList

=over 4

my @matchers = C4::Matcher::GetMatcherList();

=back

Returns an array of hashrefs list all matchers
present in the database.  Each hashref includes:

matcher_id
code
description

=cut

sub GetMatcherList {
    my $dbh = C4::Context->dbh;
    
    my $sth = $dbh->prepare_cached("SELECT matcher_id, code, description FROM marc_matchers ORDER BY matcher_id");
    $sth->execute();
    my @results = ();
    while (my $row = $sth->fetchrow_hashref) {
        push @results, $row;
    } 
    return @results;
}

=head1 METHODS

=cut

=head2 new

=over 4

my $matcher = C4::Matcher->new($record_type, $threshold);

=back

Creates a new Matcher.  C<$record_type> indicates which search
database to use, e.g., 'biblio' or 'authority' and defaults to
'biblio', while C<$threshold> is the minimum score required for a match
and defaults to 1000.

=cut

sub new {
    my $class = shift;
    my $self = {};

    $self->{'id'} = undef;

    if ($#_ > -1) {
        $self->{'record_type'} = shift;
    } else {
        $self->{'record_type'} = 'biblio';
    }

    if ($#_ > -1) {
        $self->{'threshold'} = shift;
    } else {
        $self->{'threshold'} = 1000;
    }

    $self->{'code'} = '';
    $self->{'description'} = '';

    $self->{'matchpoints'} = [];
    $self->{'required_checks'} = [];

    bless $self, $class;
    return $self;
}

=head2 fetch

=over 4

my $matcher = C4::Matcher->fetch($id);

=back

Creates a matcher object from the version stored
in the database.  If a matcher with the given
id does not exist, returns undef.

=cut

sub fetch {
    my $class = shift;
    my $id = shift;
    my $dbh = C4::Context->dbh();

    my $sth = $dbh->prepare_cached("SELECT * FROM marc_matchers WHERE matcher_id = ?");
    $sth->execute($id);
    my $row = $sth->fetchrow_hashref;
    $sth->finish();
    return undef unless defined $row;

    my $self = {};
    $self->{'id'} = $row->{'matcher_id'};
    $self->{'record_type'} = $row->{'record_type'};
    $self->{'code'} = $row->{'code'};
    $self->{'description'} = $row->{'description'};
    $self->{'threshold'} = int($row->{'threshold'});
    bless $self, $class;

    # matchpoints
    $self->{'matchpoints'} = [];
    $sth = $dbh->prepare_cached("SELECT * FROM matcher_matchpoints WHERE matcher_id = ? ORDER BY matchpoint_id");
    $sth->execute($self->{'id'});
    while (my $row = $sth->fetchrow_hashref) {
        my $matchpoint = $self->_fetch_matchpoint($row->{'matchpoint_id'});
        push @{ $self->{'matchpoints'} }, $matchpoint;
    }

    # required checks
    $self->{'required_checks'} = [];
    $sth = $dbh->prepare_cached("SELECT * FROM matchchecks WHERE matcher_id = ? ORDER BY matchcheck_id");
    $sth->execute($self->{'id'});
    while (my $row = $sth->fetchrow_hashref) {
        my $source_matchpoint = $self->_fetch_matchpoint($row->{'source_matchpoint_id'});
        my $target_matchpoint = $self->_fetch_matchpoint($row->{'target_matchpoint_id'});
        my $matchcheck = {};
        $matchcheck->{'source_matchpoint'} = $source_matchpoint;
        $matchcheck->{'target_matchpoint'} = $target_matchpoint;
        push @{ $self->{'required_checks'} }, $matchcheck;
    }

    return $self;
}

sub _fetch_matchpoint {
    my $self = shift;
    my $matchpoint_id = shift;
    
    my $dbh = C4::Context->dbh;
    my $sth = $dbh->prepare_cached("SELECT * FROM matchpoints WHERE matchpoint_id = ?");
    $sth->execute($matchpoint_id);
    my $row = $sth->fetchrow_hashref;
    my $matchpoint = {};
    $matchpoint->{'index'} = $row->{'search_index'};
    $matchpoint->{'score'} = int($row->{'score'});
    $sth->finish();

