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; |
|
|
require Unicode::Map8; |
|
|
|
|
|
my $config=XMLin(undef, forcearray => [ 'isis' ], forcecontent => 1); |
|
4 |
|
|
5 |
my %opts; |
all2all.pl - basic script for all WebPAC needs |
6 |
|
|
7 |
getopts('d:m:q', \%opts); |
=cut |
8 |
|
|
9 |
my $db_dir = $opts{d} || "ps"; # FIX |
use strict; |
10 |
|
use Data::Dumper; |
11 |
|
use Carp; |
12 |
|
|
13 |
#die "usage: $0 -d [database_dir] -m [database1,database2] " if (! %opts); |
use lib './lib'; |
14 |
|
use WebPAC; |
15 |
|
use WebPAC::jsFind; |
16 |
|
use WebPAC::Index; |
17 |
|
|
18 |
#print Dumper($config->{indexer}); |
my $webpac = new WebPAC( |
19 |
#print "-" x 70,"\n"; |
code_page => 'ISO-8859-2', |
20 |
|
limit_mfn => 500, |
21 |
|
# debug => 1, |
22 |
|
) || die; |
23 |
|
|
24 |
# how to convert isis code page to UTF8? |
my $log = $webpac->_get_logger(); |
|
my $isis_map = Unicode::Map8->new($config->{isis_codepage}) || die; |
|
25 |
|
|
26 |
sub isis2xml { |
my $index = new WebPAC::jsFind( |
27 |
|
index_path => './out/index', |
28 |
|
keys => 10, |
29 |
|
) || die; |
30 |
|
|
31 |
my $row = shift @_; |
my $thes; |
32 |
|
|
33 |
my $xml; |
$|=1; |
34 |
|
|
35 |
sub isis_sf { |
my $maxmfn = $webpac->open_isis( |
36 |
my $row = shift @_; |
filename => shift @ARGV || '/data/hidra/THS/THS', |
37 |
my $isis_id = shift @_; |
lookup => [ |
38 |
my $subfield = shift @_; |
{ 'key' => 'd:v900', 'val' => 'v250^a' }, |
39 |
if ($row->{$isis_id}->[0]) { |
# { 'eval' => '"v901^a" eq "Područje"', 'key' => 'pa:v561^4:v562^4:v461^1', 'val' => 'v900' }, |
40 |
my $sf = OpenIsis::subfields($row->{$isis_id}->[0]); |
# { 'eval '=> '"v901^a" eq "Mikrotezaurus"', 'key' => 'a:v561^4:v562^4:v461^1', 'val' => 'v900' }, |
41 |
if (! defined $subfield || length($subfield) == 0) { |
# { 'eval' => '"v901^a" eq "Deskriptor"', 'key' => 'a:v561^4:v562^4:v461^1', 'val' => 'v900' }, |
42 |
# subfield list undef, empty or no defined subfields for this record |
{ 'key' => 'a:v561^4:v562^4:v461^1', 'val' => 'v900' }, |
43 |
my $all_sf = $row->{$isis_id}->[0]; |
{ 'key' => '900_mfn:v900', 'val' => 'v000' }, |
44 |
$all_sf =~ s/\^./ /g; nuke definirions |
], |
45 |
return $all_sf; |
); |
|
} elsif ($sf->{$subfield}) { |
|
|
return $sf->{$subfield}; |
|
|
} |
|
|
} |
|
|
} |
|
46 |
|
|
47 |
foreach my $field (keys %{$config->{indexer}}) { |
$log->info("rows: $maxmfn"); |
48 |
|
|
49 |
my $display_data = ""; |
$webpac->open_import_xml(type => 'isis_hidra_ths'); |
|
my $index_data = ""; |
|
50 |
|
|
51 |
foreach my $x (@{$config->{indexer}->{$field}->{isis}}) { |
while (my $rec = $webpac->fetch_rec) { |
52 |
|
|
53 |
my $display_tmp = ""; |
my @ds = $webpac->data_structure($rec); |
|
my $index_tmp = ""; |
|
|
|
|
|
my $format = $x->{content}; |
|
|
my $i = 1; # index only |
|
|
my $d = 1; # display only |
|
|
$i = 0 if (lc($x->{type}) eq "display"); |
|
|
$d = 0 if (lc($x->{type}) eq "index"); |
