8 |
use Text::Unaccent 1.02; # 1.01 won't compile on my platform, |
use Text::Unaccent 1.02; # 1.01 won't compile on my platform, |
9 |
use Text::Iconv; |
use Text::Iconv; |
10 |
use Config::IniFiles; |
use Config::IniFiles; |
11 |
|
use Encode; |
12 |
|
|
13 |
$|=1; |
$|=1; |
14 |
|
|
18 |
|
|
19 |
my $config; |
my $config; |
20 |
|
|
21 |
use index_DBI; # there is no other, right now ;-) |
#use index_DBI; # default DBI module for index |
22 |
|
use index_DBI_cache; # faster DBI module using memory cache |
23 |
my $index = new index_DBI(); # open index |
my $index; |
24 |
|
|
25 |
my %opts; |
my %opts; |
26 |
|
|
32 |
|
|
33 |
getopts('d:m:qs', \%opts); |
getopts('d:m:qs', \%opts); |
34 |
|
|
35 |
my $db_dir; |
my $path; # this is name of database |
|
|
|
|
#die "usage: $0 -d [database_dir] -m [database1,database2] " if (! %opts); |
|
36 |
|
|
37 |
#print Dumper($config->{indexer}); |
Text::Iconv->raise_error(0); # Conversion errors don't raise exceptions |
|
#print "-" x 70,"\n"; |
|
38 |
|
|
39 |
Text::Iconv->raise_error(1); # Conversion errors raise exceptions |
# this is encoding of all files on disk, including import_xml/*.xml file and |
40 |
|
# filter/*.pm files! It will be used to store strings in perl internally! |
41 |
|
my $codepage = 'ISO-8859-2'; |
42 |
|
|
43 |
|
my $utf2cp = Text::Iconv->new('UTF-8',$codepage); |
44 |
|
# this function will convert data from XML files to local encoding |
45 |
|
sub x { |
46 |
|
return $utf2cp->convert($_[0]); |
47 |
|
} |
48 |
|
|
49 |
my $isis_codepage; |
# decode isis/excel or other import codepage |
50 |
my $index_codepage; |
my $import2cp; |
|
my $cludge_codepage = Text::Iconv->new('UTF8','ISO8859-1'); |
|
|
my $xml_codepage; |
|
51 |
|
|
52 |
my $XML_CHARSET = 'UTF8'; |
# outgoing xml must be in UTF-8 |
53 |
|
my $cp2utf = Text::Iconv->new($codepage,'UTF-8'); |
54 |
|
|
55 |
|
# mapping between data type and tag which specify |
56 |
|
# format in XML file |
57 |
|
my %type2tag = ( |
58 |
|
'isis' => 'isis', |
59 |
|
'excel' => 'column', |
60 |
|
'marc' => 'marc', |
61 |
|
); |
62 |
|
|
63 |
sub isis2xml { |
sub data2xml { |
64 |
|
|
65 |
use xmlify; |
use xmlify; |
66 |
|
|
67 |
|
my $type = shift @_; |
68 |
my $row = shift @_; |
my $row = shift @_; |
69 |
my $add_xml = shift @_; |
my $add_xml = shift @_; |
70 |
|
# needed to read values from configuration file |
71 |
|
my $cfg = shift @_; |
72 |
|
my $database = shift @_; |
73 |
|
|
74 |
my $xml; |
my $xml; |
75 |
|
|
90 |
|
|
91 |
foreach my $field (sort by_order keys %{$config->{indexer}}) { |
foreach my $field (sort by_order keys %{$config->{indexer}}) { |
92 |
|
|
93 |
|
$field=x($field); |
94 |
|
|
95 |
$field_usage{$field}++; |
$field_usage{$field}++; |
96 |
|
|
97 |
my $swish_data = ""; |
my $swish_data = ""; |
98 |
my $display_data = ""; |
my $display_data = ""; |
99 |
my $line_delimiter = ""; |
my $line_delimiter; |
100 |
|
|
101 |
my ($swish,$display); |
my ($swish,$display); |
102 |
|
|
103 |