    $matchpoint->{'components'} = [];
    $sth = $dbh->prepare_cached("SELECT * FROM matchpoint_components WHERE matchpoint_id = ? ORDER BY sequence");
    $sth->execute($matchpoint_id);
    while ($row = $sth->fetchrow_hashref) {
        my $component = {};
        $component->{'tag'} = $row->{'tag'};
        $component->{'subfields'} = { map { $_ => 1 } split(//,  $row->{'subfields'}) };
        $component->{'offset'} = int($row->{'offset'});
        $component->{'length'} = int($row->{'length'});
        $component->{'norms'} = [];
        my $sth2 = $dbh->prepare_cached("SELECT * 
                                         FROM matchpoint_component_norms 
                                         WHERE matchpoint_component_id = ? ORDER BY sequence");
        $sth2->execute($row->{'matchpoint_component_id'});
        while (my $row2 = $sth2->fetchrow_hashref) {
            push @{ $component->{'norms'} }, $row2->{'norm_routine'};
        }
        push @{ $matchpoint->{'components'} }, $component;
    }
    return $matchpoint;
}

=head2 store

=over 4

my $id = $matcher->store();

=back

Stores matcher in database.  The return value is the ID 
of the marc_matchers row.  If the matcher was 
previously retrieved from the database via the fetch()
method, the DB representation of the matcher
is replaced.

=cut

sub store {
    my $self = shift;

    if (defined $self->{'id'}) {
        # update
        $self->_del_matcher_components();
        $self->_update_marc_matchers();
    } else {
        # create new
        $self->_new_marc_matchers();
    }
    $self->_store_matcher_components();
    return $self->{'id'};
}

sub _del_matcher_components {
    my $self = shift;

    my $dbh = C4::Context->dbh();
    my $sth = $dbh->prepare_cached("DELETE FROM matchpoints WHERE matcher_id = ?");
    $sth->execute($self->{'id'});
    $sth = $dbh->prepare_cached("DELETE FROM matchchecks WHERE matcher_id = ?");
    $sth->execute($self->{'id'});
    # foreign key delete cascades take care of deleting relevant rows
    # from matcher_matchpoints, matchpoint_components, and
    # matchpoint_component_norms
}

sub _update_marc_matchers {
    my $self = shift;

    my $dbh = C4::Context->dbh();
    my $sth = $dbh->prepare_cached("UPDATE marc_matchers 
                                    SET code = ?,
                                        description = ?,
                                        record_type = ?,
                                        threshold = ?
                                    WHERE matcher_id = ?");
    $sth->execute($self->{'code'}, $self->{'description'}, $self->{'record_type'}, $self->{'threshold'}, $self->{'id'});
}

sub _new_marc_matchers {
    my $self = shift;

    my $dbh = C4::Context->dbh();
    my $sth = $dbh->prepare_cached("INSERT INTO marc_matchers
                                    (code, description, record_type, threshold)
                                    VALUES (?, ?, ?, ?)");
    $sth->execute($self->{'code'}, $self->{'description'}, $self->{'record_type'}, $self->{'threshold'});
    $self->{'id'} = $dbh->{'mysql_insertid'};
}

sub _store_matcher_components {
    my $self = shift;

    my $dbh = C4::Context->dbh();
    my $sth;
    my $matcher_id = $self->{'id'};
    foreach my $matchpoint (@{ $self->{'matchpoints'}}) {
        my $matchpoint_id = $self->_store_matchpoint($matchpoint);
        $sth = $dbh->prepare_cached("INSERT INTO matcher_matchpoints (matcher_id, matchpoint_id)
                                     VALUES (?, ?)");
        $sth->execute($matcher_id, $matchpoint_id);
    }
    foreach my $matchcheck (@{ $self->{'required_checks'} }) {
        my $source_matchpoint_id = $self->_store_matchpoint($matchcheck->{'source_matchpoint'});
        my $target_matchpoint_id = $self->_store_matchpoint($matchcheck->{'target_matchpoint'});
        $sth = $dbh->prepare_cached("INSERT INTO matchchecks
                                     (matcher_id, source_matchpoint_id, target_matchpoint_id)
                                     VALUES (?, ?, ?)");
        $sth->execute($matcher_id, $source_matchpoint_id,  $target_matchpoint_id);
    }

}

sub _store_matchpoint {
    my $self = shift;
    my $matchpoint = shift;

