1 |
#!/usr/bin/perl -w |
#!/usr/bin/perl -w |
2 |
|
|
3 |
use strict; |
=head1 NAME |
|
use OpenIsis; |
|
|
use Getopt::Std; |
|
|
use Data::Dumper; |
|
|
use XML::Simple; |
|
|
use Text::Unaccent 1.02; # 1.01 won't compile on my platform, |
|
|
use Text::Iconv; |
|
|
use Config::IniFiles; |
|
|
use Encode; |
|
|
|
|
|
$|=1; |
|
|
|
|
|
my $config_file = $0; |
|
|
$config_file =~ s/\.pl$/.conf/; |
|
|
die "FATAL: can't find configuration file '$config_file'" if (! -e $config_file); |
|
|
|
|
|
my $config; |
|
|
|
|
|
#use index_DBI; # default DBI module for index |
|
|
use index_DBI_cache; # faster DBI module using memory cache |
|
|
my $index; |
|
|
|
|
|
my %opts; |
|
|
|
|
|
# usage: |
|
|
# -d directory name |
|
|
# -m multiple directories |
|
|
# -q quiet |
|
|
# -s run swish |
|
4 |
|
|
5 |
getopts('d:m:qs', \%opts); |
all2all.pl - basic script for all WebPAC needs |
6 |
|
|
7 |
my $path; # this is name of database |
=cut |
8 |
|
|
9 |
Text::Iconv->raise_error(0); # Conversion errors don't raise exceptions |
use strict; |
10 |
|
use Data::Dumper; |
11 |
|
use Carp; |
12 |
|
|
13 |
# this is encoding of all files on disk, including import_xml/*.xml file and |
use lib './lib'; |
14 |
# filter/*.pm files! It will be used to store strings in perl internally! |
use WebPAC; |
|
my $codepage = 'ISO-8859-2'; |
|
15 |
|
|
16 |
my $utf2cp = Text::Iconv->new('UTF-8',$codepage); |
my $webpac = new WebPAC( |
17 |
# this function will convert data from XML files to local encoding |
code_page => 'ISO-8859-2', |
18 |
sub x { |
limit_mfn => 500, |
19 |
return $utf2cp->convert($_[0]); |
debug => 1, |
20 |
} |
) || die; |
21 |
|
|
22 |
# decode isis/excel or other import codepage |
$|=1; |
|
my $import2cp; |
|
|
|
|
|
# outgoing xml must be in UTF-8 |
|
|
my $cp2utf = Text::Iconv->new($codepage,'UTF-8'); |
|
23 |
|
|
24 |
# mapping between data type and tag which specify |
my $maxmfn = $webpac->open_isis( |
25 |
# format in XML file |
filename => shift @ARGV || '/data/hidra/THS/THS', |
26 |
my %type2tag = ( |
lookup => [ |
27 |
'isis' => 'isis', |
{ 'key' => 'd:v900', 'val' => 'v250^a' }, |
28 |
'excel' => 'column', |
# { 'eval' => '"v901^a" eq "Područje"', 'key' => 'pa:v561^4:v562^4:v461^1', 'val' => 'v900' }, |
29 |
'marc' => 'marc', |
# { 'eval '=> '"v901^a" eq "Mikrotezaurus"', 'key' => 'a:v561^4:v562^4:v461^1', 'val' => 'v900' }, |
30 |
'feed' => 'feed' |
# { 'eval' => '"v901^a" eq "Deskriptor"', 'key' => 'a:v561^4:v562^4:v461^1', 'val' => 'v900' }, |
31 |
|
{ 'key' => 'a:v561^4:v562^4:v461^1', 'val' => 'v900' }, |
32 |
|
], |
33 |
); |
); |
34 |
|
|
35 |
sub data2xml { |
print "rows: $maxmfn\n\n"; |
36 |
|
|
37 |
use xmlify; |
$webpac->open_import_xml(type => 'isis_hidra_ths'); |
|
|
|
|
my $type = shift @_; |
|
|
my $row = shift @_; |
|
|
my $add_xml = shift @_; |
|
|
# needed to read values from configuration file |
|
|
my $cfg = shift @_; |
|
|
my $database = shift @_; |
|
|
|
|
|
my $xml; |
|
|
|
|
|
use parse_format; |
|
|
|
|
|
my $html = ""; # html formatted display output |
|
|
|
|
|
my %field_usage; # counter for usage of each field |
|
|
|
|
|
# sort subrouting using order="" attribute |
|
|
sub by_order { |
|
|
