package C4::Biblio; # 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; require Exporter; # use utf8; use C4::Context; use MARC::Record; use MARC::File::USMARC; use MARC::File::XML; use ZOOM; use C4::Koha; use C4::Date; use C4::Log; # logaction use vars qw($VERSION @ISA @EXPORT); # set the version for version checking $VERSION = do { my @v = '$Revision$' =~ /\d+/g; shift(@v).".".join( "_", map { sprintf "%03d", $_ } @v ); }; @ISA = qw( Exporter ); # EXPORTED FUNCTIONS. # to add biblios or items push @EXPORT, qw( &AddBiblio &AddItem ); # to get something push @EXPORT, qw( &GetBiblio &GetBiblioData &GetBiblioItemData &GetBiblioItemInfosOf &GetBiblioItemByBiblioNumber &GetBiblioFromItemNumber &GetMarcItem &GetItem &GetItemInfosOf &GetItemStatus &GetItemLocation &GetLostItems &GetItemsForInventory &GetItemsCount &GetMarcNotes &GetMarcSubjects &GetMarcBiblio &GetMarcAuthors &GetMarcSeries &GetUsedMarcStructure &GetItemsInfo &GetItemnumberFromBarcode &get_itemnumbers_of &GetXmlBiblio &GetAuthorisedValueDesc &GetMarcStructure &GetMarcFromKohaField &GetFrameworkCode &GetPublisherNameFromIsbn &TransformKohaToMarc ); # To modify something push @EXPORT, qw( &ModBiblio &ModItem &ModItemTransfer &ModBiblioframework &ModZebra &ModItemInMarc &ModItemInMarconefield &ModDateLastSeen ); # To delete something push @EXPORT, qw( &DelBiblio &DelItem ); # Internal functions # those functions are exported but should not be used # they are usefull is few circumstances, so are exported. # but don't use them unless you're a core developer ;-) push @EXPORT, qw( &ModBiblioMarc &AddItemInMarc &calculatelc &itemcalculator ); # Others functions push @EXPORT, qw( &TransformMarcToKoha &TransformHtmlToMarc2 &TransformHtmlToMarc &TransformHtmlToXml &PrepareItemrecordDisplay &char_decode &GetNoZebraIndexes ); =head1 NAME C4::Biblio - cataloging management functions =head1 DESCRIPTION Biblio.pm contains functions for managing storage and editing of bibliographic data within Koha. Most of the functions in this module are used for cataloging records: adding, editing, or removing biblios, biblioitems, or items. Koha's stores bibliographic information in three places: =over 4 =item 1. in the biblio,biblioitems,items, etc tables, which are limited to a one-to-one mapping to underlying MARC data =item 2. as raw MARC in the Zebra index and storage engine =item 3. as raw MARC the biblioitems.marc and biblioitems.marcxml =back In the 3.0 version of Koha, the authoritative record-level information is in biblioitems.marcxml Because the data isn't completely normalized there's a chance for information to get out of sync. The design choice to go with a un-normalized schema was driven by performance and stability concerns. However, if this occur, it can be considered as a bug : The API is (or should be) complete & the only entry point for all biblio/items managements. =over 4 =item 1. Compared with MySQL, Zebra is slow to update an index for small data changes -- especially for proc-intensive operations like circulation =item 2. Zebra's index has been known to crash and a backup of the data is necessary to rebuild it in such cases =back Because of this design choice, the process of managing storage and editing is a bit convoluted. Historically, Biblio.pm's grown to an unmanagable size and as a result we have several types of functions currently: =over 4 =item 1. Add*/Mod*/Del*/ - high-level external functions suitable for being called from external scripts to manage the collection =item 2. _koha_* - low-level internal functions for managing the koha tables =item 3. Marc management function : as the MARC record is stored in biblioitems.marc(xml), some subs dedicated to it's management are in this package. They should be used only internally by Biblio.pm, the only official entry points being AddBiblio, AddItem, ModBiblio, ModItem. =item 4. Zebra functions used to update the Zebra index =item 5. internal helper functions such as char_decode, checkitems, etc. Some of these probably belong in Koha.pm =back The MARC record (in biblioitems.marcxml) contains the complete marc record, including items. It also contains the biblionumber. That is the reason why it is not stored directly by AddBiblio, with all other fields . To save a biblio, we need to : =over 4 =item 1. save datas in biblio and biblioitems table, that gives us a biblionumber and a biblioitemnumber =item 2. add the biblionumber and biblioitemnumber into the MARC records =item 3. save the marc record =back When dealing with items, we must : =over 4 =item 1. save the item in items table, that gives us an itemnumber =item 2. add the itemnumber to the item MARC field =item 3. overwrite the MARC record (with the added item) into biblioitems.marc(xml) When modifying a biblio or an item, the behaviour is quite similar. =back =head1 EXPORTED FUNCTIONS =head2 AddBiblio =over 4 ($biblionumber,$biblioitemnumber) = AddBiblio($record,$frameworkcode); Exported function (core API) for adding a new biblio to koha. =back =cut sub AddBiblio { my ( $record, $frameworkcode ) = @_; my $biblionumber; my $biblioitemnumber; my $dbh = C4::Context->dbh; # transform the data into koha-table style data my $olddata = TransformMarcToKoha( $dbh, $record, $frameworkcode ); $biblionumber = _koha_add_biblio( $dbh, $olddata, $frameworkcode ); $olddata->{'biblionumber'} = $biblionumber; $biblioitemnumber = _koha_add_biblioitem( $dbh, $olddata ); # we must add bibnum and bibitemnum in MARC::Record... # we build the new field with biblionumber and biblioitemnumber # we drop the original field # we add the new builded field. ( my $biblio_tag, my $biblio_subfield ) = GetMarcFromKohaField("biblio.biblionumber",$frameworkcode); ( my $biblioitem_tag, my $biblioitem_subfield ) = GetMarcFromKohaField("biblioitems.biblioitemnumber",$frameworkcode); my $newfield; # biblionumber & biblioitemnumber are in different fields if ( $biblio_tag != $biblioitem_tag ) { # deal with biblionumber if ( $biblio_tag < 10 ) { $newfield = MARC::Field->new( $biblio_tag, $biblionumber ); } else { $newfield = MARC::Field->new( $biblio_tag, '', '', "$biblio_subfield" => $biblionumber ); } # drop old field and create new one... my $old_field = $record->field($biblio_tag); $record->delete_field($old_field); $record->append_fields($newfield); # deal with biblioitemnumber if ( $biblioitem_tag < 10 ) { $newfield = MARC::Field->new( $biblioitem_tag, $biblioitemnumber, ); } else { $newfield = MARC::Field->new( $biblioitem_tag, '', '', "$biblioitem_subfield" => $biblioitemnumber, ); } # drop old field and create new one... $old_field = $record->field($biblioitem_tag); $record->delete_field($old_field); $record->insert_fields_ordered($newfield); # biblionumber & biblioitemnumber are in the same field (can't be <10 as fields <10 have only 1 value) } else { my $newfield = MARC::Field->new( $biblio_tag, '', '', "$biblio_subfield" => $biblionumber, "$biblioitem_subfield" => $biblioitemnumber ); # drop old field and create new one... my $old_field = $record->field($biblio_tag); $record->delete_field($old_field); $record->insert_fields_ordered($newfield); } # now add the record $biblionumber = ModBiblioMarc( $record, $biblionumber, $frameworkcode ); &logaction(C4::Context->userenv->{'number'},"CATALOGUING","ADD",$biblionumber,"biblio") if C4::Context->preference("CataloguingLog"); return ( $biblionumber, $biblioitemnumber ); } =head2 AddItem =over 2 $biblionumber = AddItem( $record, $biblionumber) Exported function (core API) for adding a new item to Koha =back =cut sub AddItem { my ( $record, $biblionumber ) = @_; my $dbh = C4::Context->dbh; # add item in old-DB my $frameworkcode = GetFrameworkCode( $biblionumber ); my $item = &TransformMarcToKoha( $dbh, $record, $frameworkcode ); # needs old biblionumber and biblioitemnumber $item->{'biblionumber'} = $biblionumber; my $sth = $dbh->prepare( "select biblioitemnumber,itemtype from biblioitems where biblionumber=?" ); $sth->execute( $item->{'biblionumber'} ); my $itemtype; ( $item->{'biblioitemnumber'}, $itemtype ) = $sth->fetchrow; $sth = $dbh->prepare( "select notforloan from itemtypes where itemtype='$itemtype'"); $sth->execute(); my $notforloan = $sth->fetchrow; ##Change the notforloan field if $notforloan found if ( $notforloan > 0 ) { $item->{'notforloan'} = $notforloan; &MARCitemchange( $record, "items.notforloan", $notforloan ); } if ( !$item->{'dateaccessioned'} || $item->{'dateaccessioned'} eq '' ) { # find today's date my ( $sec, $min, $hour, $mday, $mon, $year, $wday, $yday, $isdst ) = localtime(time); $year += 1900; $mon += 1; my $date = "$year-" . sprintf( "%0.2d", $mon ) . "-" . sprintf( "%0.2d", $mday ); $item->{'dateaccessioned'} = $date; &MARCitemchange( $record, "items.dateaccessioned", $date ); } my ( $itemnumber, $error ) = &_koha_new_items( $dbh, $item, $item->{barcode} ); # add itemnumber to MARC::Record before adding the item. $sth = $dbh->prepare( "select tagfield,tagsubfield from marc_subfield_structure where frameworkcode=? and kohafield=?" ); &TransformKohaToMarcOneField( $sth, $record, "items.itemnumber", $itemnumber, $frameworkcode ); # add the item &AddItemInMarc( $record, $item->{'biblionumber'},$frameworkcode ); &logaction(C4::Context->userenv->{'number'},"CATALOGUING","ADD",$itemnumber,"item") if C4::Context->preference("CataloguingLog"); return ($item->{biblionumber}, $item->{biblioitemnumber},$itemnumber); } =head2 ModBiblio ModBiblio( $record,$biblionumber,$frameworkcode); Exported function (core API) to modify a biblio =cut sub ModBiblio { my ( $record, $biblionumber, $frameworkcode ) = @_; if (C4::Context->preference("CataloguingLog")) { my $newrecord = GetMarcBiblio($biblionumber); &logaction(C4::Context->userenv->{'number'},"CATALOGUING","MODIFY",$biblionumber,"BEFORE=>".$newrecord->as_formatted); } my $dbh = C4::Context->dbh; $frameworkcode = "" unless $frameworkcode; # get the items before and append them to the biblio before updating the record, atm we just have the biblio my ( $itemtag, $itemsubfield ) = GetMarcFromKohaField("items.itemnumber",$frameworkcode); my $oldRecord = GetMarcBiblio( $biblionumber ); # parse each item, and, for an unknown reason, re-encode each subfield # if you don't do that, the record will have encoding mixed # and the biblio will be re-encoded. # strange, I (Paul P.) searched more than 1 day to understand what happends # but could only solve the problem this way... my @fields = $oldRecord->field( $itemtag ); foreach my $fielditem ( @fields ){ my $field; foreach ($fielditem->subfields()) { if ($field) { $field->add_subfields(Encode::encode('utf-8',$_->[0]) => Encode::encode('utf-8',$_->[1])); } else { $field = MARC::Field->new("$itemtag",'','',Encode::encode('utf-8',$_->[0]) => Encode::encode('utf-8',$_->[1])); } } $record->append_fields($field); } # adding biblionumber my ($tag_biblionumber, $subfield_biblionumber) = GetMarcFromKohaField('biblio.biblionumber',$frameworkcode); $record->append_fields( MARC::Field->new( $tag_biblionumber,'','',$subfield_biblionumber => $biblionumber ) ); # update the MARC record (that now contains biblio and items) with the new record data &ModBiblioMarc( $record, $biblionumber, $frameworkcode ); # load the koha-table data object my $oldbiblio = TransformMarcToKoha( $dbh, $record, $frameworkcode ); # modify the other koha tables _koha_modify_biblio( $dbh, $oldbiblio ); _koha_modify_biblioitem( $dbh, $oldbiblio ); return 1; } =head2 ModItem =over 2 Exported function (core API) for modifying an item in Koha. =back =cut sub ModItem { my ( $record, $biblionumber, $itemnumber, $delete, $new_item_hashref ) = @_; #logging &logaction(C4::Context->userenv->{'number'},"CATALOGUING","MODIFY",$itemnumber,$record->as_formatted) if C4::Context->preference("CataloguingLog"); my $dbh = C4::Context->dbh; # if we have a MARC record, we're coming from cataloging and so # we do the whole routine: update the MARC and zebra, then update the koha # tables if ($record) { my $frameworkcode = GetFrameworkCode( $biblionumber ); ModItemInMarc( $record, $biblionumber, $itemnumber, $frameworkcode ); my $olditem = TransformMarcToKoha( $dbh, $record, $frameworkcode ); _koha_modify_item( $dbh, $olditem ); return $biblionumber; } # otherwise, we're just looking to modify something quickly # (like a status) so we just update the koha tables elsif ($new_item_hashref) { _koha_modify_item( $dbh, $new_item_hashref ); } } sub ModItemTransfer { my ( $itemnumber, $frombranch, $tobranch ) = @_; my $dbh = C4::Context->dbh; #new entry in branchtransfers.... my $sth = $dbh->prepare( "INSERT INTO branchtransfers (itemnumber, frombranch, datesent, tobranch) VALUES (?, ?, now(), ?)"); $sth->execute($itemnumber, $frombranch, $tobranch); #update holdingbranch in items ..... $sth= $dbh->prepare( "UPDATE items set holdingbranch = ? WHERE items.itemnumber = ?"); $sth->execute($tobranch,$itemnumber); &ModDateLastSeen($itemnumber); $sth = $dbh->prepare( "SELECT biblionumber FROM items WHERE itemnumber=?" ); $sth->execute($itemnumber); while ( my ( $biblionumber ) = $sth->fetchrow ) { &ModItemInMarconefield( $biblionumber, $itemnumber, 'items.holdingbranch', $tobranch ); } return; } =head2 ModBiblioframework ModBiblioframework($biblionumber,$frameworkcode); Exported function to modify a biblio framework =cut sub ModBiblioframework { my ( $biblionumber, $frameworkcode ) = @_; my $dbh = C4::Context->dbh; my $sth = $dbh->prepare( "UPDATE biblio SET frameworkcode=? WHERE biblionumber=$biblionumber" ); $sth->execute($frameworkcode); return 1; } =head2 ModItemInMarconefield =over modify only 1 field in a MARC item (mainly used for holdingbranch, but could also be used for status modif - moving a book to "lost" on a long overdu