foreach my $x (@{$config->{indexer}->{$field}->{isis}}) { |
my $tag = $type2tag{$type} || die "can't find which tag to use for type $type"; |
104 |
|
foreach my $x (@{$config->{indexer}->{$field}->{$tag}}) { |
105 |
|
|
106 |
my $format = $x->{content}; |
my $format = x($x->{content}); |
107 |
my $delimiter = $x->{delimiter} || ' '; |
my $delimiter = x($x->{delimiter}) || ' '; |
108 |
|
|
109 |
# FIX: this is ugly, UGLY, cludge string is returned |
my $repeat_off = 0; # repeatable offset |
|
# in UTF8 encoding , but as if source charset |
|
|
# is ISO8859-1 and not some other. This break other |
|
|
# character encodings, so we convert it first |
|
|
# back to ISO8859-1 |
|
|
$format = $xml_codepage->convert($format); |
|
|
$delimiter = $xml_codepage->convert($delimiter) if ($delimiter); |
|
|
|
|
|
my $isis_i = 0; # isis repeatable offset |
|
110 |
|
|
111 |
my ($s,$d,$i) = (1,1,0); # swish, display default |
my ($s,$d,$i) = (1,1,0); # swish, display default |
112 |
$s = 0 if (lc($x->{type}) eq "display"); |
$s = 0 if (lc($x->{type}) eq "display"); |
113 |
$d = 0 if (lc($x->{type}) eq "swish"); |
$d = 0 if (lc($x->{type}) eq "swish"); |
114 |
($s,$d,$i) = (0,0,1) if (lc($x->{type}) eq "index"); |
($s,$d,$i) = (0,0,1) if (lc($x->{type}) eq "index"); |
|
#print STDERR "## s: $s d: $d i: $i ## $format ##\n"; |
|
115 |
|
|
116 |
if ($x->{append}) { |
# what will separate last line from this one? |
117 |
|
if ($display_data && $x->{append} && $x->{append} eq "1") { |
118 |
$line_delimiter = ' '; |
$line_delimiter = ' '; |
119 |
} else { |
} elsif ($display_data) { |
120 |
$line_delimiter = '<br/>'; |
$line_delimiter = '<br/>'; |
121 |
} |
} |
122 |
|
|
|
$display_data .= $line_delimiter if ($display_data && $display); |
|
123 |
# init vars so that we go into while... |
# init vars so that we go into while... |
124 |
($swish,$display) = (1,1); |
($swish,$display) = (1,1); |
125 |
|
|
126 |
while ($swish || $display) { |
if ($swish || $display) { |
127 |
($swish,$display) = parse_format($format,$row,$isis_i++); |
($swish,$display) = parse_format($type, $format,$row,$repeat_off++,$import2cp); |
|
#print STDERR "s: $swish\nd: $display\n" if ($swish); |
|
|
|
|
|
#print STDERR "swish: $swish<-- display: $display<--\n"; |
|
|
|
|
128 |
# filter="name" ; filter this field through |
# filter="name" ; filter this field through |
129 |
# filter/[name].pm |
# filter/[name].pm |
130 |
my $filter = $x->{filter}; |
my $filter = $x->{filter}; |
134 |
# type="swish" ; field for swish |
# type="swish" ; field for swish |
135 |
if ($s && $swish) { |
if ($s && $swish) { |
136 |
if ($filter) { |
if ($filter) { |
|
#print STDERR "using filter '$filter'\n"; |
|
137 |
no strict 'refs'; |
no strict 'refs'; |
138 |
$swish_data .= join(" ",&$filter($swish)); |
$swish_data .= join(" ",&$filter($swish)); |
139 |
} else { |
} else { |
143 |
|
|
144 |
# type="display" ; field for display |
# type="display" ; field for display |
145 |
if ($d && $display) { |
