1 |
#!/usr/bin/perl -w |
#!/usr/bin/perl -w |
2 |
|
|
3 |
use strict; |
use strict; |
4 |
use OpenIsis; |
use Biblio::Isis; |
5 |
use Getopt::Std; |
use Getopt::Std; |
6 |
use Data::Dumper; |
use Data::Dumper; |
7 |
use XML::Simple; |
use XML::Simple; |
|
use Text::Unaccent 1.02; # 1.01 won't compile on my platform, |
|
8 |
use Text::Iconv; |
use Text::Iconv; |
9 |
use Config::IniFiles; |
use Config::IniFiles; |
10 |
use Encode; |
use Encode; |
16 |
|
|
17 |
my $config_file = $0; |
my $config_file = $0; |
18 |
$config_file =~ s/\.pl$/.conf/; |
$config_file =~ s/\.pl$/.conf/; |
19 |
$config_file = $ARGV[0] if (-f $ARGV[0]); |
$config_file = $ARGV[0] if ($ARGV[0] && -f $ARGV[0]); |
20 |
die "FATAL: can't find configuration file '$config_file'" if (! -e $config_file); |
die "FATAL: can't find configuration file '$config_file'" if (! -e $config_file); |
21 |
|
|
22 |
my $config; |
my $config; |
139 |
} else { |
} else { |
140 |
print STDERR "WARNING: field '$field' doesn't have 'name' attribute!"; |
print STDERR "WARNING: field '$field' doesn't have 'name' attribute!"; |
141 |
} |
} |
142 |
|
|
143 |
if ($field_name) { |
if ($field_name) { |
144 |
|
$field_name = x($field_name); |
145 |
if (! $last_field_name) { |
if (! $last_field_name) { |
146 |
$last_field_name = x($field_name); |
$last_field_name = $field_name; |
147 |
return $last_field_name; |
return $last_field_name; |
148 |
} elsif ($field_name ne $last_field_name) { |
} elsif ($field_name ne $last_field_name) { |
149 |
$last_field_name = x($field_name); |
$last_field_name = $field_name; |
150 |
return $last_field_name; |
return $last_field_name; |
151 |
} |
} |
152 |
} |
} |
527 |
$swish_data =~ s/ +/ /g; |
$swish_data =~ s/ +/ /g; |
528 |
$swish_data =~ s/ +$//g; |
$swish_data =~ s/ +$//g; |
529 |
|
|
530 |
$xml .= xmlify($field."_swish", unac_string($codepage,$swish_data)); |
$xml .= xmlify($field."_swish", my_unac_string($codepage,$swish_data)); |
531 |
} |
} |
532 |
|
|
533 |
my $swish_exact_data = $cache->{swish_exact_data}->{$field}->[$page]; |
my $swish_exact_data = $cache->{swish_exact_data}->{$field}->[$page]; |
537 |
|
|
538 |
# add delimiters before and after word. |
# add delimiters before and after word. |
539 |
# That is required to produce exact match |
# That is required to produce exact match |
540 |
$xml .= xmlify($field."_swish_exact", unac_string($codepage,$swish_exact_data)); |
$xml .= xmlify($field."_swish_exact", my_unac_string($codepage,$swish_exact_data)); |
541 |
} |
} |
542 |
|
|
543 |
my $idel = $cache->{index_delimiter}->{$field}; |
my $idel = $cache->{index_delimiter}->{$field}; |
570 |
$swish_data =~ s/ +/ /g; |
$swish_data =~ s/ +/ /g; |
571 |
$swish_data =~ s/ +$//g; |
$swish_data =~ s/ +$//g; |
572 |
|
|
573 |
$xml .= xmlify($field."_swish", unac_string($codepage,$swish_data)); |
$xml .= xmlify($field."_swish", my_unac_string($codepage,$swish_data)); |
574 |
} |
} |
575 |
|
|
576 |
if ($swish_exact_data) { |
if ($swish_exact_data) { |
