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; |
|
|
|
|
|
|
|
|
$|=1; |
|
|
|
|
|
my $config; |
|
|
|
|
|
$config=XMLin(undef, forcearray => [ 'isis' ], forcecontent => 1); |
|
|
|
|
|
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 = $opts{d} || "ps"; # FIX |
|
|
|
|
|
#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 $cludge_codepage = Text::Iconv->new('UTF8','ISO8859-1'); |
|
|
|
|
|
sub isis2xml { |
|
|
|
|
|
use xmlify; |
|
|
|
|
|
my $row = shift @_; |
|
|
|
|
|
my $xml; |
|
|
$xml .= xmlify('db_dir',$db_dir); # FIX remove? |
|
|
|
|
|
use parse_format; |
|
|
|
|
|
|
|
|
foreach my $field (keys %{$config->{indexer}}) { |
|
|
|
|
|
my $swish_data = ""; |
|
|
my $display_data = ""; |
|
|
my $index_data = ""; |
|
|
|
|
|
foreach my $x (@{$config->{indexer}->{$field}->{isis}}) { |
|
4 |
|
|
5 |
my $format = $x->{content}; |
all2all.pl - basic script for all WebPAC needs |
|
my ($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"; |
|
6 |
|
|
7 |
$format = $cludge_codepage->convert($format); |
=cut |
|
my ($swish,$display) = parse_format($format,$row); |
|
|
#print STDERR "s: $swish\nd: $display\n" if ($swish); |
|
8 |
|
|
9 |
#print STDERR "swish: $swish<-- display: $display<--\n"; |
use strict; |
10 |
# FIX: this is ugly, UGLY, cludge: OpenIsis return |
use Data::Dumper; |
11 |
# UTF8 encoding of strings, but as if source charset |
use Carp; |
|
# 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) |
|
12 |
|
|
13 |
$swish_data .= $swish if ($s && $swish); |
use lib './lib'; |
14 |
$display_data .= $display if ($d && $display); |
use WebPAC; |
15 |
$index_data .= $display if ($i && $display); |
use WebPAC::jsFind; |
16 |
} |
use WebPAC::Index; |
17 |
|
|
18 |
|
my $webpac = new WebPAC( |
19 |
|
code_page => 'ISO-8859-2', |
20 |
|
# limit_mfn => 500, |
21 |
|
# debug => 1, |
22 |
|
) || die; |
23 |
|
|
24 |
|
my $log = $webpac->_get_logger(); |
25 |
|
|
26 |
|
my $index = new WebPAC::jsFind( |
27 |
|
index_path => './out/index', |
28 |
|
keys => 10, |
29 |
|
) || die; |
30 |
|
|
31 |
#print STDERR "s_d: $swish_data\nd_d: $display_data\n" if ($swish_data); |
my $thes; |
|
if ($display_data) { |
|
|
$display_data = $isis_codepage->convert($display_data)."##" || $display_data; |
|
|
$xml .= xmlify($field."_display", $display_data); |
|
|
} |
|
|
if ($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)); |
|
|
} |
|
32 |
|
|
33 |
# index |
$|=1; |
|
if ($index_data && $index_data ne "") { |
|
|
$index_data = $index_codepage->convert($index_data) || $index_data; |
|
|
$index->insert($field, $index_data, $db_dir); |
|
|
} |
|
34 |
|
|
35 |
} |
my $maxmfn = $webpac->open_isis( |
36 |
if ($xml) { |
filename => shift @ARGV || '/data/hidra/THS/THS', |
37 |
#print STDERR "x: $xml\n"; |
lookup => [ |
38 |
return "<xml>\n$xml</xml>\n"; |
{ 'key' => 'd:v900', 'val' => 'v250^a' }, |
39 |
|
# { 'eval' => '"v901^a" eq "Područje"', 'key' => 'pa:v561^4:v562^4:v461^1', 'val' => 'v900' }, |
40 |
|
# { 'eval '=> '"v901^a" eq "Mikrotezaurus"', 'key' => 'a:v561^4:v562^4:v461^1', 'val' => 'v900' }, |
41 |
|
# { 'eval' => '"v901^a" eq "Deskriptor"', 'key' => 'a:v561^4:v562^4:v461^1', 'val' => 'v900' }, |
42 |
|
{ 'key' => 'a:v561^4:v562^4:v461^1', 'val' => 'v900' }, |
43 |
|
{ 'key' => '900_mfn:v900', 'val' => 'v000' }, |
44 |
|
], |
45 |
|
); |
46 |
|
|
47 |
|
$log->info("rows: $maxmfn"); |
48 |
|
|
49 |
|
$webpac->open_import_xml(type => 'isis_hidra_ths'); |
50 |
|
|
51 |
|
while (my $rec = $webpac->fetch_rec) { |
52 |
|
|
53 |
|
my @ds = $webpac->data_structure($rec); |
54 |
|
|
55 |
|
if (0 && $log->is_debug) { |
56 |
|
$log->debug("rec = ",Dumper($rec)); |
57 |
|
$log->debug("ds = ",Dumper(\@ds)); |
58 |
|
} |
59 |
|
|
60 |
|
next if (! @ds); |
61 |
|
|
62 |
|
my $filename = $webpac->{'current_filename'}; |
63 |
|
|
64 |
|
if ($filename) { |
65 |
|
$webpac->output_file( |
66 |
|
file => $filename, |
67 |
|
template => 'html.tt', |
68 |
|
data => \@ds, |
69 |
|
headline => $webpac->{'headline'}, |