for example) &ModItemInMarconefield( $biblionumber, $itemnumber, $itemfield, $newvalue ) =back =cut sub ModItemInMarconefield { my ( $biblionumber, $itemnumber, $itemfield, $newvalue ) = @_; my $dbh = C4::Context->dbh; if ( !defined $newvalue ) { $newvalue = ""; } my $record = GetMarcItem( $biblionumber, $itemnumber ); my ($tagfield, $tagsubfield) = GetMarcFromKohaField( $itemfield,''); if ($tagfield && $tagsubfield) { my $tag = $record->field($tagfield); if ($tag) { # my $tagsubs = $record->field($tagfield)->subfield($tagsubfield); $tag->update( $tagsubfield => $newvalue ); $record->delete_field($tag); $record->insert_fields_ordered($tag); &ModItemInMarc( $record, $biblionumber, $itemnumber, 0 ); } } } =head2 ModItemInMarc =over &ModItemInMarc( $record, $biblionumber, $itemnumber ) =back =cut sub ModItemInMarc { my ( $ItemRecord, $biblionumber, $itemnumber, $frameworkcode) = @_; my $dbh = C4::Context->dbh; # get complete MARC record & replace the item field by the new one my $completeRecord = GetMarcBiblio($biblionumber); my ($itemtag,$itemsubfield) = GetMarcFromKohaField("items.itemnumber",$frameworkcode); my $itemField = $ItemRecord->field($itemtag); my @items = $completeRecord->field($itemtag); foreach (@items) { if ($_->subfield($itemsubfield) eq $itemnumber) { # $completeRecord->delete_field($_); $_->replace_with($itemField); } } # save the record my $sth = $dbh->prepare("update biblioitems set marc=?,marcxml=? where biblionumber=?"); $sth->execute( $completeRecord->as_usmarc(), $completeRecord->as_xml_record(),$biblionumber ); $sth->finish; ModZebra($biblionumber,"specialUpdate","biblioserver",$completeRecord); } =head2 ModDateLastSeen &ModDateLastSeen($itemnum) Mark item as seen. Is called when an item is issued, returned or manually marked during inventory/stocktaking C<$itemnum> is the item number =cut sub ModDateLastSeen { my ($itemnum) = @_; my $dbh = C4::Context->dbh; my $sth = $dbh->prepare( "update items set itemlost=0, datelastseen = now() where items.itemnumber = ?" ); $sth->execute($itemnum); return; } =head2 DelBiblio =over my $error = &DelBiblio($dbh,$biblionumber); Exported function (core API) for deleting a biblio in koha. Deletes biblio record from Zebra and Koha tables (biblio,biblioitems,items) Also backs it up to deleted* tables Checks to make sure there are not issues on any of the items return: C<$error> : undef unless an error occurs =back =cut sub DelBiblio { my ( $biblionumber ) = @_; my $dbh = C4::Context->dbh; my $error; # for error handling # First make sure there are no items with issues are still attached my $sth = $dbh->prepare( "SELECT itemnumber FROM items WHERE biblionumber=?"); $sth->execute($biblionumber); while ( my $itemnumber = $sth->fetchrow ) { my $issues = GetItemIssues($itemnumber); foreach my $issue (@$issues) { if ( ( $issue->{date_due} ) && ( $issue->{date_due} ne "Available" ) ) { #FIXME: we need a status system in Biblio like in Circ to return standard codes and messages # instead of hard-coded strings $error .= "Item is checked out to a patron -- you must return it before deleting the Biblio"; } } } return $error if $error; # Delete in Zebra. Be careful NOT to move this line after _koha_delete_biblio # for at least 2 reasons : # - we need to read the biblio if NoZebra is set (to remove it from the indexes # - if something goes wrong, the biblio may be deleted from Koha but not from zebra # and we would have no way to remove it (except manually in zebra, but I bet it would be very hard to handle the problem) ModZebra($biblionumber, "delete_record", "biblioserver", undef); # delete biblio from Koha tables and save in deletedbiblio $error = &_koha_delete_biblio( $dbh, $biblionumber ); # delete biblioitems and items from Koha tables and save in deletedbiblioitems,deleteditems $sth = $dbh->prepare( "SELECT biblioitemnumber FROM biblioitems WHERE biblionumber=?"); $sth->execute($biblionumber); while ( my $biblioitemnumber = $sth->fetchrow ) { # delete this biblioitem $error = &_koha_delete_biblioitems( $dbh, $biblioitemnumber ); return $error if $error; # delete items my $items_sth = $dbh->prepare( "SELECT itemnumber FROM items WHERE biblioitemnumber=?"); $items_sth->execute($biblioitemnumber); while ( my $itemnumber = $items_sth->fetchrow ) { $error = &_koha_delete_item( $dbh, $itemnumber ); return $error if $error; } } &logaction(C4::Context->userenv->{'number'},"CATALOGUING","DELETE",$biblionumber,"") if C4::Context->preference("CataloguingLog"); return; } =head2 DelItem =over DelItem( $biblionumber, $itemnumber ); Exported function (core API) for deleting an item record in Koha. =back =cut sub DelItem { my ( $biblionumber, $itemnumber ) = @_; my $dbh = C4::Context->dbh; &_koha_delete_item( $dbh, $itemnumber ); # get the MARC record my $record = GetMarcBiblio($biblionumber); my $frameworkcode = GetFrameworkCode($biblionumber); # backup the record my $copy2deleted = $dbh->prepare("UPDATE deleteditems SET marc=? WHERE itemnumber=?"); $copy2deleted->execute( $record->as_usmarc(), $itemnumber ); #search item field code my ( $itemtag, $itemsubfield ) = GetMarcFromKohaField("items.itemnumber",$frameworkcode); my @fields = $record->field($itemtag); # delete the item specified foreach my $field (@fields) { if ( $field->subfield($itemsubfield) eq $itemnumber ) { $record->delete_field($field); } } &ModBiblioMarc( $record, $biblionumber, $frameworkcode ); &logaction(C4::Context->userenv->{'number'},"CATALOGUING","DELETE",$itemnumber,"item") if C4::Context->preference("CataloguingLog"); } =head2 GetBiblioData =over 4 $data = &GetBiblioData($biblionumber); Returns information about the book with the given biblionumber. C<&GetBiblioData> returns a reference-to-hash. The keys are the fields in the C and C tables in the Koha database. In addition, C<$data-E{subject}> is the list of the book's subjects, separated by C<" , "> (space, comma, space). If there are multiple biblioitems with the given biblionumber, only the first one is considered. =back =cut sub GetBiblioData { my ( $bibnum ) = @_; my $dbh = C4::Context->dbh; my $query = " SELECT * , biblioitems.notes AS bnotes, biblio.notes FROM biblio LEFT JOIN biblioitems ON biblio.biblionumber = biblioitems.biblionumber LEFT JOIN itemtypes ON biblioitems.itemtype = itemtypes.itemtype WHERE biblio.biblionumber = ? AND biblioitems.biblionumber = biblio.biblionumber "; my $sth = $dbh->prepare($query); $sth->execute($bibnum); my $data; $data = $sth->fetchrow_hashref; $sth->finish; return ($data); } # sub GetBiblioData =head2 GetItemsInfo =over 4 @results = &GetItemsInfo($biblionumber, $type); Returns information about books with the given biblionumber. C<$type> may be either C or anything else. If it is not set to C, then the search will exclude lost, very overdue, and withdrawn items. C<&GetItemsInfo> returns a list of references-to-hash. Each element contains a number of keys. Most of them are table items from the C, C, C, and C tables in the Koha database. Other keys include: =over 4 =item C<$data-E{branchname}> The name (not the code) of the branch to which the book belongs. =item C<$data-E{datelastseen}> This is simply C, except that while the date is stored in YYYY-MM-DD format in the database, here it is converted to DD/MM/YYYY format. A NULL date is returned as C. =item C<$data-E{datedue}> =item C<$data-E{class}> This is the concatenation of C, the book's Dewey code, and C. =item C<$data-E{ocount}> I think this is the number of copies of the book available. =item C<$data-E{order}> If this is set, it is set to C. =back =back =cut sub GetItemsInfo { my ( $biblionumber, $type ) = @_; my $dbh = C4::Context->dbh; my $query = "SELECT *,items.notforloan as itemnotforloan FROM items, biblio, biblioitems LEFT JOIN itemtypes on biblioitems.itemtype = itemtypes.itemtype WHERE items.biblionumber = ? AND biblioitems.biblioitemnumber = items.biblioitemnumber AND biblio.biblionumber = items.biblionumber ORDER BY items.dateaccessioned desc "; my $sth = $dbh->prepare($query); $sth->execute($biblionumber); my $i = 0; my @results; my ( $date_due, $count_reserves ); while ( my $data = $sth->fetchrow_hashref ) { my $datedue = ''; my $isth = $dbh->prepare( "SELECT issues.*,borrowers.cardnumber,borrowers.surname,borrowers.firstname FROM issues, borrowers WHERE itemnumber = ? AND returndate IS NULL AND issues.borrowernumber=borrowers.borrowernumber" ); $isth->execute( $data->{'itemnumber'} ); if ( my $idata = $isth->fetchrow_hashref ) { $data->{borrowernumber} = $idata->{borrowernumber}; $data->{cardnumber} = $idata->{cardnumber}; $data->{surname} = $idata->{surname}; $data->{firstname} = $idata->{firstname}; $datedue = format_date( $idata->{'date_due'} ); } if ( $datedue eq '' ) { #$datedue="Available"; my ( $restype, $reserves ) = C4::Reserves::CheckReserves( $data->{'itemnumber'} ); if ($restype) { #$datedue=$restype; $count_reserves = $restype; } } $isth->finish; #get branch information..... my $bsth = $dbh->prepare( "SELECT * FROM branches WHERE branchcode = ? " ); $bsth->execute( $data->{'holdingbranch'} ); if ( my $bdata = $bsth->fetchrow_hashref ) { $data->{'branchname'} = $bdata->{'branchname'}; } my $date = format_date( $data->{'datelastseen'} ); $data->{'datelastseen'} = $date; $data->{'datedue'} = $datedue; $data->{'count_reserves'} = $count_reserves; # get notforloan complete status if applicable my $sthnflstatus = $dbh->prepare( 'SELECT authorised_value FROM marc_subfield_structure WHERE kohafield="items.notforloan" ' ); $sthnflstatus->execute; my ($authorised_valuecode) = $sthnflstatus->fetchrow; if ($authorised_valuecode) { $sthnflstatus = $dbh->prepare( "SELECT lib FROM authorised_values WHERE category=? AND authorised_value=?" ); $sthnflstatus->execute( $authorised_valuecode, $data->{itemnotforloan} ); my ($lib) = $sthnflstatus->fetchrow; $data->{notforloan} = $lib; } # my stack procedures my $stackstatus = $dbh->prepare( 'SELECT authorised_value FROM marc_subfield_structure WHERE kohafield="items.stack" ' ); $stackstatus->execute; ($authorised_valuecode) = $stackstatus->fetchrow; if ($authorised_valuecode) { $stackstatus = $dbh->prepare( "SELECT lib FROM authorised_values WHERE category=? AND authorised_value=? " ); $stackstatus->execute( $authorised_valuecode, $data->{stack} ); my ($lib) = $stackstatus->fetchrow; $data->{stack} = $lib; } $results[$i] = $data; $i++; } $sth->finish; return (@results); } =head2 getitemstatus =over 4 $itemstatushash = &getitemstatus($fwkcode); returns information about status. Can be MARC dependant. fwkcode is optional. But basically could be can be loan or not Create a status selector with the following code =head3 in PERL SCRIPT my $itemstatushash = getitemstatus; my @itemstatusloop; foreach my $thisstatus (keys %$itemstatushash) { my %row =(value => $thisstatus, statusname => $itemstatushash->{$thisstatus}->{'statusname'}, ); push @itemstatusloop, \%row; } $template->param(statusloop=>\@itemstatusloop); =head3 in TEMPLATE =cut sub GetItemStatus { # returns a reference to a hash of references to status... my ($fwk) = @_; my %itemstatus; my $dbh = C4::Context->dbh; my $sth; $fwk = '' unless ($fwk); my ( $tag, $subfield ) = GetMarcFromKohaField( "items.notforloan", $fwk ); if ( $tag and $subfield ) { my $sth = $dbh->prepare( "select authorised_value from marc_subfield_structure where tagfield=? and tagsubfield=? and frameworkcode=?" ); $sth->execute( $tag, $subfield, $fwk ); if ( my ($authorisedvaluecat) = $sth->fetchrow ) { my $authvalsth = $dbh->prepare( "select authorised_value, lib from authorised_values where category=? order by lib" ); $authvalsth->execute($authorisedvaluecat); while ( my ( $authorisedvalue, $lib ) = $authvalsth->fetchrow ) { $itemstatus{$authorisedvalue} = $lib; } $authvalsth->finish; return \%itemstatus; exit 1; } else { #No authvalue list # build default } $sth->finish; } #No authvalue list #build default $itemstatus{"1"} = "Not For Loan"; return \%itemstatus; } =head2 getitemlocation =over 4 $itemlochash = &getitemlocation($fwk); returns informations about location. where fwk stands for an optional framework code. Create a location selector with the following code =head3 in PERL SCRIPT my $itemlochash = getitemlocation; my @itemlocloop; foreach my $thisloc (keys %$itemlochash) { my $selected = 1 if $thisbranch eq $branch; my %row =(locval => $thisloc, selected => $selected, locname => $itemlochash->{$thisloc}, ); push @itemlocloop, \%row; } $template->param(itemlocationloop => \@itemlocloop); =head3 in TEMPLATE =back =cut sub GetItemLocation { # returns a reference to a hash of references to location... my ($fwk) = @_; my %itemlocation; my $dbh = C4::Context->dbh; my $sth; $fwk = '' unless ($fwk); my ( $tag, $subfield ) = GetMarcFromKohaField( "items.location", $fwk ); if ( $tag and $subfield ) { my $sth = $dbh->prepare( "select authorised_value from marc_subfield_structure where tagfield=? and tagsubfield=? and frameworkcode=?" ); $sth->execute( $tag, $subfield, $fwk ); if ( my ($authorisedvaluecat) = $sth->fetchrow ) { my $authvalsth = $dbh->prepare( "select authorised_value, lib from authorised_values where category=? order by lib" ); $authvalsth->execute($authorisedvaluecat); while ( my ( $authorisedvalue, $lib ) = $authvalsth->fetchrow ) { $itemlocation{$authorisedvalue} = $lib; } $authvalsth->finish; return \%itemlocation; exit 1; } else { #No authvalue list # build default } $sth->finish; } #No authvalue list #build default $itemlocation{"1"} = "Not For Loan"; return \%itemlocation; } =head2 GetLostItems $items = GetLostItems($where,$orderby); This function get the items lost into C<$items>. =over 2 =item input: C<$where> is a hashref. it containts a field of the items table as key and the value to match as value. C<$orderby> is a field of the items table. =item return: C<$items> is a reference to an array full of hasref which keys are items' table column. =item usage in the perl script: my %where; $where{barcode} = 0001548; my $items = GetLostItems( \%where, "homebranch" ); $template->param(itemsloop => $items); =back =cut sub GetLostItems { # Getting input args. my $where = shift; my $orderby = shift; my $dbh = C4::Context->dbh; my $query = " SELECT * FROM items WHERE itemlost IS NOT NULL AND itemlost <> 0 "; foreach my $key (keys %$where) { $query .