5 |
use Getopt::Std; |
use Getopt::Std; |
6 |
use Data::Dumper; |
use Data::Dumper; |
7 |
use XML::Simple; |
use XML::Simple; |
|
use Text::Unaccent 1.02; # 1.01 won't compile on my platform, |
|
8 |
use Text::Iconv; |
use Text::Iconv; |
9 |
use Config::IniFiles; |
use Config::IniFiles; |
10 |
use Encode; |
use Encode; |
16 |
|
|
17 |
my $config_file = $0; |
my $config_file = $0; |
18 |
$config_file =~ s/\.pl$/.conf/; |
$config_file =~ s/\.pl$/.conf/; |
19 |
|
$config_file = $ARGV[0] if ($ARGV[0] && -f $ARGV[0]); |
20 |
die "FATAL: can't find configuration file '$config_file'" if (! -e $config_file); |
die "FATAL: can't find configuration file '$config_file'" if (! -e $config_file); |
21 |
|
|
22 |
my $config; |
my $config; |
72 |
# if you are tight on memory, turn this off |
# if you are tight on memory, turn this off |
73 |
my $use_lhash_cache = 1; |
my $use_lhash_cache = 1; |
74 |
|
|
75 |
|
my $last_field_name; # cache to prevent repeated fields |
76 |
|
|
77 |
sub data2xml { |
sub data2xml { |
78 |
|
|
79 |
use xmlify; |
use xmlify; |
104 |
} |
} |
105 |
|
|
106 |
my @sorted_tags; |
my @sorted_tags; |
107 |
if ($cache->{tags_by_order}->{$type}) { |
if ($cache->{tags_by_order}) { |
108 |
@sorted_tags = @{$cache->{tags_by_order}->{$type}}; |
@sorted_tags = @{$cache->{tags_by_order}}; |
109 |
} else { |
} else { |
110 |
@sorted_tags = sort by_order keys %{$config->{indexer}}; |
@sorted_tags = sort by_order keys %{$config->{indexer}}; |
111 |
$cache->{tags_by_order}->{$type} = \@sorted_tags; |
$cache->{tags_by_order} = \@sorted_tags; |
112 |
} |
} |
113 |
|
|
114 |
# lookup key |
# lookup key |
115 |
my $lookup_key; |
my $lookup_key; |
116 |
|
|
117 |
# cache for field in pages |
# cache for field in pages |
118 |
$cache->{display_data} = (); |
delete $cache->{display_data}; |
119 |
$cache->{swish_data} = (); |
delete $cache->{swish_data}; |
120 |
$cache->{swish_exact_data} = (); |
delete $cache->{swish_exact_data}; |
121 |
|
delete $cache->{index_data}; |
122 |
|
delete $cache->{index_delimiter}; |
123 |
my @page_fields; # names of fields |
my @page_fields; # names of fields |
124 |
|
|
125 |
|
|
139 |
} else { |
} else { |
140 |
print STDERR "WARNING: field '$field' doesn't have 'name' attribute!"; |
print STDERR "WARNING: field '$field' doesn't have 'name' attribute!"; |
141 |
} |
} |
142 |
|
|
143 |
if ($field_name) { |
if ($field_name) { |
144 |
return x($field_name); |
$field_name = x($field_name); |
145 |
|
if (! $last_field_name) { |
146 |
|
$last_field_name = $field_name; |
147 |
|
return $last_field_name; |
148 |
|
} elsif ($field_name ne $last_field_name) { |
149 |
|
$last_field_name = $field_name; |
150 |
|
return $last_field_name; |
151 |
|
} |
152 |
} |
} |
153 |
} |
} |
154 |
|
|
155 |
|
|
156 |
|
# init variables for different types |
157 |
|
sub init_visible_type($) { |
158 |
|
my $type = shift; |
159 |
|
|
160 |
|
# swish, swish_exact, display, index, index_lookup |
161 |
|
# swish and display defaults |
162 |
|