    my $dbh = C4::Context->dbh();
    my $sth;
    my $matcher_id = $self->{'id'};
    $sth = $dbh->prepare_cached("INSERT INTO matchpoints (matcher_id, search_index, score)
                                 VALUES (?, ?, ?)");
    $sth->execute($matcher_id, $matchpoint->{'index'}, $matchpoint->{'score'});
    my $matchpoint_id = $dbh->{'mysql_insertid'};
    my $seqnum = 0;
    foreach my $component (@{ $matchpoint->{'components'} }) {
        $seqnum++;
        $sth = $dbh->prepare_cached("INSERT INTO matchpoint_components 
                                     (matchpoint_id, sequence, tag, subfields, offset, length)
                                     VALUES (?, ?, ?, ?, ?, ?)");
        $sth->bind_param(1, $matchpoint_id);
        $sth->bind_param(2, $seqnum);
        $sth->bind_param(3, $component->{'tag'});
        $sth->bind_param(4, join "", sort keys %{ $component->{'subfields'} });
        $sth->bind_param(5, $component->{'offset'});
        $sth->bind_param(6, $component->{'length'});
        $sth->execute();
        my $matchpoint_component_id = $dbh->{'mysql_insertid'};
        my $normseq = 0;
        foreach my $norm (@{ $component->{'norms'} }) {
            $normseq++;
            $sth = $dbh->prepare_cached("INSERT INTO matchpoint_component_norms
                                         (matchpoint_component_id, sequence, norm_routine)
                                         VALUES (?, ?, ?)");
            $sth->execute($matchpoint_component_id, $normseq, $norm);
        }
    }
    return $matchpoint_id;
}


=head2 delete

=over 4

C4::Matcher->delete($id);

=back

Deletes the matcher of the specified ID
from the database.

=cut

sub delete {
    my $class = shift;
    my $matcher_id = shift;

    my $dbh = C4::Context->dbh;
    my $sth = $dbh->prepare("DELETE FROM marc_matchers WHERE matcher_id = ?");
    $sth->execute($matcher_id); # relying on cascading deletes to clean up everything
}

=head2 threshold

=over 4

$matcher->threshold(1000);
my $threshold = $matcher->threshold();

=back

Accessor method.

=cut

sub threshold {
    my $self = shift;
    @_ ? $self->{'threshold'} = shift : $self->{'threshold'};
}

=head2 _id

=over 4

$matcher->_id(123);
my $id = $matcher->_id();

=back

Accessor method.  Note that using this method
to set the DB ID of the matcher should not be
done outside of the editing CGI.

=cut

sub _id {
    my $self = shift;
    @_ ? $self->{'id'} = shift : $self->{'id'};
}

=head2 code

=over 4

$matcher->code('ISBN');
my $code = $matcher->code();

=back

Accessor method.

=cut

sub code {
    my $self = shift;
    @_ ? $self->{'code'} = shift : $self->{'code'};
}

=head2 description

=over 4

$matcher->description('match on ISBN');
my $description = $matcher->description();

=back

Accessor method.

=cut

sub description {
    my $self = shift;
    @_ ? $self->{'description'} = shift : $self->{'description'};
}

=head2 add_matchpoint

=over 4

$matcher->add_matchpoint($index, $score, $matchcomponents);

=back

Adds a matchpoint that may include multiple components.  The $index
parameter identifies the index that will be searched, while $score
is the weight that will be added if a match is found.

$matchcomponents should be a reference to an array of matchpoint
compoents, each of which should be a hash containing the following 
keys:
    tag
    subfields
    offset
    length
    norms

The normalization_rules value should in turn be a reference to an
array, each element of which should be a reference to a 
normalization subroutine (under C4::Normalize) to be applied
to the source string.

=cut
    
sub add_matchpoint {
    my $self = shift;
    my ($index, $score, $matchcomponents) = @_;

    my $matchpoint = {};
    $matchpoint->{'index'} = $index;
    $matchpoint->{'score'} = $score;
    $matchpoint->{'components'} = [];
    foreach my $input_component (@{ $matchcomponents }) {
        push @{ $matchpoint->{'components'} }, _parse_match_component($input_component);
    }
    push @{ $self->{'matchpoints'} }, $matchpoint;
}

=head2 add_simple_matchpoint

=over 4

$matcher->add_simple_matchpoint($index, $score, $source_tag, $source_subfields, 
                                $source_offset, $source_length,
                                $source_normalizer);

=back

Adds a simple matchpoint rule -- after composing a key based on the source tag and subfields,
normalized per the normalization fuction, search the index.  All records retrieved
will receive the assigned score.

=cut

sub add_simple_matchpoint {
    my $self = shift;
    my ($index, $score, $source_tag, $source_subfields, $source_offset, $source_length, $source_normalizer) = @_;

    $self->add_matchpoint($index, $score, [
                          { tag => $source_tag, subfields => $source_subfields,
                            offset => $source_offset, 'length' => $source_length,
                            norms => [ $source_normalizer ]
                          }
                         ]);
}

=head2 add_required_check

=over 4

$match->add_required_check($source_matchpoint, $target_matchpoint);

=back

Adds a required check definition.  A required check means that in 
order for a match to be considered valid, the key derived from the
source (incoming) record must match the key derived from the target
(already in DB) record.

