1 |
dpavlin |
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 |
dpavlin |
5 |
use Text::Unaccent 1.02; # 1.01 won't compile on my platform, |
9 |
dpavlin |
10 |
use Text::Iconv; |
10 |
dpavlin |
13 |
use Config::IniFiles; |
11 |
dpavlin |
40 |
use Encode; |
12 |
dpavlin |
177 |
#use GDBM_File; |
13 |
|
|
use Fcntl; # for O_RDWR |
14 |
|
|
use TDB_File; |
15 |
dpavlin |
1 |
|
16 |
dpavlin |
10 |
$|=1; |
17 |
dpavlin |
9 |
|
18 |
dpavlin |
13 |
my $config_file = $0; |
19 |
|
|
$config_file =~ s/\.pl$/.conf/; |
20 |
|
|
die "FATAL: can't find configuration file '$config_file'" if (! -e $config_file); |
21 |
|
|
|
22 |
dpavlin |
10 |
my $config; |
23 |
|
|
|
24 |
dpavlin |
58 |
#use index_DBI; # default DBI module for index |
25 |
|
|
use index_DBI_cache; # faster DBI module using memory cache |
26 |
dpavlin |
50 |
my $index; |
27 |
dpavlin |
10 |
|
28 |
dpavlin |
1 |
my %opts; |
29 |
|
|
|
30 |
dpavlin |
7 |
# usage: |
31 |
|
|
# -d directory name |
32 |
|
|
# -m multiple directories |
33 |
|
|
# -q quiet |
34 |
|
|
# -s run swish |
35 |
dpavlin |
1 |
|
36 |
dpavlin |
7 |
getopts('d:m:qs', \%opts); |
37 |
|
|
|
38 |
dpavlin |
43 |
my $path; # this is name of database |
39 |
dpavlin |
1 |
|
40 |
dpavlin |
57 |
Text::Iconv->raise_error(0); # Conversion errors don't raise exceptions |
41 |
dpavlin |
1 |
|
42 |
dpavlin |
40 |
# this is encoding of all files on disk, including import_xml/*.xml file and |
43 |
|
|
# filter/*.pm files! It will be used to store strings in perl internally! |
44 |
|
|
my $codepage = 'ISO-8859-2'; |
45 |
dpavlin |
1 |
|
46 |
dpavlin |
40 |
my $utf2cp = Text::Iconv->new('UTF-8',$codepage); |
47 |
|
|
# this function will convert data from XML files to local encoding |
48 |
|
|
sub x { |
49 |
|
|
return $utf2cp->convert($_[0]); |
50 |
|
|
} |
51 |
dpavlin |
3 |
|
52 |
dpavlin |
54 |
# decode isis/excel or other import codepage |
53 |
|
|
my $import2cp; |
54 |
dpavlin |
10 |
|
55 |
dpavlin |
40 |
# outgoing xml must be in UTF-8 |
56 |
|
|
my $cp2utf = Text::Iconv->new($codepage,'UTF-8'); |
57 |
dpavlin |
29 |
|
58 |
dpavlin |
54 |
# mapping between data type and tag which specify |
59 |
|
|
# format in XML file |
60 |
|
|
my %type2tag = ( |
61 |
|
|
'isis' => 'isis', |
62 |
dpavlin |
62 |
'excel' => 'column', |
63 |
|
|
'marc' => 'marc', |
64 |
dpavlin |
67 |
'feed' => 'feed' |
65 |
dpavlin |
54 |
); |
66 |
dpavlin |
3 |
|
67 |
dpavlin |
170 |
my $cache; # for cacheing |
68 |
|
|
|
69 |
dpavlin |
177 |
# lookup hash (tied to file) |
70 |
|
|
my %lhash; |
71 |
|
|
# this option will cache all lookup entries in memory. |
72 |
|
|
# if you are tight on memory, turn this off |
73 |
|
|
my $use_lhash_cache = 1; |
74 |
|
|
|
75 |
dpavlin |
54 |
sub data2xml { |
76 |
|
|
|
77 |
dpavlin |
10 |
use xmlify; |
78 |
|
|
|
79 |
dpavlin |
54 |
my $type = shift @_; |
80 |
dpavlin |
3 |
my $row = shift @_; |
81 |
dpavlin |
13 |
my $add_xml = shift @_; |
82 |
dpavlin |
59 |
# needed to read values from configuration file |
83 |
|
|
my $cfg = shift @_; |
84 |
|
|
my $database = shift @_; |
85 |
dpavlin |
3 |
|
86 |
|
|
my $xml; |
87 |
|
|
|
88 |
dpavlin |
10 |
use parse_format; |
89 |
dpavlin |
3 |
|
90 |
dpavlin |
13 |
my $html = ""; # html formatted display output |
91 |
dpavlin |
10 |
|
92 |
dpavlin |
13 |
my %field_usage; # counter for usage of each field |
93 |
|
|
|
94 |
dpavlin |
32 |
# sort subrouting using order="" attribute |
95 |
|
|
sub by_order { |
96 |
dpavlin |
98 |
my $va = $config->{indexer}->{$a}->{order} || |
97 |
|
|
$config->{indexer}->{$a}; |
98 |
|
|
my $vb = $config->{indexer}->{$b}->{order} || |
99 |
|
|
$config->{indexer}->{$b}; |
100 |
dpavlin |
29 |
|
101 |
dpavlin |
98 |
return $va <=> $vb; |
102 |
dpavlin |
32 |
} |
103 |
dpavlin |
3 |
|
104 |
dpavlin |
170 |
my @sorted_tags; |
105 |
dpavlin |
180 |
if ($cache->{tags_by_order}) { |
106 |
|
|
@sorted_tags = @{$cache->{tags_by_order}}; |
107 |
dpavlin |
170 |
} else { |
108 |
|
|
@sorted_tags = sort by_order keys %{$config->{indexer}}; |
109 |
dpavlin |
180 |
$cache->{tags_by_order} = \@sorted_tags; |
110 |
dpavlin |
170 |
} |
111 |
dpavlin |
32 |
|
112 |
dpavlin |
177 |
# lookup key |
113 |
|
|
my $lookup_key; |
114 |
|
|
|
115 |
dpavlin |
178 |
# cache for field in pages |
116 |