my ($s,$se,$d,$i,$il) = (1,0,1,0,0); |
163 |
|
if (lc($type) eq "display") { |
164 |
|
$s = 0; |
165 |
|
} elsif (lc($type) eq "swish") { |
166 |
|
$d = 0; |
167 |
|
} elsif (lc($type) eq "index") { |
168 |
|
($s,$se,$d,$i) = (0,1,0,1); |
169 |
|
} elsif (lc($type) eq "swish_exact") { |
170 |
|
($s,$se,$d,$i) = (0,1,0,0); |
171 |
|
} elsif (lc($type) =~ /^lookup/) { |
172 |
|
($s,$se,$d,$i,$il) = (0,1,0,0,1); |
173 |
|
} elsif ($type) { |
174 |
|
print STDERR "WARNING: unknown type: $type\n"; |
175 |
|
} |
176 |
|
return ($s,$se,$d,$i,$il); |
177 |
|
} |
178 |
|
|
179 |
|
|
180 |
|
# convert |
181 |
|
# |
182 |
|
# <tag> |
183 |
|
# <delimiter>, </delimiter> |
184 |
|
# <value>200a</value> |
185 |
|
# </tag> |
186 |
|
# |
187 |
|
# to |
188 |
|
# |
189 |
|
# <tag delimiter=", ">200a</tag> |
190 |
|
# |
191 |
|
# but without loosing spaces in delimiter (becasue |
192 |
|
# new XML::Simple strips spaces in attribute values |
193 |
|
# as defined in XML specification) |
194 |
|
# |
195 |
|
sub unroll_x($) { |
196 |
|
my $x = shift; |
197 |
|
|
198 |
|
if (defined $x->{value}) { |
199 |
|
my ($v,$d) = ($x->{value}->{content}, $x->{delimiter}->{content}); |
200 |
|
delete $x->{value}; |
201 |
|
delete $x->{delimiter}; |
202 |
|
$x->{content} = $v; |
203 |
|
$x->{delimiter} = $d; |
204 |
|
} |
205 |
|
return $x; |
206 |
|
} |
207 |
|
|
208 |
# begin real work: go field by field |
# begin real work: go field by field |
209 |
foreach my $field (@sorted_tags) { |
foreach my $field (@sorted_tags) { |
210 |
|
|
214 |
my $swish_data = ""; |
my $swish_data = ""; |
215 |
my $swish_exact_data = ""; |
my $swish_exact_data = ""; |
216 |
my $display_data = ""; |
my $display_data = ""; |
217 |
|
my @index_data; |
218 |
my $line_delimiter; |
my $line_delimiter; |
219 |
|
|
220 |
my ($swish,$display); |
my ($swish,$display); |
227 |
my %page_max = (); |
my %page_max = (); |
228 |
# default line_delimiter if using |
# default line_delimiter if using |
229 |
my $page_line_delimiter = $config->{indexer}->{$field}->{page_line_delimiter} || '<br/>'; |
my $page_line_delimiter = $config->{indexer}->{$field}->{page_line_delimiter} || '<br/>'; |
230 |
|
$cache->{index_delimiter}->{$field} = $config->{indexer}->{$field}->{index_delimiter}; |
231 |
|
|
232 |
|
my $format_name = $config->{indexer}->{$field}->{format_name}; |
233 |
|
my $format_delimiter = $config->{indexer}->{$field}->{format_delimiter}; |
234 |
|
if ($format_name && $format_delimiter) { |
235 |
|
$cache->{format}->{$field}->{format_name} = $format_name; |
236 |
|
$cache->{format}->{$field}->{format_delimiter} = $format_delimiter; |
237 |
|
} |
238 |
|
|
239 |
foreach my $x (@{$config->{indexer}->{$field}->{$tag}}) { |
foreach my $x (@{$config->{indexer}->{$field}->{$tag}}) { |
240 |
|
|
241 |
|
$x = unroll_x($x); |
242 |
|
|
243 |
my $format = x($x->{content}); |
my $format = x($x->{content}); |
244 |
my $delimiter = x($x->{delimiter}) || ' '; |
my $delimiter = x($x->{delimiter}) || ' '; |
245 |
|
|
