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 <=> $vb; |
102 |
dpavlin |
32 |
} |
103 |
dpavlin |
3 |
|
104 |
dpavlin |
170 |
my @sorted_tags; |
105 |
|
|
if ($cache->{tags_by_order}->{$type}) { |
106 |
|
|
@sorted_tags = @{$cache->{tags_by_order}->{$type}}; |
107 |
|
|
} else { |
108 |
|
|
@sorted_tags = sort by_order keys %{$config->{indexer}}; |
109 |
|
|
$cache->{tags_by_order}->{$type} = \@sorted_tags; |
110 |
|
|
} |
111 |
dpavlin |
32 |
|
112 |
dpavlin |
177 |
# lookup key |
113 |
|
|
my $lookup_key; |
114 |
|
|
|
115 |
dpavlin |
170 |
foreach my $field (@sorted_tags) { |
116 |
|
|
|
117 |
dpavlin |
40 |
$field=x($field); |
118 |
dpavlin |
13 |
$field_usage{$field}++; |
119 |
|
|
|
120 |
dpavlin |
10 |
my $swish_data = ""; |
121 |
dpavlin |
163 |
my $swish_exact_data = ""; |
122 |
dpavlin |
3 |
my $display_data = ""; |
123 |
dpavlin |
35 |
my $line_delimiter; |
124 |
dpavlin |
3 |
|
125 |
dpavlin |
34 |
my ($swish,$display); |
126 |
|
|
|
127 |
dpavlin |
54 |
my $tag = $type2tag{$type} || die "can't find which tag to use for type $type"; |
128 |
|
|
foreach my $x (@{$config->{indexer}->{$field}->{$tag}}) { |
129 |
dpavlin |
3 |
|
130 |
dpavlin |
40 |
my $format = x($x->{content}); |
131 |
|
|
my $delimiter = x($x->{delimiter}) || ' '; |
132 |
dpavlin |
3 |
|
133 |
dpavlin |
54 |
my $repeat_off = 0; # repeatable offset |
134 |
dpavlin |
29 |
|
135 |
dpavlin |
177 |
# swish, swish_exact, display, index, index_lookup |
136 |
|
|
# swish and display defaults |
137 |
|
|
my ($s,$se,$d,$i,$il) = (1,0,1,0,0); |
138 |
dpavlin |
29 |
$s = 0 if (lc($x->{type}) eq "display"); |
139 |
|
|
$d = 0 if (lc($x->{type}) eq "swish"); |
140 |
dpavlin |
163 |
$se = 1 if (lc($x->{type}) eq "swish_exact"); |
141 |
dpavlin |
29 |
($s,$d,$i) = (0,0,1) if (lc($x->{type}) eq "index"); |
142 |
dpavlin |
177 |
$il = 1 if (lc($x->{type}) =~ /^lookup/); |
143 |
dpavlin |
29 |
|
144 |
dpavlin |
177 |
|
145 |
dpavlin |
35 |
# what will separate last line from this one? |
146 |
|
|
if ($display_data && $x->{append} && $x->{append} eq "1") { |
147 |
dpavlin |
34 |
$line_delimiter = ' '; |
148 |
dpavlin |
35 |
} elsif ($display_data) { |
149 |
dpavlin |
34 |
$line_delimiter = '<br/>'; |
150 |
|
|
} |
151 |
dpavlin |
29 |
|
152 |
dpavlin |
34 |
# init vars so that we go into while... |
153 |
|
|
($swish,$display) = (1,1); |
154 |
dpavlin |
29 |
|
155 |
dpavlin |
97 |
# placeholder for all repeatable entries for index |
156 |
|
|
my @index_data; |
157 |
|
|
|
158 |
dpavlin |
153 |
sub mkformat { |
159 |
|
|
my $x = shift || die "mkformat needs tag reference"; |
160 |
|
|
my $data = shift || return; |
161 |
|
|
my $format_name = x($x->{format_name}) || return $data; |
162 |
|
|
my $fmt = x($config->{format}->{$format_name}->{content}) || die "<format name=\"$format_name\"> is not defined!"; |
163 |
|
|
my $format_delimiter = x($x->{format_delimiter}); |
164 |
|
|
my @data; |
165 |
|
|
if ($format_delimiter) { |
166 |
|
|
@data = split(/$format_delimiter/,$data); |
167 |
|
|
} else { |
168 |
|
|
push @data,$data; |
169 |
|
|
} |
170 |
|
|
|
171 |
|
|
if ($fmt) { |
172 |
|
|
my $nr = scalar $fmt =~ s/%s/%s/g; |
173 |
|
|
