1 |
#!/usr/bin/perl -w |
2 |
|
3 |
use strict; |
4 |
use OpenIsis; |
5 |
use Getopt::Std; |
6 |
use Data::Dumper; |
7 |
use XML::Simple; |
8 |
use Text::Unaccent 1.02; # 1.01 won't compile on my platform, |
9 |
require Unicode::Map8; |
10 |
use DBI; |
11 |
|
12 |
my $config=XMLin(undef, forcearray => [ 'isis' ], forcecontent => 1); |
13 |
my $dbh = DBI->connect("DBI:Pg:dbname=webpac","","") || die $DBI::errstr; # FIX |
14 |
# FIX; select relname from pg_class where relname like 'index_%' ; |
15 |
$dbh->begin_work || die $dbh->errstr(); |
16 |
|
17 |
$dbh->do("delete from index_author") || die $dbh->errstr(); |
18 |
$dbh->do("delete from index_title") || die $dbh->errstr(); |
19 |
|
20 |
my %opts; |
21 |
|
22 |
# usage: |
23 |
# -d directory name |
24 |
# -m multiple directories |
25 |
# -q quiet |
26 |
# -s run swish |
27 |
|
28 |
getopts('d:m:qs', \%opts); |
29 |
|
30 |
my $db_dir = $opts{d} || "ps"; # FIX |
31 |
|
32 |
#die "usage: $0 -d [database_dir] -m [database1,database2] " if (! %opts); |
33 |
|
34 |
#print Dumper($config->{indexer}); |
35 |
#print "-" x 70,"\n"; |
36 |
|
37 |
# how to convert isis code page to UTF8? |
38 |
my $isis_map = Unicode::Map8->new($config->{isis_codepage}) || die; |
39 |
|
40 |
sub isis2xml { |
41 |
|
42 |
my $row = shift @_; |
43 |
|
44 |
my $xml; |
45 |
$xml->{db_dir} = [ $db_dir ]; # FIX remove? |
46 |
|
47 |
sub isis_sf { |
48 |
my $row = shift @_; |
49 |
my $isis_id = shift @_; |
50 |
my $subfield = shift @_; |
51 |
if ($row->{$isis_id}->[0]) { |
52 |
my $sf = OpenIsis::subfields($row->{$isis_id}->[0]); |
53 |
if (! defined $subfield || length($subfield) == 0) { |
54 |
# subfield list undef, empty or no defined subfields for this record |
55 |
my $all_sf = $row->{$isis_id}->[0]; |
56 |
$all_sf =~ s/\^./ /g; nuke definirions |
57 |
return $all_sf; |
58 |
} elsif ($sf->{$subfield}) { |
59 |
return $sf->{$subfield}; |
60 |
} |
61 |
} |
62 |
} |
63 |
|
64 |
foreach my $field (keys %{$config->{indexer}}) { |
65 |
|
66 |
my $display_data = ""; |
67 |
my $swish_data = ""; |
68 |
my $index_data = ""; |
69 |
|
70 |
foreach my $x (@{$config->{indexer}->{$field}->{isis}}) { |
71 |
|
72 |
my $display_tmp = ""; |
73 |
my $swish_tmp = ""; |
74 |
my $index_tmp = ""; |
75 |
|
76 |
my $format = $x->{content}; |
77 |
my $s = 1; # swish only |
78 |
my $d = 1; # display only |
79 |
my $i = 0; # index only |
80 |
$s = 0 if (lc($x->{type}) eq "display"); |
81 |
$d = 0 if (lc($x->{type}) eq "swish"); |
82 |
($s,$d,$i) = (0,0,1) if (lc($x->{type}) eq "index"); |
83 |
#print STDERR "## s: $s d: $d i: $i ## $format ##\n"; |
84 |
# parse format |
85 |
my $prefix = ""; |
86 |
if ($format =~ s/^([^\d]+)//) { |
87 |
$prefix = $1; |
88 |
} |
89 |
while ($format) { |
90 |