246 |
my $repeat_off = 0; # init repeatable offset |
my $repeat_off = 0; # init repeatable offset |
247 |
|
|
248 |
# swish, swish_exact, display, index, index_lookup |
my ($s,$se,$d,$i,$il) = init_visible_type($x->{type}); |
|
# swish and display defaults |
|
|
my ($s,$se,$d,$i,$il) = (1,0,1,0,0); |
|
|
$s = 0 if (lc($x->{type}) eq "display"); |
|
|
$d = 0 if (lc($x->{type}) eq "swish"); |
|
|
$se = 1 if (lc($x->{type}) eq "swish_exact"); |
|
|
($s,$d,$i) = (0,0,1) if (lc($x->{type}) eq "index"); |
|
|
$il = 1 if (lc($x->{type}) =~ /^lookup/); |
|
|
|
|
249 |
|
|
250 |
# what will separate last line from this one? |
# what will separate last line from this one? |
251 |
if ($display_data && $x->{append} && $x->{append} eq "1") { |
if ($display_data && $x->{append}) { |
252 |
$line_delimiter = ' '; |
$line_delimiter = $delimiter; |
253 |
} elsif ($display_data) { |
} elsif ($display_data) { |
254 |
$line_delimiter = '<br/>'; |
$line_delimiter = '<br/>'; |
255 |
} |
} |
258 |
($swish,$display) = (1,1); |
($swish,$display) = (1,1); |
259 |
|
|
260 |
# placeholder for all repeatable entries for index |
# placeholder for all repeatable entries for index |
|
my @index_data; |
|
261 |
|
|
262 |
sub mkformat { |
sub mkformat($$) { |
263 |
my $x = shift || die "mkformat needs tag reference"; |
my $x = shift || die "mkformat needs tag reference"; |
264 |
my $data = shift || return; |
my $data = shift || return; |
265 |
my $format_name = x($x->{format_name}) || return $data; |
my $format_name = x($x->{format_name}) || return $data; |
277 |
if (($#data+1) == $nr) { |
if (($#data+1) == $nr) { |
278 |
return sprintf($fmt,@data); |
return sprintf($fmt,@data); |
279 |
} else { |
} else { |
280 |
print STDERR "mkformat: [$data] can't be split on [$format_delimiter] to $nr fields!\n"; |
#print STDERR "mkformat: [$data] can't be split on [$format_delimiter] to $nr fields!\n"; |
281 |
return $data; |
return $data; |
282 |
} |
} |
283 |
} else { |
} else { |
306 |
$display = $new_display; |
$display = $new_display; |
307 |
$cache->{lhash}->{$display} = $new_display; |
$cache->{lhash}->{$display} = $new_display; |
308 |
} else { |
} else { |
309 |
print STDERR "WARNING: lookup for '$display' didn't find anything.\n"; |
# print STDERR "WARNING: lookup for '$display' didn't find anything.\n"; |
310 |
$display = ""; |
$display = ""; |
311 |
$cache->{lhash}->{$display} = $null; |
$cache->{lhash}->{$display} = $null; |
312 |
} |
} |
327 |
} |
} |
328 |
# type="swish" ; field for swish |
# type="swish" ; field for swish |
329 |
if ($swish) { |
if ($swish) { |
330 |
|
my $tmp = $swish; |
331 |
if ($filter && ($s || $se)) { |
if ($filter && ($s || $se)) { |
332 |
no strict 'refs'; |
no strict 'refs'; |
333 |
my $tmp = join(" ",&$filter($swish)) if ($s || $se); |
$tmp = join(" ",&$filter($tmp)) if ($s || $se); |
|
$swish_data .= $tmp if ($s); |
|
|
$swish_exact_data .= $tmp if ($se); |
|
|
|
|
|
} else { |
|
|
$swish_data .= $swish if ($s); |
|
|
