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; |
|
|
|
|
|
$|=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; # there is no other, right now ;-) |
|
|
|
|
|
my $index = new index_DBI(); # open index |
|
|
|
|
|
my %opts; |
|
|
|
|
|
# usage: |
|
|
# -d directory name |
|
|
# -m multiple directories |
|
|
# -q quiet |
|
|
# -s run swish |
|
|
|
|
|
getopts('d:m:qs', \%opts); |
|
|
|
|
|
my $db_dir; |
|
|
|
|
|
#die "usage: $0 -d [database_dir] -m [database1,database2] " if (! %opts); |
|
|
|
|
|
#print Dumper($config->{indexer}); |
|
|
#print "-" x 70,"\n"; |
|
|
|
|
|
Text::Iconv->raise_error(1); # Conversion errors raise exceptions |
|
|
|
|
|
#my $isis_codepage = Text::Iconv->new($config->{isis_codepage},'UTF8'); |
|
|
#my $index_codepage = Text::Iconv->new($config->{isis_codepage},$config->{index_codepage}); |
|
|
my $isis_codepage; |
|
|
my $index_codepage; |
|
|
my $cludge_codepage = Text::Iconv->new('UTF8','ISO8859-1'); |
|
|
my $xml_codepage; |
|
|
|
|
|
sub isis2xml { |
|
|
|
|
|
use xmlify; |
|
|
|
|
|
my $row = shift @_; |
|
|
my $add_xml = shift @_; |
|
|
|
|
|
my $xml; |
|
|
|
|
|
use parse_format; |
|
|
|
|
|
my $html = ""; # html formatted display output |
|
|
|
|
|
my %field_usage; # counter for usage of each field |
|
|
|
|
|
foreach my $field (keys %{$config->{indexer}}) { |
|
|
|
|
|
$field_usage{$field}++; |
|
|
|
|
|
my $swish_data = ""; |
|
|
my $display_data = ""; |
|
|
|
|
|
foreach my $x (@{$config->{indexer}->{$field}->{isis}}) { |
|
|
|
|
|
my $format = $x->{content}; |
|
|
my ($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"); |
|
|
#print STDERR "## s: $s d: $d i: $i ## $format ##\n"; |
|
4 |
|
|
5 |
$format = $cludge_codepage->convert($format); |
all2all.pl - basic script for all WebPAC needs |
|
my ($swish,$display) = parse_format($format,$row); |
|
|
#print STDERR "s: $swish\nd: $display\n" if ($swish); |
|
6 |
|
|
7 |
#print STDERR "swish: $swish<-- display: $display<--\n"; |
=cut |
|
# FIX: this is ugly, UGLY, cludge: OpenIsis return |
|
|
# UTF8 encoding of strings, but as if source charset |
|
|
# is ISO8859-1 and not some other. This breaks our |
|
|
# isis character encoding, so we convert it first |
|
|
# back to ISO8859-1 (which can actually be different |
|
|
# encoding in isis) |
|
8 |
|
|
9 |
$swish_data .= $swish if ($s && $swish); |
use strict; |
10 |
$display_data .= $display if ($d && $display); |
use Data::Dumper; |
11 |
|
use Carp; |
12 |
|
|
13 |
# insert into index |
use lib './lib'; |
14 |
if ($i && $display) { |
use WebPAC; |
15 |
my $index_data = $index_codepage->convert($display) || $display; |
use WebPAC::jsFind; |
16 |
$index->insert($field, $index_data, $db_dir); |
|
17 |
} |
my $webpac = new WebPAC( |
18 |
} |
code_page => 'ISO-8859-2', |
19 |
|
limit_mfn => 100, |
20 |
|
# debug => 1, |
21 |
|
) || die; |
22 |
|
|
23 |
|
my $log = $webpac->_get_logger(); |
24 |
|
|
25 |
|
my $index = new WebPAC::jsFind( |