if ($d && $display) { |
146 |
|
if ($line_delimiter && $display_data) { |
147 |
|
$display_data .= $line_delimiter; |
148 |
|
undef $line_delimiter; |
149 |
|
} |
150 |
if ($filter) { |
if ($filter) { |
151 |
no strict 'refs'; |
no strict 'refs'; |
152 |
$display_data .= join($delimiter,&$filter($display)); |
$display_data .= join($delimiter,&$filter($display)); |
161 |
|
|
162 |
# type="index" ; insert into index |
# type="index" ; insert into index |
163 |
if ($i && $display) { |
if ($i && $display) { |
164 |
my $index_data = $index_codepage->convert($display) || $display; |
my $index_data = $display; |
165 |
if ($filter) { |
if ($filter) { |
166 |
no strict 'refs'; |
no strict 'refs'; |
167 |
foreach my $d (&$filter($index_data)) { |
foreach my $d (&$filter($index_data)) { |
168 |
$index->insert($field, $d, $db_dir); |
$index->insert($field, $d, $path); |
169 |
} |
} |
170 |
} else { |
} else { |
171 |
$index->insert($field, $index_data, $db_dir); |
$index->insert($field, $index_data, $path); |
172 |
} |
} |
173 |
} |
} |
174 |
} |
} |
175 |
} |
} |
176 |
|
|
177 |
|
# now try to parse variables from configuration file |
178 |
|
foreach my $x (@{$config->{indexer}->{$field}->{'config'}}) { |
179 |
|
|
180 |
#print STDERR "s_d: $swish_data\nd_d: $display_data\n" if ($swish_data); |
my $delimiter = x($x->{delimiter}) || ' '; |
181 |
if ($display_data) { |
my $val = $cfg->val($database, x($x->{content})); |
182 |
|
|
183 |
# remove last <br> |
my ($s,$d,$i) = (1,1,0); # swish, display default |
184 |
$display_data =~ s/$line_delimiter$//; |
$s = 0 if (lc($x->{type}) eq "display"); |
185 |
|
$d = 0 if (lc($x->{type}) eq "swish"); |
186 |
|
($s,$d,$i) = (0,0,1) if (lc($x->{type}) eq "index"); |
187 |
|
|
188 |
$display_data = $isis_codepage->convert($display_data) || die "Can't convert '$display_data' !"; |
if ($val) { |
189 |
# FIX: this is removed and replaced by html tag. |
$display_data .= $delimiter.$val if ($d); |
190 |
#$xml .= xmlify($field."_display", $display_data); |
$swish_data .= $val if ($s); |
191 |
|
$index->insert($field, $val, $path) if ($i); |
192 |
|
} |
193 |
|
|
194 |
|
} |
195 |
|
|
196 |
|
|
197 |
|
if ($display_data) { |
198 |
|
|
199 |
if ($field eq "headline") { |
if ($field eq "headline") { |
200 |
$xml .= xmlify("headline", $display_data); |
$xml .= xmlify("headline", $display_data); |
212 |
print STDERR "WARNING: field '$field' doesn't have 'name' attribute!"; |
print STDERR "WARNING: field '$field' doesn't have 'name' attribute!"; |
213 |
} |
} |
214 |
if ($field_name) { |
if ($field_name) { |
215 |
$html .= $xml_codepage->convert($field_name); |
$html .= x($field_name); |
|
# $html .= "-->".$field_name."<--"; |
|
216 |
} |
} |
217 |
$html .= $display_data."###\n"; |
$html .= $display_data."###\n"; |
218 |
} |
} |
219 |
} |
} |
220 |
if ($swish_data) { |
if ($swish_data) { |
|
my $i = Text::Iconv->new($config->{isis_codepage},'ISO8859-2'); |
|
221 |
# remove extra spaces |
# remove extra spaces |
222 |