dpavlin |
180 |
delete $cache->{display_data}; |
117 |
|
|
delete $cache->{swish_data}; |
118 |
|
|
delete $cache->{swish_exact_data}; |
119 |
dpavlin |
178 |
my @page_fields; # names of fields |
120 |
|
|
|
121 |
|
|
|
122 |
|
|
# subs used to produce output |
123 |
|
|
|
124 |
|
|
sub get_field_name($$$) { |
125 |
|
|
my ($config,$field,$field_usage) = @_; |
126 |
|
|
|
127 |
|
|
# find field name (signular, plural) |
128 |
|
|
my $field_name = ""; |
129 |
|
|
if ($config->{indexer}->{$field}->{name_singular} && $field_usage == 1) { |
130 |
|
|
$field_name = $config->{indexer}->{$field}->{name_singular}; |
131 |
|
|
} elsif ($config->{indexer}->{$field}->{name_plural}) { |
132 |
|
|
$field_name = $config->{indexer}->{$field}->{name_plural}; |
133 |
|
|
} elsif ($config->{indexer}->{$field}->{name}) { |
134 |
|
|
$field_name = $config->{indexer}->{$field}->{name}; |
135 |
|
|
} else { |
136 |
|
|
print STDERR "WARNING: field '$field' doesn't have 'name' attribute!"; |
137 |
|
|
} |
138 |
|
|
if ($field_name) { |
139 |
|
|
return x($field_name); |
140 |
|
|
} |
141 |
|
|
} |
142 |
|
|
|
143 |
|
|
|
144 |
|
|
# begin real work: go field by field |
145 |
dpavlin |
170 |
foreach my $field (@sorted_tags) { |
146 |
|
|
|
147 |
dpavlin |
40 |
$field=x($field); |
148 |
dpavlin |
13 |
$field_usage{$field}++; |
149 |
|
|
|
150 |
dpavlin |
10 |
my $swish_data = ""; |
151 |
dpavlin |
163 |
my $swish_exact_data = ""; |
152 |
dpavlin |
3 |
my $display_data = ""; |
153 |
dpavlin |
35 |
my $line_delimiter; |
154 |
dpavlin |
3 |
|
155 |
dpavlin |
34 |
my ($swish,$display); |
156 |
|
|
|
157 |
dpavlin |
54 |
my $tag = $type2tag{$type} || die "can't find which tag to use for type $type"; |
158 |
dpavlin |
178 |
|
159 |
|
|
# is this field page-by-page? |
160 |
|
|
my $iterate_by_page = $config->{indexer}->{$field}->{iterate_by_page}; |
161 |
|
|
push @page_fields,$field if ($iterate_by_page); |
162 |
|
|
my %page_max = (); |
163 |
|
|
# default line_delimiter if using |
164 |
|
|
my $page_line_delimiter = $config->{indexer}->{$field}->{page_line_delimiter} || '<br/>'; |
165 |
|
|
|
166 |
dpavlin |
54 |
foreach my $x (@{$config->{indexer}->{$field}->{$tag}}) { |
167 |
dpavlin |
3 |
|
168 |
dpavlin |
40 |
my $format = x($x->{content}); |
169 |
|
|
my $delimiter = x($x->{delimiter}) || ' '; |
170 |
dpavlin |
3 |
|
171 |
dpavlin |
178 |
my $repeat_off = 0; # init repeatable offset |
172 |
dpavlin |
29 |
|
173 |
dpavlin |
177 |
# swish, swish_exact, display, index, index_lookup |
174 |
|
|
# swish and display defaults |
175 |
|
|
my ($s,$se,$d,$i,$il) = (1,0,1,0,0); |
176 |
dpavlin |
29 |
$s = 0 if (lc($x->{type}) eq "display"); |
177 |
|
|
$d = 0 if (lc($x->{type}) eq "swish"); |
178 |
dpavlin |
163 |
$se = 1 if (lc($x->{type}) eq "swish_exact"); |
179 |
dpavlin |
29 |
($s,$d,$i) = (0,0,1) if (lc($x->{type}) eq "index"); |
180 |
dpavlin |
177 |
$il = 1 if (lc($x->{type}) =~ /^lookup/); |
181 |
dpavlin |
29 |
|
182 |
dpavlin |
177 |
|
183 |
dpavlin |
35 |
# what will separate last line from this one? |
184 |
|
|
if ($display_data && $x->{append} && $x->{append} eq "1") { |
185 |
dpavlin |
34 |
$line_delimiter = ' '; |
186 |
dpavlin |
35 |
} elsif ($display_data) { |
187 |
dpavlin |
34 |
$line_delimiter = '<br/>'; |
188 |
|
|
} |
189 |
dpavlin |
29 |
|
190 |
dpavlin |
34 |
# init vars so that we go into while... |
191 |
|
|
($swish,$display) = (1,1); |
192 |
dpavlin |
29 |
|
193 |
dpavlin |
97 |
# placeholder for all repeatable entries for index |
194 |
|
|
my @index_data; |
195 |
|
|
|
196 |
dpavlin |
153 |
sub mkformat { |
197 |
|
|
my $x = shift || die "mkformat needs tag reference"; |
198 |
|
|
my $data = shift || return; |
199 |
|
|
my $format_name = x($x->{format_name}) || return $data; |
200 |
|
|
my $fmt = x($config->{format}->{$format_name}->{content}) || die "<format name=\"$format_name\"> is not defined!"; |
201 |
|
|
my $format_delimiter = x($x->{format_delimiter}); |
202 |
|
|
my @data; |
203 |
|
|
if ($format_delimiter) { |
204 |
|
|
@data = split(/$format_delimiter/,$data); |
205 |
|
|
} else { |
206 |
|
|
push @data,$data; |
207 |
|
|
} |
208 |
|
|
|
209 |
|
|
if ($fmt) { |
210 |
|
|
my $nr = scalar $fmt =~ s/%s/%s/g; |
211 |
|
|
if (($#data+1) == $nr) { |
212 |
|