$swish_exact_data .= $swish if ($se); |
|
334 |
} |
} |
335 |
|
|
336 |
|
$swish_data .= $tmp if ($s && $tmp); |
337 |
|
$swish_exact_data .= "xxbxx $tmp xxexx " if ($tmp && $tmp ne "" && $se); |
338 |
} |
} |
339 |
|
|
340 |
# type="display" ; field for display |
# type="display" ; field for display |
341 |
if ($d && $display) { |
if ($d && $display) { |
342 |
|
my $ldel = $delimiter; |
343 |
if ($line_delimiter && $display_data) { |
if ($line_delimiter && $display_data) { |
344 |
$display_data .= $line_delimiter; |
$ldel = $line_delimiter; |
345 |
} |
} |
346 |
if ($filter) { |
if ($filter) { |
347 |
no strict 'refs'; |
no strict 'refs'; |
348 |
if ($display_data) { |
my @arr; |
349 |
$display_data .= $delimiter.join($delimiter,mkformat($x,&$filter($display))); |
foreach my $tmp (&$filter($display)) { |
350 |
} else { |
my $tmp2 = mkformat($x,$tmp); |
351 |
$display_data = join($delimiter,mkformat($x,&$filter($display))); |
push @arr,$tmp2 if ($tmp2); |
352 |
} |
} |
353 |
|
$display_data .= $ldel if ($display_data && @arr); |
354 |
|
$display_data .= join($delimiter,@arr); |
355 |
} else { |
} else { |
356 |
if ($display_data) { |
$display_data .= $ldel if ($display_data); |
357 |
$display_data .= $delimiter.mkformat($x,$display); |
my $tmp = mkformat($x,$display); |
358 |
} else { |
$display_data .= $tmp if ($tmp); |
|
$display_data = mkformat($x,$display); |
|
|
} |
|
359 |
} |
} |
360 |
} |
} |
361 |
|
|
362 |
# type="index" ; insert into index |
# type="index" ; insert into index |
363 |
|
my $idisplay; |
364 |
if ($i && $display) { |
if ($i && $display) { |
365 |
|
$idisplay = $display; |
366 |
if ($filter) { |
if ($filter) { |
367 |
no strict 'refs'; |
no strict 'refs'; |
368 |
$display = &$filter($display); |
$idisplay = &$filter($idisplay); |
|
} |
|
|
if ($x->{append} && @index_data) { |
|
|
$index_data[$#index_data].=$display; |
|
|
} else { |
|
|
push @index_data, $display; |
|
369 |
} |
} |
370 |
|
push @index_data, $idisplay if ($idisplay && !$iterate_by_page); |
371 |
} |
} |
372 |
|
|
373 |
# store fields in lookup |
# store fields in lookup |
376 |
if ($lookup_key) { |
if ($lookup_key) { |
377 |
print STDERR "WARNING: try to redefine lookup_key (keys shouldn't be repeatable fields!)"; |
print STDERR "WARNING: try to redefine lookup_key (keys shouldn't be repeatable fields!)"; |
378 |
} else { |
} else { |
379 |
$lookup_key = $display; |
if ($filter) { |
380 |
|
no strict 'refs'; |
381 |
|
$lookup_key = &$filter($display); |
382 |
|
} else { |
383 |
|
$lookup_key = $display; |
384 |
|
} |
385 |
} |
} |
386 |
} elsif (lc($x->{type}) eq "lookup_val") { |
} elsif (lc($x->{type}) eq "lookup_val") { |
387 |
if ($lookup_key) { |
if ($lookup_key) { |
388 |
$lhash{$lookup_key} = $display; |
if ($filter) { |
389 |
|
no strict 'refs'; |
390 |
|
$lhash{$lookup_key} = &$filter($display); |
391 |
|
} else { |
392 |
|
$lhash{$lookup_key} = $display; |
393 |
|
} |
394 |
} else { |
} else { |
395 |
print STDERR "WARNING: no lookup_key defined for '$display'?"; |