= " AND " . $key . " LIKE '%" . $where->{$key} . "%'"; } $query .= " ORDER BY ".$orderby if defined $orderby; my $sth = $dbh->prepare($query); $sth->execute; my @items; while ( my $row = $sth->fetchrow_hashref ){ push @items, $row; } return \@items; } =head2 GetItemsForInventory $itemlist = GetItemsForInventory($minlocation,$maxlocation,$datelastseen,$offset,$size) Retrieve a list of title/authors/barcode/callnumber, for biblio inventory. The sub returns a list of hashes, containing itemnumber, author, title, barcode & item callnumber. It is ordered by callnumber,title. The minlocation & maxlocation parameters are used to specify a range of item callnumbers the datelastseen can be used to specify that you want to see items not seen since a past date only. offset & size can be used to retrieve only a part of the whole listing (defaut behaviour) =cut sub GetItemsForInventory { my ( $minlocation, $maxlocation,$location, $datelastseen, $branch, $offset, $size ) = @_; my $dbh = C4::Context->dbh; my $sth; if ($datelastseen) { $datelastseen=format_date_in_iso($datelastseen); my $query = "SELECT itemnumber,barcode,itemcallnumber,title,author,biblio.biblionumber,datelastseen FROM items LEFT JOIN biblio ON items.biblionumber=biblio.biblionumber WHERE itemcallnumber>= ? AND itemcallnumber <=? AND (datelastseen< ? OR datelastseen IS NULL)"; $query.= " AND items.location=".$dbh->quote($location) if $location; $query.= " AND items.homebranch=".$dbh->quote($branch) if $branch; $query .= " ORDER BY itemcallnumber,title"; $sth = $dbh->prepare($query); $sth->execute( $minlocation, $maxlocation, $datelastseen ); } else { my $query =" SELECT itemnumber,barcode,itemcallnumber,biblio.biblionumber,title,author,datelastseen FROM items LEFT JOIN biblio ON items.biblionumber=biblio.biblionumber WHERE itemcallnumber>= ? AND itemcallnumber <=?"; $query.= " AND items.location=".$dbh->quote($location) if $location; $query.= " AND items.homebranch=".$dbh->quote($branch) if $branch; $query .= " ORDER BY itemcallnumber,title"; $sth = $dbh->prepare($query); $sth->execute( $minlocation, $maxlocation ); } my @results; while ( my $row = $sth->fetchrow_hashref ) { $offset-- if ($offset); $row->{datelastseen}=format_date($row->{datelastseen}); if ( ( !$offset ) && $size ) { push @results, $row; $size--; } } return \@results; } =head2 &GetBiblioItemData =over 4 $itemdata = &GetBiblioItemData($biblioitemnumber); Looks up the biblioitem with the given biblioitemnumber. Returns a reference-to-hash. The keys are the fields from the C, C, and C tables in the Koha database, except that C is given as C<$itemdata-E{bnotes}>. =back =cut #' sub GetBiblioItemData { my ($bibitem) = @_; my $dbh = C4::Context->dbh; my $sth = $dbh->prepare( "Select *,biblioitems.notes as bnotes from biblioitems, biblio,itemtypes where biblio.biblionumber = biblioitems.biblionumber and biblioitemnumber = ? and biblioitems.itemtype = itemtypes.itemtype" ); my $data; $sth->execute($bibitem); $data = $sth->fetchrow_hashref; $sth->finish; return ($data); } # sub &GetBiblioItemData =head2 GetItemnumberFromBarcode =over 4 $result = GetItemnumberFromBarcode($barcode); =back =cut sub GetItemnumberFromBarcode { my ($barcode) = @_; my $dbh = C4::Context->dbh; my $rq = $dbh->prepare("SELECT itemnumber from items where items.barcode=?"); $rq->execute($barcode); my ($result) = $rq->fetchrow; return ($result); } =head2 GetBiblioItemByBiblioNumber =over 4 NOTE : This function has been copy/paste from C4/Biblio.pm from head before zebra integration. =back =cut sub GetBiblioItemByBiblioNumber { my ($biblionumber) = @_; my $dbh = C4::Context->dbh; my $sth = $dbh->prepare("Select * from biblioitems where biblionumber = ?"); my $count = 0; my @results; $sth->execute($biblionumber); while ( my $data = $sth->fetchrow_hashref ) { push @results, $data; } $sth->finish; return @results; } =head2 GetBiblioFromItemNumber =over 4 $item = &GetBiblioFromItemNumber($itemnumber); Looks up the item with the given itemnumber. C<&itemnodata> returns a reference-to-hash whose keys are the fields from the C, C, and C tables in the Koha database. =back =cut #' sub GetBiblioFromItemNumber { my ( $itemnumber ) = @_; my $dbh = C4::Context->dbh; my $sth = $dbh->prepare( "SELECT * FROM biblio,items,biblioitems WHERE items.itemnumber = ? AND biblio.biblionumber = items.biblionumber AND biblioitems.biblioitemnumber = items.biblioitemnumber" ); $sth->execute($itemnumber); my $data = $sth->fetchrow_hashref; $sth->finish; return ($data); } =head2 GetBiblio =over 4 ( $count, @results ) = &GetBiblio($biblionumber); =back =cut sub GetBiblio { my ($biblionumber) = @_; my $dbh = C4::Context->dbh; my $sth = $dbh->prepare("Select * from biblio where biblionumber = ?"); my $count = 0; my @results; $sth->execute($biblionumber); while ( my $data = $sth->fetchrow_hashref ) { $results[$count] = $data; $count++; } # while $sth->finish; return ( $count, @results ); } # sub GetBiblio =head2 GetItem =over 4 $data = &GetItem($itemnumber,$barcode); return Item information, for a given itemnumber or barcode =back =cut sub GetItem { my ($itemnumber,$barcode) = @_; my $dbh = C4::Context->dbh; if ($itemnumber) { my $sth = $dbh->prepare(" SELECT * FROM items WHERE itemnumber = ?"); $sth->execute($itemnumber); my $data = $sth->fetchrow_hashref; return $data; } else { my $sth = $dbh->prepare(" SELECT * FROM items WHERE barcode = ?" ); $sth->execute($barcode); my $data = $sth->fetchrow_hashref; return $data; } } # sub GetItem =head2 get_itemnumbers_of =over 4 my @itemnumbers_of = get_itemnumbers_of(@biblionumbers); Given a list of biblionumbers, return the list of corresponding itemnumbers for each biblionumber. Return a reference on a hash where keys are biblionumbers and values are references on array of itemnumbers. =back =cut sub get_itemnumbers_of { my @biblionumbers = @_; my $dbh = C4::Context->dbh; my $query = ' SELECT itemnumber, biblionumber FROM items WHERE biblionumber IN (?' . ( ',?' x scalar @biblionumbers - 1 ) . ') '; my $sth = $dbh->prepare($query); $sth->execute(@biblionumbers); my %itemnumbers_of; while ( my ( $itemnumber, $biblionumber ) = $sth->fetchrow_array ) { push @{ $itemnumbers_of{$biblionumber} }, $itemnumber; } return \%itemnumbers_of; } =head2 GetItemInfosOf =over 4 GetItemInfosOf(@itemnumbers); =back =cut sub GetItemInfosOf { my @itemnumbers = @_; my $query = ' SELECT * FROM items WHERE itemnumber IN (' . join( ',', @itemnumbers ) . ') '; return get_infos_of( $query, 'itemnumber' ); } =head2 GetBiblioItemInfosOf =over 4 GetBiblioItemInfosOf(@biblioitemnumbers); =back =cut sub GetBiblioItemInfosOf { my @biblioitemnumbers = @_; my $query = ' SELECT biblioitemnumber, publicationyear, itemtype FROM biblioitems WHERE biblioitemnumber IN (' . join( ',', @biblioitemnumbers ) . ') '; return get_infos_of( $query, 'biblioitemnumber' ); } =head1 FUNCTIONS FOR HANDLING MARC MANAGEMENT =head2 GetMarcStructure =over 4 $res = GetMarcStructure($forlibrarian,$frameworkcode); Returns a reference to a big hash of hash, with the Marc structure for the given frameworkcode $forlibrarian :if set to 1, the MARC descriptions are the librarians ones, otherwise it's the public (OPAC) ones $frameworkcode : the framework code to read =back =cut sub GetMarcStructure { my ( $forlibrarian, $frameworkcode ) = @_; my $dbh=C4::Context->dbh; $frameworkcode = "" unless $frameworkcode; my $sth; my $libfield = ( $forlibrarian eq 1 ) ? 'liblibrarian' : 'libopac'; # check that framework exists $sth = $dbh->prepare( "select count(*) from marc_tag_structure where frameworkcode=?"); $sth->execute($frameworkcode); my ($total) = $sth->fetchrow; $frameworkcode = "" unless ( $total > 0 ); $sth = $dbh->prepare( "select tagfield,liblibrarian,libopac,mandatory,repeatable from marc_tag_structure where frameworkcode=? order by tagfield" ); $sth->execute($frameworkcode); my ( $liblibrarian, $libopac, $tag, $res, $tab, $mandatory, $repeatable ); while ( ( $tag, $liblibrarian, $libopac, $mandatory, $repeatable ) = $sth->fetchrow ) { $res->{$tag}->{lib} = ( $forlibrarian or !$libopac ) ? $liblibrarian : $libopac; # why the hell do we need to explicitly decode utf8 ? # that's a good question, but we must do it... $res->{$tab}->{tab} = ""; # XXX $res->{$tag}->{mandatory} = $mandatory; $res->{$tag}->{repeatable} = $repeatable; } $sth = $dbh->prepare( "select tagfield,tagsubfield,liblibrarian,libopac,tab, mandatory, repeatable,authorised_value,authtypecode,value_builder,kohafield,seealso,hidden,isurl,link,defaultvalue from marc_subfield_structure where frameworkcode=? order by tagfield,tagsubfield" ); $sth->execute($frameworkcode); my $subfield; my $authorised_value; my $authtypecode; my $value_builder; my $kohafield; my $seealso; my $hidden; my $isurl; my $link; my $defaultvalue; while ( ( $tag, $subfield, $liblibrarian, , $libopac, $tab, $mandatory, $repeatable, $authorised_value, $authtypecode, $value_builder, $kohafield, $seealso, $hidden, $isurl, $link,$defaultvalue ) = $sth->fetchrow ) { $res->{$tag}->{$subfield}->{lib} = ( $forlibrarian or !$libopac ) ? $liblibrarian : $libopac; $res->{$tag}->{$subfield}->{tab} = $tab; $res->{$tag}->{$subfield}->{mandatory} = $mandatory; $res->{$tag}->{$subfield}->{repeatable} = $repeatable; $res->{$tag}->{$subfield}->{authorised_value} = $authorised_value; $res->{$tag}->{$subfield}->{authtypecode} = $authtypecode; $res->{$tag}->{$subfield}->{value_builder} = $value_builder; $res->{$tag}->{$subfield}->{kohafield} = $kohafield; $res->{$tag}->{$subfield}->{seealso} = $seealso; $res->{$tag}->{$subfield}->{hidden} = $hidden; $res->{$tag}->{$subfield}->{isurl} = $isurl; $res->{$tag}->{$subfield}->{'link'} = $link; $res->{$tag}->{$subfield}->{defaultvalue} = $defaultvalue; } return $res; } =head2 GetUsedMarcStructure the same function as GetMarcStructure expcet it just take field in tab 0-9. (used field) my $results = GetUsedMarcStructure($frameworkcode); L<$results> is a ref to an array which each case containts a ref to a hash which each keys is the columns from marc_subfield_structure L<$frameworkcode> is the framework code. =cut sub GetUsedMarcStructure($){ my $frameworkcode = shift || ''; my $dbh = C4::Context->dbh; my $query = qq/ SELECT * FROM marc_subfield_structure WHERE tab > -1 AND frameworkcode = ? /; my @results; my $sth = $dbh->prepare($query); $sth->execute($frameworkcode); while (my $row = $sth->fetchrow_hashref){ push @results,$row; } return \@results; } =head2 GetMarcFromKohaField =over 4 ($MARCfield,$MARCsubfield)=GetMarcFromKohaField($kohafield,$frameworkcode); Returns the MARC fields & subfields mapped to the koha field for the given frameworkcode =back =cut sub GetMarcFromKohaField { my ( $kohafield, $frameworkcode ) = @_; return 0, 0 unless $kohafield; my $relations = C4::Context->marcfromkohafield; return ( $relations->{$frameworkcode}->{$kohafield}->[0], $relations->{$frameworkcode}->{$kohafield}->[1] ); } =head2 GetMarcBiblio =over 4 Returns MARC::Record of the biblionumber passed in parameter. the marc record contains both biblio & item datas =back =cut sub GetMarcBiblio { my $biblionumber = shift; my $dbh = C4::Context->dbh; my $sth = $dbh->prepare("select marcxml from biblioitems where biblionumber=? "); $sth->execute($biblionumber); my ($marcxml) = $sth->fetchrow; MARC::File::XML->default_record_format(C4::Context->preference('marcflavour')); $marcxml =~ s/\x1e//g; $marcxml =~ s/\x1f//g; $marcxml =~ s/\x1d//g; $marcxml =~ s/\x0f//g; $marcxml =~ s/\x0c//g; # warn $marcxml; my $record = MARC::Record->new(); $record = MARC::Record::new_from_xml( $marcxml, "utf8",C4::Context->preference('marcflavour')) if $marcxml; # $record = MARC::Record::new_from_usmarc( $marc) if $marc; return $record; } =head2 GetXmlBiblio =over 4 my $marcxml = GetXmlBiblio($biblionumber); Returns biblioitems.marcxml of the biblionumber passed in parameter. The XML contains both biblio & item datas =back =cut sub GetXmlBiblio { my ( $biblionumber ) = @_; my $dbh = C4::Context->dbh; my $sth = $dbh->prepare("select marcxml from biblioitems where biblionumber=? "); $sth->execute($biblionumber); my ($marcxml) = $sth->fetchrow; return $marcxml; } =head2 GetAuthorisedValueDesc =over 4 my $subfieldvalue =get_authorised_value_desc( $tag, $subf[$i][0],$subf[$i][1], '', $taglib); Retrieve the complete description for a given authorised value. =back =cut sub GetAuthorisedValueDesc { my ( $tag, $subfield, $value, $framework, $tagslib ) = @_; my $dbh = C4::Context->dbh; #---- branch if ( $tagslib->{$tag}->{$subfield}->{'authorised_value'} eq "branches" ) { return C4::Branch::GetBranchName($value); } #---- itemtypes if ( $tagslib->{$tag}->{$subfield}->{'authorised_value'} eq "itemtypes" ) { return getitemtypeinfo($value)->{description}; } #---- "true" authorized value my $category = $tagslib->{$tag}->{$subfield}->{'authorised_value'}; if ( $category ne "" ) { my $sth = $dbh->prepare( "select lib from authorised_values where category = ? and authorised_value = ?" ); $sth->execute( $category, $value ); my $data = $sth->fetchrow_hashref; return $data->{'lib'}; } else { return $value; # if nothing is found return the original value } } =head2 