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 |
|
|
34 |
|
|
35 |
my $path; # this is name of database |
my $path; # this is name of database |
36 |
|
|
37 |
Text::Iconv->raise_error(1); # Conversion errors raise exceptions |
Text::Iconv->raise_error(0); # Conversion errors don't raise exceptions |
38 |
|
|
39 |
# this is encoding of all files on disk, including import_xml/*.xml file and |
# 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! |
# filter/*.pm files! It will be used to store strings in perl internally! |
46 |
return $utf2cp->convert($_[0]); |
return $utf2cp->convert($_[0]); |
47 |
} |
} |
48 |
|
|
49 |
# decode isis import codepage |
# decode isis/excel or other import codepage |
50 |
my $isis2cp; |
my $import2cp; |
51 |
|
|
52 |
# outgoing xml must be in UTF-8 |
# outgoing xml must be in UTF-8 |
53 |
my $cp2utf = Text::Iconv->new($codepage,'UTF-8'); |
my $cp2utf = Text::Iconv->new($codepage,'UTF-8'); |
54 |
|
|
55 |
sub isis2xml { |
# 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 |
|
'feed' => 'feed' |
62 |
|
); |
63 |
|
|
64 |
|
sub data2xml { |
65 |
|
|
66 |
use xmlify; |
use xmlify; |
67 |
|
|
68 |
|
my $type = shift @_; |
69 |
my $row = shift @_; |
my $row = shift @_; |
70 |
my $add_xml = shift @_; |
my $add_xml = shift @_; |
71 |
|
# needed to read values from configuration file |
72 |
|
my $cfg = shift @_; |
73 |
|
my $database = shift @_; |
74 |
|
|
75 |
my $xml; |
my $xml; |
76 |
|
|
82 |
|
|
83 |
# sort subrouting using order="" attribute |
# sort subrouting using order="" attribute |
84 |
sub by_order { |
sub by_order { |
85 |
return 0 if (! $config->{indexer}->{$a}->{order}); |
my $va = $config->{indexer}->{$a}->{order} || |
86 |
return 0 if (! $config->{indexer}->{$b}->{order}); |
$config->{indexer}->{$a}; |
87 |
|
my $vb = $config->{indexer}->{$b}->{order} || |
88 |
|
$config->{indexer}->{$b}; |
89 |
|
|
90 |
return $config->{indexer}->{$a}->{order} <=> |
return $va <=> $vb; |
|
$config->{indexer}->{$b}->{order} ; |
|
91 |
} |
} |
92 |
|
|
93 |
foreach my $field (sort by_order keys %{$config->{indexer}}) { |
foreach my $field (sort by_order keys %{$config->{indexer}}) { |
94 |
|
|
95 |
$field=x($field); |
$field=x($field); |
|
|
|
96 |
$field_usage{$field}++; |
$field_usage{$field}++; |
97 |
|
|
98 |
my $swish_data = ""; |
my $swish_data = ""; |
101 |
|
|
102 |
my ($swish,$display); |
my ($swish,$display); |
103 |
|
|
104 |
foreach my $x (@{$config->{indexer}->{$field}->{isis}}) { |
my $tag = $type2tag{$type} || die "can't find which tag to use for type $type"; |
105 |
|
foreach my $x (@{$config->{indexer}->{$field}->{$tag}}) { |
106 |
|
|
107 |
my $format = x($x->{content}); |
my $format = x($x->{content}); |
108 |
my $delimiter = x($x->{delimiter}) || ' '; |
my $delimiter = x($x->{delimiter}) || ' '; |
109 |
|
|
110 |
my $isis_i = 0; # isis repeatable offset |
my $repeat_off = 0; # repeatable offset |
111 |
|
|
112 |
my ($s,$d,$i) = (1,1,0); # swish, display default |
my ($s,$d,$i) = (1,1,0); # swish, display default |
113 |
$s = 0 if (lc($x->{type}) eq "display"); |