return 0 if (! $config->{indexer}->{$a}->{order}); |
|
|
return 0 if (! $config->{indexer}->{$b}->{order}); |
|
|
|
|
|
return $config->{indexer}->{$a}->{order} <=> |
|
|
$config->{indexer}->{$b}->{order} ; |
|
|
} |
|
|
|
|
|
foreach my $field (sort by_order keys %{$config->{indexer}}) { |
|
|
|
|
|
$field=x($field); |
|
|
|
|
|
$field_usage{$field}++; |
|
|
|
|
|
my $swish_data = ""; |
|
|
my $display_data = ""; |
|
|
my $line_delimiter; |
|
|
|
|
|
my ($swish,$display); |
|
|
|
|
|
my $tag = $type2tag{$type} || die "can't find which tag to use for type $type"; |
|
|
foreach my $x (@{$config->{indexer}->{$field}->{$tag}}) { |
|
|
|
|
|
my $format = x($x->{content}); |
|
|
my $delimiter = x($x->{delimiter}) || ' '; |
|
|
|
|
|
my $repeat_off = 0; # repeatable offset |
|
|
|
|
|
my ($s,$d,$i) = (1,1,0); # swish, display default |
|
|
$s = 0 if (lc($x->{type}) eq "display"); |
|
|
$d = 0 if (lc($x->{type}) eq "swish"); |
|
|
($s,$d,$i) = (0,0,1) if (lc($x->{type}) eq "index"); |
|
|
|
|
|
# what will separate last line from this one? |
|
|
if ($display_data && $x->{append} && $x->{append} eq "1") { |
|
|
$line_delimiter = ' '; |
|
|
} elsif ($display_data) { |
|
|
$line_delimiter = '<br/>'; |
|
|
} |
|
|
|
|
|
# init vars so that we go into while... |
|
|
($swish,$display) = (1,1); |
|
|
|
|
|
if ($swish || $display) { |
|
|
($swish,$display) = parse_format($type, $format,$row,$repeat_off++,$import2cp); |
|
|
# filter="name" ; filter this field through |
|
|
# filter/[name].pm |
|
|
my $filter = $x->{filter}; |
|
|
if ($filter) { |
|
|
require "filter/".$filter.".pm"; |
|
|
} |
|
|
# type="swish" ; field for swish |
|
|
if ($s && $swish) { |
|
|
if ($filter) { |
|
|
no strict 'refs'; |
|
|
$swish_data .= join(" ",&$filter($swish)); |
|
|
} else { |
|
|
$swish_data .= $swish; |
|
|
} |
|
|
} |
|
|
|
|
|
# type="display" ; field for display |
|
|
if ($d && $display) { |
|
|
if ($line_delimiter && $display_data) { |
|
|
$display_data .= $line_delimiter; |
|
|
undef $line_delimiter; |
|
|
} |
|
|
if ($filter) { |
|
|
no strict 'refs'; |
|
|
$display_data .= join($delimiter,&$filter($display)); |
|
|
} else { |
|
|
if ($display_data) { |
|
|
$display_data .= $delimiter.$display; |
|
|
} else { |
|
|
$display_data .= $display; |
|
|
} |
|
|
} |
|
|
} |
|
|
|
|
|
# type="index" ; insert into index |
|
|
if ($i && $display) { |
|
|
my $index_data = $display; |
|
|
if ($filter) { |
|
|
no strict 'refs'; |
|
|
foreach my $d (&$filter($index_data)) { |
|
|
$index->insert($field, $d, $path); |
|
|
} |
|
|
} else { |
|
|
$index->insert($field, $index_data, $path); |
|
|
} |
|
|
} |
|
|
} |
|
|
} |
|
|
|
|
|
# now try to parse variables from configuration file |
|
|
foreach my $x (@{$config->{indexer}->{$field}->{'config'}}) { |
|
|
|
|
|
my $delimiter = x($x->{delimiter}) || ' '; |
|
|
my $val = $cfg->val($database, x($x->{content})); |
|
|
|
|
|
my ($s,$d,$i) = (1,1,0); # swish, display default |
|
|
$s = 0 if (lc($x->{type}) eq "display"); |
|
|
$d = 0 if (lc($x->{type}) eq "swish"); |
|
|
($s,$d,$i) = (0,0,1) if (lc($x->{type}) eq "index"); |
|
|
|
|
|
if ($val) { |
|
|
$display_data .= $delimiter.$val if ($d); |
|
|
$swish_data .= $val if ($s); |
|
|