GetMarcItem =over 4 Returns MARC::Record of the item passed in parameter. =back =cut sub GetMarcItem { my ( $biblionumber, $itemnumber ) = @_; my $dbh = C4::Context->dbh; my $newrecord = MARC::Record->new(); my $marcflavour = C4::Context->preference('marcflavour'); my $marcxml = GetXmlBiblio($biblionumber); my $record = MARC::Record->new(); $record = MARC::Record::new_from_xml( $marcxml, "utf8", $marcflavour ); # now, find where the itemnumber is stored & extract only the item my ( $itemnumberfield, $itemnumbersubfield ) = GetMarcFromKohaField( 'items.itemnumber', '' ); my @fields = $record->field($itemnumberfield); foreach my $field (@fields) { if ( $field->subfield($itemnumbersubfield) eq $itemnumber ) { $newrecord->insert_fields_ordered($field); } } return $newrecord; } =head2 GetMarcNotes =over 4 $marcnotesarray = GetMarcNotes( $record, $marcflavour ); Get all notes from the MARC record and returns them in an array. The note are stored in differents places depending on MARC flavour =back =cut sub GetMarcNotes { my ( $record, $marcflavour ) = @_; my $scope; if ( $marcflavour eq "MARC21" ) { $scope = '5..'; } else { # assume unimarc if not marc21 $scope = '3..'; } my @marcnotes; my $note = ""; my $tag = ""; my $marcnote; foreach my $field ( $record->field($scope) ) { my $value = $field->as_string(); if ( $note ne "" ) { $marcnote = { marcnote => $note, }; push @marcnotes, $marcnote; $note = $value; } if ( $note ne $value ) { $note = $note . " " . $value; } } if ( $note ) { $marcnote = { marcnote => $note }; push @marcnotes, $marcnote; #load last tag into array } return \@marcnotes; } # end GetMarcNotes =head2 GetMarcSubjects =over 4 $marcsubjcts = GetMarcSubjects($record,$marcflavour); Get all subjects from the MARC record and returns them in an array. The subjects are stored in differents places depending on MARC flavour =back =cut sub GetMarcSubjects { my ( $record, $marcflavour ) = @_; my ( $mintag, $maxtag ); if ( $marcflavour eq "MARC21" ) { $mintag = "600"; $maxtag = "699"; } else { # assume unimarc if not marc21 $mintag = "600"; $maxtag = "611"; } my @marcsubjcts; foreach my $field ( $record->fields ) { next unless $field->tag() >= $mintag && $field->tag() <= $maxtag; my @subfields = $field->subfields(); my $link = "su:"; my $label; my $flag = 0; my $authoritysep=C4::Context->preference("authoritysep"); for my $subject_subfield ( @subfields ) { if ( $marcflavour ne 'MARC21' and ( ($subject_subfield->[0] eq '3') or ($subject_subfield->[0] eq '4') or ($subject_subfield->[0] eq '5') ) ) { next; } my $code = $subject_subfield->[0]; $label .= $subject_subfield->[1].$authoritysep unless ( $code == 9 ); $link .= " and su-to:".$subject_subfield->[1] unless ( $code == 9 ); if ( $code == 9 ) { $link = "an:".$subject_subfield->[1]; $flag = 1; } elsif ( ! $flag ) { $link =~ s/ and\ssu-to:$//; } } $label =~ s/$authoritysep$//; push @marcsubjcts, { label => $label, link => $link } } return \@marcsubjcts; } #end GetMarcSubjects =head2 GetMarcAuthors =over 4 authors = GetMarcAuthors($record,$marcflavour); Get all authors from the MARC record and returns them in an array. The authors are stored in differents places depending on MARC flavour =back =cut sub GetMarcAuthors { my ( $record, $marcflavour ) = @_; my ( $mintag, $maxtag ); # tagslib useful for UNIMARC author reponsabilities my $tagslib = &GetMarcStructure( 1, '' ); # FIXME : we don't have the framework available, we take the default framework. May be bugguy on some setups, will be usually correct. if ( $marcflavour eq "MARC21" ) { $mintag = "100"; $maxtag = "111"; } else { # assume unimarc if not marc21 $mintag = "701"; $maxtag = "712"; } my @marcauthors; foreach my $field ( $record->fields ) { next unless $field->tag() >= $mintag && $field->tag() <= $maxtag; my %hash; my @subfields = $field->subfields(); my $count_auth = 0; my $and ; for my $authors_subfield (@subfields) { if ( $marcflavour ne 'MARC21' and ( ($authors_subfield->[0] eq '3') or ($authors_subfield->[0] eq '5') ) ) { next; } if ($count_auth ne '0'){ $and = " and au:"; } $count_auth++; my $subfieldcode = $authors_subfield->[0]; my $value; # deal with UNIMARC author responsability if ( $marcflavour ne 'MARC21' and ($authors_subfield->[0] eq '4') ) { $value = "(".GetAuthorisedValueDesc( $field->tag(), $authors_subfield->[0], $authors_subfield->[1], '', $tagslib ).")"; } else { $value = $authors_subfield->[1]; } $hash{tag} = $field->tag; $hash{value} .= $value . " " if ($subfieldcode != 9) ; $hash{link} .= $value if ($subfieldcode eq 9); } push @marcauthors, \%hash; } return \@marcauthors; } =head2 GetMarcSeries =over 4 $marcseriessarray = GetMarcSeries($record,$marcflavour); Get all series from the MARC record and returns them in an array. The series are stored in differents places depending on MARC flavour =back =cut sub GetMarcSeries { my ($record, $marcflavour) = @_; my ($mintag, $maxtag); if ($marcflavour eq "MARC21") { $mintag = "440"; $maxtag = "490"; } else { # assume unimarc if not marc21 $mintag = "600"; $maxtag = "619"; } my @marcseries; my $subjct = ""; my $subfield = ""; my $marcsubjct; foreach my $field ($record->field('440'), $record->field('490')) { my @subfields_loop; #my $value = $field->subfield('a'); #$marcsubjct = {MARCSUBJCT => $value,}; my @subfields = $field->subfields(); #warn "subfields:".join " ", @$subfields; my $counter = 0; my @link_loop; for my $series_subfield (@subfields) { my $volume_number; undef $volume_number; # see if this is an instance of a volume if ($series_subfield->[0] eq 'v') { $volume_number=1; } my $code = $series_subfield->[0]; my $value = $series_subfield->[1]; my $linkvalue = $value; $linkvalue =~ s/(\(|\))//g; my $operator = " and " unless $counter==0; push @link_loop, {link => $linkvalue, operator => $operator }; my $separator = C4::Context->preference("authoritysep") unless $counter==0; if ($volume_number) { push @subfields_loop, {volumenum => $value}; } else { push @subfields_loop, {code => $code, value => $value, link_loop => \@link_loop, separator => $separator, volumenum => $volume_number}; } $counter++; } push @marcseries, { MARCSERIES_SUBFIELDS_LOOP => \@subfields_loop }; #$marcsubjct = {MARCSUBJCT => $field->as_string(),}; #push @marcsubjcts, $marcsubjct; #$subjct = $value; } my $marcseriessarray=\@marcseries; return $marcseriessarray; } #end getMARCseriess =head2 GetFrameworkCode =over 4 $frameworkcode = GetFrameworkCode( $biblionumber ) =back =cut sub GetFrameworkCode { my ( $biblionumber ) = @_; my $dbh = C4::Context->dbh; my $sth = $dbh->prepare("select frameworkcode from biblio where biblionumber=?"); $sth->execute($biblionumber); my ($frameworkcode) = $sth->fetchrow; return $frameworkcode; } =head2 GetPublisherNameFromIsbn $name = GetPublishercodeFromIsbn($isbn); if(defined $name){ ... } =cut sub GetPublisherNameFromIsbn($){ my $isbn = shift; $isbn =~ s/[- _]//g; $isbn =~ s/^0*//; my @codes = (split '-', DisplayISBN($isbn)); my $code = $codes[0].$codes[1].$codes[2]; my $dbh = C4::Context->dbh; my $query = qq{ SELECT distinct publishercode FROM biblioitems WHERE isbn LIKE ? AND publishercode IS NOT NULL LIMIT 1 }; my $sth = $dbh->prepare($query); $sth->execute("$code%"); my $name = $sth->fetchrow; return $name if length $name; return undef; } =head2 TransformKohaToMarc =over 4 $record = TransformKohaToMarc( $hash ) This function builds partial MARC::Record from a hash Hash entries can be from biblio or biblioitems. This function is called in acquisition module, to create a basic catalogue entry from user entry =back =cut sub TransformKohaToMarc { my ( $hash ) = @_; my $dbh = C4::Context->dbh; my $sth = $dbh->prepare( "select tagfield,tagsubfield from marc_subfield_structure where frameworkcode=? and kohafield=?" ); my $record = MARC::Record->new(); foreach (keys %{$hash}) { &TransformKohaToMarcOneField( $sth, $record, $_, $hash->{$_}, '' ); } return $record; } =head2 TransformKohaToMarcOneField =over 4 $record = TransformKohaToMarcOneField( $sth, $record, $kohafieldname, $value, $frameworkcode ); =back =cut sub TransformKohaToMarcOneField { my ( $sth, $record, $kohafieldname, $value, $frameworkcode ) = @_; $frameworkcode='' unless $frameworkcode; my $tagfield; my $tagsubfield; if ( !defined $sth ) { my $dbh = C4::Context->dbh; $sth = $dbh->prepare( "select tagfield,tagsubfield from marc_subfield_structure where frameworkcode=? and kohafield=?" ); } $sth->execute( $frameworkcode, $kohafieldname ); if ( ( $tagfield, $tagsubfield ) = $sth->fetchrow ) { my $tag = $record->field($tagfield); if ($tag) { $tag->update( $tagsubfield => $value ); $record->delete_field($tag); $record->insert_fields_ordered($tag); } else { $record->add_fields( $tagfield, " ", " ", $tagsubfield => $value ); } } return $record; } =head2 TransformHtmlToXml =over 4 $xml = TransformHtmlToXml( $tags, $subfields, $values, $indicator, $ind_tag, $auth_type ) $auth_type contains : - nothing : rebuild a biblio, un UNIMARC the encoding is in 100$a pos 26/27 - UNIMARCAUTH : rebuild an authority. In UNIMARC, the encoding is in 100$a pos 13/14 - ITEM : rebuild an item : in UNIMARC, 100$a, it's in the biblio ! (otherwise, we would get 2 100 fields !) =back =cut sub TransformHtmlToXml { my ( $tags, $subfields, $values, $indicator, $ind_tag, $auth_type ) = @_; my $xml = MARC::File::XML::header('UTF-8'); $auth_type = C4::Context->preference('marcflavour') unless $auth_type; MARC::File::XML->default_record_format($auth_type); # in UNIMARC, field 100 contains the encoding # check that there is one, otherwise the # MARC::Record->new_from_xml will fail (and Koha will die) my $unimarc_and_100_exist=0; $unimarc_and_100_exist=1 if $auth_type eq 'ITEM'; # if we rebuild an item, no need of a 100 field my $prevvalue; my $prevtag = -1; my $first = 1; my $j = -1; for ( my $i = 0 ; $i <= @$tags ; $i++ ) { if (C4::Context->preference('marcflavour') eq 'UNIMARC' and @$tags[$i] eq "100" and @$subfields[$i] eq "a") { # if we have a 100 field and it's values are not correct, skip them. # if we don't have any valid 100 field, we will create a default one at the end my $enc = substr( @$values[$i], 26, 2 ); if ($enc eq '01' or $enc eq '50' or $enc eq '03') { $unimarc_and_100_exist=1; } else { next; } } @$values[$i] =~ s/&/&/g; @$values[$i] =~ s//>/g; @$values[$i] =~ s/"/"/g; @$values[$i] =~ s/'/'/g; # if ( !utf8::is_utf8( @$values[$i] ) ) { # utf8::decode( @$values[$i] ); # } if ( ( @$tags[$i] ne $prevtag ) ) { $j++ unless ( @$tags[$i] eq "" ); if ( !$first ) { $xml .= "\n"; if ( ( @$tags[$i] && @$tags[$i] > 10 ) && ( @$values[$i] ne "" ) ) { my $ind1 = substr( @$indicator[$j], 0, 1 ); my $ind2; if ( @$indicator[$j] ) { $ind2 = substr( @$indicator[$j], 1, 1 ); } else { warn "Indicator in @$tags[$i] is empty"; $ind2 = " "; } $xml .= "\n"; $xml .= "@$values[$i]\n"; $first = 0; } else { $first = 1; } } else { if ( @$values[$i] ne "" ) { # leader if ( @$tags[$i] eq "000" ) { $xml .= "@$values[$i]\n"; $first = 1; # rest of the fixed fields } elsif ( @$tags[$i] < 10 ) { $xml .= "@$values[$i]\n"; $first = 1; } else { my $ind1 = substr( @$indicator[$j], 0, 1 ); my $ind2 = substr( @$indicator[$j], 1, 1 ); $xml .= "\n"; $xml .= "@$values[$i]\n"; $first = 0; } } } } else { # @$tags[$i] eq $prevtag if ( @$values[$i] eq "" ) { } else { if ($first) { my $ind1 = substr( @$indicator[$j], 0, 1 ); my $ind2 = substr( @$indicator[$j], 1, 1 ); $xml .= "\n"; $first = 0; } $xml .= "@$values[$i]\n"; } } $prevtag = @$tags[$i]; } if (C4::Context->preference('marcflavour') and !$unimarc_and_100_exist) { # warn "SETTING 100 for $auth_type"; use POSIX qw(strftime); my $string = strftime( "%Y%m%d", localtime(time) ); # set 50 to position 26 is biblios, 13 if authorities my $pos=26; $pos=13 if $auth_type eq 'UNIMARCAUTH'; $string = sprintf( "%-*s", 35, $string ); substr( $string, $pos , 6, "50" ); $xml .= "\n"; $xml .= "$string\n"; $xml .= "\n"; } $xml .= MARC::File::XML::footer(); return $xml; } =head2 TransformHtmlToMarc L<$record> = TransformHtmlToMarc(L<$params>,L<$cgi>) L<$params> is a ref to an array as below: { 'tag_010_indicator_531951' , 'tag_010_code_a_531951_145735' , 'tag_010_subfield_a_531951_145735' , 'tag_200_indicator_873510' , 'tag_200_code_a_873510_673465' , 'tag_200_subfield_a_873510_673465' , 'tag_200_code_b_873510_704318' , 'tag_200_subfield_b_873510_704318' , 'tag_200_code_e_873510_280822' , 'tag_200_subfield_e_873510_280822' , 'tag_200_code_f_873510_110730' , 'tag_200_subfield_f_873510_110730' , } L<$cgi> is the CGI object which containts the value. L<$record> is the MARC::Record object. =cut sub TransformHtmlToMarc { my $params = shift; my $cgi = shift; # creating a new record my $record = MARC::Record->new(); my $i=0; my @fields; while ($params->[$i]){ # browse all CGI params my $param = $params->[$i]; my $newfield=0; # if we are on biblionumber, store it in the MARC::Record (it may not be in the edited fields) if ($param eq 'biblionumber') { my ( $biblionumbertagfield, $biblionumbertagsubfield ) = &GetMarcFromKohaField( "biblio.biblionumber", '' ); if ($biblionumbertagfield < 10) { $newfield = MARC::Field->new( $biblionumbertagfield, $cgi->param($param), ); } else { $newfield = MARC::Field->new( $biblionumbertagfield, '', '', "$biblionumbertagsubfield" => $cgi->param($param), ); } push @fields,$newfield if($newfield); } elsif ($param =~ /^tag_(\d*)_indicator_/){ # new field start when having 'input name="..._indicator_..." my $tag = $1; my $ind1 = substr($cgi->param($param),0,1); my $ind2 = substr($cgi->param($param),1,1); $newfield=0; my $j=$i+1; if($tag < 10){ # no code for theses fields # in MARC editor, 000 contains the leader. if ($tag eq '000' ) { $record->leader($cgi->param($params->[$j+1])) if length($cgi->param($params->[$j+1]))==24; # between 001 and 009 (included) } else { $newfield = MARC::Field->new( $tag, $cgi->param($params->[$j+1]), ); } # > 009, deal with subfields } else { while($params->[$j] =~ /_code_/){ # browse all it's subfield my $inner_param = $params->[$j]; if ($newfield){ if($cgi->param($params->[$j+1])){ # only if there is a value (code => value) $newfield->add_subfields( $cgi->param($inner_param) => $cgi->param($params->[$j+1]) ); } } else { if ( $cgi->param($params->[$j+1]) ) { # creating only if there is a value (code => value) $newfield = MARC::Field->new( $tag, ''.