if ($format =~ s/^(\d\d\d)(\w?)//) { |
91 |
my $isis_tmp = isis_sf($row,$1,$2); |
92 |
if ($isis_tmp) { |
93 |
# $display_tmp .= $prefix . "/$1/$2/".$isis_tmp if ($d); |
94 |
$display_tmp .= $prefix . $isis_tmp if ($d); |
95 |
$swish_tmp .= $isis_tmp." " if ($s); |
96 |
$index_tmp .= $prefix . $isis_tmp if ($i); |
97 |
#print STDERR " $isis_tmp <--\n"; |
98 |
} |
99 |
$prefix = ""; |
100 |
} elsif ($format =~ s/^([^\d]+)//) { |
101 |
$prefix = $1; |
102 |
} else { |
103 |
print STDERR "WARNING: unparsed format '$format'\n"; |
104 |
last; |
105 |
}; |
106 |
} |
107 |
# add suffix |
108 |
$display_tmp .= $prefix if ($display_tmp); |
109 |
$index_tmp .= $prefix if ($index_tmp); |
110 |
|
111 |
# $display_data .= $display_tmp if ($display_tmp ne ""); |
112 |
# $swish_data .= $swish_tmp if ($swish_tmp ne ""); |
113 |
$display_data .= $display_tmp; |
114 |
$swish_data .= $swish_tmp; |
115 |
$index_data .= $index_tmp; |
116 |
|
117 |
} |
118 |
#print "--display:$display_data\n--swish:$swish_data\n"; |
119 |
#$xml->{$field."_display"} = $isis_map->tou($display_data)->utf8 if ($display_data); |
120 |
#$xml->{$field."_swish"} = unac_string($config->{isis_codepage},$swish_data) if ($swish_data); |
121 |
$xml->{$field."_display" } = [ $isis_map->tou($display_data)->utf8 ] if ($display_data); |
122 |
$xml->{$field."_swish"} = [ unac_string($config->{isis_codepage},$swish_data) ] if ($swish_data); |
123 |
|
124 |
# index |
125 |
if ($index_data && $index_data ne "") { |
126 |
my $sql = "select $field from index_$field where upper($field)=upper(?)"; |
127 |
my $sth = $dbh->prepare($sql) || die $dbh->errstr(); |
128 |
$sth->execute($index_data) || die "SQL: $sql; ".$dbh->errstr(); |
129 |
#print STDERR "--->$index_data<---\n"; |
130 |
if (! $sth->fetchrow_hashref) { |
131 |
my $sql = "insert into index_$field values (?)"; |
132 |
my $sth = $dbh->prepare($sql) || die $dbh->errstr(); |
133 |
#print STDERR "$sql: $index_data<!----\n"; |
134 |
$sth->execute($index_data) || die "SQL: $sql; ".$dbh->errstr(); |
135 |
} |
136 |
} |
137 |
|
138 |
} |
139 |
if ($xml) { |
140 |
return XMLout($xml, rootname => 'xml', keeproot => 0, noattr => 0 ); |
141 |
} else { |
142 |
return; |
143 |
} |
144 |
} |
145 |
|
146 |
########################################################################## |
147 |
|
148 |
my $last_tell=0; |
149 |
|
150 |
my @isis_dirs = ( '.' ); # use dirname as database name |
151 |
|
152 |
if ($opts{m}) { |
153 |
@isis_dirs = split(/,/,$opts{m}); |
154 |
} |
155 |
|
156 |
my @isis_dbs; |
157 |
|
158 |
foreach (@isis_dirs) { |
159 |
if (-e $config->{isis_data}."/$db_dir/$_/LIBRI") { |
160 |
push @isis_dbs,$config->{isis_data}."/$db_dir/$_/LIBRI/LIBRI"; |
161 |
} |
162 |
if (-e $config->{isis_data}."/$db_dir/$_/PERI") { |