Unlike a regular matchpoint, only the first repeat of each tag 
in the source and target match criteria are considered.

A typical example of a required check would be verifying that the
titles and publication dates match.

$source_matchpoint and $target_matchpoint are each a reference to
an array of hashes, where each hash follows the same definition
as the matchpoint component specification in add_matchpoint, i.e.,

    tag
    subfields
    offset
    length
    norms

The normalization_rules value should in turn be a reference to an
array, each element of which should be a reference to a 
normalization subroutine (under C4::Normalize) to be applied
to the source string.

=cut

sub add_required_check {
    my $self = shift;
    my ($source_matchpoint, $target_matchpoint) = @_;

    my $matchcheck = {};
    $matchcheck->{'source_matchpoint'}->{'index'} = '';
    $matchcheck->{'source_matchpoint'}->{'score'} = 0;
    $matchcheck->{'source_matchpoint'}->{'components'} = [];
    $matchcheck->{'target_matchpoint'}->{'index'} = '';
    $matchcheck->{'target_matchpoint'}->{'score'} = 0;
    $matchcheck->{'target_matchpoint'}->{'components'} = [];
    foreach my $input_component (@{ $source_matchpoint }) {
        push @{ $matchcheck->{'source_matchpoint'}->{'components'} }, _parse_match_component($input_component);
    }
    foreach my $input_component (@{ $target_matchpoint }) {
        push @{ $matchcheck->{'target_matchpoint'}->{'components'} }, _parse_match_component($input_component);
    }
    push @{ $self->{'required_checks'} }, $matchcheck;
}

=head2 add_simple_required_check

$matcher->add_simple_required_check($source_tag, $source_subfields, $source_offset, $source_length, $source_normalizer,
                                    $target_tag, $target_subfields, $target_offset, $target_length, $target_normalizer);

=over 4

Adds a required check, which requires that the normalized keys made from the source and targets
must match for a match to be considered valid.

=back

=cut

sub add_simple_required_check {
    my $self = shift;
    my ($source_tag, $source_subfields, $source_offset, $source_length, $source_normalizer,
        $target_tag, $target_subfields, $target_offset, $target_length, $target_normalizer) = @_;

    $self->add_required_check(
      [ { tag => $source_tag, subfields => $source_subfields, offset => $source_offset, 'length' => $source_length,
          norms => [ $source_normalizer ] } ],
      [ { tag => $target_tag, subfields => $target_subfields, offset => $target_offset, 'length' => $target_length,
          norms => [ $target_normalizer ] } ]
    );
}

=head2 find_matches

=over 4

my @matches = $matcher->get_matches($marc_record, $max_matches);
foreach $match (@matches) {
  # matches already sorted in order of
  # decreasing score
  print "record ID: $match->{'record_id'};
  print "score:     $match->{'score'};
}

=back

Identifies all of the records matching the given MARC record.  For a record already 
in the database to be considered a match, it must meet the following criteria:

=over 2

=item 1. Total score from its matching field must exceed the supplied threshold.

=item 2. It must pass all required checks.

=back

Only the top $max_matches matches are returned.  The returned array is sorted
in order of decreasing score, i.e., the best match is first.

=cut

sub get_matches {
    my $self = shift;
    my ($source_record, $max_matches) = @_;

    my %matches = ();

    foreach my $matchpoint (@{ $self->{'matchpoints'} }) {
        my @source_keys = _get_match_keys($source_record, $matchpoint);
        next if scalar(@source_keys) == 0;
        # build query
        my $query = join(" or ", map { "$matchpoint->{'index'}=$_" } @source_keys);
        # FIXME only searching biblio index at the moment
        my ($error, $searchresults) = SimpleSearch($query);

        warn "search failed ($query) $error" if $error;
        foreach my $matched (@$searchresults) {
            $matches{$matched} += $matchpoint->{'score'};
        }
    }

    # get rid of any that don't meet the threshold
    %matches = map { ($matches{$_} >= $self->{'threshold'}) ? ($_ => $matches{$_}) : () } keys %matches;

    # get rid of any that don't meet the required checks
    %matches = map { _passes_required_checks($source_record, $_, $self->{'required_checks'}) ?  ($_ => $matches{$_}) : () } 
                keys %matches;

    my @results = ();
    foreach my $marcblob (keys %matches) {
        my $target_record = MARC::Record->new_from_usmarc($marcblob);
        my $result = TransformMarcToKoha(C4::Context->dbh, $target_record, '');
        # FIXME - again, bibliospecific
        # also, can search engine be induced to give just the number in the first place?
        my $record_number = $result->{'biblionumber'};
        push @results, { 'record_id' => $record_number, 'score' => $matches{$marcblob} };
    }
    @results = sort { $b->{'score'} cmp $a->{'score'} } @results;
    if (scalar(@results) > $max_matches) {
        @results = @results[0..$max_matches-1];
    }
    return @results;

}

=head2 dump

=over 4

$description = $matcher->dump();

=back

Returns a reference to a structure containing all of the information
in the matcher object.  This is mainly a convenience method to
aid setting up a HTML editing form.