579 |
|
|
580 |
# add delimiters before and after word. |
# add delimiters before and after word. |
581 |
# That is required to produce exact match |
# That is required to produce exact match |
582 |
$xml .= xmlify($field."_swish_exact", unac_string($codepage,$swish_exact_data)); |
$xml .= xmlify($field."_swish_exact", my_unac_string($codepage,$swish_exact_data)); |
583 |
} |
} |
584 |
} |
} |
585 |
} |
} |
616 |
|
|
617 |
my $show_progress = $cfg_global->val('global', 'show_progress'); |
my $show_progress = $cfg_global->val('global', 'show_progress'); |
618 |
|
|
619 |
my $unac_filter = $cfg_global->val('global', 'unac_filter'); |
my $my_unac_filter = $cfg_global->val('global', 'my_unac_filter'); |
620 |
if ($unac_filter) { |
if ($my_unac_filter) { |
621 |
require $unac_filter; |
print STDERR "using $my_unac_filter to filter characters for search\n"; |
622 |
|
require $my_unac_filter; |
623 |
|
} else { |
624 |
|
print STDERR "### fallback to default my_unac_string!\n"; |
625 |
|
eval q{ |
626 |
|
sub main::my_unac_string($$) { |
627 |
|
my ($charset, $string) = (@_); |
628 |
|
return $string; |
629 |
|
} |
630 |
|
}; |
631 |
} |
} |
632 |
|
|
633 |
foreach my $database ($cfg->Sections) { |
foreach my $database ($cfg->Sections) { |
639 |
my $lookup_file = $cfg -> val($database, 'lookup_newfile'); # optional |
my $lookup_file = $cfg -> val($database, 'lookup_newfile'); # optional |
640 |
if ($lookup_file) { |
if ($lookup_file) { |
641 |
#tie %lhash, 'GDBM_File', $lookup_file, &GDBM_NEWDB, 0644; |
#tie %lhash, 'GDBM_File', $lookup_file, &GDBM_NEWDB, 0644; |
642 |
|
if (! -e $lookup_file) { |
643 |
|
open(LOOKUP, "> $lookup_file") || die "can't create $lookup_file': $!"; |
644 |
|
close(LOOKUP); |
645 |
|
} |
646 |
tie %lhash, 'TDB_File', $lookup_file, TDB_CLEAR_IF_FIRST, O_RDWR, 0644; |
tie %lhash, 'TDB_File', $lookup_file, TDB_CLEAR_IF_FIRST, O_RDWR, 0644; |
647 |
print STDERR "creating lookup file '$lookup_file'\n"; |
print STDERR "creating lookup file '$lookup_file'\n"; |
648 |
# delete memory cache for lookup file |
# delete memory cache for lookup file |
665 |
|
|
666 |
$config=XMLin("./import_xml/$type.xml", ForceArray => [ $type2tag{$type_base}, 'config', 'format' ], ForceContent => 1 ); |
$config=XMLin("./import_xml/$type.xml", ForceArray => [ $type2tag{$type_base}, 'config', 'format' ], ForceContent => 1 ); |
667 |
|
|
668 |
|
# helper for progress bar |
669 |
|
sub fmt_time { |
670 |
|
my $t = shift || 0; |
671 |
|
my $out = ""; |
672 |
|
|
673 |
|
my ($ss,$mm,$hh) = gmtime($t); |
674 |
|
$out .= "${hh}h" if ($hh); |
675 |
|
$out .= sprintf("%02d:%02d", $mm,$ss); |
676 |
|
$out .= " " if ($hh == 0); |
677 |
|
return $out; |
678 |
|
} |
679 |
|
|
680 |
# output current progress indicator |
# output current progress indicator |
681 |
my $last_p = 0; |
my $last_p = 0; |
682 |
|
my $start_t = time(); |
683 |
sub progress { |
sub progress { |
684 |
return if (! $show_progress); |
return if (! $show_progress); |
685 |