|
|
#print "## i: $i d: $d ## $format ##"; |
|
|
# parse format |
|
|
my $prefix = ""; |
|
|
if ($format =~ s/^([^\d]+)//) { |
|
|
$prefix = $1; |
|
|
} |
|
|
while ($format) { |
|
|
if ($format =~ s/^(\d\d\d)(\w?)//) { |
|
|
my $isis_tmp = isis_sf($row,$1,$2); |
|
|
if ($isis_tmp) { |
|
|
# $display_tmp .= $prefix . "/$1/$2/".$isis_tmp if ($d); |
|
|
$display_tmp .= $prefix . $isis_tmp if ($d); |
|
|
$index_tmp .= $isis_tmp." " if ($i); |
|
|
#print " $isis_tmp <--\n"; |
|
|
} |
|
|
$prefix = ""; |
|
|
} elsif ($format =~ s/^([^\d]+)//) { |
|
|
$prefix = $1; |
|
|
} else { |
|
|
print STDERR "WARNING: unparsed format '$format'\n"; |
|
|
last; |
|
|
}; |
|
|
} |
|
|
# add suffix |
|
|
$display_tmp .= $prefix if ($display_tmp); |
|
|
|
|
|
# $display_data .= $display_tmp if ($display_tmp ne ""); |
|
|
# $index_data .= $index_tmp if ($index_tmp ne ""); |
|
|
$display_data .= $display_tmp; |
|
|
$index_data .= $index_tmp; |
|
54 |
|
|
55 |
} |
if (0 && $log->is_debug) { |
56 |
#print "--display:$display_data\n--index:$index_data\n"; |
$log->debug("rec = ",Dumper($rec)); |
57 |
$xml->{$field}->{display} .= $isis_map->tou($display_data)->utf8 if ($display_data); |
$log->debug("ds = ",Dumper(\@ds)); |
|
$xml->{$field}->{index} .= unac_string($config->{isis_codepage},$index_data) if ($index_data); |
|
|
|
|
58 |
} |
} |
|
if ($xml) { |
|
|
return XMLout($xml, rootname => 'xml', noattr => 1 ); |
|
|
} else { |
|
|
return; |
|
|
} |
|
|
} |
|
|
|
|
|
########################################################################## |
|
|
|
|
|
my $last_tell=0; |
|
59 |
|
|
60 |
my @isis_dirs = ( '.' ); # use dirname as database name |
next if (! @ds); |
61 |
|
|
62 |
if ($opts{m}) { |
my $filename = $webpac->{'current_filename'}; |
|
@isis_dirs = split(/,/,$opts{m}); |
|
|
} |
|
|
|
|
|
my @isis_dbs; |
|
63 |
|
|
64 |
foreach (@isis_dirs) { |
if ($filename) { |
65 |
if (-e $config->{isis_data}."/$db_dir/$_/LIBRI") { |
$webpac->output_file( |
66 |
push @isis_dbs,$config->{isis_data}."/$db_dir/$_/LIBRI/LIBRI"; |
file => $filename, |
67 |
} |
template => 'html.tt', |
68 |
if (-e $config->{isis_data}."/$db_dir/$_/PERI") { |
data => \@ds, |
69 |
push @isis_dbs,$config->{isis_data}."/$db_dir/$_/PERI/PERI"; |
headline => $webpac->{'headline'}, |
70 |
} |
); |
71 |
if (-e $config->{isis_data}."/$db_dir/$_/AMS") { |
} else { |
72 |
push @isis_dbs,$config->{isis_data}."/$db_dir/$_/AMS/AMS"; |
print $webpac->output( |
73 |
} |
template => 'text.tt', |
74 |
if (-e $config->{isis_data}."/$db_dir/$_/ARTI") { |
data => \@ds, |
75 |
# push @isis_dbs,$config->{isis_data}."/$db_dir/$_/ARTI/ARTI"; |
headline => $webpac->{'headline'}, |
76 |
|
); |
77 |
} |
} |
|
} |
|
|
|
|
|
print STDERR "FATAL: Can't find isis database.\nPerhaps isis_data (".$config->{isis_data}.") has wrong value?\n" if (! @isis_dbs); |
|
78 |
|
|
79 |
my $db; |
my $headline = $webpac->{'headline'}; |
80 |
|
|
81 |