if (($#data+1) == $nr) { |
174 |
|
|
return sprintf($fmt,@data); |
175 |
|
|
} else { |
176 |
|
|
print STDERR "mkformat: [$data] can't be split on [$format_delimiter] to $nr fields!\n"; |
177 |
|
|
return $data; |
178 |
|
|
} |
179 |
|
|
} else { |
180 |
|
|
print STDERR "usage of link '$format_name' without defined format (<link> tag)\n"; |
181 |
|
|
} |
182 |
|
|
} |
183 |
|
|
|
184 |
dpavlin |
90 |
# while because of repeatable fields |
185 |
|
|
while ($swish || $display) { |
186 |
dpavlin |
54 |
($swish,$display) = parse_format($type, $format,$row,$repeat_off++,$import2cp); |
187 |
dpavlin |
90 |
if ($repeat_off > 1000) { |
188 |
|
|
print STDERR "loop (more than 1000 repeatable fields) deteced in $row, $format\n"; |
189 |
|
|
last; |
190 |
|
|
} |
191 |
dpavlin |
177 |
|
192 |
|
|
# is this field is lookup? |
193 |
|
|
if ($display && $x->{lookup}) { |
194 |
|
|
if ($use_lhash_cache) { |
195 |
|
|
if (!defined($cache->{lhash}->{$display})) { |
196 |
|
|
my $new_display = $lhash{$display}; |
197 |
|
|
if ($new_display) { |
198 |
|
|
#print STDERR "lookup cache store '$display' = '$new_display'\n"; |
199 |
|
|
$display = $new_display; |
200 |
|
|
$cache->{lhash}->{$display} = $new_display; |
201 |
|
|
} else { |
202 |
|
|
print STDERR "WARNING: lookup for '$display' didn't find anything.\n"; |
203 |
|
|
$display = ""; |
204 |
|
|
$cache->{lhash}->{$display} = ""; |
205 |
|
|
} |
206 |
|
|
} else { |
207 |
|
|
$display = $cache->{lhash}->{$display}; |
208 |
|
|
} |
209 |
|
|
} else { |
210 |
|
|
$display = $lhash{$display} || ""; |
211 |
|
|
} |
212 |
|
|
} |
213 |
|
|
|
214 |
dpavlin |
29 |
# filter="name" ; filter this field through |
215 |
|
|
# filter/[name].pm |
216 |
|
|
my $filter = $x->{filter}; |
217 |
dpavlin |
170 |
if ($filter && !$cache->{filter_loaded}->{$filter}) { |
218 |
dpavlin |
29 |
require "filter/".$filter.".pm"; |
219 |
dpavlin |
170 |
$cache->{filter_loaded}->{$filter}++; |
220 |
dpavlin |
20 |
} |
221 |
dpavlin |
29 |
# type="swish" ; field for swish |
222 |
dpavlin |
163 |
if ($swish) { |
223 |
|
|
if ($filter && ($s || $se)) { |
224 |
dpavlin |
29 |
no strict 'refs'; |
225 |
dpavlin |
163 |
my $tmp = join(" ",&$filter($swish)) if ($s || $se); |
226 |
|
|
$swish_data .= $tmp if ($s); |
227 |
|
|
$swish_exact_data .= $tmp if ($se); |
228 |
|
|
|
229 |
dpavlin |
29 |
} else { |
230 |
dpavlin |
163 |
$swish_data .= $swish if ($s); |
231 |
|
|
$swish_exact_data .= $swish if ($se); |
232 |
dpavlin |
29 |
} |
233 |
|
|
} |
234 |
dpavlin |
17 |
|
235 |
dpavlin |
29 |
# type="display" ; field for display |
236 |
|
|
if ($d && $display) { |
237 |
dpavlin |
35 |
if ($line_delimiter && $display_data) { |
238 |
|
|
$display_data .= $line_delimiter; |
239 |
|
|
undef $line_delimiter; |
240 |
|
|
} |
241 |
dpavlin |
29 |
if ($filter) { |
242 |
|
|
no strict 'refs'; |
243 |
dpavlin |
138 |
if ($display_data) { |
244 |
dpavlin |
153 |
$display_data .= $delimiter.join($delimiter,mkformat($x,&$filter($display))); |
245 |
dpavlin |
138 |
} else { |
246 |
dpavlin |
153 |
$display_data = join($delimiter,mkformat($x,&$filter($display))); |
247 |
dpavlin |
138 |
} |
248 |
dpavlin |
29 |