163 |
push @isis_dbs,$config->{isis_data}."/$db_dir/$_/PERI/PERI"; |
164 |
} |
165 |
if (-e $config->{isis_data}."/$db_dir/$_/AMS") { |
166 |
push @isis_dbs,$config->{isis_data}."/$db_dir/$_/AMS/AMS"; |
167 |
} |
168 |
if (-e $config->{isis_data}."/$db_dir/$_/ARTI") { |
169 |
# push @isis_dbs,$config->{isis_data}."/$db_dir/$_/ARTI/ARTI"; |
170 |
} |
171 |
} |
172 |
|
173 |
print STDERR "FATAL: Can't find isis database.\nPerhaps isis_data (".$config->{isis_data}.") has wrong value?\n" if (! @isis_dbs); |
174 |
|
175 |
my $db; |
176 |
|
177 |
foreach my $isis_db (@isis_dbs) { |
178 |
|
179 |
|
180 |
my $db = OpenIsis::open( $isis_db ); |
181 |
if (0) { |
182 |
# # FIX |
183 |
# if (! $db ) { |
184 |
print STDERR "WARNING: can't open '$isis_db'\n"; |
185 |
next ; |
186 |
} |
187 |
|
188 |
my $max_rowid = OpenIsis::maxRowid( $db ); |
189 |
|
190 |
print STDERR "Reading database: $isis_db [$max_rowid rows]\n"; |
191 |
|
192 |
my $last_p = 0; |
193 |
|
194 |
# { my $row_id = 1; |
195 |
# FIX |
196 |
for (my $row_id = 1; $row_id <= $max_rowid; $row_id++ ) { |
197 |
my $row = OpenIsis::read( $db, $row_id ); |
198 |
if ($row && $row->{mfn}) { |
199 |
|
200 |
# output current process indicator |
201 |
my $p = int($row->{mfn} * 100 / $max_rowid); |
202 |
if ($p != $last_p) { |
203 |
printf STDERR ("%5d / %5d [%-51s] %-2d %% \r",$row->{mfn},$max_rowid,"=" x ($p/2).">", $p ) if (! $opts{q}); |
204 |
$last_p = $p; |
205 |
} |
206 |
|
207 |
if (my $xml = isis2xml($row)) { |
208 |
my $path = $isis_db; |
209 |
$path =~ s#$config->{isis_data}/*##g; |
210 |
my $out = "Path-Name: $path#".$row->{mfn}."\n"; |
211 |
$out .= "Content-Length: ".(length($xml)+1)."\n"; |
212 |
$out .= "Document-Type: XML\n\n$xml\n"; |
213 |
print $out; |
214 |
} |
215 |
} |
216 |
} |
217 |
print STDERR "\n"; |
218 |
} |
219 |
|
220 |
$dbh->commit || die $dbh->errstr(); |
221 |
|
222 |
1; |
223 |
__END__ |
224 |
########################################################################## |
225 |
|
226 |
=head1 NAME |
227 |
|
228 |
isis2xml.pl - read isis file and dump XML |
229 |
|
230 |
=head1 DESCRIPTION |
231 |
|
232 |
This command will read ISIS data file using OpenIsis perl module and |
233 |
create XML file for usage with I<SWISH-E> |
234 |
indexer. Dispite it's name, this script B<isn't general xml generator> |
235 |
from isis files (isis allready has something like that). Output of this |
236 |
script is tailor-made for SWISH-E. |
237 |
|
238 |
=head1 AUTHOR |
239 |
|
240 |
Dobrica Pavlinusic <dpavlin@rot13.org> |
241 |
|
242 |
=head1 COPYRIGHT |
243 |
|
244 |
GNU Public License (GPL) v2 or later |
245 |
|
246 |
=head1 SEE ALSO |
247 |
|
248 |
SWISH-E web site at http://www.swish-e.org |
249 |
|
250 |
=cut |