print STDERR "WARNING: no lookup_key defined for '$display'?"; |
396 |
} |
} |
403 |
sub iterate_fld($$$$$$) { |
sub iterate_fld($$$$$$) { |
404 |
my ($cache,$what,$field,$page,$data,$append) = @_; |
my ($cache,$what,$field,$page,$data,$append) = @_; |
405 |
return if (!$data); |
return if (!$data); |
406 |
my $line_delimiter = $page_line_delimiter; |
|
407 |
$line_delimiter = " " if ($append); |
my $ldel = $page_line_delimiter; |
408 |
|
$ldel = " " if ($append); |
409 |
|
#print STDERR "line delimiter: ",Dumper($ldel) if ($ldel); |
410 |
if (! $cache->{$what}->{$field}->[$page]) { |
if (! $cache->{$what}->{$field}->[$page]) { |
411 |
$cache->{$what}->{$field}->[$page] = $data; |
$cache->{$what}->{$field}->[$page] = $data; |
412 |
} else { |
} else { |
413 |
$cache->{$what}->{$field}->[$page] .= $line_delimiter.$data; |
$cache->{$what}->{$field}->[$page] .= $ldel.$data; |
414 |
} |
} |
415 |
} |
} |
416 |
|
|
417 |
if ($display_data) { |
if ($display_data) { |
|
print STDERR "line delimiter: ",Dumper($line_delimiter) if ($line_delimiter); |
|
418 |
iterate_fld($cache,'display_data',$field,$page,$display_data,$x->{append}); |
iterate_fld($cache,'display_data',$field,$page,$display_data,$x->{append}); |
419 |
} |
} |
420 |
$display_data = ""; |
$display_data = ""; |
426 |
iterate_fld($cache,'swish_exact_data',$field,$page,$swish_exact_data,$x->{append}); |
iterate_fld($cache,'swish_exact_data',$field,$page,$swish_exact_data,$x->{append}); |
427 |
$swish_exact_data = ""; |
$swish_exact_data = ""; |
428 |
} |
} |
429 |
|
|
430 |
|
if ($idisplay) { |
431 |
|
my $ldel=$page_line_delimiter; |
432 |
|
my @index_data; |
433 |
|
if ($cache->{index_data}->{$field}->[$page]) { |
434 |
|
|
435 |
|
@index_data = @{$cache->{index_data}->{$field}->[$page]}; |
436 |
|
} |
437 |
|
if ($x->{append}) { |
438 |
|
if (@index_data) { |
439 |
|
$index_data[$#index_data] .= $idisplay; |
440 |
|
} else { |
441 |
|
push @index_data, $idisplay; |
442 |
|
} |
443 |
|
} else { |
444 |
|
push @index_data, $idisplay; |
445 |
|
} |
446 |
|
$idisplay = ""; |
447 |
|
@{$cache->{index_data}->{$field}->[$page]} = @index_data; |
448 |
|
} |
449 |
} |
} |
450 |
} |
} |
451 |
|
|
452 |
# fill data in index |
if (! $iterate_by_page) { |
453 |
foreach my $d (@index_data) { |
my $idel = $x->{index_delimiter}; |
454 |
$index->insert($field, $d, $path); |
# fill data in index |
455 |
|
foreach my $tmp (@index_data) { |
456 |
|
my $i = $d = $tmp; |
457 |
|
if ($idel && $tmp =~ m/$idel/) { |
458 |
|
($i,$d) = split(/$idel/,$tmp); |
459 |
|
} |
460 |
|
$index->insert($field, $i, $d, $path); |
461 |
|
} |
462 |
|
@index_data = (); |
463 |
} |
} |
464 |
} |
} |
465 |
|
|
466 |
# now try to parse variables from configuration file |
# now try to parse variables from configuration file |
467 |
foreach my $x (@{$config->{indexer}->{$field}->{'config'}}) { |
foreach my $x (@{$config->{indexer}->{$field}->{'config'}}) { |
468 |
|
|
469 |
|
$x = unroll_x($x); |
470 |
|
|
471 |