} else { |
249 |
dpavlin |
138 |
if ($display_data) { |
250 |
dpavlin |
153 |
$display_data .= $delimiter.mkformat($x,$display); |
251 |
dpavlin |
138 |
} else { |
252 |
dpavlin |
153 |
$display_data = mkformat($x,$display); |
253 |
dpavlin |
138 |
} |
254 |
dpavlin |
29 |
} |
255 |
dpavlin |
20 |
} |
256 |
dpavlin |
29 |
|
257 |
|
|
# type="index" ; insert into index |
258 |
|
|
if ($i && $display) { |
259 |
dpavlin |
177 |
if ($filter) { |
260 |
|
|
no strict 'refs'; |
261 |
|
|
$display = &$filter($display); |
262 |
|
|
} |
263 |
|
|
if ($x->{append} && @index_data) { |
264 |
|
|
$index_data[$#index_data].=$display; |
265 |
|
|
} else { |
266 |
|
|
push @index_data, $display; |
267 |
|
|
} |
268 |
dpavlin |
97 |
} |
269 |
|
|
|
270 |
dpavlin |
177 |
# store fields in lookup |
271 |
|
|
if ($il && $display) { |
272 |
|
|
if (lc($x->{type}) eq "lookup_key") { |
273 |
|
|
if ($lookup_key) { |
274 |
|
|
print STDERR "WARNING: try to redefine lookup_key (keys shouldn't be repeatable fields!)"; |
275 |
|
|
} else { |
276 |
|
|
$lookup_key = $display; |
277 |
|
|
} |
278 |
|
|
} elsif (lc($x->{type}) eq "lookup_val") { |
279 |
|
|
if ($lookup_key) { |
280 |
|
|
$lhash{$lookup_key} = $display; |
281 |
|
|
} else { |
282 |
|
|
print STDERR "WARNING: no lookup_key defined for '$display'?"; |
283 |
|
|
} |
284 |
dpavlin |
20 |
} |
285 |
|
|
} |
286 |
dpavlin |
17 |
} |
287 |
dpavlin |
177 |
|
288 |
|
|
# fill data in index |
289 |
|
|
foreach my $d (@index_data) { |
290 |
|
|
$index->insert($field, $d, $path); |
291 |
|
|
} |
292 |
dpavlin |
3 |
} |
293 |
dpavlin |
9 |
|
294 |
dpavlin |
59 |
# now try to parse variables from configuration file |
295 |
|
|
foreach my $x (@{$config->{indexer}->{$field}->{'config'}}) { |
296 |
dpavlin |
13 |
|
297 |
dpavlin |
62 |
my $delimiter = x($x->{delimiter}) || ' '; |
298 |
dpavlin |
59 |
my $val = $cfg->val($database, x($x->{content})); |
299 |
|
|
|
300 |
|
|
my ($s,$d,$i) = (1,1,0); # swish, display default |
301 |
|
|
$s = 0 if (lc($x->{type}) eq "display"); |
302 |
|
|
$d = 0 if (lc($x->{type}) eq "swish"); |
303 |
dpavlin |
163 |
# no support for swish exact in config. |
304 |
|
|
# IMHO, it's useless |
305 |
dpavlin |
59 |
($s,$d,$i) = (0,0,1) if (lc($x->{type}) eq "index"); |
306 |
|
|
|
307 |
|
|
if ($val) { |
308 |
dpavlin |
62 |
$display_data .= $delimiter.$val if ($d); |
309 |
dpavlin |
59 |
$swish_data .= $val if ($s); |
310 |
|
|
$index->insert($field, $val, $path) if ($i); |
311 |
|
|
} |
312 |
|
|
|
313 |
|
|
} |
314 |
|
|
|
315 |
|
|
|
316 |
dpavlin |
10 |
if ($display_data) { |
317 |
dpavlin |
29 |
|
318 |
dpavlin |
13 |
if ($field eq "headline") { |
319 |
|
|
$xml .= xmlify("headline", $display_data); |
320 |
|
|
} else { |
321 |
|
|
|
322 |
|
|
# find field name (signular, plural) |
323 |
|
|
my $field_name = ""; |
324 |
|
|
if ($config->{indexer}->{$field}->{name_singular} && $field_usage{$field} == 1) { |
325 |
|
|
$field_name = $config->{indexer}->{$field}->{name_singular}."#-#"; |
326 |
|
|
} elsif ($config->{indexer}->{$field}->{name_plural}) { |
327 |
|
|
$field_name = $config->{indexer}->{$field}->{name_plural}."#-#"; |
328 |
dpavlin |
20 |