26 |
|
index_path => './out/index', |
27 |
|
) || die; |
28 |
|
|
29 |
|
$|=1; |
30 |
|
|
31 |
#print STDERR "s_d: $swish_data\nd_d: $display_data\n" if ($swish_data); |
my $maxmfn = $webpac->open_isis( |
32 |
if ($display_data) { |
filename => shift @ARGV || '/data/hidra/THS/THS', |
33 |
$display_data = $isis_codepage->convert($display_data) || die "Can't convert '$display_data' !"; |
lookup => [ |
34 |
# FIX: this is removed and replaced by html tag. |
{ 'key' => 'd:v900', 'val' => 'v250^a' }, |
35 |
#$xml .= xmlify($field."_display", $display_data); |
# { 'eval' => '"v901^a" eq "Područje"', 'key' => 'pa:v561^4:v562^4:v461^1', 'val' => 'v900' }, |
36 |
|
# { 'eval '=> '"v901^a" eq "Mikrotezaurus"', 'key' => 'a:v561^4:v562^4:v461^1', 'val' => 'v900' }, |
37 |
if ($field eq "headline") { |
# { 'eval' => '"v901^a" eq "Deskriptor"', 'key' => 'a:v561^4:v562^4:v461^1', 'val' => 'v900' }, |
38 |
$xml .= xmlify("headline", $display_data); |
{ 'key' => 'a:v561^4:v562^4:v461^1', 'val' => 'v900' }, |
39 |
} else { |
{ 'key' => '900_mfn:v900', 'val' => 'v000' }, |
40 |
|
], |
41 |
# find field name (signular, plural) |
); |
42 |
my $field_name = ""; |
|
43 |
if ($config->{indexer}->{$field}->{name_singular} && $field_usage{$field} == 1) { |
$log->info("rows: $maxmfn"); |
44 |
$field_name = $config->{indexer}->{$field}->{name_singular}."#-#"; |
|
45 |
} elsif ($config->{indexer}->{$field}->{name_plural}) { |
$webpac->open_import_xml(type => 'isis_hidra_ths'); |
46 |
$field_name = $config->{indexer}->{$field}->{name_plural}."#-#"; |
|
47 |
} else { |
while (my $rec = $webpac->fetch_rec) { |
48 |
$field_name = $config->{indexer}->{$field}->{name}."#-#"; |
|
49 |
} |
my @ds = $webpac->data_structure($rec); |
50 |
if ($field_name) { |
|
51 |
$html .= $xml_codepage->convert($field_name); |
if ($log->is_debug) { |
52 |
} |
$log->debug("rec = ",Dumper($rec)); |
53 |
$html .= $display_data."###\n"; |
$log->debug("ds = ",Dumper(\@ds)); |
54 |
} |
} |
|
} |
|
|
if ($swish_data) { |
|
|
my $i = Text::Iconv->new($config->{isis_codepage},'ISO8859-2'); |
|
|
$swish_data = $i->convert($swish_data); |
|
|
$xml .= xmlify($field."_swish",unac_string('ISO8859-2',$swish_data)); |
|
|
#$swish_data = $isis_codepage->convert($swish_data)."##" || $swish_data; |
|
|
#$xml .= xmlify($field."_swish",unac_string($config->{isis_codepage},$swish_data)); |
|
|
} |
|
55 |
|
|
56 |
|
next if (! @ds); |
57 |
|
|
58 |
} |
my $filename = $webpac->{'current_filename'}; |
59 |
|
|
60 |
# dump formatted output in <html> |
if ($filename) { |
61 |
if ($html) { |
open(OUT,"> $filename") || $log->logdie("can't open output '$filename': $!"); |
62 |
$xml .= xmlify("html",$html); |
print OUT $webpac->output( |
63 |
} |
template => 'html.tt', |
64 |
|
data => \@ds, |
65 |
if ($xml) { |
); |
66 |
#print STDERR "x: $xml\n"; |
close(OUT); |
|
$xml .= $add_xml if ($add_xml); |
|
|
return "<xml>\n$xml</xml>\n"; |
|
67 |