$ind1, ''.$ind2, $cgi->param($inner_param) => $cgi->param($params->[$j+1]), ); } } $j+=2; } } push @fields,$newfield if($newfield); } $i++; } $record->append_fields(@fields); return $record; } =head2 TransformMarcToKoha =over 4 $result = TransformMarcToKoha( $dbh, $record, $frameworkcode ) =back =cut sub TransformMarcToKoha { my ( $dbh, $record, $frameworkcode ) = @_; # FIXME :: This query is unused.. # my $sth = # $dbh->prepare( #"select tagfield,tagsubfield from marc_subfield_structure where frameworkcode=? and kohafield=?" # ); my $result; my $sth2 = $dbh->prepare("SHOW COLUMNS from biblio"); $sth2->execute; my $field; while ( ($field) = $sth2->fetchrow ) { $result = &TransformMarcToKohaOneField( "biblio", $field, $record, $result, $frameworkcode ); } my $sth2 = $dbh->prepare("SHOW COLUMNS from biblioitems"); $sth2->execute; while ( ($field) = $sth2->fetchrow ) { if ( $field eq 'notes' ) { $field = 'bnotes'; } $result = &TransformMarcToKohaOneField( "biblioitems", $field, $record, $result, $frameworkcode ); } $sth2 = $dbh->prepare("SHOW COLUMNS from items"); $sth2->execute; while ( ($field) = $sth2->fetchrow ) { $result = &TransformMarcToKohaOneField( "items", $field, $record, $result, $frameworkcode ); } # # modify copyrightdate to keep only the 1st year found my $temp = $result->{'copyrightdate'}; $temp =~ m/c(\d\d\d\d)/; # search cYYYY first if ( $1 > 0 ) { $result->{'copyrightdate'} = $1; } else { # if no cYYYY, get the 1st date. $temp =~ m/(\d\d\d\d)/; $result->{'copyrightdate'} = $1; } # modify publicationyear to keep only the 1st year found $temp = $result->{'publicationyear'}; $temp =~ m/c(\d\d\d\d)/; # search cYYYY first if ( $1 > 0 ) { $result->{'publicationyear'} = $1; } else { # if no cYYYY, get the 1st date. $temp =~ m/(\d\d\d\d)/; $result->{'publicationyear'} = $1; } return $result; } =head2 TransformMarcToKohaOneField =over 4 $result = TransformMarcToKohaOneField( $kohatable, $kohafield, $record, $result, $frameworkcode ) =back =cut sub TransformMarcToKohaOneField { # FIXME ? if a field has a repeatable subfield that is used in old-db, # only the 1st will be retrieved... my ( $kohatable, $kohafield, $record, $result, $frameworkcode ) = @_; my $res = ""; my ( $tagfield, $subfield ) = GetMarcFromKohaField( $kohatable . "." . $kohafield, $frameworkcode ); foreach my $field ( $record->field($tagfield) ) { if ( $field->tag() < 10 ) { if ( $result->{$kohafield} ) { $result->{$kohafield} .= " | " . $field->data(); } else { $result->{$kohafield} = $field->data(); } } else { if ( $field->subfields ) { my @subfields = $field->subfields(); foreach my $subfieldcount ( 0 .. $#subfields ) { if ( $subfields[$subfieldcount][0] eq $subfield ) { if ( $result->{$kohafield} ) { $result->{$kohafield} .= " | " . $subfields[$subfieldcount][1]; } else { $result->{$kohafield} = $subfields[$subfieldcount][1]; } } } } } } return $result; } =head1 OTHER FUNCTIONS =head2 char_decode =over 4 my $string = char_decode( $string, $encoding ); converts ISO 5426 coded string to UTF-8 sloppy code : should be improved in next issue =back =cut sub char_decode { my ( $string, $encoding ) = @_; $_ = $string; $encoding = C4::Context->preference("marcflavour") unless $encoding; if ( $encoding eq "UNIMARC" ) { # s/\xe1/Æ/gm; s/\xe2/Ğ/gm; s/\xe9/Ø/gm; s/\xec/ş/gm; s/\xf1/æ/gm; s/\xf3/ğ/gm; s/\xf9/ø/gm; s/\xfb/ß/gm; s/\xc1\x61/à/gm; s/\xc1\x65/è/gm; s/\xc1\x69/ì/gm; s/\xc1\x6f/ò/gm; s/\xc1\x75/ù/gm; s/\xc1\x41/À/gm; s/\xc1\x45/È/gm; s/\xc1\x49/Ì/gm; s/\xc1\x4f/Ò/gm; s/\xc1\x55/Ù/gm; s/\xc2\x41/Á/gm; s/\xc2\x45/É/gm; s/\xc2\x49/Í/gm; s/\xc2\x4f/Ó/gm; s/\xc2\x55/Ú/gm; s/\xc2\x59/İ/gm; s/\xc2\x61/á/gm; s/\xc2\x65/é/gm; s/\xc2\x69/í/gm; s/\xc2\x6f/ó/gm; s/\xc2\x75/ú/gm; s/\xc2\x79/ı/gm; s/\xc3\x41/Â/gm; s/\xc3\x45/Ê/gm; s/\xc3\x49/Î/gm; s/\xc3\x4f/Ô/gm; s/\xc3\x55/Û/gm; s/\xc3\x61/â/gm; s/\xc3\x65/ê/gm; s/\xc3\x69/î/gm; s/\xc3\x6f/ô/gm; s/\xc3\x75/û/gm; s/\xc4\x41/Ã/gm; s/\xc4\x4e/Ñ/gm; s/\xc4\x4f/Õ/gm; s/\xc4\x61/ã/gm; s/\xc4\x6e/ñ/gm; s/\xc4\x6f/õ/gm; s/\xc8\x41/Ä/gm; s/\xc8\x45/Ë/gm; s/\xc8\x49/Ï/gm; s/\xc8\x61/ä/gm; s/\xc8\x65/ë/gm; s/\xc8\x69/ï/gm; s/\xc8\x6F/ö/gm; s/\xc8\x75/ü/gm; s/\xc8\x76/ÿ/gm; s/\xc9\x41/Ä/gm; s/\xc9\x45/Ë/gm; s/\xc9\x49/Ï/gm; s/\xc9\x4f/Ö/gm; s/\xc9\x55/Ü/gm; s/\xc9\x61/ä/gm; s/\xc9\x6f/ö/gm; s/\xc9\x75/ü/gm; s/\xca\x41/Å/gm; s/\xca\x61/å/gm; s/\xd0\x43/Ç/gm; s/\xd0\x63/ç/gm; # this handles non-sorting blocks (if implementation requires this) $string = nsb_clean($_); } elsif ( $encoding eq "USMARC" || $encoding eq "MARC21" ) { ##MARC-8 to UTF-8 s/\xe1\x61/à/gm; s/\xe1\x65/è/gm; s/\xe1\x69/ì/gm; s/\xe1\x6f/ò/gm; s/\xe1\x75/ù/gm; s/\xe1\x41/À/gm; s/\xe1\x45/È/gm; s/\xe1\x49/Ì/gm; s/\xe1\x4f/Ò/gm; s/\xe1\x55/Ù/gm; s/\xe2\x41/Á/gm; s/\xe2\x45/É/gm; s/\xe2\x49/Í/gm; s/\xe2\x4f/Ó/gm; s/\xe2\x55/Ú/gm; s/\xe2\x59/İ/gm; s/\xe2\x61/á/gm; s/\xe2\x65/é/gm; s/\xe2\x69/í/gm; s/\xe2\x6f/ó/gm; s/\xe2\x75/ú/gm; s/\xe2\x79/ı/gm; s/\xe3\x41/Â/gm; s/\xe3\x45/Ê/gm; s/\xe3\x49/Î/gm; s/\xe3\x4f/Ô/gm; s/\xe3\x55/Û/gm; s/\xe3\x61/â/gm; s/\xe3\x65/ê/gm; s/\xe3\x69/î/gm; s/\xe3\x6f/ô/gm; s/\xe3\x75/û/gm; s/\xe4\x41/Ã/gm; s/\xe4\x4e/Ñ/gm; s/\xe4\x4f/Õ/gm; s/\xe4\x61/ã/gm; s/\xe4\x6e/ñ/gm; s/\xe4\x6f/õ/gm; s/\xe6\x41/Ă/gm; s/\xe6\x45/Ĕ/gm; s/\xe6\x65/ĕ/gm; s/\xe6\x61/ă/gm; s/\xe8\x45/Ë/gm; s/\xe8\x49/Ï/gm; s/\xe8\x65/ë/gm; s/\xe8\x69/ï/gm; s/\xe8\x76/ÿ/gm; s/\xe9\x41/A/gm; s/\xe9\x4f/O/gm; s/\xe9\x55/U/gm; s/\xe9\x61/a/gm; s/\xe9\x6f/o/gm; s/\xe9\x75/u/gm; s/\xea\x41/A/gm; s/\xea\x61/a/gm; #Additional Turkish characters s/\x1b//gm; s/\x1e//gm; s/(\xf0)s/\xc5\x9f/gm; s/(\xf0)S/\xc5\x9e/gm; s/(\xf0)c/ç/gm; s/(\xf0)C/Ç/gm; s/\xe7\x49/\\xc4\xb0/gm; s/(\xe6)G/\xc4\x9e/gm; s/(\xe6)g/ğ\xc4\x9f/gm; s/\xB8/ı/gm; s/\xB9/£/gm; s/(\xe8|\xc8)o/ö/gm; s/(\xe8|\xc8)O/Ö/gm; s/(\xe8|\xc8)u/ü/gm; s/(\xe8|\xc8)U/Ü/gm; s/\xc2\xb8/\xc4\xb1/gm; s/¸/\xc4\xb1/gm; # this handles non-sorting blocks (if implementation requires this) $string = nsb_clean($_); } return ($string); } =head2 nsb_clean =over 4 my $string = nsb_clean( $string, $encoding ); =back =cut sub nsb_clean { my $NSB = '\x88'; # NSB : begin Non Sorting Block my $NSE = '\x89'; # NSE : Non Sorting Block end # handles non sorting blocks my ($string) = @_; $_ = $string; s/$NSB/(/gm; s/[ ]{0,1}$NSE/) /gm; $string = $_; return ($string); } =head2 PrepareItemrecordDisplay =over 4 PrepareItemrecordDisplay($itemrecord,$bibnum,$itemumber); Returns a hash with all the fields for Display a given item data in a template =back =cut sub PrepareItemrecordDisplay { my ( $bibnum, $itemnum ) = @_; my $dbh = C4::Context->dbh; my $frameworkcode = &GetFrameworkCode( $bibnum ); my ( $itemtagfield, $itemtagsubfield ) = &GetMarcFromKohaField( "items.itemnumber", $frameworkcode ); my $tagslib = &GetMarcStructure( 1, $frameworkcode ); my $itemrecord = GetMarcItem( $bibnum, $itemnum) if ($itemnum); my @loop_data; my $authorised_values_sth = $dbh->prepare( "select authorised_value,lib from authorised_values where category=? order by lib" ); foreach my $tag ( sort keys %{$tagslib} ) { my $previous_tag = ''; if ( $tag ne '' ) { # loop through each subfield my $cntsubf; foreach my $subfield ( sort keys %{ $tagslib->{$tag} } ) { next if ( subfield_is_koha_internal_p($subfield) ); next if ( $tagslib->{$tag}->{$subfield}->{'tab'} ne "10" ); my %subfield_data; $subfield_data{tag} = $tag; $subfield_data{subfield} = $subfield; $subfield_data{countsubfield} = $cntsubf++; $subfield_data{kohafield} = $tagslib->{$tag}->{$subfield}->{'kohafield'}; # $subfield_data{marc_lib}=$tagslib->{$tag}->{$subfield}->{lib}; $subfield_data{marc_lib} = "{$tag}->{$subfield}->{lib} . "\">" . substr( $tagslib->{$tag}->{$subfield}->{lib}, 0, 12 ) . ""; $subfield_data{mandatory} = $tagslib->{$tag}->{$subfield}->{mandatory}; $subfield_data{repeatable} = $tagslib->{$tag}->{$subfield}->{repeatable}; $subfield_data{hidden} = "display:none" if $tagslib->{$tag}->{$subfield}->{hidden}; my ( $x, $value ); ( $x, $value ) = _find_value( $tag, $subfield, $itemrecord ) if ($itemrecord); $value =~ s/"/"/g; # search for itemcallnumber if applicable if ( $tagslib->{$tag}->{$subfield}->{kohafield} eq 'items.itemcallnumber' && C4::Context->preference('itemcallnumber') ) { my $CNtag = substr( C4::Context->preference('itemcallnumber'), 0, 3 ); my $CNsubfield = substr( C4::Context->preference('itemcallnumber'), 3, 1 ); my $temp = $itemrecord->field($CNtag) if ($itemrecord); if ($temp) { $value = $temp->subfield($CNsubfield); } } if ( $tagslib->{$tag}->{$subfield}->{authorised_value} ) { my @authorised_values; my %authorised_lib; # builds list, depending on authorised value... #---- branch if ( $tagslib->{$tag}->{$subfield}->{'authorised_value'} eq "branches" ) { if ( ( C4::Context->preference("IndependantBranches") ) && ( C4::Context->userenv->{flags} != 1 ) ) { my $sth = $dbh->prepare( "select branchcode,branchname from branches where branchcode = ? order by branchname" ); $sth->execute( C4::Context->userenv->{branch} ); push @authorised_values, "" unless ( $tagslib->{$tag}->{$subfield}->{mandatory} ); while ( my ( $branchcode, $branchname ) = $sth->fetchrow_array ) { push @authorised_values, $branchcode; $authorised_lib{$branchcode} = $branchname; } } else { my $sth = $dbh->prepare( "select branchcode,branchname from branches order by branchname" ); $sth->execute; push @authorised_values, "" unless ( $tagslib->{$tag}->{$subfield}->{mandatory} ); while ( my ( $branchcode, $branchname ) = $sth->fetchrow_array ) { push @authorised_values, $branchcode; $authorised_lib{$branchcode} = $branchname; } } #----- itemtypes } elsif ( $tagslib->{$tag}->{$subfield}->{authorised_value} eq "itemtypes" ) { my $sth = $dbh->prepare( "select itemtype,description from itemtypes order by description" ); $sth->execute; push @authorised_values, "" unless ( $tagslib->{$tag}->{$subfield}->{mandatory} ); while ( my ( $itemtype, $description ) = $sth->fetchrow_array ) { push @authorised_values, $itemtype; $authorised_lib{$itemtype} = $description; } #---- "true" authorised value } else { $authorised_values_sth->execute( $tagslib->{$tag}->{$subfield}->{authorised_value} ); push @authorised_values, "" unless ( $tagslib->{$tag}->{$subfield}->{mandatory} ); while ( my ( $value, $lib ) = $authorised_values_sth->fetchrow_array ) { push @authorised_values, $value; $authorised_lib{$value} = $lib; } } $subfield_data{marc_value} = CGI::scrolling_list( -name => 'field_value', -values => \@authorised_values, -default => "$value", -labels => \%authorised_lib, -size => 1, -tabindex => '', -multiple => 0, ); } elsif ( $tagslib->{$tag}->{$subfield}->{thesaurus_category} ) { $subfield_data{marc_value} = " {$tag}->{$subfield}->{thesaurus_category}&index=',)\">..."; #" # COMMENTED OUT because No $i is provided with this API. # And thus, no value_builder can be activated. # BUT could be thought over. # } elsif ($tagslib->{$tag}->{$subfield}->{'value_builder'}) { # my $plugin="value_builder/".$tagslib->{$tag}->{$subfield}->{'value_builder'}; # require $plugin; # my $extended_param = plugin_parameters($dbh,$itemrecord,$tagslib,$i,0); # my ($function_name,$javascript) = plugin_javascript($dbh,$record,$tagslib,$i,0); # $subfield_data{marc_value}=" ... $javascript"; } else { $subfield_data{marc_value} = ""; } push( @loop_data, \%subfield_data ); } } } my $itemnumber = $itemrecord->subfield( $itemtagfield, $itemtagsubfield ) if ( $itemrecord && $itemrecord->field($itemtagfield) ); return { 'itemtagfield' => $itemtagfield, 'itemtagsubfield' => $itemtagsubfield, 'itemnumber' => $itemnumber, 'iteminformation' => \@loop_data }; } #" # # true ModZebra commented until indexdata fixes zebraDB crashes (it seems they occur on multiple updates # at the same time # replaced by a zebraqueue table, that is filled with ModZebra to run. # the table is emptied by misc/cronjobs/zebraqueue_start.pl script # =head2 ModZebrafiles # # &ModZebrafiles( $dbh, $biblionumber, $record, $folder, $server ); # # =cut # # sub ModZebrafiles { # # my ( $dbh, $biblionumber, $record, $folder, $server ) = @_; # # my $op; # my $zebradir = # C4::Context->zebraconfig($server)->{directory} . "/" . $folder . "/"; # unless ( opendir( DIR, "$zebradir" ) ) { # warn "$zebradir not found"; # return; # } # closedir DIR; # my $filename = $zebradir . $biblionumber; # # if ($record) { # open( OUTPUT, ">", $filename . ".xml" ); # print OUTPUT $record; # close OUTPUT; # } # } =head2 ModZebra =over 4 ModZebra( $biblionumber, $op, $server, $newRecord ); $biblionumber is the biblionumber we want to index $op is specialUpdate or delete, and is used to know what we want to do $server is the server that we want to update $newRecord is the MARC::Record containing the new record. It is usefull only when NoZebra=1, and is used to know what to add to the nozebra database. (the record in mySQL being, if it exist, the previous record, the one just before the modif. We need both : the previous and the new one. =back =cut sub ModZebra { ###Accepts a $server variable thus we can use it for biblios authorities or other zebra dbs my ( $biblionumber, $op, $server, $newRecord ) = @_; my $dbh=C4::Context->dbh; # true ModZebra commented until indexdata fixes zebraDB crashes (it seems they occur on multiple updates # at the same time # replaced by a zebraqueue table, that is filled with ModZebra to run. # the table is emptied by misc/cronjobs/zebraqueue_start.pl script if (C4::Context->preference("NoZebra")) { # lock the nozebra table : we will read index lines, update them in Perl process # and write everything in 1 transaction. # lock the table to avoid someone else overwriting what we are doing $dbh->do('LOCK TABLES nozebra WRITE,biblio WRITE,biblioitems WRITE, systempreferences WRITE, auth_types WRITE, auth_header WRITE'); my %result; # the result hash that will be builded by deletion / add, and written on mySQL at the end, to improve speed my $record; if ($server eq 'biblioserver') { $record= GetMarcBiblio($biblionumber); } else { $record= C4::AuthoritiesMarc::GetAuthority($biblionumber); } if ($op eq 'specialUpdate') { # OK, we have to add or update the record # 1st delete (virtually, in indexes) ... %result = _DelBiblioNoZebra($biblionumber,$record,$server); # ... add the record %result=_AddBiblioNoZebra($biblionumber,$newRecord, $server, %result); } else { # it's a deletion, delete the record... # warn "DELETE the record $biblionumber on $server".