} elsif ($config->{indexer}->{$field}->{name}) { |
329 |
|
|
$field_name = $config->{indexer}->{$field}->{name}."#-#"; |
330 |
dpavlin |
13 |
} else { |
331 |
dpavlin |
20 |
print STDERR "WARNING: field '$field' doesn't have 'name' attribute!"; |
332 |
dpavlin |
13 |
} |
333 |
|
|
if ($field_name) { |
334 |
dpavlin |
40 |
$html .= x($field_name); |
335 |
dpavlin |
13 |
} |
336 |
|
|
$html .= $display_data."###\n"; |
337 |
|
|
} |
338 |
dpavlin |
10 |
} |
339 |
|
|
if ($swish_data) { |
340 |
dpavlin |
20 |
# remove extra spaces |
341 |
|
|
$swish_data =~ s/ +/ /g; |
342 |
|
|
$swish_data =~ s/ +$//g; |
343 |
|
|
|
344 |
dpavlin |
40 |
$xml .= xmlify($field."_swish", unac_string($codepage,$swish_data)); |
345 |
dpavlin |
10 |
} |
346 |
|
|
|
347 |
dpavlin |
163 |
if ($swish_exact_data) { |
348 |
|
|
$swish_exact_data =~ s/ +/ /g; |
349 |
|
|
$swish_exact_data =~ s/ +$//g; |
350 |
dpavlin |
9 |
|
351 |
dpavlin |
163 |
# add delimiters before and after word. |
352 |
|
|
# That is required to produce exact match |
353 |
|
|
$xml .= xmlify($field."_swish_exact", unac_string($codepage,'xxbxx '.$swish_exact_data.' xxexx')); |
354 |
|
|
} |
355 |
|
|
|
356 |
|
|
|
357 |
dpavlin |
3 |
} |
358 |
dpavlin |
13 |
|
359 |
|
|
# dump formatted output in <html> |
360 |
|
|
if ($html) { |
361 |
dpavlin |
81 |
#$xml .= xmlify("html",$html); |
362 |
|
|
$xml .= "<html><![CDATA[ $html ]]></html>"; |
363 |
dpavlin |
13 |
} |
364 |
|
|
|
365 |
dpavlin |
3 |
if ($xml) { |
366 |
dpavlin |
13 |
$xml .= $add_xml if ($add_xml); |
367 |
dpavlin |
10 |
return "<xml>\n$xml</xml>\n"; |
368 |
dpavlin |
3 |
} else { |
369 |
|
|
return; |
370 |
|
|
} |
371 |
|
|
} |
372 |
|
|
|
373 |
|
|
########################################################################## |
374 |
|
|
|
375 |
dpavlin |
54 |
# read configuration for this script |
376 |
dpavlin |
13 |
my $cfg = new Config::IniFiles( -file => $config_file ); |
377 |
dpavlin |
1 |
|
378 |
dpavlin |
54 |
# read global.conf configuration |
379 |
|
|
my $cfg_global = new Config::IniFiles( -file => 'global.conf' ); |
380 |
|
|
|
381 |
dpavlin |
50 |
# open index |
382 |
|
|
$index = new index_DBI( |
383 |
dpavlin |
54 |
$cfg_global->val('global', 'dbi_dbd'), |
384 |
|
|
$cfg_global->val('global', 'dbi_dsn'), |
385 |
|
|
$cfg_global->val('global', 'dbi_user'), |
386 |
|
|
$cfg_global->val('global', 'dbi_passwd') || '', |
387 |
dpavlin |
50 |
); |
388 |
|
|
|
389 |
dpavlin |
97 |
my $show_progress = $cfg_global->val('global', 'show_progress'); |
390 |
|
|
|
391 |
dpavlin |
164 |
my $unac_filter = $cfg_global->val('global', 'unac_filter'); |
392 |
|
|
if ($unac_filter) { |
393 |
|
|
require $unac_filter; |
394 |
|
|
} |
395 |
|
|
|
396 |
dpavlin |
13 |
foreach my $database ($cfg->Sections) { |
397 |
dpavlin |
1 |
|
398 |
dpavlin |
54 |
my $type = lc($cfg -> val($database, 'type')) || die "$database doesn't have 'type' defined"; |
399 |
dpavlin |
40 |
my $add_xml = $cfg -> val($database, 'xml'); # optional |
400 |
dpavlin |
1 |
|
401 |
dpavlin |
177 |
# create new lookup file |
402 |
|
|
my $lookup_file = $cfg -> val($database, 'lookup_newfile'); # optional |
403 |
|
|
if ($lookup_file) { |