$s = 0 if (lc($x->{type}) eq "display"); |
124 |
# init vars so that we go into while... |
# init vars so that we go into while... |
125 |
($swish,$display) = (1,1); |
($swish,$display) = (1,1); |
126 |
|
|
127 |
while ($swish || $display) { |
# placeholder for all repeatable entries for index |
128 |
($swish,$display) = parse_format($format,$row,$isis_i++,$isis2cp); |
my @index_data; |
129 |
|
my $index_filter; |
130 |
|
|
131 |
|
# while because of repeatable fields |
132 |
|
while ($swish || $display) { |
133 |
|
($swish,$display) = parse_format($type, $format,$row,$repeat_off++,$import2cp); |
134 |
|
if ($repeat_off > 1000) { |
135 |
|
print STDERR "loop (more than 1000 repeatable fields) deteced in $row, $format\n"; |
136 |
|
last; |
137 |
|
} |
138 |
|
|
139 |
# filter="name" ; filter this field through |
# filter="name" ; filter this field through |
140 |
# filter/[name].pm |
# filter/[name].pm |
141 |
my $filter = $x->{filter}; |
my $filter = $x->{filter}; |
172 |
|
|
173 |
# type="index" ; insert into index |
# type="index" ; insert into index |
174 |
if ($i && $display) { |
if ($i && $display) { |
175 |
my $index_data = $display; |
push @index_data, $display; |
176 |
if ($filter) { |
$index_filter = $filter if ($filter); |
177 |
no strict 'refs'; |
} |
178 |
foreach my $d (&$filter($index_data)) { |
} |
179 |
$index->insert($field, $d, $path); |
|
180 |
} |
# fill data in index |
181 |
} else { |
if (@index_data) { |
182 |
$index->insert($field, $index_data, $path); |
if ($index_filter) { |
183 |
|
no strict 'refs'; |
184 |
|
foreach my $d (&$index_filter(@index_data)) { |
185 |
|
$index->insert($field, $d, $path); |
186 |
|
} |
187 |
|
} else { |
188 |
|
foreach my $d (@index_data) { |
189 |
|
$index->insert($field, $d, $path); |
190 |
} |
} |
191 |
} |
} |
192 |
} |
} |
193 |
} |
} |
194 |
|
|
195 |
|
# now try to parse variables from configuration file |
196 |
|
foreach my $x (@{$config->{indexer}->{$field}->{'config'}}) { |
197 |
|
|
198 |
|
my $delimiter = x($x->{delimiter}) || ' '; |
199 |
|
my $val = $cfg->val($database, x($x->{content})); |
200 |
|
|
201 |
|
my ($s,$d,$i) = (1,1,0); # swish, display default |
202 |
|
$s = 0 if (lc($x->{type}) eq "display"); |
203 |
|
$d = 0 if (lc($x->{type}) eq "swish"); |
204 |
|
($s,$d,$i) = (0,0,1) if (lc($x->{type}) eq "index"); |
205 |
|
|
206 |
|
if ($val) { |
207 |
|
$display_data .= $delimiter.$val if ($d); |
208 |
|
$swish_data .= $val if ($s); |
209 |
|
$index->insert($field, $val, $path) if ($i); |
210 |
|
} |
211 |
|
|
212 |
|
} |
213 |
|
|
214 |
|
|
215 |
if ($display_data) { |
if ($display_data) { |
216 |
|
|
248 |
|
|
249 |
# dump formatted output in <html> |
# dump formatted output in <html> |
250 |
if ($html) { |
if ($html) { |
251 |
$xml .= xmlify("html",$html); |
#$xml .= xmlify("html",$html); |
252 |
|
$xml .= "<html><![CDATA[ $html ]]></html>"; |
253 |
} |
} |
254 |
|
|
255 |
if ($xml) { |
if ($xml) { |
262 |
|
|
263 |
########################################################################## |