$record->as_formatted; %result=_DelBiblioNoZebra($biblionumber,$record,$server); } # ok, now update the database... my $sth = $dbh->prepare("UPDATE nozebra SET biblionumbers=? WHERE server=? AND indexname=? AND value=?"); foreach my $key (keys %result) { foreach my $index (keys %{$result{$key}}) { $sth->execute($result{$key}->{$index}, $server, $key, $index); } } $dbh->do('UNLOCK TABLES'); } else { # # we use zebra, just fill zebraqueue table # my $sth=$dbh->prepare("insert into zebraqueue (biblio_auth_number ,server,operation) values(?,?,?)"); $sth->execute($biblionumber,$server,$op); $sth->finish; } } =head2 GetNoZebraIndexes %indexes = GetNoZebraIndexes; return the data from NoZebraIndexes syspref. =cut sub GetNoZebraIndexes { my $index = C4::Context->preference('NoZebraIndexes'); my %indexes; foreach my $line (split /('|"),/,$index) { $line =~ /(.*)=>(.*)/; my $index = substr($1,1); # get the index, don't forget to remove initial ' or " my $fields = $2; $index =~ s/'|"| //g; $fields =~ s/'|"| //g; $indexes{$index}=$fields; } return %indexes; } =head1 INTERNAL FUNCTIONS =head2 _DelBiblioNoZebra($biblionumber,$record,$server); function to delete a biblio in NoZebra indexes This function does NOT delete anything in database : it reads all the indexes entries that have to be deleted & delete them in the hash The SQL part is done either : - after the Add if we are modifying a biblio (delete + add again) - immediatly after this sub if we are doing a true deletion. $server can be 'biblioserver' or 'authorityserver' : it indexes biblios or authorities (in the same table, $server being part of the table itself =cut sub _DelBiblioNoZebra { my ($biblionumber, $record, $server)=@_; # Get the indexes my $dbh = C4::Context->dbh; # Get the indexes my %index; my $title; if ($server eq 'biblioserver') { %index=GetNoZebraIndexes; # get title of the record (to store the 10 first letters with the index) my ($titletag,$titlesubfield) = GetMarcFromKohaField('biblio.title'); $title = lc($record->subfield($titletag,$titlesubfield)); } else { # for authorities, the "title" is the $a mainentry my $authref = C4::AuthoritiesMarc::GetAuthType($record->subfield(152,'b')); warn "ERROR : authtype undefined for ".$record->as_formatted unless $authref; $title = $record->subfield($authref->{auth_tag_to_report},'a'); $index{'mainmainentry'}= $authref->{'auth_tag_to_report'}.'a'; $index{'mainentry'} = $authref->{'auth_tag_to_report'}.'*'; $index{'auth_type'} = '152b'; } my %result; # remove blancks comma (that could cause problem when decoding the string for CQL retrieval) and regexp specific values $title =~ s/ |,|;|\[|\]|\(|\)|\*|-|'|=//g; # limit to 10 char, should be enough, and limit the DB size $title = substr($title,0,10); #parse each field my $sth2=$dbh->prepare('SELECT biblionumbers FROM nozebra WHERE server=? AND indexname=? AND value=?'); foreach my $field ($record->fields()) { #parse each subfield next if $field->tag <10; foreach my $subfield ($field->subfields()) { my $tag = $field->tag(); my $subfieldcode = $subfield->[0]; my $indexed=0; # check each index to see if the subfield is stored somewhere # otherwise, store it in __RAW__ index foreach my $key (keys %index) { # warn "examining $key index : ".$index{$key}." for $tag $subfieldcode"; if ($index{$key} =~ /$tag\*/ or $index{$key} =~ /$tag$subfieldcode/) { $indexed=1; my $line= lc $subfield->[1]; # remove meaningless value in the field... $line =~ s/-|\.|\?|,|;|!|'|\(|\)|\[|\]|{|}|"|<|>|&|\+|\*|\/|=|:/ /g; # ... and split in words foreach (split / /,$line) { next unless $_; # skip empty values (multiple spaces) # if the entry is already here, do nothing, the biblionumber has already be removed unless ($result{$key}->{$_} =~ /$biblionumber,$title\-(\d);/) { # get the index value if it exist in the nozebra table and remove the entry, otherwise, do nothing $sth2->execute($server,$key,$_); my $existing_biblionumbers = $sth2->fetchrow; # it exists if ($existing_biblionumbers) { # warn " existing for $key $_: $existing_biblionumbers"; $result{$key}->{$_} =$existing_biblionumbers; $result{$key}->{$_} =~ s/$biblionumber,$title\-(\d);//; } } } } } # the subfield is not indexed, store it in __RAW__ index anyway unless ($indexed) { my $line= lc $subfield->[1]; $line =~ s/-|\.|\?|,|;|!|'|\(|\)|\[|\]|{|}|"|<|>|&|\+|\*|\/|=|:/ /g; # ... and split in words foreach (split / /,$line) { next unless $_; # skip empty values (multiple spaces) # if the entry is already here, do nothing, the biblionumber has already be removed unless ($result{'__RAW__'}->{$_} =~ /$biblionumber,$title\-(\d);/) { # get the index value if it exist in the nozebra table and remove the entry, otherwise, do nothing $sth2->execute($server,'__RAW__',$_); my $existing_biblionumbers = $sth2->fetchrow; # it exists if ($existing_biblionumbers) { $result{'__RAW__'}->{$_} =$existing_biblionumbers; $result{'__RAW__'}->{$_} =~ s/$biblionumber,$title\-(\d);//; } } } } } } return %result; } =head2 _AddBiblioNoZebra($biblionumber, $record, $server, %result); function to add a biblio in NoZebra indexes =cut sub _AddBiblioNoZebra { my ($biblionumber, $record, $server, %result)=@_; my $dbh = C4::Context->dbh; # Get the indexes my %index; my $title; if ($server eq 'biblioserver') { %index=GetNoZebraIndexes; # get title of the record (to store the 10 first letters with the index) my ($titletag,$titlesubfield) = GetMarcFromKohaField('biblio.title'); $title = lc($record->subfield($titletag,$titlesubfield)); } else { # warn "server : $server"; # for authorities, the "title" is the $a mainentry my $authref = C4::AuthoritiesMarc::GetAuthType($record->subfield(152,'b')); warn "ERROR : authtype undefined for ".$record->as_formatted unless $authref; $title = $record->subfield($authref->{auth_tag_to_report},'a'); $index{'mainmainentry'} = $authref->{auth_tag_to_report}.'a'; $index{'mainentry'} = $authref->{auth_tag_to_report}.'*'; $index{'auth_type'} = '152b'; } # remove blancks comma (that could cause problem when decoding the string for CQL retrieval) and regexp specific values $title =~ s/ |,|;|\[|\]|\(|\)|\*|-|'|=//g; # limit to 10 char, should be enough, and limit the DB size $title = substr($title,0,10); #parse each field my $sth2=$dbh->prepare('SELECT biblionumbers FROM nozebra WHERE server=? AND indexname=? AND value=?'); foreach my $field ($record->fields()) { #parse each subfield next if $field->tag <10; foreach my $subfield ($field->subfields()) { my $tag = $field->tag(); my $subfieldcode = $subfield->[0]; my $indexed=0; # check each index to see if the subfield is stored somewhere # otherwise, store it in __RAW__ index foreach my $key (keys %index) { # warn "examining $key index : ".$index{$key}." for $tag $subfieldcode"; if ($index{$key} =~ /$tag\*/ or $index{$key} =~ /$tag$subfieldcode/) { $indexed=1; my $line= lc $subfield->[1]; # remove meaningless value in the field... $line =~ s/-|\.|\?|,|;|!|'|\(|\)|\[|\]|{|}|"|<|>|&|\+|\*|\/|=|:/ /g; # ... and split in words foreach (split / /,$line) { next unless $_; # skip empty values (multiple spaces) # if the entry is already here, improve weight # warn "managing $_"; if ($result{$key}->{"$_"} =~ /$biblionumber,$title\-(\d);/) { my $weight=$1+1; $result{$key}->{"$_"} =~ s/$biblionumber,$title\-(\d);//; $result{$key}->{"$_"} .= "$biblionumber,$title-$weight;"; } else { # get the value if it exist in the nozebra table, otherwise, create it $sth2->execute($server,$key,$_); my $existing_biblionumbers = $sth2->fetchrow; # it exists if ($existing_biblionumbers) { $result{$key}->{"$_"} =$existing_biblionumbers; my $weight=$1+1; $result{$key}->{"$_"} =~ s/$biblionumber,$title\-(\d);//; $result{$key}->{"$_"} .= "$biblionumber,$title-$weight;"; # create a new ligne for this entry } else { # warn "INSERT : $server / $key / $_"; $dbh->do('INSERT INTO nozebra SET server='.$dbh->quote($server).', indexname='.$dbh->quote($key).',value='.$dbh->quote($_)); $result{$key}->{"$_"}.="$biblionumber,$title-1;"; } } } } } # the subfield is not indexed, store it in __RAW__ index anyway unless ($indexed) { my $line= lc $subfield->[1]; $line =~ s/-|\.|\?|,|;|!|'|\(|\)|\[|\]|{|}|"|<|>|&|\+|\*|\/|=|:/ /g; # ... and split in words foreach (split / /,$line) { next unless $_; # skip empty values (multiple spaces) # if the entry is already here, improve weight if ($result{'__RAW__'}->{"$_"} =~ /$biblionumber,$title\-(\d);/) { my $weight=$1+1; $result{'__RAW__'}->{"$_"} =~ s/$biblionumber,$title\-(\d);//; $result{'__RAW__'}->{"$_"} .= "$biblionumber,$title-$weight;"; } else { # get the value if it exist in the nozebra table, otherwise, create it $sth2->execute($server,'__RAW__',$_); my $existing_biblionumbers = $sth2->fetchrow; # it exists if ($existing_biblionumbers) { $result{'__RAW__'}->{"$_"} =$existing_biblionumbers; my $weight=$1+1; $result{'__RAW__'}->{"$_"} =~ s/$biblionumber,$title\-(\d);//; $result{'__RAW__'}->{"$_"} .= "$biblionumber,$title-$weight;"; # create a new ligne for this entry } else { $dbh->do('INSERT INTO nozebra SET server='.$dbh->quote($server).', indexname="__RAW__",value='.$dbh->quote($_)); $result{'__RAW__'}->{"$_"}.="$biblionumber,$title-1;"; } } } } } } return %result; } =head2 MARCitemchange =over 4 &MARCitemchange( $record, $itemfield, $newvalue ) Function to update a single value in an item field. Used twice, could probably be replaced by something else, but works well... =back =back =cut sub MARCitemchange { my ( $record, $itemfield, $newvalue ) = @_; my $dbh = C4::Context->dbh; my ( $tagfield, $tagsubfield ) = GetMarcFromKohaField( $itemfield, "" ); if ( ($tagfield) && ($tagsubfield) ) { my $tag = $record->field($tagfield); if ($tag) { $tag->update( $tagsubfield => $newvalue ); $record->delete_field($tag); $record->insert_fields_ordered($tag); } } } =head2 _koha_add_biblio =over 4 _koha_add_biblio($dbh,$biblioitem); Internal function to add a biblio ($biblio is a hash with the values) =back =cut sub _koha_add_biblio { my ( $dbh, $biblio, $frameworkcode ) = @_; my $sth = $dbh->prepare("Select max(biblionumber) from biblio"); $sth->execute; my $data = $sth->fetchrow_arrayref; my $biblionumber = $$data[0] + 1; my $series = 0; if ( $biblio->{'seriestitle'} ) { $series = 1 } $sth->finish; $sth = $dbh->prepare( "INSERT INTO biblio SET biblionumber = ?, title = ?, author = ?, copyrightdate = ?, serial = ?, seriestitle = ?, notes = ?, abstract = ?, unititle = ?, frameworkcode = ? " ); $sth->execute( $biblionumber, $biblio->{'title'}, $biblio->{'author'}, $biblio->{'copyrightdate'}, $biblio->{'serial'}, $biblio->{'seriestitle'}, $biblio->{'notes'}, $biblio->{'abstract'}, $biblio->{'unititle'}, $frameworkcode ); $sth->finish; return ($biblionumber); } =head2 _find_value =over 4 ($indicators, $value) = _find_value($tag, $subfield, $record,$encoding); Find the given $subfield in the given $tag in the given MARC::Record $record. If the subfield is found, returns the (indicators, value) pair; otherwise, (undef, undef) is returned. PROPOSITION : Such a function is used in addbiblio AND additem and serial-edit and maybe could be used in Authorities. I suggest we export it from this module. =back =cut sub _find_value { my ( $tagfield, $insubfield, $record, $encoding ) = @_; my @result; my $indicator; if ( $tagfield < 10 ) { if ( $record->field($tagfield) ) { push @result, $record->field($tagfield)->data(); } else { push @result, ""; } } else { foreach my $field ( $record->field($tagfield) ) { my @subfields = $field->subfields(); foreach my $subfield (@subfields) { if ( @$subfield[0] eq $insubfield ) { push @result, @$subfield[1]; $indicator = $field->indicator(1) . $field->indicator(2); } } } } return ( $indicator, @result ); } =head2 _koha_modify_biblio =over 4 $biblionumber = _koha_modify_biblio($dbh,$biblio); Internal function for updating the biblio table =back =cut sub _koha_modify_biblio { my ( $dbh, $biblio ) = @_; # FIXME: this code could be made more portable by not hard-coding # the values that are supposed to be in biblio table my $query = qq{ UPDATE biblio SET title = ?, author = ?, abstract = ?, copyrightdate = ?, seriestitle = ?, serial = ?, unititle = ?, notes = ? WHERE biblionumber = ? }; my $sth = $dbh->prepare($query); $sth->execute( $biblio->{'title'}, $biblio->{'author'}, $biblio->{'abstract'}, $biblio->{'copyrightdate'}, $biblio->{'seriestitle'}, $biblio->{'serial'}, $biblio->{'unititle'}, $biblio->{'notes'}, $biblio->{'biblionumber'} ) if $biblio->{'biblionumber'}; warn $sth->err if $sth->err; warn "BIG ERROR :: No biblionumber for $biblio->{title}" if $biblio->{biblionumber} !