|
return sprintf($fmt,@data); |
213 |
|
|
} else { |
214 |
|
|
print STDERR "mkformat: [$data] can't be split on [$format_delimiter] to $nr fields!\n"; |
215 |
|
|
return $data; |
216 |
|
|
} |
217 |
|
|
} else { |
218 |
|
|
print STDERR "usage of link '$format_name' without defined format (<link> tag)\n"; |
219 |
|
|
} |
220 |
|
|
} |
221 |
|
|
|
222 |
dpavlin |
90 |
# while because of repeatable fields |
223 |
|
|
while ($swish || $display) { |
224 |
dpavlin |
178 |
my $page = $repeat_off; |
225 |
|
|
$page_max{$field} = $page if ($iterate_by_page && $page > ($page_max{$field} || 0)); |
226 |
dpavlin |
54 |
($swish,$display) = parse_format($type, $format,$row,$repeat_off++,$import2cp); |
227 |
dpavlin |
90 |
if ($repeat_off > 1000) { |
228 |
|
|
print STDERR "loop (more than 1000 repeatable fields) deteced in $row, $format\n"; |
229 |
|
|
last; |
230 |
|
|
} |
231 |
dpavlin |
177 |
|
232 |
|
|
# is this field is lookup? |
233 |
|
|
if ($display && $x->{lookup}) { |
234 |
dpavlin |
178 |
my $null = "<!-- null -->"; |
235 |
dpavlin |
177 |
if ($use_lhash_cache) { |
236 |
|
|
if (!defined($cache->{lhash}->{$display})) { |
237 |
|
|
my $new_display = $lhash{$display}; |
238 |
dpavlin |
178 |
if (defined($new_display)) { |
239 |
dpavlin |
177 |
#print STDERR "lookup cache store '$display' = '$new_display'\n"; |
240 |
|
|
$display = $new_display; |
241 |
|
|
$cache->{lhash}->{$display} = $new_display; |
242 |
|
|
} else { |
243 |
|
|
print STDERR "WARNING: lookup for '$display' didn't find anything.\n"; |
244 |
|
|
$display = ""; |
245 |
dpavlin |
178 |
$cache->{lhash}->{$display} = $null; |
246 |
dpavlin |
177 |
} |
247 |
|
|
} else { |
248 |
|
|
$display = $cache->{lhash}->{$display}; |
249 |
|
|
} |
250 |
|
|
} else { |
251 |
dpavlin |
178 |
$display = $lhash{$display} || $null; |
252 |
dpavlin |
177 |
} |
253 |
|
|
} |
254 |
|
|
|
255 |
dpavlin |
29 |
# filter="name" ; filter this field through |
256 |
|
|
# filter/[name].pm |
257 |
|
|
my $filter = $x->{filter}; |
258 |
dpavlin |
170 |
if ($filter && !$cache->{filter_loaded}->{$filter}) { |
259 |
dpavlin |
29 |
require "filter/".$filter.".pm"; |
260 |
dpavlin |
170 |
$cache->{filter_loaded}->{$filter}++; |
261 |
dpavlin |
20 |
} |
262 |
dpavlin |
29 |
# type="swish" ; field for swish |
263 |
dpavlin |
163 |
if ($swish) { |
264 |
|
|
if ($filter && ($s || $se)) { |
265 |
dpavlin |
29 |
no strict 'refs'; |
266 |
dpavlin |
163 |
my $tmp = join(" ",&$filter($swish)) if ($s || $se); |
267 |
|
|
$swish_data .= $tmp if ($s); |
268 |
|
|
$swish_exact_data .= $tmp if ($se); |
269 |
|
|
|
270 |
dpavlin |
29 |
} else { |
271 |
dpavlin |
163 |
$swish_data .= $swish if ($s); |
272 |
|
|
$swish_exact_data .= $swish if ($se); |
273 |
dpavlin |
29 |
} |
274 |
|
|
} |
275 |
dpavlin |
17 |
|
276 |
dpavlin |
29 |
# type="display" ; field for display |
277 |
|
|
if ($d && $display) { |
278 |
dpavlin |
35 |
if ($line_delimiter && $display_data) { |
279 |
|
|
$display_data .= $line_delimiter; |
280 |
|
|
} |
281 |
dpavlin |
29 |
if ($filter) { |
282 |
|
|
no strict 'refs'; |
283 |
dpavlin |
138 |
if ($display_data) { |
284 |
dpavlin |
153 |
$display_data .= $delimiter.join($delimiter,mkformat($x,&$filter($display))); |
285 |
dpavlin |
138 |
} else { |
286 |
dpavlin |
153 |
$display_data = join($delimiter,mkformat($x,&$filter($display))); |
287 |
dpavlin |
138 |
} |
288 |
dpavlin |
29 |
} else { |
289 |
dpavlin |
138 |
if ($display_data) { |
290 |
dpavlin |
153 |
$display_data .= $delimiter.mkformat($x,$display); |
291 |
dpavlin |
138 |
} else { |
292 |
dpavlin |
153 |
$display_data = mkformat($x,$display); |
293 |
dpavlin |
138 |
} |
294 |
dpavlin |
29 |
} |
295 |
dpavlin |
20 |
} |
296 |
dpavlin |
29 |
|
297 |
|
|
# type="index" ; insert into index |
298 |
|
|
if ($i && $display) { |
299 |
dpavlin |
177 |
if ($filter) { |
300 |
|
|
no strict 'refs'; |
301 |
|
|
$display = &$filter($display); |
302 |
|
|
} |
303 |
|
|
if ($x->{append} && @index_data) { |
304 |
|
|
$index_data[$#index_data].=$display; |
305 |
|
|
} else { |
306 |
|
|
push @index_data, $display; |
307 |
|
|
} |
308 |
dpavlin |
97 |
} |
309 |
|
|
|
310 |
dpavlin |
177 |
# store fields in lookup |
311 |
|
|
if ($il && $display) { |
312 |
|
|
if (lc($x->{type}) eq "lookup_key") { |
