package C4::Breeding; # Copyright 2000-2002 Katipo Communications # # 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::Biblio; use C4::Koha; use MARC::File::USMARC; use C4::ImportBatch; require Exporter; use vars qw($VERSION @ISA @EXPORT @EXPORT_OK %EXPORT_TAGS); # set the version for version checking $VERSION = 0.01; =head1 NAME C4::Breeding : module to add biblios to import_records via the breeding/reservoir API. =head1 SYNOPSIS use C4::Scan; &ImportBreeding($marcrecords,$overwrite_biblio,$filename,$z3950random,$batch_type); C<$marcrecord> => the MARC::Record C<$overwrite_biblio> => if set to 1 a biblio with the same ISBN will be overwritted. if set to 0 a biblio with the same isbn will be ignored (the previous will be kept) if set to -1 the biblio will be added anyway (more than 1 biblio with the same ISBN possible in the breeding C<$encoding> => USMARC or UNIMARC. used for char_decoding. If not present, the parameter marcflavour is used instead C<$z3950random> => the random value created during a z3950 search result. =head1 DESCRIPTION ImportBreeding import MARC records in the reservoir (import_records/import_batches tables). the records can be properly encoded or not, we try to reencode them in utf-8 if needed. works perfectly with BNF server, that sends UNIMARC latin1 records. Should work with other servers too. the FixEncoding sub is in Koha.pm, as it's a general usage sub. =cut @ISA = qw(Exporter); @EXPORT = qw(&ImportBreeding &BreedingSearch); =head2 ImportBreeding ImportBreeding($marcrecords,$overwrite_biblio,$filename,$encoding,$z3950random,$batch_type); TODO description =cut sub ImportBreeding { my ($marcrecords,$overwrite_biblio,$filename,$encoding,$z3950random,$batch_type) = @_; my @marcarray = split /\x1D/, $marcrecords; my $dbh = C4::Context->dbh; my $batch_id = 0; if ($batch_type eq 'z3950') { $batch_id = GetZ3950BatchId($filename); } else { # create a new one $batch_id = AddImportBatch('create_new', 'staging', 'batch', $filename, ''); } my $searchisbn = $dbh->prepare("select biblioitemnumber from biblioitems where isbn=?"); my $searchissn = $dbh->prepare("select biblioitemnumber from biblioitems where issn=?"); # FIXME -- not sure that this kind of checking is actually needed my $searchbreeding = $dbh->prepare("select import_record_id from import_biblios where isbn=? and title=?"); $encoding = C4::Context->preference("marcflavour") unless $encoding; # fields used for import results my $imported=0; my $alreadyindb = 0; my $alreadyinfarm = 0; my $notmarcrecord = 0; my $breedingid; for (my $i=0;$i<=$#marcarray;$i++) { my $marcrecord = FixEncoding($marcarray[$i]."\x1D"); # FIXME - currently this does nothing my @warnings = $marcrecord->warnings(); if (scalar($marcrecord->fields()) == 0) { $notmarcrecord++; } else { my $oldbiblio = TransformMarcToKoha($dbh,$marcrecord,''); # if isbn found and biblio does not exist, add it. If isbn found and biblio exists, # overwrite or ignore depending on user choice # drop every "special" char : spaces, - ... $oldbiblio->{isbn} =~ s/\(.*$//; $oldbiblio->{isbn} =~ tr/ -_//; $oldbiblio->{isbn} = uc $oldbiblio->{isbn}; # search if biblio exists my $biblioitemnumber; if ($oldbiblio->{isbn}) { $searchisbn->execute($oldbiblio->{isbn}); ($biblioitemnumber) = $searchisbn->fetchrow; } else { if ($oldbiblio->{issn}) { $searchissn->execute($oldbiblio->{issn}); ($biblioitemnumber) = $searchissn->fetchrow; } } if ($biblioitemnumber && $overwrite_biblio ne 2) { $alreadyindb++; } else { # FIXME - in context of batch load, # rejecting records because already present in the reservoir # not correct in every case. # search in breeding farm if ($oldbiblio->{isbn}) { $searchbreeding->execute($oldbiblio->{isbn},$oldbiblio->{title}); ($breedingid) = $searchbreeding->fetchrow; } elsif ($oldbiblio->{issn}){ $searchbreeding->execute($oldbiblio->{issn},$oldbiblio->{title}); ($breedingid) = $searchbreeding->fetchrow; } if ($breedingid && $overwrite_biblio eq '0') { $alreadyinfarm++; } else { if ($breedingid && $overwrite_biblio eq '1') { ModBiblioInBatch($breedingid, $marcrecord); } else { my $import_id = AddBiblioToBatch($batch_id, $imported, $marcrecord, $encoding, $z3950random); $breedingid = $import_id; } $imported++; } } } } return ($notmarcrecord,$alreadyindb,$alreadyinfarm,$imported,$breedingid); } =head2 BreedingSearch ($count, @results) = &BreedingSearch($title,$isbn,$random); C<$title> contains the title, C<$isbn> contains isbn or issn, C<$random> contains the random seed from a z3950 search. C<$count> is the number of items in C<@results>. C<@results> is an array of references-to-hash; the keys are the items from the C and C tables of the Koha database. =cut sub BreedingSearch { my ($title,$isbn,$z3950random) = @_; my $dbh = C4::Context->dbh; my $count = 0; my ($query,@bind); my $sth; my @results; $query = "SELECT import_record_id, file_name, isbn, title, author FROM import_biblios JOIN import_records USING (import_record_id) JOIN import_batches USING (import_batch_id) WHERE "; if ($z3950random) { $query .= "z3950random = ?"; @bind=($z3950random); } else { @bind=(); if ($title) { $query .= "title like ?"; push(@bind,"$title%"); } if ($title && $isbn) { $query .= " and "; } if ($isbn) { $query .= "isbn like ?"; push(@bind,"$isbn%"); } } $sth = $dbh->prepare($query); $sth->execute(@bind); while (my $data = $sth->fetchrow_hashref) { $results[$count] = $data; # FIXME - hack to reflect difference in name # of columns in old marc_breeding and import_records # There needs to be more separation between column names and # field names used in the templates $data->{'file'} = $data->{'file_name'}; $data->{'id'} = $data->{'import_record_id'}; $count++; } # while $sth->finish; return($count, @results); } # sub breedingsearch END { } # module clean-up code here (global destructor)