$swish_data =~ s/ +/ /g; |
$swish_data =~ s/ +/ /g; |
223 |
$swish_data =~ s/ +$//g; |
$swish_data =~ s/ +$//g; |
224 |
|
|
225 |
$swish_data = $i->convert($swish_data); |
$xml .= xmlify($field."_swish", unac_string($codepage,$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},$swish_data)); |
|
226 |
} |
} |
227 |
|
|
228 |
|
|
234 |
} |
} |
235 |
|
|
236 |
if ($xml) { |
if ($xml) { |
|
#print STDERR "x: $xml\n"; |
|
237 |
$xml .= $add_xml if ($add_xml); |
$xml .= $add_xml if ($add_xml); |
238 |
return "<xml>\n$xml</xml>\n"; |
return "<xml>\n$xml</xml>\n"; |
239 |
} else { |
} else { |
243 |
|
|
244 |
########################################################################## |
########################################################################## |
245 |
|
|
246 |
|
# read configuration for this script |
247 |
my $cfg = new Config::IniFiles( -file => $config_file ); |
my $cfg = new Config::IniFiles( -file => $config_file ); |
248 |
|
|
249 |
|
# read global.conf configuration |
250 |
|
my $cfg_global = new Config::IniFiles( -file => 'global.conf' ); |
251 |
|
|
252 |
|
# open index |
253 |
|
$index = new index_DBI( |
254 |
|
$cfg_global->val('global', 'dbi_dbd'), |
255 |
|
$cfg_global->val('global', 'dbi_dsn'), |
256 |
|
$cfg_global->val('global', 'dbi_user'), |
257 |
|
$cfg_global->val('global', 'dbi_passwd') || '', |
258 |
|
); |
259 |
|
|
260 |
foreach my $database ($cfg->Sections) { |
foreach my $database ($cfg->Sections) { |
261 |
|
|
262 |
my $isis_db = $cfg -> val($database, 'isis_db'); |
my $type = lc($cfg -> val($database, 'type')) || die "$database doesn't have 'type' defined"; |
263 |
my $type = $cfg -> val($database, 'type'); |
my $add_xml = $cfg -> val($database, 'xml'); # optional |
|
my $add_xml = $cfg -> val($database, 'xml'); |
|
|
|
|
|
# read configuration for this type |
|
|
$config=XMLin("./import_xml/$type.xml", forcearray => [ 'isis' ], forcecontent => 1); |
|
|
$isis_codepage = Text::Iconv->new($config->{isis_codepage},$XML_CHARSET); |
|
|
$index_codepage = Text::Iconv->new($config->{isis_codepage},$config->{index_codepage}); |
|
|
$xml_codepage = Text::Iconv->new($cfg->val($database,'xml_codepage'),$XML_CHARSET); |
|
|
|
|
|
my $db = OpenIsis::open( $isis_db ); |
|
|
if (0) { |
|
|
# # FIX |
|
|
# if (! $db ) { |
|
|
print STDERR "WARNING: can't open '$isis_db'\n"; |
|
|
next ; |
|
|
} |
|
264 |
|
|
265 |
my $max_rowid = OpenIsis::maxRowid( $db ); |
print STDERR "reading ./import_xml/$type.xml\n"; |
266 |
|
|
267 |
print STDERR "Reading database: $isis_db [$max_rowid rows]\n"; |
# extract just type basic |
268 |
|
my $type_base = $type; |
269 |
|
$type_base =~ s/_.+$//g; |
270 |
|
|
271 |
my $path = $database; # was $isis_db |
$config=XMLin("./import_xml/$type.xml", forcearray => [ $type2tag{$type_base}, 'config' ], forcecontent => 1); |
272 |
|
|
273 |
|
# output current progress indicator |
274 |
my $last_p = 0; |
my $last_p = 0; |
275 |
|
sub progress { |
276 |
|