my $current = shift; |
my $current = shift; |
686 |
my $total = shift || 1; |
my $total = shift || 1; |
687 |
my $p = int($current * 100 / $total); |
my $p = int($current * 100 / $total); |
688 |
if ($p != $last_p) { |
if ($p < $last_p || $current == 1) { |
689 |
printf STDERR ("%5d / %5d [%-51s] %-2d %% \r",$current,$total,"=" x ($p/2).">", $p ); |
$start_t = time(); |
690 |
|
$last_p = 0; |
691 |
|
} elsif ($p != $last_p) { |
692 |
|
my $rate = ($current / (time() - $start_t || 1)); |
693 |
|
my $eta = ($total-$current) / ($rate || 1); |
694 |
|
printf STDERR ("%5d [%-38s] %-5d %0.1f/s %s\r",$current,"=" x ($p/3)."$p%>", $total, $rate, fmt_time($eta)); |
695 |
$last_p = $p; |
$last_p = $p; |
696 |
} |
} |
697 |
} |
} |
698 |
|
|
699 |
my $fake_dir = 1; |
my $fake_dir = 1; |
700 |
|
my $fake_pos = 0; |
701 |
|
my $last_fake_t = time(); |
702 |
sub fakeprogress { |
sub fakeprogress { |
703 |
return if (! $show_progress); |
return if (! $show_progress); |
704 |
my $current = shift @_; |
my $current = shift @_; |
705 |
|
|
706 |
my @ind = ('-','\\','|','/','-','\\','|','/', '-'); |
my @ind = ('-','\\','|','/','-','\\','|','/'); |
707 |
|
|
708 |
$last_p += $fake_dir; |
if ($current < $fake_pos) { |
709 |
$fake_dir = -$fake_dir if ($last_p > 1000 || $last_p < 0); |
$start_t = time(); |
710 |
if ($last_p % 10 == 0) { |
$last_fake_t = 0; |
711 |
printf STDERR ("%5d / %5s [%-51s]\r",$current,"?"," " x ($last_p/20).$ind[($last_p/20) % $#ind]); |
$fake_dir = 1; |
712 |
|
$fake_pos = 0; |
713 |
|
} |
714 |
|
|
715 |
|
if (time()-$last_fake_t >= 1) { |
716 |
|
$last_fake_t = time(); |
717 |
|
$fake_pos += $fake_dir; |
718 |
|
$fake_dir = -$fake_dir if ($fake_pos > 38); |
719 |
|
} |
720 |
|
|
721 |
|
if ($current % 10 == 0) { |
722 |
|
my $rate = ($current / (time() - $start_t || 1)); |
723 |
|
printf STDERR ("%5d [%-38s] %0.1f/s\r",$current, " " x $fake_pos .$ind[($current / 10) % 8], $rate); |
724 |
} |
} |
725 |
} |
} |
726 |
|
|
735 |
my $isis_db = $cfg -> val($database, 'isis_db') || die "$database doesn't have 'isis_db' defined!"; |
my $isis_db = $cfg -> val($database, 'isis_db') || die "$database doesn't have 'isis_db' defined!"; |
736 |
|
|
737 |
$import2cp = Text::Iconv->new($config->{isis_codepage},$codepage); |
$import2cp = Text::Iconv->new($config->{isis_codepage},$codepage); |
738 |
my $db = OpenIsis::open( $isis_db ); |
my $db = new Biblio::Isis( isisdb => $isis_db ); |
|
|
|
|
# check if .txt database for OpenIsis is zero length, |
|
|
# if so, erase it and re-open database |
|
|
sub check_txt_db { |
|
|
my $isis_db = shift || die "need isis database name"; |
|
|
my $reopen = 0; |
|
|
|
|
|
if (-e $isis_db.".TXT") { |
|
|
print STDERR "WARNING: removing $isis_db.TXT OpenIsis database...\n"; |
|
|
unlink $isis_db.".TXT" || warn "FATAL: unlink error on '$isis_db.TXT': $!"; |
|
|
$reopen++; |
|
|
} |
|
|
if (-e $isis_db.".PTR") { |
|
|
print STDERR "WARNING: removing $isis_db.PTR OpenIsis database...\n"; |
|
|