my $delimiter = x($x->{delimiter}) || ' '; |
my $delimiter = x($x->{delimiter}) || ' '; |
472 |
my $val = $cfg->val($database, x($x->{content})); |
my $val = $cfg->val($database, x($x->{content})); |
473 |
|
|
474 |
my ($s,$d,$i) = (1,1,0); # swish, display default |
# FIXME index_lookup is not supported! |
475 |
$s = 0 if (lc($x->{type}) eq "display"); |
my ($s,$se,$d,$i,$il) = init_visible_type($x->{type}); |
|
$d = 0 if (lc($x->{type}) eq "swish"); |
|
|
# no support for swish exact in config. |
|
|
# IMHO, it's useless |
|
|
($s,$d,$i) = (0,0,1) if (lc($x->{type}) eq "index"); |
|
476 |
|
|
477 |
if ($val) { |
if ($val) { |
478 |
$display_data .= $delimiter.$val if ($d); |
$display_data .= $delimiter.$val if ($d); |
479 |
$swish_data .= $val if ($s); |
$swish_data .= " ".$val if ($s); |
480 |
$index->insert($field, $val, $path) if ($i); |
$index->insert($field, $val, $path) if ($i); |
481 |
} |
} |
482 |
|
|
505 |
#print STDERR "field '$field' iterate over ",($nr_pages || 0)," pages...\n"; |
#print STDERR "field '$field' iterate over ",($nr_pages || 0)," pages...\n"; |
506 |
#print STDERR Dumper($cache->{display_data}); |
#print STDERR Dumper($cache->{display_data}); |
507 |
for (my $page=0; $page <= $nr_pages; $page++) { |
for (my $page=0; $page <= $nr_pages; $page++) { |
508 |
|
my $display_data; |
509 |
my $display_data = $cache->{display_data}->{$field}->[$page]; |
if ($cache->{format}->{$field}) { |
510 |
|
my $tmp = mkformat($cache->{format}->{$field},$cache->{display_data}->{$field}->[$page]); |
511 |
|
$display_data=$tmp if ($tmp); |
512 |
|
} else { |
513 |
|
$display_data = $cache->{display_data}->{$field}->[$page]; |
514 |
|
} |
515 |
if ($display_data) { # default |
if ($display_data) { # default |
516 |
if ($field eq "headline") { |
if ($field eq "headline") { |
517 |
$xml .= xmlify("headline", $display_data); |
$xml .= xmlify("headline", $display_data); |
529 |
$swish_data =~ s/ +/ /g; |
$swish_data =~ s/ +/ /g; |
530 |
$swish_data =~ s/ +$//g; |
$swish_data =~ s/ +$//g; |
531 |
|
|
532 |
$xml .= xmlify($field."_swish", unac_string($codepage,$swish_data)); |
$xml .= xmlify($field."_swish", my_unac_string($codepage,$swish_data)); |
533 |
} |
} |
534 |
|
|
535 |
my $swish_exact_data = $cache->{swish_exact_data}->{$field}->[$page]; |
my $swish_exact_data = $cache->{swish_exact_data}->{$field}->[$page]; |
539 |
|
|
540 |
# add delimiters before and after word. |
# add delimiters before and after word. |
541 |
# That is required to produce exact match |
# That is required to produce exact match |
542 |
$xml .= xmlify($field."_swish_exact", unac_string($codepage,'xxbxx '.$swish_exact_data.' xxexx')); |
$xml .= xmlify($field."_swish_exact", my_unac_string($codepage,$swish_exact_data)); |
543 |
|
} |
544 |
|
|
545 |
|
my $idel = $cache->{index_delimiter}->{$field}; |
546 |
|
foreach my $tmp (@{$cache->{index_data}->{$field}->[$page]}) { |
547 |
|
my $i = $tmp; |
548 |
|
my $d = $tmp; |
549 |
|
if ($idel && $tmp =~ m/$idel/) { |