########################################################################## |
264 |
|
|
265 |
|
# read configuration for this script |
266 |
my $cfg = new Config::IniFiles( -file => $config_file ); |
my $cfg = new Config::IniFiles( -file => $config_file ); |
267 |
|
|
268 |
|
# read global.conf configuration |
269 |
|
my $cfg_global = new Config::IniFiles( -file => 'global.conf' ); |
270 |
|
|
271 |
|
# open index |
272 |
|
$index = new index_DBI( |
273 |
|
$cfg_global->val('global', 'dbi_dbd'), |
274 |
|
$cfg_global->val('global', 'dbi_dsn'), |
275 |
|
$cfg_global->val('global', 'dbi_user'), |
276 |
|
$cfg_global->val('global', 'dbi_passwd') || '', |
277 |
|
); |
278 |
|
|
279 |
|
my $show_progress = $cfg_global->val('global', 'show_progress'); |
280 |
|
|
281 |
foreach my $database ($cfg->Sections) { |
foreach my $database ($cfg->Sections) { |
282 |
|
|
283 |
my $isis_db = $cfg -> val($database, 'isis_db') || die "$database doesn't have 'isis_db' defined!"; |
my $type = lc($cfg -> val($database, 'type')) || die "$database doesn't have 'type' defined"; |
|
my $type = $cfg -> val($database, 'type') || die "$database doesn't have 'type' defined"; |
|
284 |
my $add_xml = $cfg -> val($database, 'xml'); # optional |
my $add_xml = $cfg -> val($database, 'xml'); # optional |
285 |
|
|
286 |
$config=XMLin("./import_xml/$type.xml", forcearray => [ 'isis' ], forcecontent => 1); |
print STDERR "reading ./import_xml/$type.xml\n"; |
287 |
|
|
288 |
$isis2cp = Text::Iconv->new($config->{isis_codepage},$codepage); |
# extract just type basic |
289 |
|
my $type_base = $type; |
290 |
|
$type_base =~ s/_.+$//g; |
291 |
|
|
292 |
my $db = OpenIsis::open( $isis_db ); |
$config=XMLin("./import_xml/$type.xml", forcearray => [ $type2tag{$type_base}, 'config' ], forcecontent => 1); |
293 |
if (0) { |
|
294 |
# # FIX |
# output current progress indicator |
295 |
# if (! $db ) { |
my $last_p = 0; |
296 |
print STDERR "WARNING: can't open '$isis_db'\n"; |
sub progress { |
297 |
next ; |
return if (! $show_progress); |
298 |
|
my $current = shift; |
299 |
|
my $total = shift || 1; |
300 |
|
my $p = int($current * 100 / $total); |
301 |
|
if ($p != $last_p) { |
302 |
|
printf STDERR ("%5d / %5d [%-51s] %-2d %% \r",$current,$total,"=" x ($p/2).">", $p ); |
303 |
|
$last_p = $p; |
304 |
|
} |
305 |
} |
} |
306 |
|
|
307 |
my $max_rowid = OpenIsis::maxRowid( $db ); |
my $fake_dir = 1; |
308 |
|
sub fakeprogress { |
309 |
|
return if (! $show_progress); |
310 |
|
my $current = shift @_; |
311 |
|
|
312 |
|
my @ind = ('-','\\','|','/','-','\\','|','/', '-'); |
313 |
|
|
314 |
|
$last_p += $fake_dir; |
315 |
|
$fake_dir = -$fake_dir if ($last_p > 1000 || $last_p < 0); |
316 |
|
if ($last_p % 10 == 0) { |
317 |
|
printf STDERR ("%5d / %5s [%-51s]\r",$current,"?"," " x ($last_p/20).$ind[($last_p/20) % $#ind]); |
318 |
|
} |
319 |
|
} |
320 |
|
|
321 |
print STDERR "Reading database: $isis_db [$max_rowid rows]\n"; |
# now read database |
322 |
|
print STDERR "using: $type...\n"; |
323 |
|
|
324 |
my $path = $database; # was $isis_db |
if ($type_base eq "isis") { |