#return if (! $opts{q}); # FIXME |
277 |
|
my $current = shift; |
278 |
|
my $total = shift || 1; |
279 |
|
my $p = int($current * 100 / $total); |
280 |
|
if ($p != $last_p) { |
281 |
|
printf STDERR ("%5d / %5d [%-51s] %-2d %% \r",$current,$total,"=" x ($p/2).">", $p ); |
282 |
|
$last_p = $p; |
283 |
|
} |
284 |
|
} |
285 |
|
|
286 |
|
# now read database |
287 |
|
print STDERR "using: $type...\n"; |
288 |
|
|
289 |
|
if ($type_base eq "isis") { |
290 |
|
|
291 |
|
my $isis_db = $cfg -> val($database, 'isis_db') || die "$database doesn't have 'isis_db' defined!"; |
292 |
|
|
293 |
|
$import2cp = Text::Iconv->new($config->{isis_codepage},$codepage); |
294 |
|
my $db = OpenIsis::open( $isis_db ); |
295 |
|
|
296 |
# { my $row_id = 4514; |
my $max_rowid = OpenIsis::maxRowid( $db ); |
297 |
# FIX |
|
298 |
for (my $row_id = 1; $row_id <= $max_rowid; $row_id++ ) { |
print STDERR "Reading database: $isis_db [$max_rowid rows]\n"; |
299 |
my $row = OpenIsis::read( $db, $row_id ); |
|
300 |
if ($row && $row->{mfn}) { |
my $path = $database; |
301 |
#print STDERR "mfn: ",$row->{mfn},"\n"; |
|
302 |
# output current process indicator |
for (my $row_id = 1; $row_id <= $max_rowid; $row_id++ ) { |
303 |
my $p = int($row->{mfn} * 100 / $max_rowid); |
my $row = OpenIsis::read( $db, $row_id ); |
304 |
if ($p != $last_p) { |
if ($row && $row->{mfn}) { |
305 |
printf STDERR ("%5d / %5d [%-51s] %-2d %% \r",$row->{mfn},$max_rowid,"=" x ($p/2).">", $p ) if (! $opts{q}); |
|
306 |
$last_p = $p; |
progress($row->{mfn}, $max_rowid); |
307 |
|
|
308 |
|
my $swishpath = $path."#".int($row->{mfn}); |
309 |
|
|
310 |
|
if (my $xml = data2xml($type_base,$row,$add_xml,$cfg,$database)) { |
311 |
|
$xml = $cp2utf->convert($xml); |
312 |
|
use bytes; # as opposed to chars |
313 |
|
print "Path-Name: $swishpath\n"; |
314 |
|
print "Content-Length: ".(length($xml)+1)."\n"; |
315 |
|
print "Document-Type: XML\n\n$xml\n"; |
316 |
|
} |
317 |
} |
} |
318 |
|
} |
319 |
|
print STDERR "\n"; |
320 |
|
|
321 |
|
} elsif ($type_base eq "excel") { |
322 |
|
use Spreadsheet::ParseExcel; |
323 |
|
use Spreadsheet::ParseExcel::Utility qw(int2col); |
324 |
|
|
325 |
|
$import2cp = Text::Iconv->new($config->{excel_codepage},$codepage); |
326 |
|
my $excel_file = $cfg -> val($database, 'excel_file') || die "$database doesn't have 'excel_file' defined!"; |
327 |
|
|
328 |
|
my $sheet = x($config->{sheet}) || die "no sheet in $type.xml"; |
329 |
|
my $start_row = x($config->{start_row}) - 1 || die "no start_row in $type.xml"; |
330 |
|
|
331 |
|
my $oBook = Spreadsheet::ParseExcel::Workbook->Parse($excel_file) || die "can't open Excel file '$excel_file'"; |
332 |
|
|
333 |
|
my $sheet_nr = 0; |
334 |
|
foreach my $oWks (@{$oBook->{Worksheet}}) { |
335 |
|
#print STDERR "-- SHEET $sheet_nr:", $oWks->{Name}, "\n"; |
336 |
|
last if ($oWks->{Name} eq $sheet); |
337 |
|
$sheet_nr++; |
338 |
|
} |
339 |
|
|
340 |
|
my $oWorksheet = $oBook->{Worksheet}[$sheet_nr]; |
341 |
|
|
342 |
|