$index->insert($field, $val, $path) if ($i); |
|
|
} |
|
|
|
|
|
} |
|
|
|
|
|
|
|
|
if ($display_data) { |
|
|
|
|
|
if ($field eq "headline") { |
|
|
$xml .= xmlify("headline", $display_data); |
|
|
} else { |
|
|
|
|
|
# find field name (signular, plural) |
|
|
my $field_name = ""; |
|
|
if ($config->{indexer}->{$field}->{name_singular} && $field_usage{$field} == 1) { |
|
|
$field_name = $config->{indexer}->{$field}->{name_singular}."#-#"; |
|
|
} elsif ($config->{indexer}->{$field}->{name_plural}) { |
|
|
$field_name = $config->{indexer}->{$field}->{name_plural}."#-#"; |
|
|
} elsif ($config->{indexer}->{$field}->{name}) { |
|
|
$field_name = $config->{indexer}->{$field}->{name}."#-#"; |
|
|
} else { |
|
|
print STDERR "WARNING: field '$field' doesn't have 'name' attribute!"; |
|
|
} |
|
|
if ($field_name) { |
|
|
$html .= x($field_name); |
|
|
} |
|
|
$html .= $display_data."###\n"; |
|
|
} |
|
|
} |
|
|
if ($swish_data) { |
|
|
# remove extra spaces |
|
|
$swish_data =~ s/ +/ /g; |
|
|
$swish_data =~ s/ +$//g; |
|
|
|
|
|
$xml .= xmlify($field."_swish", unac_string($codepage,$swish_data)); |
|
|
} |
|
|
|
|
|
|
|
|
} |
|
|
|
|
|
# dump formatted output in <html> |
|
|
if ($html) { |
|
|
$xml .= xmlify("html",$html); |
|
|
} |
|
|
|
|
|
if ($xml) { |
|
|
$xml .= $add_xml if ($add_xml); |
|
|
return "<xml>\n$xml</xml>\n"; |
|
|
} else { |
|
|
return; |
|
|
} |
|
|
} |
|
38 |
|
|
39 |
########################################################################## |
while (my $rec = $webpac->fetch_rec) { |
40 |
|
|
41 |
# read configuration for this script |
my @ds = $webpac->data_structure($rec); |
|
my $cfg = new Config::IniFiles( -file => $config_file ); |
|
42 |
|
|
43 |
# read global.conf configuration |
print $webpac->output( |
44 |
my $cfg_global = new Config::IniFiles( -file => 'global.conf' ); |
template => 'text.tt', |
45 |
|
data => \@ds, |
46 |
|
) if (@ds); |
47 |
|
|
|
# open index |
|
|
$index = new index_DBI( |
|
|
$cfg_global->val('global', 'dbi_dbd'), |
|
|
$cfg_global->val('global', 'dbi_dsn'), |
|
|
$cfg_global->val('global', 'dbi_user'), |
|
|
$cfg_global->val('global', 'dbi_passwd') || '', |
|
|
); |
|
|
|
|
|
foreach my $database ($cfg->Sections) { |
|
|
|
|
|
my $type = lc($cfg -> val($database, 'type')) || die "$database doesn't have 'type' defined"; |
|
|
my $add_xml = $cfg -> val($database, 'xml'); # optional |
|
|
|
|
|
print STDERR "reading ./import_xml/$type.xml\n"; |
|
|
|
|
|
# extract just type basic |
|
|
my $type_base = $type; |
|
|
$type_base =~ s/_.+$//g; |
|
|
|
|
|
$config=XMLin("./import_xml/$type.xml", forcearray => [ $type2tag{$type_base}, 'config' ], forcecontent => 1); |
|
|
|
|
|
# output current progress indicator |
|
|
my $last_p = 0; |
|
|
sub progress { |
|
|
#return if (! $opts{q}); # FIXME |
|
|
my $current = shift; |
|
|
my $total = shift || 1; |
|
|
my $p = int($current * 100 / $total); |
|
|
if ($p != $last_p) { |
|
|
printf STDERR ("%5d / %5d [%-51s] %-2d %% \r",$current,$total,"=" x ($p/2).">", $p ); |
|
|
$last_p = $p; |
|
|
} |
|
|
} |
|
|
|
|
|
my $fake_dir = 1; |
|
|
sub fakeprogress { |
|
|
my $current = shift @_; |
|
|
|
|
|
my @ind = ('-','\\','|','/','-','\\','|','/', '-'); |