325 |
|
|
326 |
my $last_p = 0; |
my $isis_db = $cfg -> val($database, 'isis_db') || die "$database doesn't have 'isis_db' defined!"; |
327 |
|
|
328 |
# { my $row_id = 4514; |
$import2cp = Text::Iconv->new($config->{isis_codepage},$codepage); |
329 |
# FIX |
my $db = OpenIsis::open( $isis_db ); |
330 |
for (my $row_id = 1; $row_id <= $max_rowid; $row_id++ ) { |
|
331 |
my $row = OpenIsis::read( $db, $row_id ); |
# check if .txt database for OpenIsis is zero length, |
332 |
if ($row && $row->{mfn}) { |
# if so, erase it and re-open database |
333 |
# output current process indicator |
sub check_txt_db { |
334 |
my $p = int($row->{mfn} * 100 / $max_rowid); |
my $isis_db = shift || die "need isis database name"; |
335 |
if ($p != $last_p) { |
if (-e $isis_db.".TXT") { |
336 |
printf STDERR ("%5d / %5d [%-51s] %-2d %% \r",$row->{mfn},$max_rowid,"=" x ($p/2).">", $p ) if (! $opts{q}); |
print STDERR "WARNING: removing .txt OpenIsis database...\n"; |
337 |
$last_p = $p; |
unlink $isis_db.".TXT" || warn "unlink error on '$isis_db.TXT': $!"; |
338 |
|
my $db = OpenIsis::open( $isis_db ); |
339 |
|
return $db; |
340 |
} |
} |
341 |
|
} |
342 |
|
|
343 |
|
# EOF error |
344 |
|
if ($db == -1) { |
345 |
|
$db = check_txt_db($isis_db); |
346 |
|
if (! $db) { |
347 |
|
print STDERR "FATAL: OpenIsis can't open zero size file $isis_db\n"; |
348 |
|
next; |
349 |
|
} |
350 |
|
} |
351 |
|
|
352 |
|
# OpenIsis::ERR_BADF |
353 |
|
if ($db == -4) { |
354 |
|
print STDERR "FATAL: OpenIsis can't find file $isis_db\n"; |
355 |
|
next; |
356 |
|
# OpenIsis::ERR_IO |
357 |
|
} elsif ($db == -5) { |
358 |
|
print STDERR "FATAL: OpenIsis can't access file $isis_db\n"; |
359 |
|
next; |
360 |
|
} elsif ($db < 0) { |
361 |
|
print STDERR "FATAL: OpenIsis unknown error $db with file $isis_db\n"; |
362 |
|
next; |
363 |
|
} |
364 |
|
|
365 |
my $swishpath = $path."#".int($row->{mfn}); |
my $max_rowid = OpenIsis::maxRowid( $db ); |
366 |
|
|
367 |
if (my $xml = $cp2utf->convert(isis2xml($row,$add_xml."<path>$swishpath</path>"))) { |
# if 0 records, try to rease isis .txt database |
368 |
|
if ($max_rowid == 0) { |
369 |
|
# force removal of database |
370 |
|
$db = check_txt_db($isis_db); |
371 |
|
$max_rowid = OpenIsis::maxRowid( $db ); |
372 |
|
} |
373 |
|
|
374 |
|
print STDERR "Reading database: $isis_db [$max_rowid rows]\n"; |
375 |
|
|
376 |
|
my $path = $database; |
377 |
|
|
378 |
|
for (my $row_id = 1; $row_id <= $max_rowid; $row_id++ ) { |
379 |
|
my $row = OpenIsis::read( $db, $row_id ); |
380 |
|
if ($row && $row->{mfn}) { |
381 |
|
|
382 |
|
progress($row->{mfn}, $max_rowid); |
383 |
|
|
384 |
|
my $swishpath = $path."#".int($row->{mfn}); |
385 |
|
|
386 |
|
if (my $xml = data2xml($type_base,$row,$add_xml,$cfg,$database)) { |
387 |
|
$xml = $cp2utf->convert($xml); |
388 |
|
use bytes; # as opposed to chars |
389 |
|
print "Path-Name: $swishpath\n"; |
390 |
|
print "Content-Length: ".(length($xml)+1)."\n"; |
391 |
|
print "Document-Type: XML\n\n$xml\n"; |
392 |
|
} |
393 |
|
} |
394 |
|
} |
395 |
|
print STDERR "\n"; |
396 |