unlink $isis_db.".PTR" || warn "FATAL: unlink error on '$isis_db.PTR': $!"; |
|
|
$reopen++; |
|
|
} |
|
|
return OpenIsis::open( $isis_db ) if ($reopen); |
|
|
} |
|
|
|
|
|
# EOF error |
|
|
if ($db == -1) { |
|
|
$db = check_txt_db($isis_db); |
|
|
if ($db == -1) { |
|
|
print STDERR "FATAL: OpenIsis can't open zero size file $isis_db\n"; |
|
|
next; |
|
|
} |
|
|
} |
|
739 |
|
|
740 |
# OpenIsis::ERR_BADF |
my $max_rowid = $db->count || die "can't find maxmfn"; |
|
if ($db == -4) { |
|
|
print STDERR "FATAL: OpenIsis can't find file $isis_db\n"; |
|
|
next; |
|
|
# OpenIsis::ERR_IO |
|
|
} elsif ($db == -5) { |
|
|
print STDERR "FATAL: OpenIsis can't access file $isis_db\n"; |
|
|
next; |
|
|
} elsif ($db < 0) { |
|
|
print STDERR "FATAL: OpenIsis unknown error $db with file $isis_db\n"; |
|
|
next; |
|
|
} |
|
|
|
|
|
my $max_rowid = OpenIsis::maxRowid( $db ); |
|
|
|
|
|
# if 0 records, try to rease isis .txt database |
|
|
if ($max_rowid == 0) { |
|
|
# force removal of database |
|
|
$db = check_txt_db($isis_db); |
|
|
$max_rowid = OpenIsis::maxRowid( $db ); |
|
|
} |
|
741 |
|
|
742 |
print STDERR "Reading database: $isis_db [$max_rowid rows]\n"; |
print STDERR "Reading database: $isis_db [$max_rowid rows]\n"; |
743 |
|
|
744 |
my $path = $database; |
my $path = $database; |
745 |
|
|
746 |
for (my $row_id = 1; $row_id <= $max_rowid; $row_id++ ) { |
for (my $row_id = 1; $row_id <= $max_rowid; $row_id++ ) { |
747 |
my $row = OpenIsis::read( $db, $row_id ); |
my $row = $db->to_hash( $row_id ); |
748 |
if ($row && $row->{mfn}) { |
if ($row) { |
749 |
|
|
750 |
|
$row->{mfn} = $row_id; |
751 |
|
$row->{record} = $db->{record}; |
752 |
|
|
753 |
progress($row->{mfn}, $max_rowid); |
progress($row->{mfn}, $max_rowid); |
754 |
|
|
755 |
my $swishpath = $path."#".int($row->{mfn}); |
my $swishpath = $path."#".int($row->{mfn}); |
763 |
} |
} |
764 |
} |
} |
765 |
} |
} |
|
# for this to work with current version of OpenIsis (0.9.0) |
|
|
# you might need my patch from |
|
|
# http://www.rot13.org/~dpavlin/projects/openisis-0.9.0-perl_close.diff |
|
|
OpenIsis::close($db); |
|
766 |
print STDERR "\n"; |
print STDERR "\n"; |
767 |
|
|
768 |
} elsif ($type_base eq "excel") { |
} elsif ($type_base eq "excel") { |
796 |
for(my $iC = $oWorksheet->{MinCol} ; defined $oWorksheet->{MaxCol} && $iC <= $oWorksheet->{MaxCol} ; $iC++) { |
for(my $iC = $oWorksheet->{MinCol} ; defined $oWorksheet->{MaxCol} && $iC <= $oWorksheet->{MaxCol} ; $iC++) { |
797 |
my $cell = $oWorksheet->{Cells}[$iR][$iC]; |
my $cell = $oWorksheet->{Cells}[$iR][$iC]; |
798 |
if ($cell) { |
if ($cell) { |
799 |
$row->{int2col($iC)} = $cell->Value; |
# this conversion is a cludge. |
800 |
|
# Files from Excell could have |
801 |
|
# characters which don't fit into |
802 |
|
# destination encoding. |
803 |
|
$row->{int2col($iC)} = $utf2cp->convert($cell->Value) || $cell->Value; |
804 |
} |
} |
805 |
} |
} |
806 |
|
|
826 |
} |
} |
827 |
} elsif ($type_base eq "marc") { |
} elsif ($type_base eq "marc") { |