404 |
|
|
#tie %lhash, 'GDBM_File', $lookup_file, &GDBM_NEWDB, 0644; |
405 |
|
|
tie %lhash, 'TDB_File', $lookup_file, TDB_CLEAR_IF_FIRST, O_RDWR, 0644; |
406 |
|
|
print STDERR "creating lookup file '$lookup_file'\n"; |
407 |
|
|
} |
408 |
|
|
|
409 |
|
|
# open existing lookup file |
410 |
|
|
$lookup_file = $cfg -> val($database, 'lookup_open'); # optional |
411 |
|
|
if ($lookup_file) { |
412 |
|
|
#tie %lhash, 'GDBM_File', $lookup_file, &GDBM_READER, 0644; |
413 |
|
|
tie %lhash, 'TDB_File', $lookup_file, TDB_DEFAULT, O_RDWR, 0644; |
414 |
|
|
print STDERR "opening lookup file '$lookup_file'\n"; |
415 |
|
|
} |
416 |
|
|
|
417 |
dpavlin |
54 |
print STDERR "reading ./import_xml/$type.xml\n"; |
418 |
dpavlin |
1 |
|
419 |
dpavlin |
58 |
# extract just type basic |
420 |
|
|
my $type_base = $type; |
421 |
|
|
$type_base =~ s/_.+$//g; |
422 |
dpavlin |
40 |
|
423 |
dpavlin |
153 |
$config=XMLin("./import_xml/$type.xml", forcearray => [ $type2tag{$type_base}, 'config', 'format' ], forcecontent => 1); |
424 |
dpavlin |
58 |
|
425 |
dpavlin |
54 |
# output current progress indicator |
426 |
|
|
my $last_p = 0; |
427 |
|
|
sub progress { |
428 |
dpavlin |
101 |
return if (! $show_progress); |
429 |
dpavlin |
54 |
my $current = shift; |
430 |
dpavlin |
57 |
my $total = shift || 1; |
431 |
dpavlin |
54 |
my $p = int($current * 100 / $total); |
432 |
|
|
if ($p != $last_p) { |
433 |
|
|
printf STDERR ("%5d / %5d [%-51s] %-2d %% \r",$current,$total,"=" x ($p/2).">", $p ); |
434 |
|
|
$last_p = $p; |
435 |
|
|
} |
436 |
dpavlin |
3 |
} |
437 |
|
|
|
438 |
dpavlin |
67 |
my $fake_dir = 1; |
439 |
|
|
sub fakeprogress { |
440 |
dpavlin |
104 |
return if (! $show_progress); |
441 |
dpavlin |
67 |
my $current = shift @_; |
442 |
|
|
|
443 |
|
|
my @ind = ('-','\\','|','/','-','\\','|','/', '-'); |
444 |
|
|
|
445 |
|
|
$last_p += $fake_dir; |
446 |
|
|
$fake_dir = -$fake_dir if ($last_p > 1000 || $last_p < 0); |
447 |
|
|
if ($last_p % 10 == 0) { |
448 |
|
|
printf STDERR ("%5d / %5s [%-51s]\r",$current,"?"," " x ($last_p/20).$ind[($last_p/20) % $#ind]); |
449 |
|
|
} |
450 |
|
|
} |
451 |
|
|
|
452 |
dpavlin |
54 |
# now read database |
453 |
|
|
print STDERR "using: $type...\n"; |
454 |
dpavlin |
1 |
|
455 |
dpavlin |
58 |
if ($type_base eq "isis") { |
456 |
|
|
|
457 |
dpavlin |
54 |
my $isis_db = $cfg -> val($database, 'isis_db') || die "$database doesn't have 'isis_db' defined!"; |
458 |
dpavlin |
1 |
|
459 |
dpavlin |
54 |
$import2cp = Text::Iconv->new($config->{isis_codepage},$codepage); |
460 |
|
|
my $db = OpenIsis::open( $isis_db ); |
461 |
dpavlin |
10 |
|
462 |
dpavlin |
106 |
# check if .txt database for OpenIsis is zero length, |
463 |
|
|
# if so, erase it and re-open database |
464 |
|
|
sub check_txt_db { |
465 |
|
|
my $isis_db = shift || die "need isis database name"; |
466 |
dpavlin |
109 |
my $reopen = 0; |
467 |
|
|
|
468 |
dpavlin |
106 |
if (-e $isis_db.".TXT") { |
469 |
dpavlin |
109 |
print STDERR "WARNING: removing $isis_db.TXT OpenIsis database...\n"; |
470 |
dpavlin |
108 |
unlink $isis_db.".TXT" || warn "FATAL: unlink error on '$isis_db.TXT': $!"; |
471 |
dpavlin |
109 |