foreach my $isis_db (@isis_dbs) { |
my $f = $filename; |
82 |
|
$f =~ s!out/!!; |
83 |
|
|
84 |
|
# save into index |
85 |
|
foreach my $ds (@ds) { |
86 |
|
next if (! $ds->{'swish'}); |
87 |
|
|
88 |
my $db = OpenIsis::open( $isis_db ); |
$index->insert( |
89 |
if (0) { |
index_name => $ds->{'tag'}, |
90 |
# # FIX |
path => $f, |
91 |
# if (! $db ) { |
headline => $headline, |
92 |
print STDERR "WARNING: can't open '$isis_db'\n"; |
words => join(" ",@{$ds->{'swish'}}) |
93 |
next ; |
); |
94 |
} |
} |
95 |
|
|
96 |
my $max_rowid = OpenIsis::maxRowid( $db ); |
# save into sorted index (thesaurus) |
97 |
|
foreach my $ds (@ds) { |
98 |
print STDERR "Reading database: $isis_db [$max_rowid rows]\n"; |
next if (! $ds->{'index'}); |
|
|
|
|
my $last_p = 0; |
|
99 |
|
|
100 |
# { my $row_id = 1; |
$thes->{$ds->{'tag'}} ||= new WebPAC::Index; |
|
# FIX |
|
|
for (my $row_id = 1; $row_id <= $max_rowid; $row_id++ ) { |
|
|
my $row = OpenIsis::read( $db, $row_id ); |
|
|
if ($row && $row->{mfn}) { |
|
101 |
|
|
102 |
# output current process indicator |
foreach my $h (@{$ds->{'index'}}) { |
103 |
my $p = int($row->{mfn} * 100 / $max_rowid); |
$thes->{$ds->{'tag'}}->insert( |
104 |
if ($p != $last_p) { |
path => $f, |
105 |
printf STDERR ("%5d / %5d [%-51s] %-2d %% \r",$row->{mfn},$max_rowid,"=" x ($p/2).">", $p ) if (! $opts{q}); |
headline => $h, |
106 |
$last_p = $p; |
); |
|
} |
|
|
|
|
|
if (my $xml = isis2xml($row)) { |
|
|
print "Path-Name: $isis_db#".$row->{mfn}."\n"; |
|
|
print "Content-Length: ".(length($xml)+1)."\n"; |
|
|
print "Document-Type: XML\n\n$xml\n"; |
|
|
} |
|
107 |
} |
} |
108 |
} |
} |
|
print STDERR "\n"; |
|
|
} |
|
|
|
|
109 |
|
|
110 |
1; |
# print Dumper(\@ds); |
|
__END__ |
|
|
########################################################################## |
|
|
|
|
|
=head1 NAME |
|
111 |
|
|
112 |
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> |
|
113 |
|
|
114 |
=head1 COPYRIGHT |
foreach my $t (keys %{$thes}) { |
115 |
|
|
116 |
GNU Public License (GPL) v2 or later |
my @e = $thes->{$t}->elements; |
117 |
|
if (! @e) { |
118 |
|
$log->logwarn("no elements in sorted index $t?"); |
119 |
|
next; |
120 |
|
} |
121 |
|
|
122 |
|
my $file = "./out/bfilter/$t.txt"; |
123 |
|
$log->info("saving sorted index $t to '$file' [".scalar(@e)." elements]"); |
124 |
|
|
125 |
|
$webpac->output_file( |
126 |
|
file => $file, |
127 |
|
template => 'index.tt', |
128 |
|
data => \@e, |
129 |
|
index_name => $t, |
130 |
|
); |
131 |
|
} |
132 |
|
|
133 |
=head1 SEE ALSO |
if (0 && $log->is_debug) { |
134 |
|
$log->debug("lookup hash: ",Dumper($webpac->{'lookup'})); |
135 |
|
$log->debug("data hash: ",Dumper($webpac->{'data'})); |
136 |
|
foreach my $t (keys %{$thes}) { |
137 |
|
$log->debug("thesaurus $t hash: ",Dumper($thes->{$t})); |
138 |
|
} |
139 |
|
} |
140 |
|
|
141 |
SWISH-E web site at http://www.swish-e.org |
$index->close; |
142 |
|
|
|
=cut |
|