313 |
|
|
if ($lookup_key) { |
314 |
|
|
print STDERR "WARNING: try to redefine lookup_key (keys shouldn't be repeatable fields!)"; |
315 |
|
|
} else { |
316 |
|
|
$lookup_key = $display; |
317 |
|
|
} |
318 |
|
|
} elsif (lc($x->{type}) eq "lookup_val") { |
319 |
|
|
if ($lookup_key) { |
320 |
|
|
$lhash{$lookup_key} = $display; |
321 |
|
|
} else { |
322 |
|
|
print STDERR "WARNING: no lookup_key defined for '$display'?"; |
323 |
|
|
} |
324 |
dpavlin |
20 |
} |
325 |
dpavlin |
178 |
|
326 |
dpavlin |
20 |
} |
327 |
dpavlin |
178 |
|
328 |
|
|
# store data for page-by-page repeatable fields |
329 |
|
|
if ($iterate_by_page) { |
330 |
|
|
sub iterate_fld($$$$$$) { |
331 |
|
|
my ($cache,$what,$field,$page,$data,$append) = @_; |
332 |
|
|
return if (!$data); |
333 |
|
|
my $line_delimiter = $page_line_delimiter; |
334 |
|
|
$line_delimiter = " " if ($append); |
335 |
|
|
if (! $cache->{$what}->{$field}->[$page]) { |
336 |
|
|
$cache->{$what}->{$field}->[$page] = $data; |
337 |
|
|
} else { |
338 |
|
|
$cache->{$what}->{$field}->[$page] .= $line_delimiter.$data; |
339 |
|
|
} |
340 |
|
|
} |
341 |
|
|
|
342 |
|
|
if ($display_data) { |
343 |
|
|
print STDERR "line delimiter: ",Dumper($line_delimiter) if ($line_delimiter); |
344 |
|
|
iterate_fld($cache,'display_data',$field,$page,$display_data,$x->{append}); |
345 |
|
|
} |
346 |
|
|
$display_data = ""; |
347 |
|
|
if ($swish_data) { |
348 |
|
|
iterate_fld($cache,'swish_data',$field,$page,$swish_data,$x->{append}); |
349 |
|
|
$swish_data = ""; |
350 |
|
|
} |
351 |
|
|
if ($swish_exact_data) { |
352 |
|
|
iterate_fld($cache,'swish_exact_data',$field,$page,$swish_exact_data,$x->{append}); |
353 |
|
|
$swish_exact_data = ""; |
354 |
|
|
} |
355 |
|
|
} |
356 |
dpavlin |
17 |
} |
357 |
dpavlin |
177 |
|
358 |
|
|
# fill data in index |
359 |
|
|
foreach my $d (@index_data) { |
360 |
|
|
$index->insert($field, $d, $path); |
361 |
|
|
} |
362 |
dpavlin |
3 |
} |
363 |
dpavlin |
9 |
|
364 |
dpavlin |
59 |
# now try to parse variables from configuration file |
365 |
|
|
foreach my $x (@{$config->{indexer}->{$field}->{'config'}}) { |
366 |
dpavlin |
13 |
|
367 |
dpavlin |
62 |
my $delimiter = x($x->{delimiter}) || ' '; |
368 |
dpavlin |
59 |
my $val = $cfg->val($database, x($x->{content})); |
369 |
|
|
|
370 |
|
|
my ($s,$d,$i) = (1,1,0); # swish, display default |
371 |
|
|
$s = 0 if (lc($x->{type}) eq "display"); |
372 |
|
|
$d = 0 if (lc($x->{type}) eq "swish"); |
373 |
dpavlin |
163 |
# no support for swish exact in config. |
374 |
|
|
# IMHO, it's useless |
375 |
dpavlin |
59 |
($s,$d,$i) = (0,0,1) if (lc($x->{type}) eq "index"); |
376 |
|
|
|
377 |
|
|
if ($val) { |
378 |
dpavlin |
62 |
$display_data .= $delimiter.$val if ($d); |
379 |
dpavlin |
59 |
$swish_data .= $val if ($s); |
380 |
|
|
$index->insert($field, $val, $path) if ($i); |
381 |
|
|
} |
382 |
|
|
|
383 |
dpavlin |
178 |
if ($iterate_by_page) { |
384 |
|
|
# FIXME data from config tag will appear just |
385 |
|
|
# on first page!!! |
386 |
|
|
my $page = 0; |
387 |
|
|
if ($display_data) { |
388 |
|
|
$cache->{display_data}->{$field}->[$page] = $display_data; |
389 |
|
|
$display_data = ""; |
390 |
|
|
} |
391 |
|
|
if ($swish_data) { |
392 |
|
|
$cache->{swish_data}->{$field}->[$page] = $swish_data; |
393 |
|
|
$swish_data = ""; |
394 |
|
|
} |
395 |
|
|
if ($swish_exact_data) { |
396 |
|
|
$cache->{swish_exact_data}->{$field}->[$page] = $swish_exact_data; |
397 |
|
|
$swish_exact_data = ""; |
398 |
|
|
} |
399 |
|
|
} |
400 |
dpavlin |
59 |
} |
401 |
|
|
|
402 |
dpavlin |
178 |
# save data page-by-page |
403 |
|
|
foreach my $field (@page_fields) { |
404 |
|
|
my $nr_pages = $page_max{$field} || next; |
405 |
|
|
#print STDERR "field '$field' iterate over ",($nr_pages || 0)," pages...\n"; |
406 |
|
|
#print STDERR Dumper($cache->{display_data}); |
407 |
|
|
for (my $page=0; $page <= $nr_pages; $page++) { |
408 |
dpavlin |
59 |
|
409 |
dpavlin |
178 |
my $display_data = $cache->{display_data}->{$field}->[$page]; |
410 |
|
|
if ($display_data) { # default |
411 |
|
|
if ($field eq "headline") { |
412 |
|
|
$xml .= xmlify("headline", $display_data); |
413 |
|
|
} else { |
414 |
dpavlin |
29 |
|
415 |
dpavlin |
178 |