|
|
397 |
|
} elsif ($type_base eq "excel") { |
398 |
|
use Spreadsheet::ParseExcel; |
399 |
|
use Spreadsheet::ParseExcel::Utility qw(int2col); |
400 |
|
|
401 |
|
$import2cp = Text::Iconv->new($config->{excel_codepage},$codepage); |
402 |
|
my $excel_file = $cfg -> val($database, 'excel_file') || die "$database doesn't have 'excel_file' defined!"; |
403 |
|
|
404 |
|
my $sheet = x($config->{sheet}) || die "no sheet in $type.xml"; |
405 |
|
my $start_row = x($config->{start_row}) - 1 || die "no start_row in $type.xml"; |
406 |
|
|
407 |
|
my $oBook = Spreadsheet::ParseExcel::Workbook->Parse($excel_file) || die "can't open Excel file '$excel_file'"; |
408 |
|
|
409 |
|
my $sheet_nr = 0; |
410 |
|
foreach my $oWks (@{$oBook->{Worksheet}}) { |
411 |
|
#print STDERR "-- SHEET $sheet_nr:", $oWks->{Name}, "\n"; |
412 |
|
last if ($oWks->{Name} eq $sheet); |
413 |
|
$sheet_nr++; |
414 |
|
} |
415 |
|
|
416 |
|
my $oWorksheet = $oBook->{Worksheet}[$sheet_nr]; |
417 |
|
|
418 |
|
print STDERR "using sheet: ",$oWorksheet->{Name},"\n"; |
419 |
|
defined ($oWorksheet) || die "can't find sheet '$sheet' in $excel_file"; |
420 |
|
my $end_row = x($config->{end_row}) || $oWorksheet->{MaxRow}; |
421 |
|
|
422 |
|
for(my $iR = $start_row ; defined $end_row && $iR <= $end_row ; $iR++) { |
423 |
|
my $row; |
424 |
|
for(my $iC = $oWorksheet->{MinCol} ; defined $oWorksheet->{MaxCol} && $iC <= $oWorksheet->{MaxCol} ; $iC++) { |
425 |
|
my $cell = $oWorksheet->{Cells}[$iR][$iC]; |
426 |
|
if ($cell) { |
427 |
|
$row->{int2col($iC)} = $cell->Value; |
428 |
|
} |
429 |
|
} |
430 |
|
|
431 |
|
progress($iR, $end_row); |
432 |
|
|
433 |
|
# print "row[$iR/$end_row] "; |
434 |
|
# foreach (keys %{$row}) { |
435 |
|
# print "$_: ",$row->{$_},"\t"; |
436 |
|
# } |
437 |
|
# print "\n"; |
438 |
|
|
439 |
|
my $swishpath = $database."#".$iR; |
440 |
|
|
441 |
|
next if (! $row); |
442 |
|
|
443 |
|
if (my $xml = data2xml($type_base,$row,$add_xml,$cfg,$database)) { |
444 |
|
$xml = $cp2utf->convert($xml); |
445 |
|
use bytes; # as opposed to chars |
446 |
|
print "Path-Name: $swishpath\n"; |
447 |
|
print "Content-Length: ".(length($xml)+1)."\n"; |
448 |
|
print "Document-Type: XML\n\n$xml\n"; |
449 |
|
} |
450 |
|
} |
451 |
|
} elsif ($type_base eq "marc") { |
452 |
|
|
453 |
|
use MARC; |
454 |
|
|
455 |
|
$import2cp = Text::Iconv->new($config->{marc_codepage},$codepage); |
456 |
|
my $marc_file = $cfg -> val($database, 'marc_file') || die "$database doesn't have 'marc_file' defined!"; |
457 |
|
|
458 |
|
# optional argument is format |
459 |
|
my $format = x($config->{format}) || 'usmarc'; |
460 |
|
|
461 |
|
print STDERR "Reading MARC file '$marc_file'\n"; |
462 |
|
|
463 |
|
my $marc = new MARC; |
464 |
|
my $nr = $marc->openmarc({ |
465 |
|
file=>$marc_file, format=>$format |
466 |
|
}) || die "Can't open MARC file '$marc_file'"; |
467 |
|
|
468 |
|
my $i=0; # record nr. |
469 |
|
|
470 |
|
my $rec; |
471 |
|
|
472 |
|
while ($marc->nextmarc(1)) { |
473 |
|
|
474 |
|
# XXX |
475 |
|
fakeprogress($i++); |
476 |
|
|
477 |
|