=cut

sub dump {
    my $self = shift;
   
    my $result = {};

    $result->{'matcher_id'} = $self->{'id'};
    $result->{'code'} = $self->{'code'};
    $result->{'description'} = $self->{'description'};

    $result->{'matchpoints'} = [];
    foreach my $matchpoint (@{ $self->{'matchpoints'} }) {
        push @{  $result->{'matchpoints'} }, $matchpoint;
    }
    $result->{'matchchecks'} = [];
    foreach my $matchcheck (@{ $self->{'required_checks'} }) {
        push @{  $result->{'matchchecks'} }, $matchcheck;
    }

    return $result;
}

sub _passes_required_checks {
    my ($source_record, $target_blob, $matchchecks) = @_;
    my $target_record = MARC::Record->new_from_usmarc($target_blob); # FIXME -- need to avoid parsing record twice

    # no checks supplied == automatic pass
    return 1 if $#{ $matchchecks } == -1;

    foreach my $matchcheck (@{ $matchchecks }) {
        my $source_key = join "", _get_match_keys($source_record, $matchcheck->{'source_matchpoint'});
        my $target_key = join "", _get_match_keys($target_record, $matchcheck->{'target_matchpoint'});
        return 0 unless $source_key eq $target_key;
    }
    return 1;
}

sub _get_match_keys {
    my $source_record = shift;
    my $matchpoint = shift;
    my $check_only_first_repeat = @_ ? shift : 0;

    # If there is more than one component to the matchpoint (e.g.,
    # matchpoint includes both 003 and 001), any repeats
    # of the first component's tag are identified; repeats
    # of the subsequent components' tags are appended to
    # each parallel key dervied from the first component,
    # up to the number of repeats of the first component's tag.
    #
    # For example, if the record has one 003 and two 001s, only
    # one key is retrieved because there is only one 003.  The key
    # will consist of the contents of the first 003 and first 001.
    #
    # If there are two 003s and two 001s, there will be two keys:
    #    first 003 + first 001
    #    second 003 + second 001
    
    my @keys = ();
    for (my $i = 0; $i <= $#{ $matchpoint->{'components'} }; $i++) {
        my $component = $matchpoint->{'components'}->[$i];
        my $j = -1;
        FIELD: foreach my $field ($source_record->field($component->{'tag'})) {
            $j++;
            last FIELD if $j > 0 and $check_only_first_repeat;
            last FIELD if $i > 0 and $j > $#keys;
            my $key = "";
            if ($field->is_control_field()) {
                if ($component->{'length'}) {
                    $key = _normalize(substr($field->data(), $component->{'offset'}, $component->{'length'}))
                            # FIXME normalize, substr
                } else {
                    $key = _normalize($field->data());
                }
            } else {
                foreach my $subfield ($field->subfields()) {
                    if (exists $component->{'subfields'}->{$subfield->[0]}) {
                        $key .= " " . $subfield->[1];
                    }
                }
                $key = _normalize($key);
            }
            if ($i == 0) {
                push @keys, $key if $key;
            } else {
                $keys[$j] .= " $key" if $key;
            }
        }
    }
    return @keys;
    
}


sub _parse_match_component {
    my $input_component = shift;

    my $component = {};
    $component->{'tag'} = $input_component->{'tag'};
    $component->{'subfields'} = { map { $_ => 1 } split(//, $input_component->{'subfields'}) };
    $component->{'offset'} = exists($input_component->{'offset'}) ? $input_component->{'offset'} : -1;
    $component->{'length'} = $input_component->{'length'} ? $input_component->{'length'} : 0;
    $component->{'norms'} =  $input_component->{'norms'} ? $input_component->{'norms'} : [];

    return $component;
}

# FIXME - default normalizer
sub _normalize {
    my $value = uc shift;
    $value =~ s/^\s+//;
    $value =~ s/^\s+$//;
    $value =~ s/\s+/ /g;
    $value =~ s/[.;,\]\[\)\(\/"']//g;
    return $value;
}

1;
__END__

=head1 AUTHOR

Koha Development Team <info@koha.org>

Galen Charlton <galen.charlton@liblime.com>

=cut