# fallback to empty field name if needed |
416 |
|
|
$html .= get_field_name($config,$field,$field_usage{$field}) || ''; |
417 |
|
|
$html .= "#-#".$display_data."###\n"; |
418 |
|
|
} |
419 |
|
|
} |
420 |
|
|
|
421 |
|
|
my $swish_data = $cache->{swish_data}->{$field}->[$page]; |
422 |
|
|
if ($swish_data) { |
423 |
|
|
# remove extra spaces |
424 |
|
|
$swish_data =~ s/ +/ /g; |
425 |
|
|
$swish_data =~ s/ +$//g; |
426 |
dpavlin |
13 |
|
427 |
dpavlin |
178 |
$xml .= xmlify($field."_swish", unac_string($codepage,$swish_data)); |
428 |
dpavlin |
13 |
} |
429 |
dpavlin |
178 |
|
430 |
|
|
my $swish_exact_data = $cache->{swish_exact_data}->{$field}->[$page]; |
431 |
|
|
if ($swish_exact_data) { |
432 |
|
|
$swish_exact_data =~ s/ +/ /g; |
433 |
|
|
$swish_exact_data =~ s/ +$//g; |
434 |
|
|
|
435 |
|
|
# add delimiters before and after word. |
436 |
|
|
# That is required to produce exact match |
437 |
|
|
$xml .= xmlify($field."_swish_exact", unac_string($codepage,'xxbxx '.$swish_exact_data.' xxexx')); |
438 |
dpavlin |
13 |
} |
439 |
|
|
} |
440 |
dpavlin |
178 |
|
441 |
dpavlin |
10 |
} |
442 |
dpavlin |
178 |
|
443 |
|
|
if (! $iterate_by_page) { |
444 |
|
|
if ($display_data) { |
445 |
|
|
if ($field eq "headline") { |
446 |
|
|
$xml .= xmlify("headline", $display_data); |
447 |
|
|
} else { |
448 |
dpavlin |
20 |
|
449 |
dpavlin |
178 |
# fallback to empty field name if needed |
450 |
|
|
$html .= get_field_name($config,$field,$field_usage{$field}) || ''; |
451 |
|
|
$html .= "#-#".$display_data."###\n"; |
452 |
|
|
} |
453 |
|
|
} |
454 |
|
|
if ($swish_data) { |
455 |
|
|
# remove extra spaces |
456 |
|
|
$swish_data =~ s/ +/ /g; |
457 |
|
|
$swish_data =~ s/ +$//g; |
458 |
dpavlin |
10 |
|
459 |
dpavlin |
178 |
$xml .= xmlify($field."_swish", unac_string($codepage,$swish_data)); |
460 |
|
|
} |
461 |
dpavlin |
9 |
|
462 |
dpavlin |
178 |
if ($swish_exact_data) { |
463 |
|
|
$swish_exact_data =~ s/ +/ /g; |
464 |
|
|
$swish_exact_data =~ s/ +$//g; |
465 |
|
|
|
466 |
|
|
# add delimiters before and after word. |
467 |
|
|
# That is required to produce exact match |
468 |
|
|
$xml .= xmlify($field."_swish_exact", unac_string($codepage,'xxbxx '.$swish_exact_data.' xxexx')); |
469 |
|
|
} |
470 |
dpavlin |
163 |
} |
471 |
dpavlin |
3 |
} |
472 |
dpavlin |
13 |
|
473 |
|
|
# dump formatted output in <html> |
474 |
|
|
if ($html) { |
475 |
dpavlin |
81 |
#$xml .= xmlify("html",$html); |
476 |
|
|
$xml .= "<html><![CDATA[ $html ]]></html>"; |
477 |
dpavlin |
13 |
} |
478 |
|
|
|
479 |
dpavlin |
3 |
if ($xml) { |
480 |
dpavlin |
13 |
$xml .= $add_xml if ($add_xml); |
481 |
dpavlin |
10 |
return "<xml>\n$xml</xml>\n"; |
482 |
dpavlin |
3 |
} else { |
483 |
|
|
return; |
484 |
|
|
} |
485 |
|
|
} |
486 |
|
|
|
487 |
|
|
########################################################################## |
488 |
|
|
|
489 |
dpavlin |
54 |
# read configuration for this script |
490 |
dpavlin |
13 |
my $cfg = new Config::IniFiles( -file => $config_file ); |
491 |
dpavlin |
1 |
|
492 |
dpavlin |
54 |
# read global.conf configuration |
493 |
|
|
my $cfg_global = new Config::IniFiles( -file => 'global.conf' ); |
494 |
|
|
|
495 |
dpavlin |
50 |
# open index |
496 |
|
|
$index = new index_DBI( |
497 |
dpavlin |
54 |
$cfg_global->val('global', 'dbi_dbd'), |
498 |
|
|
$cfg_global->val('global', 'dbi_dsn'), |
499 |
|
|
$cfg_global->val('global', 'dbi_user'), |
500 |
|
|
$cfg_global->val('global', 'dbi_passwd') || '', |
501 |
dpavlin |
50 |
); |
502 |
|
|
|
503 |
dpavlin |
97 |
my $show_progress = $cfg_global->val('global', 'show_progress'); |
504 |
|
|
|
505 |
dpavlin |
164 |
my $unac_filter = $cfg_global->val('global', 'unac_filter'); |
506 |
|
|
if ($unac_filter) { |
507 |
|
|
require $unac_filter; |
508 |
|
|
} |
509 |
|
|
|
510 |
dpavlin |
13 |
foreach my $database ($cfg->Sections) { |
511 |
dpavlin |
1 |
|
512 |
dpavlin |
54 |
my $type = lc($cfg -> val($database, 'type')) || die "$database doesn't have 'type' defined"; |
513 |
dpavlin |
40 |
my $add_xml = $cfg -> val($database, 'xml'); # optional |
514 |
dpavlin |
1 |
|
515 |
dpavlin |
177 |
# create new lookup file |
516 |
|
|
my $lookup_file = $cfg -> val($database, 'lookup_newfile'); # optional |
517 |
|
|
if ($lookup_file) { |
518 |
|
|