|
|
|
|
|
$last_p += $fake_dir; |
|
|
$fake_dir = -$fake_dir if ($last_p > 1000 || $last_p < 0); |
|
|
if ($last_p % 10 == 0) { |
|
|
printf STDERR ("%5d / %5s [%-51s]\r",$current,"?"," " x ($last_p/20).$ind[($last_p/20) % $#ind]); |
|
|
} |
|
|
} |
|
|
|
|
|
# now read database |
|
|
print STDERR "using: $type...\n"; |
|
|
|
|
|
if ($type_base eq "isis") { |
|
|
|
|
|
my $isis_db = $cfg -> val($database, 'isis_db') || die "$database doesn't have 'isis_db' defined!"; |
|
|
|
|
|
$import2cp = Text::Iconv->new($config->{isis_codepage},$codepage); |
|
|
my $db = OpenIsis::open( $isis_db ); |
|
|
|
|
|
my $max_rowid = OpenIsis::maxRowid( $db ); |
|
|
|
|
|
print STDERR "Reading database: $isis_db [$max_rowid rows]\n"; |
|
|
|
|
|
my $path = $database; |
|
|
|
|
|
for (my $row_id = 1; $row_id <= $max_rowid; $row_id++ ) { |
|
|
my $row = OpenIsis::read( $db, $row_id ); |
|
|
if ($row && $row->{mfn}) { |
|
|
|
|
|
progress($row->{mfn}, $max_rowid); |
|
|
|
|
|
my $swishpath = $path."#".int($row->{mfn}); |
|
|
|
|
|
if (my $xml = data2xml($type_base,$row,$add_xml,$cfg,$database)) { |
|
|
$xml = $cp2utf->convert($xml); |
|
|
use bytes; # as opposed to chars |
|
|
print "Path-Name: $swishpath\n"; |
|
|
print "Content-Length: ".(length($xml)+1)."\n"; |
|
|
print "Document-Type: XML\n\n$xml\n"; |
|
|
} |
|
|
} |
|
|
} |
|
|
print STDERR "\n"; |
|
|
|
|
|
} elsif ($type_base eq "excel") { |
|
|
use Spreadsheet::ParseExcel; |
|
|
use Spreadsheet::ParseExcel::Utility qw(int2col); |
|
|
|
|
|
$import2cp = Text::Iconv->new($config->{excel_codepage},$codepage); |
|
|
my $excel_file = $cfg -> val($database, 'excel_file') || die "$database doesn't have 'excel_file' defined!"; |
|
|
|
|
|
my $sheet = x($config->{sheet}) || die "no sheet in $type.xml"; |
|
|
my $start_row = x($config->{start_row}) - 1 || die "no start_row in $type.xml"; |
|
|
|
|
|
my $oBook = Spreadsheet::ParseExcel::Workbook->Parse($excel_file) || die "can't open Excel file '$excel_file'"; |
|
|
|
|
|
my $sheet_nr = 0; |
|
|
foreach my $oWks (@{$oBook->{Worksheet}}) { |
|
|
#print STDERR "-- SHEET $sheet_nr:", $oWks->{Name}, "\n"; |
|
|
last if ($oWks->{Name} eq $sheet); |
|
|
$sheet_nr++; |
|
|
} |
|
|
|
|
|
my $oWorksheet = $oBook->{Worksheet}[$sheet_nr]; |
|
|
|
|
|
print STDERR "using sheet: ",$oWorksheet->{Name},"\n"; |
|
|
defined ($oWorksheet) || die "can't find sheet '$sheet' in $excel_file"; |
|
|
my $end_row = x($config->{end_row}) || $oWorksheet->{MaxRow}; |
|
|
|
|
|
for(my $iR = $start_row ; defined $end_row && $iR <= $end_row ; $iR++) { |
|
|
my $row; |
|
|
for(my $iC = $oWorksheet->{MinCol} ; defined $oWorksheet->{MaxCol} && $iC <= $oWorksheet->{MaxCol} ; $iC++) { |
|
|
my $cell = $oWorksheet->{Cells}[$iR][$iC]; |
|
|
if ($cell) { |
|
|
$row->{int2col($iC)} = $cell->Value; |
|
|
} |
|
|
} |
|
|
|
|
|
progress($iR, $end_row); |
|
|
|
|
|
# print "row[$iR/$end_row] "; |
|
|
# foreach (keys %{$row}) { |
|
|
# print "$_: ",$row->{$_},"\t"; |
|
|
# } |
|
|
# print "\n"; |
|
|
|
|
|
my $swishpath = $database."#".$iR; |
|
|
|
|
|
next if (! $row); |
|
|
|
|
|
if (my $xml = data2xml($type_base,$row,$add_xml,$cfg,$database)) { |
|
|
$xml = $cp2utf->convert($xml); |
|
|