$reopen++; |
472 |
dpavlin |
106 |
} |
473 |
dpavlin |
109 |
if (-e $isis_db.".PTR") { |
474 |
|
|
print STDERR "WARNING: removing $isis_db.PTR OpenIsis database...\n"; |
475 |
|
|
unlink $isis_db.".PTR" || warn "FATAL: unlink error on '$isis_db.PTR': $!"; |
476 |
|
|
$reopen++; |
477 |
|
|
} |
478 |
|
|
return OpenIsis::open( $isis_db ) if ($reopen); |
479 |
dpavlin |
106 |
} |
480 |
|
|
|
481 |
|
|
# EOF error |
482 |
|
|
if ($db == -1) { |
483 |
|
|
$db = check_txt_db($isis_db); |
484 |
dpavlin |
108 |
if ($db == -1) { |
485 |
dpavlin |
106 |
print STDERR "FATAL: OpenIsis can't open zero size file $isis_db\n"; |
486 |
|
|
next; |
487 |
|
|
} |
488 |
|
|
} |
489 |
|
|
|
490 |
dpavlin |
101 |
# OpenIsis::ERR_BADF |
491 |
|
|
if ($db == -4) { |
492 |
|
|
print STDERR "FATAL: OpenIsis can't find file $isis_db\n"; |
493 |
|
|
next; |
494 |
|
|
# OpenIsis::ERR_IO |
495 |
|
|
} elsif ($db == -5) { |
496 |
|
|
print STDERR "FATAL: OpenIsis can't access file $isis_db\n"; |
497 |
|
|
next; |
498 |
|
|
} elsif ($db < 0) { |
499 |
|
|
print STDERR "FATAL: OpenIsis unknown error $db with file $isis_db\n"; |
500 |
|
|
next; |
501 |
|
|
} |
502 |
|
|
|
503 |
dpavlin |
54 |
my $max_rowid = OpenIsis::maxRowid( $db ); |
504 |
dpavlin |
3 |
|
505 |
dpavlin |
106 |
# if 0 records, try to rease isis .txt database |
506 |
|
|
if ($max_rowid == 0) { |
507 |
|
|
# force removal of database |
508 |
|
|
$db = check_txt_db($isis_db); |
509 |
|
|
$max_rowid = OpenIsis::maxRowid( $db ); |
510 |
|
|
} |
511 |
|
|
|
512 |
dpavlin |
54 |
print STDERR "Reading database: $isis_db [$max_rowid rows]\n"; |
513 |
|
|
|
514 |
|
|
my $path = $database; |
515 |
|
|
|
516 |
|
|
for (my $row_id = 1; $row_id <= $max_rowid; $row_id++ ) { |
517 |
|
|
my $row = OpenIsis::read( $db, $row_id ); |
518 |
|
|
if ($row && $row->{mfn}) { |
519 |
|
|
|
520 |
|
|
progress($row->{mfn}, $max_rowid); |
521 |
|
|
|
522 |
|
|
my $swishpath = $path."#".int($row->{mfn}); |
523 |
|
|
|
524 |
dpavlin |
59 |
if (my $xml = data2xml($type_base,$row,$add_xml,$cfg,$database)) { |
525 |
dpavlin |
54 |
$xml = $cp2utf->convert($xml); |
526 |
|
|
use bytes; # as opposed to chars |
527 |
|
|
print "Path-Name: $swishpath\n"; |
528 |
|
|
print "Content-Length: ".(length($xml)+1)."\n"; |
529 |
|
|
print "Document-Type: XML\n\n$xml\n"; |
530 |
|
|
} |
531 |
dpavlin |
3 |
} |
532 |
dpavlin |
54 |
} |
533 |
dpavlin |
108 |
# for this to work with current version of OpenIsis (0.9.0) |
534 |
|
|
# you might need my patch from |
535 |
|
|
# http://www.rot13.org/~dpavlin/projects/openisis-0.9.0-perl_close.diff |
536 |
|
|
OpenIsis::close($db); |
537 |
dpavlin |
54 |
print STDERR "\n"; |
538 |
dpavlin |
3 |
|
539 |
dpavlin |
58 |
} elsif ($type_base eq "excel") { |
540 |
dpavlin |
54 |
use Spreadsheet::ParseExcel; |
541 |
|
|
use Spreadsheet::ParseExcel::Utility qw(int2col); |
542 |
|
|
|
543 |
|
|
$import2cp = Text::Iconv->new($config->{excel_codepage},$codepage); |
544 |
|
|
my $excel_file = $cfg -> val($database, 'excel_file') || die "$database doesn't have 'excel_file' defined!"; |
545 |
dpavlin |
43 |
|
546 |
dpavlin |
54 |
my $sheet = x($config->{sheet}) || die "no sheet in $type.xml"; |