print STDERR "using sheet: ",$oWorksheet->{Name},"\n"; |
343 |
|
defined ($oWorksheet) || die "can't find sheet '$sheet' in $excel_file"; |
344 |
|
my $end_row = x($config->{end_row}) || $oWorksheet->{MaxRow}; |
345 |
|
|
346 |
|
for(my $iR = $start_row ; defined $end_row && $iR <= $end_row ; $iR++) { |
347 |
|
my $row; |
348 |
|
for(my $iC = $oWorksheet->{MinCol} ; defined $oWorksheet->{MaxCol} && $iC <= $oWorksheet->{MaxCol} ; $iC++) { |
349 |
|
my $cell = $oWorksheet->{Cells}[$iR][$iC]; |
350 |
|
if ($cell) { |
351 |
|
$row->{int2col($iC)} = $cell->Value; |
352 |
|
} |
353 |
|
} |
354 |
|
|
355 |
|
progress($iR, $end_row); |
356 |
|
|
357 |
|
# print "row[$iR/$end_row] "; |
358 |
|
# foreach (keys %{$row}) { |
359 |
|
# print "$_: ",$row->{$_},"\t"; |
360 |
|
# } |
361 |
|
# print "\n"; |
362 |
|
|
363 |
|
my $swishpath = $database."#".$iR; |
364 |
|
|
365 |
|
next if (! $row); |
366 |
|
|
367 |
|
if (my $xml = data2xml($type_base,$row,$add_xml,$cfg,$database)) { |
368 |
|
$xml = $cp2utf->convert($xml); |
369 |
|
use bytes; # as opposed to chars |
370 |
|
print "Path-Name: $swishpath\n"; |
371 |
|
print "Content-Length: ".(length($xml)+1)."\n"; |
372 |
|
print "Document-Type: XML\n\n$xml\n"; |
373 |
|
} |
374 |
|
} |
375 |
|
} elsif ($type_base eq "marc") { |
376 |
|
## XXX |
377 |
|
use MARC; |
378 |
|
|
379 |
|
$import2cp = Text::Iconv->new($config->{marc_codepage},$codepage); |
380 |
|
my $marc_file = $cfg -> val($database, 'marc_file') || die "$database doesn't have 'marc_file' defined!"; |
381 |
|
|
382 |
|
# optional argument is format |
383 |
|
my $format = x($config->{format}) || 'usmarc'; |
384 |
|
|
385 |
|
my %id_stored; # to aviod duplicates |
386 |
|
|
387 |
|
print STDERR "Reading MARC file '$marc_file'\n"; |
388 |
|
|
389 |
|
my $marc = new MARC; |
390 |
|
my $nr = $marc->openmarc({ |
391 |
|
file=>$marc_file, format=>$format |
392 |
|
}) || die "Can't open MARC file '$marc_file'"; |
393 |
|
|
394 |
|
my $i=0; # record nr. |
395 |
|
my $inc=1; |
396 |
|
my $max_i=1000; |
397 |
|
|
398 |
|
my $rec; |
399 |
|
|
400 |
|
while ($marc->nextmarc(1)) { |
401 |
|
|
402 |
|
# XXX |
403 |
|
progress($i, $max_i); |
404 |
|
$i += $inc; |
405 |
|
$inc = -$inc if ($i > $max_i || $i < 0); |
406 |
|
|
407 |
|
my $swishpath = $database."#".$i; |
408 |
|
|
409 |
if (my $xml = isis2xml($row,$add_xml)) { |
if (my $xml = data2xml($type_base,$marc,$add_xml,$cfg,$database)) { |
410 |
#print STDERR "--ret-->$xml\n"; |
$xml = $cp2utf->convert($xml); |
411 |
print "Path-Name: $path#".int($row->{mfn})."\n"; |
use bytes; # as opposed to chars |
412 |
|
print "Path-Name: $swishpath\n"; |
413 |
print "Content-Length: ".(length($xml)+1)."\n"; |
print "Content-Length: ".(length($xml)+1)."\n"; |
414 |
print "Document-Type: XML\n\n$xml\n"; |
print "Document-Type: XML\n\n$xml\n"; |
415 |
} |
} |
416 |
} |
} |
417 |
} |
} |
|
print STDERR "\n"; |
|
418 |
} |
} |
419 |
|
|
420 |
# call this to commit index |
# call this to commit index |