#tie %lhash, 'GDBM_File', $lookup_file, &GDBM_NEWDB, 0644; |
519 |
|
|
tie %lhash, 'TDB_File', $lookup_file, TDB_CLEAR_IF_FIRST, O_RDWR, 0644; |
520 |
|
|
print STDERR "creating lookup file '$lookup_file'\n"; |
521 |
|
|
} |
522 |
|
|
|
523 |
|
|
# open existing lookup file |
524 |
|
|
$lookup_file = $cfg -> val($database, 'lookup_open'); # optional |
525 |
|
|
if ($lookup_file) { |
526 |
|
|
#tie %lhash, 'GDBM_File', $lookup_file, &GDBM_READER, 0644; |
527 |
|
|
tie %lhash, 'TDB_File', $lookup_file, TDB_DEFAULT, O_RDWR, 0644; |
528 |
|
|
print STDERR "opening lookup file '$lookup_file'\n"; |
529 |
|
|
} |
530 |
|
|
|
531 |
dpavlin |
54 |
print STDERR "reading ./import_xml/$type.xml\n"; |
532 |
dpavlin |
1 |
|
533 |
dpavlin |
58 |
# extract just type basic |
534 |
|
|
my $type_base = $type; |
535 |
|
|
$type_base =~ s/_.+$//g; |
536 |
dpavlin |
40 |
|
537 |
dpavlin |
153 |
$config=XMLin("./import_xml/$type.xml", forcearray => [ $type2tag{$type_base}, 'config', 'format' ], forcecontent => 1); |
538 |
dpavlin |
58 |
|
539 |
dpavlin |
54 |
# output current progress indicator |
540 |
|
|
my $last_p = 0; |
541 |
|
|
sub progress { |
542 |
dpavlin |
101 |
return if (! $show_progress); |
543 |
dpavlin |
54 |
my $current = shift; |
544 |
dpavlin |
57 |
my $total = shift || 1; |
545 |
dpavlin |
54 |
my $p = int($current * 100 / $total); |
546 |
|
|
if ($p != $last_p) { |
547 |
|
|
printf STDERR ("%5d / %5d [%-51s] %-2d %% \r",$current,$total,"=" x ($p/2).">", $p ); |
548 |
|
|
$last_p = $p; |
549 |
|
|
} |
550 |
dpavlin |
3 |
} |
551 |
|
|
|
552 |
dpavlin |
67 |
my $fake_dir = 1; |
553 |
|
|
sub fakeprogress { |
554 |
dpavlin |
104 |
return if (! $show_progress); |
555 |
dpavlin |
67 |
my $current = shift @_; |
556 |
|
|
|
557 |
|
|
my @ind = ('-','\\','|','/','-','\\','|','/', '-'); |
558 |
|
|
|
559 |
|
|
$last_p += $fake_dir; |
560 |
|
|
$fake_dir = -$fake_dir if ($last_p > 1000 || $last_p < 0); |
561 |
|
|
if ($last_p % 10 == 0) { |
562 |
|
|
printf STDERR ("%5d / %5s [%-51s]\r",$current,"?"," " x ($last_p/20).$ind[($last_p/20) % $#ind]); |
563 |
|
|
} |
564 |
|
|
} |
565 |
|
|
|
566 |
dpavlin |
54 |
# now read database |
567 |
|
|
print STDERR "using: $type...\n"; |
568 |
dpavlin |
1 |
|
569 |
dpavlin |
180 |
# erase cache for tags by order in this database |
570 |
|
|
delete $cache->{tags_by_order}; |
571 |
|
|
|
572 |
dpavlin |
58 |
if ($type_base eq "isis") { |
573 |
|
|
|
574 |
dpavlin |
54 |
my $isis_db = $cfg -> val($database, 'isis_db') || die "$database doesn't have 'isis_db' defined!"; |
575 |
dpavlin |
1 |
|
576 |
dpavlin |
54 |
$import2cp = Text::Iconv->new($config->{isis_codepage},$codepage); |
577 |
|
|
my $db = OpenIsis::open( $isis_db ); |
578 |
dpavlin |
10 |
|
579 |
dpavlin |
106 |
# check if .txt database for OpenIsis is zero length, |
580 |
|
|
# if so, erase it and re-open database |
581 |
|
|
sub check_txt_db { |
582 |
|
|
my $isis_db = shift || die "need isis database name"; |
583 |
dpavlin |
109 |
my $reopen = 0; |
584 |
|
|
|
585 |
dpavlin |
106 |
if (-e $isis_db.".TXT") { |
586 |
dpavlin |
109 |
print STDERR "WARNING: removing $isis_db.TXT OpenIsis database...\n"; |
587 |
dpavlin |
108 |
unlink $isis_db.".TXT" || warn "FATAL: unlink error on '$isis_db.TXT': $!"; |
588 |
dpavlin |
109 |
$reopen++; |
589 |
dpavlin |
106 |
} |
590 |
dpavlin |
109 |
if (-e $isis_db.".PTR") { |
591 |
|
|
print STDERR "WARNING: removing $isis_db.PTR OpenIsis database...\n"; |
592 |
|
|
unlink $isis_db.".PTR" || warn "FATAL: unlink error on '$isis_db.PTR': $!"; |
593 |
|
|
$reopen++; |
594 |
|
|
} |
595 |
|
|
return OpenIsis::open( $isis_db ) if ($reopen); |
596 |
dpavlin |
106 |
} |
597 |
|
|
|
598 |
|
|
# EOF error |
599 |
|
|
if ($db == -1) { |
600 |
|
|
$db = check_txt_db($isis_db); |
601 |
dpavlin |
108 |
if ($db == -1) { |
602 |
dpavlin |
106 |
print STDERR "FATAL: OpenIsis can't open zero size file $isis_db\n"; |
603 |
|
|
next; |
604 |
|
|
} |
605 |
|
|
} |
606 |
|
|
|
607 |
dpavlin |
101 |
# OpenIsis::ERR_BADF |
608 |
|
|
if ($db == -4) { |
609 |
|
|
print STDERR "FATAL: OpenIsis can't find file $isis_db\n"; |
610 |
|
|
next; |
611 |
|
|
# OpenIsis::ERR_IO |
612 |
|
|
} elsif ($db == -5) { |
613 |
|
|
print STDERR "FATAL: OpenIsis can't access file $isis_db\n"; |