~ /\d+/; # if it is not a number return ( $biblio->{'biblionumber'} ); } =head2 _koha_modify_biblioitem =over 4 _koha_modify_biblioitem( $dbh, $biblioitem ); =back =cut sub _koha_modify_biblioitem { my ( $dbh, $biblioitem ) = @_; my $query; ##Recalculate LC in case it changed --TG $biblioitem->{'itemtype'} = $dbh->quote( $biblioitem->{'itemtype'} ); $biblioitem->{'url'} = $dbh->quote( $biblioitem->{'url'} ); $biblioitem->{'isbn'} = $dbh->quote( $biblioitem->{'isbn'} ); $biblioitem->{'issn'} = $dbh->quote( $biblioitem->{'issn'} ); $biblioitem->{'publishercode'} = $dbh->quote( $biblioitem->{'publishercode'} ); $biblioitem->{'publicationyear'} = $dbh->quote( $biblioitem->{'publicationyear'} ); $biblioitem->{'classification'} = $dbh->quote( $biblioitem->{'classification'} ); $biblioitem->{'dewey'} = $dbh->quote( $biblioitem->{'dewey'} ); $biblioitem->{'subclass'} = $dbh->quote( $biblioitem->{'subclass'} ); $biblioitem->{'illus'} = $dbh->quote( $biblioitem->{'illus'} ); $biblioitem->{'pages'} = $dbh->quote( $biblioitem->{'pages'} ); $biblioitem->{'volumeddesc'} = $dbh->quote( $biblioitem->{'volumeddesc'} ); $biblioitem->{'bnotes'} = $dbh->quote( $biblioitem->{'bnotes'} ); $biblioitem->{'size'} = $dbh->quote( $biblioitem->{'size'} ); $biblioitem->{'place'} = $dbh->quote( $biblioitem->{'place'} ); $biblioitem->{'collectiontitle'} = $dbh->quote( $biblioitem->{'collectiontitle'} ); $biblioitem->{'collectionissn'} = $dbh->quote( $biblioitem->{'collectionissn'} ); $biblioitem->{'collectionvolume'} = $dbh->quote( $biblioitem->{'collectionvolume'} ); $biblioitem->{'editionstatement'} = $dbh->quote( $biblioitem->{'editionstatement'} ); $biblioitem->{'editionresponsability'} = $dbh->quote( $biblioitem->{'editionresponsability'} ); $biblioitem->{'ccode'} = $dbh->quote( $biblioitem->{'ccode'} ); $biblioitem->{'biblionumber'} = $dbh->quote( $biblioitem->{'biblionumber'} ); $query = "UPDATE biblioitems SET itemtype = $biblioitem->{'itemtype'}, url = $biblioitem->{'url'}, isbn = $biblioitem->{'isbn'}, issn = $biblioitem->{'issn'}, publishercode = $biblioitem->{'publishercode'}, publicationyear = $biblioitem->{'publicationyear'}, classification = $biblioitem->{'classification'}, dewey = $biblioitem->{'dewey'}, subclass = $biblioitem->{'subclass'}, illus = $biblioitem->{'illus'}, pages = $biblioitem->{'pages'}, volumeddesc = $biblioitem->{'volumeddesc'}, notes = $biblioitem->{'bnotes'}, size = $biblioitem->{'size'}, place = $biblioitem->{'place'}, collectiontitle = $biblioitem->{'collectiontitle'}, collectionissn = $biblioitem->{'collectionissn'}, collectionvolume= $biblioitem->{'collectionvolume'}, editionstatement= $biblioitem->{'editionstatement'}, editionresponsability= $biblioitem->{'editionresponsability'}, ccode = $biblioitem->{'ccode'} where biblionumber = $biblioitem->{'biblionumber'}"; $dbh->do($query); if ( $dbh->errstr ) { warn "ERROR in _koha_modify_biblioitem $query"; } } =head2 _koha_add_biblioitem =over 4 _koha_add_biblioitem( $dbh, $biblioitem ); Internal function to add a biblioitem =back =cut sub _koha_add_biblioitem { my ( $dbh, $biblioitem ) = @_; # my $dbh = C4Connect; my $sth = $dbh->prepare("SELECT max(biblioitemnumber) FROM biblioitems"); my $data; my $bibitemnum; $sth->execute; $data = $sth->fetchrow_arrayref; $bibitemnum = $$data[0] + 1; $sth->finish; $sth = $dbh->prepare( "INSERT INTO biblioitems SET biblioitemnumber = ?, biblionumber = ?, volume = ?, number = ?, classification = ?, itemtype = ?, url = ?, isbn = ?, issn = ?, dewey = ?, subclass = ?, publicationyear = ?, publishercode = ?, volumedate = ?, volumeddesc = ?, illus = ?, pages = ?, notes = ?, size = ?, lccn = ?, marc = ?, lcsort = ?, place = ?, ccode = ?, collectiontitle = ?, collectionissn = ?, collectionvolume = ?, editionstatement= ?, editionresponsability= ? " ); my ($lcsort) = calculatelc( $biblioitem->{'classification'} ) . $biblioitem->{'subclass'}; $sth->execute( $bibitemnum, $biblioitem->{'biblionumber'}, $biblioitem->{'volume'}, $biblioitem->{'number'}, $biblioitem->{'classification'}, $biblioitem->{'itemtype'}, $biblioitem->{'url'}, $biblioitem->{'isbn'}, $biblioitem->{'issn'}, $biblioitem->{'dewey'}, $biblioitem->{'subclass'}, $biblioitem->{'publicationyear'}, $biblioitem->{'publishercode'}, $biblioitem->{'volumedate'}, $biblioitem->{'volumeddesc'}, $biblioitem->{'illus'}, $biblioitem->{'pages'}, $biblioitem->{'bnotes'}, $biblioitem->{'size'}, $biblioitem->{'lccn'}, $biblioitem->{'marc'}, $biblioitem->{'place'}, $lcsort, $biblioitem->{'ccode'}, $biblioitem->{'collectiontitle'},$biblioitem->{'collectionissn'}, $biblioitem->{'collectionvolume'},$biblioitem->{'editionstatement'}, $biblioitem->{'editionresponsability'} ); $sth->finish; return ($bibitemnum); } =head2 _koha_new_items =over 4 _koha_new_items( $dbh, $item, $barcode ); =back =cut sub _koha_new_items { my ( $dbh, $item, $barcode ) = @_; # my $dbh = C4Connect; my $sth = $dbh->prepare("Select max(itemnumber) from items"); my $data; my $itemnumber; my $error = ""; $sth->execute; $data = $sth->fetchrow_hashref; $itemnumber = $data->{'max(itemnumber)'} + 1; $sth->finish; ## Now calculate lccalnumber my ($cutterextra) = itemcalculator( $dbh, $item->{'biblioitemnumber'}, $item->{'itemcallnumber'} ); # FIXME the "notforloan" field seems to be named "loan" in some places. workaround bugfix. if ( $item->{'loan'} ) { $item->{'notforloan'} = $item->{'loan'}; } # if dateaccessioned is provided, use it. Otherwise, set to NOW() if ( $item->{'dateaccessioned'} eq '' || !$item->{'dateaccessioned'} ) { $sth = $dbh->prepare( "Insert into items set itemnumber = ?, biblionumber = ?, multivolumepart = ?, biblioitemnumber = ?, barcode = ?, booksellerid = ?, dateaccessioned = NOW(), homebranch = ?, holdingbranch = ?, price = ?, replacementprice = ?, replacementpricedate = NOW(), datelastseen = NOW(), multivolume = ?, stack = ?, itemlost = ?, wthdrawn = ?, paidfor = ?, itemnotes = ?, itemcallnumber =?, notforloan = ?, location = ?, Cutterextra = ? " ); $sth->execute( $itemnumber, $item->{'biblionumber'}, $item->{'multivolumepart'}, $item->{'biblioitemnumber'}, $barcode, $item->{'booksellerid'}, $item->{'homebranch'}, $item->{'holdingbranch'}, $item->{'price'}, $item->{'replacementprice'}, $item->{multivolume}, $item->{stack}, $item->{itemlost}, $item->{wthdrawn}, $item->{paidfor}, $item->{'itemnotes'}, $item->{'itemcallnumber'}, $item->{'notforloan'}, $item->{'location'}, $cutterextra ); } else { $sth = $dbh->prepare( "INSERT INTO items SET itemnumber = ?, biblionumber = ?, multivolumepart = ?, biblioitemnumber = ?, barcode = ?, booksellerid = ?, dateaccessioned = ?, homebranch = ?, holdingbranch = ?, price = ?, replacementprice = ?, replacementpricedate = NOW(), datelastseen = NOW(), multivolume = ?, stack = ?, itemlost = ?, wthdrawn = ?, paidfor = ?, itemnotes = ?, itemcallnumber = ?, notforloan = ?, location = ?, Cutterextra = ? " ); $sth->execute( $itemnumber, $item->{'biblionumber'}, $item->{'multivolumepart'}, $item->{'biblioitemnumber'}, $barcode, $item->{'booksellerid'}, $item->{'dateaccessioned'}, $item->{'homebranch'}, $item->{'holdingbranch'}, $item->{'price'}, $item->{'replacementprice'}, $item->{multivolume}, $item->{stack}, $item->{itemlost}, $item->{wthdrawn}, $item->{paidfor}, $item->{'itemnotes'}, $item->{'itemcallnumber'}, $item->{'notforloan'}, $item->{'location'}, $cutterextra ); } if ( defined $sth->errstr ) { $error .= $sth->errstr; } return ( $itemnumber, $error ); } =head2 _koha_modify_item =over 4 _koha_modify_item( $dbh, $item, $op ); =back =cut sub _koha_modify_item { my ( $dbh, $item, $op ) = @_; $item->{'itemnum'} = $item->{'itemnumber'} unless $item->{'itemnum'}; # if all we're doing is setting statuses, just update those and get out if ( $op eq "setstatus" ) { my $query = "UPDATE items SET itemlost=?,wthdrawn=?,binding=? WHERE itemnumber=?"; my @bind = ( $item->{'itemlost'}, $item->{'wthdrawn'}, $item->{'binding'}, $item->{'itemnumber'} ); my $sth = $dbh->prepare($query); $sth->execute(@bind); $sth->finish; return undef; } ## Now calculate lccalnumber my ($cutterextra) = itemcalculator( $dbh, $item->{'bibitemnum'}, $item->{'itemcallnumber'} ); my $query = "UPDATE items SET barcode=?,itemnotes=?,itemcallnumber=?,notforloan=?,location=?,multivolumepart=?,multivolume=?,stack=?,wthdrawn=?,holdingbranch=?,homebranch=?,cutterextra=?, onloan=?, binding=?"; my @bind = ( $item->{'barcode'}, $item->{'notes'}, $item->{'itemcallnumber'}, $item->{'notforloan'}, $item->{'location'}, $item->{multivolumepart}, $item->{multivolume}, $item->{stack}, $item->{wthdrawn}, $item->{holdingbranch}, $item->{homebranch}, $cutterextra, $item->{onloan}, $item->{binding} ); if ( $item->{'lost'} ne '' ) { $query = "update items set biblioitemnumber=?,barcode=?,itemnotes=?,homebranch=?, itemlost=?,wthdrawn=?,itemcallnumber=?,notforloan=?, location=?,multivolumepart=?,multivolume=?,stack=?,wthdrawn=?,holdingbranch=?,cutterextra=?,onloan=?, binding=?"; @bind = ( $item->{'bibitemnum'}, $item->{'barcode'}, $item->{'notes'}, $item->{'homebranch'}, $item->{'lost'}, $item->{'wthdrawn'}, $item->{'itemcallnumber'}, $item->{'notforloan'}, $item->{'location'}, $item->{multivolumepart}, $item->{multivolume}, $item->{stack}, $item->{wthdrawn}, $item->{holdingbranch}, $cutterextra, $item->{onloan}, $item->{binding} ); if ( $item->{homebranch} ) { $query .= ",homebranch=?"; push @bind, $item->{homebranch}; } if ( $item->{holdingbranch} ) { $query .= ",holdingbranch=?"; push @bind, $item->{holdingbranch}; } } $query .= " where itemnumber=?"; push @bind, $item->{'itemnum'}; if ( $item->{'replacement'} ne '' ) { $query =~ s/ where/,replacementprice='$item->{'replacement'}' where/; } my $sth = $dbh->prepare($query); $sth->execute(@bind); $sth->finish; } =head2 _koha_delete_biblio =over 4 $error = _koha_delete_biblio($dbh,$biblionumber); Internal sub for deleting from biblio table -- also saves to deletedbiblio C<$dbh> - the database handle C<$biblionumber> - the biblionumber of the biblio to be deleted =back =cut # FIXME: add error handling sub _koha_delete_biblio { my ( $dbh, $biblionumber ) = @_; # get all the data for this biblio my $sth = $dbh->prepare("SELECT * FROM biblio WHERE biblionumber=?"); $sth->execute($biblionumber); if ( my $data = $sth->fetchrow_hashref ) { # save the record in deletedbiblio # find the fields to save my $query = "INSERT INTO deletedbiblio SET "; my @bind = (); foreach my $temp ( keys %$data ) { $query .= "$temp = ?,"; push( @bind, $data->{$temp} ); } # replace the last , by ",?)" $query =~ s/\,$//; my $bkup_sth = $dbh->prepare($query); $bkup_sth->execute(@bind); $bkup_sth->finish; # delete the biblio my $del_sth = $dbh->prepare("DELETE FROM biblio WHERE biblionumber=?"); $del_sth->execute($biblionumber); $del_sth->finish; } $sth->finish; return undef; } =head2 _koha_delete_biblioitems =over 4 $error = _koha_delete_biblioitems($dbh,$biblioitemnumber); Internal sub for deleting from biblioitems table -- also saves to deletedbiblioitems C<$dbh> - the database handle C<$biblionumber> - the biblioitemnumber of the biblioitem to be deleted =back =cut # FIXME: add error handling sub _koha_delete_biblioitems { my ( $dbh, $biblioitemnumber ) = @_; # get all the data for this biblioitem my $sth = $dbh->prepare("SELECT * FROM biblioitems WHERE biblioitemnumber=?"); $sth->execute($biblioitemnumber); if ( my $data = $sth->fetchrow_hashref ) { # save the record in deletedbiblioitems # find the fields to save my $query = "INSERT INTO deletedbiblioitems SET "; my @bind = (); foreach my $temp ( keys %$data ) { $query .= "$temp = ?,"; push( @bind, $data->{$temp} ); } # replace the last , by ",?)" $query =~ s/\,$//; my $bkup_sth = $dbh->prepare($query); $bkup_sth->execute(@bind); $bkup_sth->finish; # delete the biblioitem my $del_sth = $dbh->prepare("DELETE FROM biblioitems WHERE biblioitemnumber=?"); $del_sth->execute($biblioitemnumber); $del_sth->finish; } $sth->finish; return undef; } =head2 _koha_delete_item =over 4 _koha_delete_item( $dbh, $itemnum ); Internal function to delete an item record from the koha tables =back =cut sub _koha_delete_item { my ( $dbh, $itemnum ) = @_; my $sth = $dbh->prepare("select * from items where itemnumber=?"); $sth->execute($itemnum); my $data = $sth->fetchrow_hashref; $sth->finish; my $query = "Insert into deleteditems set "; my @bind = (); foreach my $temp ( keys %$data ) { $query .= "$temp = ?,"; push( @bind, $data->{$temp} ); } $query =~ s/\,$//; # print $query; $sth = $dbh->prepare($query); $sth->execute(@bind); $sth->finish; $sth = $dbh->prepare("Delete from items where itemnumber=?"); $sth->execute($itemnum); $sth->finish; } =head1 UNEXPORTED FUNCTIONS =over 4 =head2 calculatelc $lc = calculatelc($classification); =back =cut sub calculatelc { my ($classification) = @_; $classification =~ s/^\s+|\s+$//g; my $i = 0; my $lc2; my $lc1; for ( $i = 0 ; $i < length($classification) ; $i++ ) { my $c = ( substr( $classification, $i, 1 ) ); if ( $c ge '0' && $c le '9' ) { $lc2 = substr( $classification, $i ); last; } else { $lc1 .= substr( $classification, $i, 1 ); } } #while my $other = length($lc1); if ( !$lc1 ) { $other = 0; } my $extras; if ( $other < 4 ) { for ( 1 .. ( 4 - $other ) ) { $extras .= "0"; } } $lc1 .= $extras; $lc2 =~ s/^ //g; $lc2 =~ s/ //g; $extras = ""; ##Find the decimal part of $lc2 my $pos = index( $lc2, "." ); if ( $pos < 0 ) { $pos = length($lc2); } if ( $pos >= 0 && $pos < 5 ) { ##Pad lc2 with zeros to create a 5digit decimal needed in marc record to sort as numeric for ( 1 .. ( 5 - $pos ) ) { $extras .