use bytes; # as opposed to chars |
|
|
print "Path-Name: $swishpath\n"; |
|
|
print "Content-Length: ".(length($xml)+1)."\n"; |
|
|
print "Document-Type: XML\n\n$xml\n"; |
|
|
} |
|
|
} |
|
|
} elsif ($type_base eq "marc") { |
|
|
|
|
|
use MARC; |
|
|
|
|
|
$import2cp = Text::Iconv->new($config->{marc_codepage},$codepage); |
|
|
my $marc_file = $cfg -> val($database, 'marc_file') || die "$database doesn't have 'marc_file' defined!"; |
|
|
|
|
|
# optional argument is format |
|
|
my $format = x($config->{format}) || 'usmarc'; |
|
|
|
|
|
print STDERR "Reading MARC file '$marc_file'\n"; |
|
|
|
|
|
my $marc = new MARC; |
|
|
my $nr = $marc->openmarc({ |
|
|
file=>$marc_file, format=>$format |
|
|
}) || die "Can't open MARC file '$marc_file'"; |
|
|
|
|
|
my $i=0; # record nr. |
|
|
|
|
|
my $rec; |
|
|
|
|
|
while ($marc->nextmarc(1)) { |
|
|
|
|
|
# XXX |
|
|
fakeprogress($i++); |
|
|
|
|
|
my $swishpath = $database."#".$i; |
|
|
|
|
|
if (my $xml = data2xml($type_base,$marc,$add_xml,$cfg,$database)) { |
|
|
$xml = $cp2utf->convert($xml); |
|
|
use bytes; # as opposed to chars |
|
|
print "Path-Name: $swishpath\n"; |
|
|
print "Content-Length: ".(length($xml)+1)."\n"; |
|
|
print "Document-Type: XML\n\n$xml\n"; |
|
|
} |
|
|
} |
|
|
} elsif ($type_base eq "feed") { |
|
|
|
|
|
$import2cp = Text::Iconv->new($config->{feed_codepage},$codepage); |
|
|
my $prog = x($config->{prog}) || die "$database doesn't have 'prog' defined!"; |
|
|
|
|
|
print STDERR "Reading feed from program '$prog'\n"; |
|
|
|
|
|
open(FEED,"feeds/$prog |") || die "can't start $prog: $!"; |
|
|
|
|
|
my $i=0; # record nr. |
|
|
|
|
|
my $data; |
|
|
my $line=1; |
|
|
|
|
|
while (<FEED>) { |
|
|
chomp; |
|
|
|
|
|
if (/^$/) { |
|
|
my $swishpath = $database."#".$i; |
|
|
|
|
|
if (my $xml = data2xml($type_base,$data,$add_xml,$cfg,$database)) { |
|
|
$xml = $cp2utf->convert($xml); |
|
|
use bytes; # as opposed to chars |
|
|
print "Path-Name: $swishpath\n"; |
|
|
print "Content-Length: ".(length($xml)+1)."\n"; |
|
|
print "Document-Type: XML\n\n$xml\n"; |
|
|
} |
|
|
$line = 1; |
|
|
$data = {}; |
|
|
next; |
|
|
} |
|
|
|
|
|
$data->{$line++} = $_; |
|
|
|
|
|
fakeprogress($i++); |
|
|
|
|
|
} |
|
|
} |
|
48 |
} |
} |
49 |
|
|
50 |
# call this to commit index |
print "## lookup ",Dumper($webpac->{'lookup'}); |
51 |
$index->close; |
print "## data ",Dumper($webpac->{'data'}); |
|
|
|
|
1; |
|
|
__END__ |
|
|
########################################################################## |
|
|
|
|
|
=head1 NAME |
|
|
|
|
|
isis2xml.pl - read isis file and dump XML |
|
|
|
|
|
=head1 DESCRIPTION |
|
|
|
|
|
This command will read ISIS data file using OpenIsis perl module and |
|
|
create XML file for usage with I<SWISH-E> |
|
|
indexer. Dispite it's name, this script B<isn't general xml generator> |
|
|
from isis files (isis allready has something like that). Output of this |
|
|
script is tailor-made for SWISH-E. |
|
|
|
|
|
=head1 AUTHOR |
|
|
|
|
|
Dobrica Pavlinusic <dpavlin@rot13.org> |
|
|
|
|
|
=head1 COPYRIGHT |
|
|
|
|
|
GNU Public License (GPL) v2 or later |
|
|
|
|
|
=head1 SEE ALSO |
|
|
|
|
|
SWISH-E web site at http://www.swish-e.org |
|
|
|
|
|
=cut |
|