70 |
|
); |
71 |
} else { |
} else { |
72 |
return; |
print $webpac->output( |
73 |
|
template => 'text.tt', |
74 |
|
data => \@ds, |
75 |
|
headline => $webpac->{'headline'}, |
76 |
|
); |
77 |
} |
} |
|
} |
|
78 |
|
|
79 |
########################################################################## |
my $headline = $webpac->{'headline'}; |
80 |
|
|
81 |
my $last_tell=0; |
my $f = $filename; |
82 |
|
$f =~ s!out/!!; |
83 |
|
|
84 |
my @isis_dirs = ( '.' ); # use dirname as database name |
# save into index |
85 |
|
foreach my $ds (@ds) { |
86 |
|
next if (! $ds->{'swish'}); |
87 |
|
|
88 |
if ($opts{m}) { |
$index->insert( |
89 |
@isis_dirs = split(/,/,$opts{m}); |
index_name => $ds->{'tag'}, |
90 |
} |
path => $f, |
91 |
|
headline => $headline, |
92 |
my @isis_dbs; |
words => join(" ",@{$ds->{'swish'}}) |
93 |
|
); |
|
foreach (@isis_dirs) { |
|
|
if (-e $config->{isis_data}."/$db_dir/$_/LIBRI") { |
|
|
push @isis_dbs,$config->{isis_data}."/$db_dir/$_/LIBRI/LIBRI"; |
|
|
} |
|
|
if (-e $config->{isis_data}."/$db_dir/$_/PERI") { |
|
|
push @isis_dbs,$config->{isis_data}."/$db_dir/$_/PERI/PERI"; |
|
94 |
} |
} |
|
if (-e $config->{isis_data}."/$db_dir/$_/AMS") { |
|
|
push @isis_dbs,$config->{isis_data}."/$db_dir/$_/AMS/AMS"; |
|
|
} |
|
|
if (-e $config->{isis_data}."/$db_dir/$_/ARTI") { |
|
|
# push @isis_dbs,$config->{isis_data}."/$db_dir/$_/ARTI/ARTI"; |
|
|
} |
|
|
} |
|
|
|
|
|
print STDERR "FATAL: Can't find isis database.\nPerhaps isis_data (".$config->{isis_data}.") has wrong value?\n" if (! @isis_dbs); |
|
95 |
|
|
96 |
my $db; |
# save into sorted index (thesaurus) |
97 |
|
foreach my $ds (@ds) { |
98 |
|
next if (! $ds->{'index'}); |
99 |
|
|
100 |
foreach my $isis_db (@isis_dbs) { |
$thes->{$ds->{'tag'}} ||= new WebPAC::Index; |
101 |
|
|
102 |
|
$thes->{$ds->{'tag'}}->insert( |
103 |
my $db = OpenIsis::open( $isis_db ); |
path => $f, |
104 |
if (0) { |
headline => join(" ",@{$ds->{'index'}}), |
105 |
# # FIX |
); |
|
# if (! $db ) { |
|
|
print STDERR "WARNING: can't open '$isis_db'\n"; |
|
|
next ; |
|
106 |
} |
} |
107 |
|
|
108 |
my $max_rowid = OpenIsis::maxRowid( $db ); |
# print Dumper(\@ds); |
|
|
|
|
print STDERR "Reading database: $isis_db [$max_rowid rows]\n"; |
|
109 |
|
|
110 |
my $path = $isis_db; |
} |
|
$path =~ s#$config->{isis_data}/*##g; |
|
111 |
|
|
112 |
my $last_p = 0; |
foreach my $t (keys %{$thes}) { |
113 |
|
|
114 |
# { my $row_id = 4514; |
my @e = $thes->{$t}->elements; |
115 |
# FIX |
if (! @e) { |
116 |
for (my $row_id = 1; $row_id <= $max_rowid; $row_id++ ) { |
$log->logwarn("no elements in sorted index $t?"); |
117 |
my $row = OpenIsis::read( $db, $row_id ); |
next; |
118 |
if ($row && $row->{mfn}) { |
} |
119 |
#print STDERR "mfn: ",$row->{mfn},"\n"; |
|
120 |
# output current process indicator |
$log->info("saving sorted index $t [".scalar(@e)." elements]"); |
121 |
my $p = int($row->{mfn} * 100 / $max_rowid); |
|
122 |
if ($p != $last_p) { |
$webpac->output_file( |
123 |
printf STDERR ("%5d / %5d [%-51s] %-2d %% \r",$row->{mfn},$max_rowid,"=" x ($p/2).">", $p ) if (! $opts{q}); |
file => "./out/bfilter/$t.txt", |
124 |
$last_p = $p; |
template => 'index.tt', |
125 |
} |
data => \@e, |
126 |
|
index_name => $t, |
127 |
|
); |
128 |
|
} |
129 |
|
|
130 |
if (my $xml = isis2xml($row)) { |
if (0 && $log->is_debug) { |
131 |
#print STDERR "--ret-->$xml\n"; |
$log->debug("lookup hash: ",Dumper($webpac->{'lookup'})); |
132 |
print "Path-Name: $path#".int($row->{mfn})."\n"; |
$log->debug("data hash: ",Dumper($webpac->{'data'})); |
133 |
print "Content-Length: ".(length($xml)+1)."\n"; |
foreach my $t (keys %{$thes}) { |
134 |
print "Document-Type: XML\n\n$xml\n"; |
$log->debug("thesaurus $t hash: ",Dumper($thes->{$t})); |
|
} |
|
|
} |
|
135 |
} |
} |
|
print STDERR "\n"; |
|
136 |
} |
} |
137 |
|
|
|
# call this to commit index |
|
138 |
$index->close; |
$index->close; |
139 |
|
|
|
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 |
|