547 |
dpavlin |
56 |
my $start_row = x($config->{start_row}) - 1 || die "no start_row in $type.xml"; |
548 |
dpavlin |
54 |
|
549 |
|
|
my $oBook = Spreadsheet::ParseExcel::Workbook->Parse($excel_file) || die "can't open Excel file '$excel_file'"; |
550 |
|
|
|
551 |
|
|
my $sheet_nr = 0; |
552 |
|
|
foreach my $oWks (@{$oBook->{Worksheet}}) { |
553 |
|
|
#print STDERR "-- SHEET $sheet_nr:", $oWks->{Name}, "\n"; |
554 |
|
|
last if ($oWks->{Name} eq $sheet); |
555 |
|
|
$sheet_nr++; |
556 |
|
|
} |
557 |
|
|
|
558 |
|
|
my $oWorksheet = $oBook->{Worksheet}[$sheet_nr]; |
559 |
|
|
|
560 |
|
|
print STDERR "using sheet: ",$oWorksheet->{Name},"\n"; |
561 |
|
|
defined ($oWorksheet) || die "can't find sheet '$sheet' in $excel_file"; |
562 |
|
|
my $end_row = x($config->{end_row}) || $oWorksheet->{MaxRow}; |
563 |
|
|
|
564 |
dpavlin |
56 |
for(my $iR = $start_row ; defined $end_row && $iR <= $end_row ; $iR++) { |
565 |
dpavlin |
54 |
my $row; |
566 |
|
|
for(my $iC = $oWorksheet->{MinCol} ; defined $oWorksheet->{MaxCol} && $iC <= $oWorksheet->{MaxCol} ; $iC++) { |
567 |
|
|
my $cell = $oWorksheet->{Cells}[$iR][$iC]; |
568 |
|
|
if ($cell) { |
569 |
|
|
$row->{int2col($iC)} = $cell->Value; |
570 |
|
|
} |
571 |
|
|
} |
572 |
|
|
|
573 |
|
|
progress($iR, $end_row); |
574 |
|
|
|
575 |
|
|
# print "row[$iR/$end_row] "; |
576 |
|
|
# foreach (keys %{$row}) { |
577 |
|
|
# print "$_: ",$row->{$_},"\t"; |
578 |
|
|
# } |
579 |
|
|
# print "\n"; |
580 |
|
|
|
581 |
|
|
my $swishpath = $database."#".$iR; |
582 |
|
|
|
583 |
|
|
next if (! $row); |
584 |
|
|
|
585 |
dpavlin |
59 |
if (my $xml = data2xml($type_base,$row,$add_xml,$cfg,$database)) { |
586 |
dpavlin |
44 |
$xml = $cp2utf->convert($xml); |
587 |
dpavlin |
35 |
use bytes; # as opposed to chars |
588 |
dpavlin |
43 |
print "Path-Name: $swishpath\n"; |
589 |
dpavlin |
10 |
print "Content-Length: ".(length($xml)+1)."\n"; |
590 |
|
|
print "Document-Type: XML\n\n$xml\n"; |
591 |
dpavlin |
3 |
} |
592 |
dpavlin |
1 |
} |
593 |
dpavlin |
62 |
} elsif ($type_base eq "marc") { |
594 |
dpavlin |
67 |
|
595 |
dpavlin |
62 |
use MARC; |
596 |
|
|
|
597 |
|
|
$import2cp = Text::Iconv->new($config->{marc_codepage},$codepage); |
598 |
|
|
my $marc_file = $cfg -> val($database, 'marc_file') || die "$database doesn't have 'marc_file' defined!"; |
599 |
|
|
|
600 |
|
|
# optional argument is format |
601 |
|
|
my $format = x($config->{format}) || 'usmarc'; |
602 |
|
|
|
603 |
|
|
print STDERR "Reading MARC file '$marc_file'\n"; |
604 |
|
|
|
605 |
|
|
my $marc = new MARC; |
606 |
|
|
my $nr = $marc->openmarc({ |
607 |
|
|
file=>$marc_file, format=>$format |
608 |
|
|
}) || die "Can't open MARC file '$marc_file'"; |
609 |
|
|
|
610 |
|
|
my $i=0; # record nr. |
611 |
|
|
|
612 |
|
|
my $rec; |
613 |
|
|
|
614 |
|
|
while ($marc->nextmarc(1)) { |
615 |
|
|
|
616 |
|
|
# XXX |
617 |
dpavlin |
67 |
fakeprogress($i++); |
618 |
dpavlin |
62 |
|
619 |
|
|
my $swishpath = $database."#".$i; |
620 |
|
|
|
621 |
|
|
if (my $xml = data2xml($type_base,$marc,$add_xml,$cfg,$database)) { |
622 |
|
|
$xml = $cp2utf->convert($xml); |
623 |
|
|
use bytes; # as opposed to chars |