828 |
|
|
829 |
require MARC; |
require MARC::File::USMARC; |
830 |
|
|
831 |
$import2cp = Text::Iconv->new($config->{marc_codepage},$codepage); |
$import2cp = Text::Iconv->new($config->{marc_codepage},$codepage); |
832 |
my $marc_file = $cfg -> val($database, 'marc_file') || die "$database doesn't have 'marc_file' defined!"; |
my $marc_file = $cfg -> val($database, 'marc_file') || die "$database doesn't have 'marc_file' defined!"; |
833 |
|
|
834 |
# optional argument is format |
# optional argument is format |
835 |
my $format = x($config->{marc_format}) || 'usmarc'; |
warn "marc_format is no longer used!" if ($config->{marc_format}); |
|
|
|
836 |
print STDERR "Reading MARC file '$marc_file'\n"; |
print STDERR "Reading MARC file '$marc_file'\n"; |
837 |
|
|
838 |
my $marc = new MARC; |
my $marc = MARC::File::USMARC->in( $marc_file ) |
839 |
my $nr = $marc->openmarc({ |
|| die "Can't open MARC file '$marc_file': ".$MARC::File::ERROR; |
|
file=>$marc_file, format=>$format |
|
|
}) || die "Can't open MARC file '$marc_file' with format '$format'"; |
|
840 |
|
|
841 |
# read MARC file in memory |
# count records in MARC file |
842 |
$marc->nextmarc(-1); |
sub marc_count { |
843 |
|
my $filename = shift || die; |
844 |
|
my $file = MARC::File::USMARC->in($filename) || die $MARC::File::ERROR; |
845 |
|
my $count = 0; |
846 |
|
while ($file->skip()) { |
847 |
|
$count++; |
848 |
|
} |
849 |
|
return $count; |
850 |
|
} |
851 |
|
|
852 |
my $max_rec = $marc->marc_count(); |
my $count = marc_count($marc_file) || warn "no records in '$marc_file'?"; |
853 |
|
|
854 |
for(my $i=1; $i<=$max_rec; $i++) { |
my $i = 0; |
855 |
|
|
856 |
progress($i,$max_rec); |
while( my $rec = $marc->next() ) { |
857 |
|
|
858 |
# store value for marc_sf.pm |
progress($i++,$count); |
|
$main::cache->{marc_record} = $i; |
|
859 |
|
|
860 |
my $swishpath = $database."#".$i; |
my $swishpath = $database."#".$i; |
861 |
|
|
862 |
if (my $xml = data2xml($type_base,$marc,$add_xml,$cfg,$database)) { |
if (my $xml = data2xml($type_base,$rec,$add_xml,$cfg,$database)) { |
863 |
$xml = $cp2utf->convert($xml); |
$xml = $cp2utf->convert($xml); |
864 |
use bytes; # as opposed to chars |
use bytes; # as opposed to chars |
865 |
print "Path-Name: $swishpath\n"; |
print "Path-Name: $swishpath\n"; |
930 |
|
|
931 |
=head1 DESCRIPTION |
=head1 DESCRIPTION |
932 |
|
|
933 |
This command will read ISIS data file using OpenIsis perl module, MARC |
This command will read ISIS data file using IsisDB perl module, MARC |
934 |
records using MARC module and optionally Micro$oft Excel files to |
records using MARC module and optionally Micro$oft Excel files to |
935 |
create one XML file for usage with I<SWISH-E> indexer. Dispite it's name, |
create one XML file for usage with I<SWISH-E> indexer. Dispite it's name, |
936 |
this script B<isn't general xml generator> from isis files (isis allready |
this script B<isn't general xml generator> from isis files (isis allready |