= "0"; } } $lc2 = $extras . $lc2; return ( $lc1 . $lc2 ); } =head2 itemcalculator =over 4 $cutterextra = itemcalculator( $dbh, $biblioitem, $callnumber ); =back =cut sub itemcalculator { my ( $dbh, $biblioitem, $callnumber ) = @_; my $sth = $dbh->prepare( "select classification, subclass from biblioitems where biblioitemnumber=?" ); $sth->execute($biblioitem); my ( $classification, $subclass ) = $sth->fetchrow; my $all = $classification . " " . $subclass; my $total = length($all); my $cutterextra = substr( $callnumber, $total - 1 ); return $cutterextra; } =head2 ModBiblioMarc &ModBiblioMarc($newrec,$biblionumber,$frameworkcode); Add MARC data for a biblio to koha Function exported, but should NOT be used, unless you really know what you're doing =cut sub ModBiblioMarc { # pass the MARC::Record to this function, and it will create the records in the marc field my ( $record, $biblionumber, $frameworkcode ) = @_; my $dbh = C4::Context->dbh; my @fields = $record->fields(); if ( !$frameworkcode ) { $frameworkcode = ""; } my $sth = $dbh->prepare("UPDATE biblio SET frameworkcode=? WHERE biblionumber=?"); $sth->execute( $frameworkcode, $biblionumber ); $sth->finish; my $encoding = C4::Context->preference("marcflavour"); # deal with UNIMARC field 100 (encoding) : create it if needed & set encoding to unicode if ( $encoding eq "UNIMARC" ) { my $string; if ( length($record->subfield( 100, "a" )) == 35 ) { $string = $record->subfield( 100, "a" ); my $f100 = $record->field(100); $record->delete_field($f100); } else { $string = POSIX::strftime( "%Y%m%d", localtime ); $string =~ s/\-//g; $string = sprintf( "%-*s", 35, $string ); } substr( $string, 22, 6, "frey50" ); unless ( $record->subfield( 100, "a" ) ) { $record->insert_grouped_field( MARC::Field->new( 100, "", "", "a" => $string ) ); } } ModZebra($biblionumber,"specialUpdate","biblioserver",$record); $sth = $dbh->prepare( "update biblioitems set marc=?,marcxml=? where biblionumber=?"); $sth->execute( $record->as_usmarc(), $record->as_xml_record($encoding), $biblionumber ); $sth->finish; return $biblionumber; } =head2 AddItemInMarc =over 4 $newbiblionumber = AddItemInMarc( $record, $biblionumber, $frameworkcode ); Add an item in a MARC record and save the MARC record Function exported, but should NOT be used, unless you really know what you're doing =back =cut sub AddItemInMarc { # pass the MARC::Record to this function, and it will create the records in the marc tables my ( $record, $biblionumber, $frameworkcode ) = @_; my $newrec = &GetMarcBiblio($biblionumber); # create it my @fields = $record->fields(); foreach my $field (@fields) { $newrec->append_fields($field); } # FIXME: should we be making sure the biblionumbers are the same? my $newbiblionumber = &ModBiblioMarc( $newrec, $biblionumber, $frameworkcode ); return $newbiblionumber; } =head2 z3950_extended_services z3950_extended_services($serviceType,$serviceOptions,$record); z3950_extended_services is used to handle all interactions with Zebra's extended serices package, which is employed to perform all management of the MARC data stored in Zebra. C<$serviceType> one of: itemorder,create,drop,commit,update,xmlupdate C<$serviceOptions> a has of key/value pairs. For instance, if service_type is 'update', $service_options should contain: action => update action, one of specialUpdate, recordInsert, recordReplace, recordDelete, elementUpdate. and maybe recordidOpaque => Opaque Record ID (user supplied) or recordidNumber => Record ID number (system number). syntax => the record syntax (transfer syntax) databaseName = Database from connection object To set serviceOptions, call set_service_options($serviceType) C<$record> the record, if one is needed for the service type A record should be in XML. You can convert it to XML from MARC by running it through marc2xml(). =cut sub z3950_extended_services { my ( $server, $serviceType, $action, $serviceOptions ) = @_; # get our connection object my $Zconn = C4::Context->Zconn( $server, 0, 1 ); # create a new package object my $Zpackage = $Zconn->package(); # set our options $Zpackage->option( action => $action ); if ( $serviceOptions->{'databaseName'} ) { $Zpackage->option( databaseName => $serviceOptions->{'databaseName'} ); } if ( $serviceOptions->{'recordIdNumber'} ) { $Zpackage->option( recordIdNumber => $serviceOptions->{'recordIdNumber'} ); } if ( $serviceOptions->{'recordIdOpaque'} ) { $Zpackage->option( recordIdOpaque => $serviceOptions->{'recordIdOpaque'} ); } # this is an ILL request (Zebra doesn't support it, but Koha could eventually) #if ($serviceType eq 'itemorder') { # $Zpackage->option('contact-name' => $serviceOptions->{'contact-name'}); # $Zpackage->option('contact-phone' => $serviceOptions->{'contact-phone'}); # $Zpackage->option('contact-email' => $serviceOptions->{'contact-email'}); # $Zpackage->option('itemorder-item' => $serviceOptions->{'itemorder-item'}); #} if ( $serviceOptions->{record} ) { $Zpackage->option( record => $serviceOptions->{record} ); # can be xml or marc if ( $serviceOptions->{'syntax'} ) { $Zpackage->option( syntax => $serviceOptions->{'syntax'} ); } } # send the request, handle any exception encountered eval { $Zpackage->send($serviceType) }; if ( $@ && $@->isa("ZOOM::Exception") ) { return "error: " . $@->code() . " " . $@->message() . "\n"; } # free up package resources $Zpackage->destroy(); } =head2 set_service_options my $serviceOptions = set_service_options($serviceType); C<$serviceType> itemorder,create,drop,commit,update,xmlupdate Currently, we only support 'create', 'commit', and 'update'. 'drop' support will be added as soon as Zebra supports it. =cut sub set_service_options { my ($serviceType) = @_; my $serviceOptions; # FIXME: This needs to be an OID ... if we ever need 'syntax' this sub will need to change # $serviceOptions->{ 'syntax' } = ''; #zebra doesn't support syntaxes other than xml if ( $serviceType eq 'commit' ) { # nothing to do } if ( $serviceType eq 'create' ) { # nothing to do } if ( $serviceType eq 'drop' ) { die "ERROR: 'drop' not currently supported (by Zebra)"; } return $serviceOptions; } =head2 GetItemsCount $count = &GetItemsCount( $biblionumber); this function return count of item with $biblionumber =cut sub GetItemsCount { my ( $biblionumber ) = @_; my $dbh = C4::Context->dbh; my $query = qq|SELECT count(*) FROM items WHERE biblionumber=?|; my $sth = $dbh->prepare($query); $sth->execute($biblionumber); my $count = $sth->fetchrow; $sth->finish; return ($count); } END { } # module clean-up code here (global destructor) 1; __END__ =head1 AUTHOR Koha Developement team Paul POULAIN paul.poulain@free.fr Joshua Ferraro jmf@liblime.com =cut # $Id$ # $Log$ # Revision 1.221 2007/07/31 16:01:11 toins # Some new functions. # TransformHTMLtoMarc rewrited. # # Revision 1.220 2007/07/20 15:43:16 hdl # Bug Fixing GetMarcSubjects. # Links parameters were mixed. # # Revision 1.218 2007/07/19 07:40:08 hdl # Adding selection by location for inventory # # Revision 1.217 2007/07/03 13:47:44 tipaul # fixing some display bugs (itemtype not properly returned and a html table bug that makes items appear strangely # # Revision 1.216 2007/07/03 09:40:58 tipaul # return itemtype description properly # # Revision 1.215 2007/07/03 09:33:05 tipaul # if you just replace su by a space in subjects, you'll replace jesus by je s, which is strange for users. this fix solves the problem and introduces authoritysep systempref as separator of subfields, for a better identification of where the authority starts and end # # Revision 1.214 2007/07/02 09:13:22 tipaul # unimarc bugfix : the encoding is in field 100 in UNIMARC. when TransformHTMLtoXML on an item, you must not automatically add a 100 field in items, otherwise there will be 2 100 fields in the biblio, which is wrong # # Revision 1.213 2007/06/25 15:01:45 tipaul # bugfixes on unimarc 100 handling (the field used for encoding) # # Revision 1.212 2007/06/15 13:44:44 tipaul # some fixes (and only fixes) # # Revision 1.211 2007/06/15 09:40:06 toins # do not get $3 $4 and $5 on GetMarcSubjects GetMarcAuthors on unimarc flavour. # # Revision 1.210 2007/06/13 13:03:34 toins # removing warn compilation. # # Revision 1.209 2007/05/23 16:19:40 tipaul # various bugfixes (minor) and french translation updated # # Revision 1.208 2007/05/22 09:13:54 tipaul # Bugfixes & improvements (various and minor) : # - updating templates to have tmpl_process3.pl running without any errors # - adding a drupal-like css for prog templates (with 3 small images) # - fixing some bugs in circulation & other scripts # - updating french translation # - fixing some typos in templates # # Revision 1.207 2007/05/22 08:51:19 hdl # Changing GetMarcStructure signature. # Deleting first parameter $dbh # # Revision 1.206 2007/05/21 08:44:17 btoumi # add security when u delete biblio : # u must delete linked items before delete biblio # # Revision 1.205 2007/05/11 16:04:03 btoumi # bug fix: # problem in displayed label link with subject in detail.tmpl # ex: label random => rdom # # Revision 1.204 2007/05/10 14:45:15 tipaul # Koha NoZebra : # - support for authorities # - some bugfixes in ordering and "CCL" parsing # - support for authorities <=> biblios walking # # Seems I can do what I want now, so I consider its done, except for bugfixes that will be needed i m sure ! # # Revision 1.203 2007/05/03 15:16:02 tipaul # BUGFIX for : NoZebra # - NoZebra features : seems they work fine now (adding, modifying, deleting) # - Biblio edition major bugfix : before this commit editing a biblio resulted in an item removal in marcxml field # # Revision 1.202 2007/05/02 16:44:31 tipaul # NoZebra SQL index management : # * adding 3 subs in Biblio.pm # - GetNoZebraIndexes, that get the index structure in a new systempreference (added with this commit) # - _DelBiblioNoZebra, that retrieve all index entries for a biblio and remove in a variable the biblio reference # - _AddBiblioNoZebra, that add index entries for a biblio. # Note that the 2 _Add and _Del subs work only in a hash variable, to speed up things in case of a modif (ie : delete+add). The effective SQL update is done in the ModZebra sub (that existed before, and dealed with zebra index). # I think the code has to be more deeply tested, but it works at least partially. # # Revision 1.201 2007/04/27 14:00:49 hdl # Removing $dbh from GetMarcFromKohaField (dbh is not used in this function.) # # Revision 1.200 2007/04/25 16:26:42 tipaul # Koha 3.0 nozebra 1st commit : the script misc/migration_tools/rebuild_nozebra.pl build the nozebra table, and, if you set NoZebra to Yes, queries will be done through zebra. TODO : # - add nozebra table management on biblio editing # - the index table content is hardcoded. I still have to add some specific systempref to let the library update it # - manage pagination (next/previous) # - manage facets # WHAT works : # - NZgetRecords : has exactly the same API & returns as zebra getQuery, except that some parameters are unused # - search & sort works quite good # - CQL parser is better that what I thought I could do : title="harry and sally" and publicationyear>2000 not itemtype=LIVR should work fine # # Revision 1.199 2007/04/24 09:07:53 tipaul # moving dotransfer to Biblio.pm::ModItemTransfer + some CheckReserves fixes # # Revision 1.198 2007/04/23 15:21:17 tipaul # renaming currenttransfers to transferstoreceive # # Revision 1.197 2007/04/18 17:00:14 tipaul # removing all useless %env / $env # # Revision 1.196 2007/04/17 08:48:00 tipaul # circulation cleaning continued: bufixing # # Revision 1.195 2007/04/04 16:46:22 tipaul # HUGE COMMIT : code cleaning circulation. # # some stuff to do, i'll write a mail on koha-devel NOW ! # # Revision 1.194 2007/03/30 12:00:42 tipaul # why the hell do we need to explicitly utf8 decode this string ? I really don't know, but it seems it's mandatory, otherwise, tag descriptions are not properly encoded... # # Revision 1.193 2007/03/29 16:45:53 tipaul # Code cleaning of Biblio.pm (continued) # # All subs have be cleaned : # - removed useless # - merged some # - reordering Biblio.pm completly # - using only naming conventions # # Seems to have broken nothing, but it still has to be heavily tested. # Note that Biblio.pm is now much more efficient than previously & probably more reliable as well. # # Revision 1.192 2007/03/29 13:30:31 tipaul # Code cleaning : # == Biblio.pm cleaning (useless) == # * some sub declaration dropped # * removed modbiblio sub # * removed moditem sub # * removed newitems. It was used only in finishrecieve. Replaced by a TransformKohaToMarc+AddItem, that is better. # * removed MARCkoha2marcItem # * removed MARCdelsubfield declaration # * removed MARCkoha2marcBiblio # # == Biblio.pm cleaning (naming conventions) == # * MARCgettagslib renamed to GetMarcStructure # * MARCgetitems renamed to GetMarcItem # * MARCfind_frameworkcode renamed to GetFrameworkCode # * MARCmarc2koha renamed to TransformMarcToKoha # * MARChtml2marc renamed to TransformHtmlToMarc # * MARChtml2xml renamed to TranformeHtmlToXml # * zebraop renamed to ModZebra # # == MARC=OFF == # * removing MARC=OFF related scripts (in cataloguing directory) # * removed checkitems (function related to MARC=off feature, that is completly broken in head. If someone want to reintroduce it, hard work coming...) # * removed getitemsbybiblioitem (used only by MARC=OFF scripts, that is removed as well) # # Revision 1.191 2007/03/29 09:42:13 tipaul # adding default value new feature into cataloguing. The system (definition) part has already been added by toins # # Revision 1.190 2007/03/29 08:45:19 hdl # Deleting ignore_errors(1) pour MARC::Charset # # Revision 1.189 2007/03/28 10:39:16 hdl # removing $dbh as a parameter in AuthoritiesMarc functions # And reporting all differences into the scripts taht relies on those functions.