624 |
|
|
print "Path-Name: $swishpath\n"; |
625 |
|
|
print "Content-Length: ".(length($xml)+1)."\n"; |
626 |
|
|
print "Document-Type: XML\n\n$xml\n"; |
627 |
|
|
} |
628 |
|
|
} |
629 |
dpavlin |
67 |
} elsif ($type_base eq "feed") { |
630 |
|
|
|
631 |
|
|
$import2cp = Text::Iconv->new($config->{feed_codepage},$codepage); |
632 |
|
|
my $prog = x($config->{prog}) || die "$database doesn't have 'prog' defined!"; |
633 |
|
|
|
634 |
|
|
print STDERR "Reading feed from program '$prog'\n"; |
635 |
|
|
|
636 |
|
|
open(FEED,"feeds/$prog |") || die "can't start $prog: $!"; |
637 |
|
|
|
638 |
dpavlin |
74 |
my $i=1; # record nr. |
639 |
dpavlin |
67 |
|
640 |
|
|
my $data; |
641 |
|
|
my $line=1; |
642 |
|
|
|
643 |
|
|
while (<FEED>) { |
644 |
|
|
chomp; |
645 |
|
|
|
646 |
|
|
if (/^$/) { |
647 |
dpavlin |
74 |
my $swishpath = $database."#".$i++; |
648 |
dpavlin |
67 |
|
649 |
|
|
if (my $xml = data2xml($type_base,$data,$add_xml,$cfg,$database)) { |
650 |
|
|
$xml = $cp2utf->convert($xml); |
651 |
|
|
use bytes; # as opposed to chars |
652 |
|
|
print "Path-Name: $swishpath\n"; |
653 |
|
|
print "Content-Length: ".(length($xml)+1)."\n"; |
654 |
|
|
print "Document-Type: XML\n\n$xml\n"; |
655 |
|
|
} |
656 |
|
|
$line = 1; |
657 |
|
|
$data = {}; |
658 |
|
|
next; |
659 |
|
|
} |
660 |
|
|
|
661 |
dpavlin |
74 |
$line = $1 if (s/^(\d+):\s*//); |
662 |
dpavlin |
67 |
$data->{$line++} = $_; |
663 |
|
|
|
664 |
dpavlin |
74 |
fakeprogress($i); |
665 |
dpavlin |
67 |
|
666 |
|
|
} |
667 |
dpavlin |
177 |
# close lookup |
668 |
|
|
untie %lhash if (%lhash); |
669 |
dpavlin |
1 |
} |
670 |
|
|
} |
671 |
dpavlin |
3 |
|
672 |
dpavlin |
10 |
# call this to commit index |
673 |
|
|
$index->close; |
674 |
dpavlin |
3 |
|
675 |
|
|
1; |
676 |
|
|
__END__ |
677 |
|
|
########################################################################## |
678 |
|
|
|
679 |
|
|
=head1 NAME |
680 |
|
|
|
681 |
dpavlin |
81 |
all2xml.pl - read various file formats and dump XML for SWISH-E |
682 |
dpavlin |
3 |
|
683 |
|
|
=head1 DESCRIPTION |
684 |
|
|
|
685 |
dpavlin |
81 |
This command will read ISIS data file using OpenIsis perl module, MARC |
686 |
|
|
records using MARC module and optionally Micro$oft Excel files to |
687 |
|
|
create one XML file for usage with I<SWISH-E> indexer. Dispite it's name, |
688 |
|
|
this script B<isn't general xml generator> from isis files (isis allready |
689 |
|
|
has something like that). Output of this script is tailor-made for SWISH-E. |
690 |
dpavlin |
3 |
|
691 |
dpavlin |
81 |
=head1 BUGS |
692 |
|
|
|
693 |
|
|
Documentation is really lacking. However, in true Open Source spirit, source |
694 |
|
|
is best documentation. I even made considerable effort to comment parts |
695 |
|
|
which are not intuitively clear, so... |
696 |
|
|
|
697 |
dpavlin |
3 |
=head1 AUTHOR |
698 |
|
|
|
699 |
|
|
Dobrica Pavlinusic <dpavlin@rot13.org> |
700 |
|
|
|
701 |
|
|
=head1 COPYRIGHT |
702 |
|
|
|
703 |
|
|
GNU Public License (GPL) v2 or later |
704 |
|
|
|
705 |
|
|
=head1 SEE ALSO |
706 |
|
|
|
707 |
|
|
SWISH-E web site at http://www.swish-e.org |
708 |
|
|
|
709 |
|
|
=cut |