} else { |
} else { |
68 |
return; |
print $webpac->output( |
69 |
|
template => 'text.tt', |
70 |
|
data => \@ds, |
71 |
|
); |
72 |
} |
} |
|
} |
|
|
|
|
|
########################################################################## |
|
|
|
|
|
my $cfg = new Config::IniFiles( -file => $config_file ); |
|
|
|
|
|
foreach my $database ($cfg->Sections) { |
|
73 |
|
|
74 |
my $isis_db = $cfg -> val($database, 'isis_db'); |
my $headline; |
75 |
my $type = $cfg -> val($database, 'type'); |
foreach my $ds (@ds) { |
76 |
my $add_xml = $cfg -> val($database, 'xml'); |
if ($ds->{'tag'} eq 'headline') { |
77 |
|
$headline = join(" ",@{$ds->{'display'}}); |
78 |
# read configuration for this type |
last; |
79 |
$config=XMLin("./import_xml/$type.xml", forcearray => [ 'isis' ], forcecontent => 1); |
} |
|
$isis_codepage = Text::Iconv->new($config->{isis_codepage},'UTF8'); |
|
|
$index_codepage = Text::Iconv->new($config->{isis_codepage},$config->{index_codepage}); |
|
|
$xml_codepage = Text::Iconv->new($cfg->val($database,'xml_codepage'),'UTF8'); |
|
|
|
|
|
my $db = OpenIsis::open( $isis_db ); |
|
|
if (0) { |
|
|
# # FIX |
|
|
# if (! $db ) { |
|
|
print STDERR "WARNING: can't open '$isis_db'\n"; |
|
|
next ; |
|
80 |
} |
} |
81 |
|
|
82 |
my $max_rowid = OpenIsis::maxRowid( $db ); |
my $f = $filename; |
83 |
|
$f =~ s!out/!!; |
84 |
|
|
85 |
print STDERR "Reading database: $isis_db [$max_rowid rows]\n"; |
# save into index |
86 |
|
foreach my $ds (@ds) { |
87 |
|
next if (! $ds->{'swish'}); |
88 |
|
|
89 |
|
$index->insert( |
90 |
|
index_name => $ds->{'tag'}, |
91 |
|
path => $f, |
92 |
|
headline => $headline, |
93 |
|
words => join(" ",@{$ds->{'swish'}}) |
94 |
|
); |
95 |
|
} |
96 |
|
|
97 |
my $path = $database; # was $isis_db |
# print Dumper(\@ds); |
98 |
|
|
99 |
my $last_p = 0; |
} |
100 |
|
|
101 |
# { my $row_id = 4514; |
if ($log->is_debug) { |
102 |
# FIX |
$log->debug("lookup hash: ",Dumper($webpac->{'lookup'})); |
103 |
for (my $row_id = 1; $row_id <= $max_rowid; $row_id++ ) { |
$log->debug("data hash: ",Dumper($webpac->{'data'})); |
|
my $row = OpenIsis::read( $db, $row_id ); |
|
|
if ($row && $row->{mfn}) { |
|
|
#print STDERR "mfn: ",$row->{mfn},"\n"; |
|
|
# output current process indicator |
|
|
my $p = int($row->{mfn} * 100 / $max_rowid); |
|
|
if ($p != $last_p) { |
|
|
printf STDERR ("%5d / %5d [%-51s] %-2d %% \r",$row->{mfn},$max_rowid,"=" x ($p/2).">", $p ) if (! $opts{q}); |
|
|
$last_p = $p; |
|
|
} |
|
|
|
|
|
if (my $xml = isis2xml($row,$add_xml)) { |
|
|
#print STDERR "--ret-->$xml\n"; |
|
|
print "Path-Name: $path#".int($row->{mfn})."\n"; |
|
|
print "Content-Length: ".(length($xml)+1)."\n"; |
|
|
print "Document-Type: XML\n\n$xml\n"; |
|
|
} |
|
|
} |
|
|
} |
|
|
print STDERR "\n"; |
|
104 |
} |
} |
105 |
|
|
|
# call this to commit index |
|
106 |
$index->close; |
$index->close; |
107 |
|
|
|
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 |
|