my $swishpath = $database."#".$i; |
478 |
|
|
479 |
|
if (my $xml = data2xml($type_base,$marc,$add_xml,$cfg,$database)) { |
480 |
|
$xml = $cp2utf->convert($xml); |
481 |
use bytes; # as opposed to chars |
use bytes; # as opposed to chars |
482 |
print "Path-Name: $swishpath\n"; |
print "Path-Name: $swishpath\n"; |
483 |
print "Content-Length: ".(length($xml)+1)."\n"; |
print "Content-Length: ".(length($xml)+1)."\n"; |
484 |
print "Document-Type: XML\n\n$xml\n"; |
print "Document-Type: XML\n\n$xml\n"; |
485 |
} |
} |
486 |
} |
} |
487 |
|
} elsif ($type_base eq "feed") { |
488 |
|
|
489 |
|
$import2cp = Text::Iconv->new($config->{feed_codepage},$codepage); |
490 |
|
my $prog = x($config->{prog}) || die "$database doesn't have 'prog' defined!"; |
491 |
|
|
492 |
|
print STDERR "Reading feed from program '$prog'\n"; |
493 |
|
|
494 |
|
open(FEED,"feeds/$prog |") || die "can't start $prog: $!"; |
495 |
|
|
496 |
|
my $i=1; # record nr. |
497 |
|
|
498 |
|
my $data; |
499 |
|
my $line=1; |
500 |
|
|
501 |
|
while (<FEED>) { |
502 |
|
chomp; |
503 |
|
|
504 |
|
if (/^$/) { |
505 |
|
my $swishpath = $database."#".$i++; |
506 |
|
|
507 |
|
if (my $xml = data2xml($type_base,$data,$add_xml,$cfg,$database)) { |
508 |
|
$xml = $cp2utf->convert($xml); |
509 |
|
use bytes; # as opposed to chars |
510 |
|
print "Path-Name: $swishpath\n"; |
511 |
|
print "Content-Length: ".(length($xml)+1)."\n"; |
512 |
|
print "Document-Type: XML\n\n$xml\n"; |
513 |
|
} |
514 |
|
$line = 1; |
515 |
|
$data = {}; |
516 |
|
next; |
517 |
|
} |
518 |
|
|
519 |
|
$line = $1 if (s/^(\d+):\s*//); |
520 |
|
$data->{$line++} = $_; |
521 |
|
|
522 |
|
fakeprogress($i); |
523 |
|
|
524 |
|
} |
525 |
} |
} |
|
print STDERR "\n"; |
|
526 |
} |
} |
527 |
|
|
528 |
# call this to commit index |
# call this to commit index |
534 |
|
|
535 |
=head1 NAME |
=head1 NAME |
536 |
|
|
537 |
isis2xml.pl - read isis file and dump XML |
all2xml.pl - read various file formats and dump XML for SWISH-E |
538 |
|
|
539 |
=head1 DESCRIPTION |
=head1 DESCRIPTION |
540 |
|
|
541 |
This command will read ISIS data file using OpenIsis perl module and |
This command will read ISIS data file using OpenIsis perl module, MARC |
542 |
create XML file for usage with I<SWISH-E> |
records using MARC module and optionally Micro$oft Excel files to |
543 |
indexer. Dispite it's name, this script B<isn't general xml generator> |
create one XML file for usage with I<SWISH-E> indexer. Dispite it's name, |
544 |
from isis files (isis allready has something like that). Output of this |
this script B<isn't general xml generator> from isis files (isis allready |
545 |
script is tailor-made for SWISH-E. |
has something like that). Output of this script is tailor-made for SWISH-E. |
546 |
|
|
547 |
|
=head1 BUGS |
548 |
|
|
549 |
|
Documentation is really lacking. However, in true Open Source spirit, source |
550 |
|
is best documentation. I even made considerable effort to comment parts |
551 |
|
which are not intuitively clear, so... |
552 |
|
|
553 |
=head1 AUTHOR |
=head1 AUTHOR |
554 |
|
|