614 |
|
|
next; |
615 |
|
|
} elsif ($db < 0) { |
616 |
|
|
print STDERR "FATAL: OpenIsis unknown error $db with file $isis_db\n"; |
617 |
|
|
next; |
618 |
|
|
} |
619 |
|
|
|
620 |
dpavlin |
54 |
my $max_rowid = OpenIsis::maxRowid( $db ); |
621 |
dpavlin |
3 |
|
622 |
dpavlin |
106 |
# if 0 records, try to rease isis .txt database |
623 |
|
|
if ($max_rowid == 0) { |
624 |
|
|
# force removal of database |
625 |
|
|
$db = check_txt_db($isis_db); |
626 |
|
|
$max_rowid = OpenIsis::maxRowid( $db ); |
627 |
|
|
} |
628 |
|
|
|
629 |
dpavlin |
54 |
print STDERR "Reading database: $isis_db [$max_rowid rows]\n"; |
630 |
|
|
|
631 |
|
|
my $path = $database; |
632 |
|
|
|
633 |
|
|
for (my $row_id = 1; $row_id <= $max_rowid; $row_id++ ) { |
634 |
|
|
my $row = OpenIsis::read( $db, $row_id ); |
635 |
|
|
if ($row && $row->{mfn}) { |
636 |
|
|
|
637 |
|
|
progress($row->{mfn}, $max_rowid); |
638 |
|
|
|
639 |
|
|
my $swishpath = $path."#".int($row->{mfn}); |
640 |
|
|
|
641 |
dpavlin |
59 |
if (my $xml = data2xml($type_base,$row,$add_xml,$cfg,$database)) { |
642 |
dpavlin |
54 |
$xml = $cp2utf->convert($xml); |
643 |
|
|
use bytes; # as opposed to chars |
644 |
|
|
print "Path-Name: $swishpath\n"; |
645 |
|
|
print "Content-Length: ".(length($xml)+1)."\n"; |
646 |
|
|
print "Document-Type: XML\n\n$xml\n"; |
647 |
|
|
} |
648 |
dpavlin |
3 |
} |
649 |
dpavlin |
54 |
} |
650 |
dpavlin |
108 |
# for this to work with current version of OpenIsis (0.9.0) |
651 |
|
|
# you might need my patch from |
652 |
|
|
# https://www.rot13.org/~dpavlin/projects/openisis-0.9.0-perl_close.diff |
653 |
|
|
OpenIsis::close($db); |
654 |
dpavlin |
54 |
print STDERR "\n"; |
655 |
dpavlin |
3 |
|
656 |
dpavlin |
58 |
} elsif ($type_base eq "excel") { |
657 |
dpavlin |
54 |
use Spreadsheet::ParseExcel; |
658 |
|
|
use Spreadsheet::ParseExcel::Utility qw(int2col); |
659 |
|
|
|
660 |
|
|
$import2cp = Text::Iconv->new($config->{excel_codepage},$codepage); |
661 |
|
|
my $excel_file = $cfg -> val($database, 'excel_file') || die "$database doesn't have 'excel_file' defined!"; |
662 |
dpavlin |
43 |
|
663 |
dpavlin |
54 |
my $sheet = x($config->{sheet}) || die "no sheet in $type.xml"; |
664 |
dpavlin |
56 |
my $start_row = x($config->{start_row}) - 1 || die "no start_row in $type.xml"; |
665 |
dpavlin |
54 |
|
666 |
|
|
my $oBook = Spreadsheet::ParseExcel::Workbook->Parse($excel_file) || die "can't open Excel file '$excel_file'"; |
667 |
|
|
|
668 |
|
|
my $sheet_nr = 0; |
669 |
|
|
foreach my $oWks (@{$oBook->{Worksheet}}) { |
670 |
|
|
#print STDERR "-- SHEET $sheet_nr:", $oWks->{Name}, "\n"; |
671 |
|
|
last if ($oWks->{Name} eq $sheet); |
672 |
|
|
$sheet_nr++; |
673 |
|
|
} |
674 |
|
|
|
675 |
|
|
my $oWorksheet = $oBook->{Worksheet}[$sheet_nr]; |
676 |
|
|
|
677 |
|
|
print STDERR "using sheet: ",$oWorksheet->{Name},"\n"; |
678 |
|
|
defined ($oWorksheet) || die "can't find sheet '$sheet' in $excel_file"; |
679 |
|
|
my $end_row = x($config->{end_row}) || $oWorksheet->{MaxRow}; |
680 |
|
|
|
681 |
dpavlin |
56 |
for(my $iR = $start_row ; defined $end_row && $iR <= $end_row ; $iR++) { |
682 |
dpavlin |
54 |
my $row; |
683 |
|
|
for(my $iC = $oWorksheet->{MinCol} ; defined $oWorksheet->{MaxCol} && $iC <= $oWorksheet->{MaxCol} ; $iC++) { |
684 |
|
|
my $cell = $oWorksheet->{Cells}[$iR][$iC]; |
685 |
|
|
if ($cell) { |
686 |
|
|
$row->{int2col($iC)} = $cell->Value; |
687 |
|
|
} |
688 |
|
|
} |
689 |
|
|
|
690 |
|
|
progress($iR, $end_row); |
691 |
|
|
|
692 |
|
|
# print "row[$iR/$end_row] "; |
693 |
|
|
# foreach (keys %{$row}) { |
694 |
|
|
# print "$_: ",$row->{$_},"\t"; |
695 |
|
|
# } |
696 |
|
|
# print "\n"; |
697 |
|
|
|
698 |
|
|
my $swishpath = $database."#".$iR; |
699 |
|
|
|
700 |
|
|
next if (! $row); |
701 |
|
|
|
702 |
dpavlin |
59 |
if (my $xml = data2xml($type_base,$row,$add_xml,$cfg,$database)) { |
703 |
dpavlin |
44 |
$xml = $cp2utf->convert($xml); |
704 |
dpavlin |
35 |
use bytes; # as opposed to chars |
705 |
dpavlin |
43 |
print "Path-Name: $swishpath\n"; |
706 |
dpavlin |
10 |
print "Content-Length: ".(length($xml)+1)."\n"; |
707 |
|
|
print "Document-Type: XML\n\n$xml\n"; |
708 |
dpavlin |
3 |
} |
709 |
dpavlin |
1 |
} |
710 |
dpavlin |
62 |
} elsif ($type_base eq "marc") { |
711 |
dpavlin |
67 |
|
712 |
dpavlin |
62 |
use MARC; |
713 |
|
|
|
714 |
|
|
$import2cp = Text::Iconv->new($config->{marc_codepage},$codepage); |
715 |
|
|
my $marc_file = $cfg -> val($database, 'marc_file') || die "$database doesn't have 'marc_file' defined!"; |
716 |
|
|
|
717 |
|
|
# optional argument is format |
718 |
|
|
my $format = x($config->{format}) || 'usmarc'; |
719 |
|
|
|
720 |
|
|
print STDERR "Reading MARC file '$marc_file'\n"; |
721 |
|
|
|
722 |
|
|
my $marc = new MARC; |
723 |
|
|
my $nr = $marc->openmarc({ |
724 |
|
|
file=>$marc_file, format=>$format |
725 |
|
|
}) || die "Can't open MARC file '$marc_file'"; |
726 |
|
|
|
727 |
|
|
my $i=0; # record nr. |
728 |
|
|
|
729 |
|
|
my $rec; |
730 |
|
|
|
731 |
|
|
while ($marc->nextmarc(1)) { |
732 |
|
|
|
733 |
|
|
# XXX |
734 |
dpavlin |
67 |
fakeprogress($i++); |
735 |
dpavlin |
62 |
|
736 |
|
|
my $swishpath = $database."#".$i; |
737 |
|
|
|
738 |
|
|
if (my $xml = data2xml($type_base,$marc,$add_xml,$cfg,$database)) { |
739 |
|
|
$xml = $cp2utf->convert($xml); |
740 |
|
|
use bytes; # as opposed to chars |
741 |
|
|
print "Path-Name: $swishpath\n"; |
742 |
|
|
print "Content-Length: ".(length($xml)+1)."\n"; |
743 |
|
|
print "Document-Type: XML\n\n$xml\n"; |
744 |
|
|
} |
745 |
|
|
} |
746 |
dpavlin |
67 |
} elsif ($type_base eq "feed") { |
747 |
|
|
|
748 |
|
|
$import2cp = Text::Iconv->new($config->{feed_codepage},$codepage); |
749 |
|
|
my $prog = x($config->{prog}) || die "$database doesn't have 'prog' defined!"; |
750 |
|
|
|
751 |
|
|
print STDERR "Reading feed from program '$prog'\n"; |
752 |
|
|
|
753 |
|
|
open(FEED,"feeds/$prog |") || die "can't start $prog: $!"; |
754 |
|
|
|
755 |
dpavlin |
74 |
my $i=1; # record nr. |
756 |
dpavlin |
67 |
|
757 |
|
|
my $data; |
758 |
|
|
my $line=1; |
759 |
|
|
|
760 |
|
|
while (<FEED>) { |
761 |
|
|
chomp; |
762 |
|
|
|
763 |
|
|
if (/^$/) { |
764 |
dpavlin |
74 |
my $swishpath = $database."#".$i++; |
765 |
dpavlin |
67 |
|
766 |
|
|
if (my $xml = data2xml($type_base,$data,$add_xml,$cfg,$database)) { |
767 |
|
|
$xml = $cp2utf->convert($xml); |
768 |
|
|
use bytes; # as opposed to chars |
769 |
|
|
print "Path-Name: $swishpath\n"; |
770 |
|
|
print "Content-Length: ".(length($xml)+1)."\n"; |
771 |
|
|
print "Document-Type: XML\n\n$xml\n"; |
772 |
|
|
} |
773 |
|
|
$line = 1; |
774 |
|
|
$data = {}; |
775 |
|
|
next; |
776 |
|
|
} |
777 |
|
|
|
778 |
dpavlin |
74 |
$line = $1 if (s/^(\d+):\s*//); |
779 |
dpavlin |
67 |
$data->{$line++} = $_; |
780 |
|
|
|
781 |
dpavlin |
74 |
fakeprogress($i); |
782 |
dpavlin |
67 |
|
783 |
|
|
} |
784 |
dpavlin |
177 |
# close lookup |
785 |
|
|
untie %lhash if (%lhash); |
786 |
dpavlin |
1 |
} |
787 |
|
|
} |
788 |
dpavlin |
3 |
|
789 |
dpavlin |
10 |
# call this to commit index |
790 |
|
|
$index->close; |
791 |
dpavlin |
3 |
|
792 |
|
|
1; |
793 |
|
|
__END__ |
794 |
|
|
########################################################################## |
795 |
|
|
|
796 |
|
|
=head1 NAME |
797 |
|
|
|
798 |
dpavlin |
81 |
all2xml.pl - read various file formats and dump XML for SWISH-E |
799 |
dpavlin |
3 |
|
800 |
|
|
=head1 DESCRIPTION |
801 |
|
|
|
802 |
dpavlin |
81 |
This command will read ISIS data file using OpenIsis perl module, MARC |
803 |
|
|
records using MARC module and optionally Micro$oft Excel files to |
804 |
|
|
create one XML file for usage with I<SWISH-E> indexer. Dispite it's name, |
805 |
|
|
this script B<isn't general xml generator> from isis files (isis allready |
806 |
|
|
has something like that). Output of this script is tailor-made for SWISH-E. |
807 |
dpavlin |
3 |
|
808 |
dpavlin |
81 |
=head1 BUGS |
809 |
|
|
|
810 |
|
|
Documentation is really lacking. However, in true Open Source spirit, source |
811 |
|
|
is best documentation. I even made considerable effort to comment parts |
812 |
|
|
which are not intuitively clear, so... |
813 |
|
|
|
814 |
dpavlin |
3 |
=head1 AUTHOR |
815 |
|
|
|
816 |
|
|
Dobrica Pavlinusic <dpavlin@rot13.org> |
817 |
|
|
|
818 |
|
|
=head1 COPYRIGHT |
819 |
|
|
|
820 |
|
|
GNU Public License (GPL) v2 or later |
821 |
|
|
|
822 |
|
|
=head1 SEE ALSO |
823 |
|
|
|
824 |
|
|
SWISH-E web site at http://www.swish-e.org |
825 |
|
|
|
826 |
|
|
=cut |