550 |
|
($i,$d) = split(/$idel/,$tmp); |
551 |
|
} |
552 |
|
$index->insert($field, $i, $d, $path); |
553 |
|
#print STDERR "index [$idel] $field: $i --> $d [$path]\n"; |
554 |
} |
} |
555 |
} |
} |
556 |
|
|
572 |
$swish_data =~ s/ +/ /g; |
$swish_data =~ s/ +/ /g; |
573 |
$swish_data =~ s/ +$//g; |
$swish_data =~ s/ +$//g; |
574 |
|
|
575 |
$xml .= xmlify($field."_swish", unac_string($codepage,$swish_data)); |
$xml .= xmlify($field."_swish", my_unac_string($codepage,$swish_data)); |
576 |
} |
} |
577 |
|
|
578 |
if ($swish_exact_data) { |
if ($swish_exact_data) { |
581 |
|
|
582 |
# add delimiters before and after word. |
# add delimiters before and after word. |
583 |
# That is required to produce exact match |
# That is required to produce exact match |
584 |
$xml .= xmlify($field."_swish_exact", unac_string($codepage,'xxbxx '.$swish_exact_data.' xxexx')); |
$xml .= xmlify($field."_swish_exact", my_unac_string($codepage,$swish_exact_data)); |
585 |
} |
} |
586 |
} |
} |
587 |
} |
} |
618 |
|
|
619 |
my $show_progress = $cfg_global->val('global', 'show_progress'); |
my $show_progress = $cfg_global->val('global', 'show_progress'); |
620 |
|
|
621 |
my $unac_filter = $cfg_global->val('global', 'unac_filter'); |
my $my_unac_filter = $cfg_global->val('global', 'my_unac_filter'); |
622 |
if ($unac_filter) { |
if ($my_unac_filter) { |
623 |
require $unac_filter; |
print STDERR "using $my_unac_filter to filter characters for search\n"; |
624 |
|
require $my_unac_filter; |
625 |
|
} else { |
626 |
|
print STDERR "### fallback to default my_unac_string!\n"; |
627 |
|
eval q{ |
628 |
|
sub main::my_unac_string($$) { |
629 |
|
my ($charset, $string) = (@_); |
630 |
|
return $string; |
631 |
|
} |
632 |
|
}; |
633 |
} |
} |
634 |
|
|
635 |
foreach my $database ($cfg->Sections) { |
foreach my $database ($cfg->Sections) { |
641 |
my $lookup_file = $cfg -> val($database, 'lookup_newfile'); # optional |
my $lookup_file = $cfg -> val($database, 'lookup_newfile'); # optional |
642 |
if ($lookup_file) { |
if ($lookup_file) { |
643 |
#tie %lhash, 'GDBM_File', $lookup_file, &GDBM_NEWDB, 0644; |
#tie %lhash, 'GDBM_File', $lookup_file, &GDBM_NEWDB, 0644; |
644 |
|
if (! -e $lookup_file) { |
645 |
|
open(LOOKUP, "> $lookup_file") || die "can't create $lookup_file': $!"; |
646 |
|
close(LOOKUP); |
647 |
|
} |
648 |
tie %lhash, 'TDB_File', $lookup_file, TDB_CLEAR_IF_FIRST, O_RDWR, 0644; |
tie %lhash, 'TDB_File', $lookup_file, TDB_CLEAR_IF_FIRST, O_RDWR, 0644; |
649 |
print STDERR "creating lookup file '$lookup_file'\n"; |
print STDERR "creating lookup file '$lookup_file'\n"; |
650 |
|
# delete memory cache for lookup file |
651 |
|
delete $cache->{lhash}; |
652 |
} |
} |
653 |
|
|
654 |
# open existing lookup file |
# open existing lookup file |
665 |
my $type_base = $type; |
my $type_base = $type; |
666 |
$type_base =~ s/_.+$//g; |
$type_base =~ s/_.+$//g; |
667 |
|
|
668 |
$config=XMLin("./import_xml/$type.xml", forcearray => [ $type2tag{$type_base}, 'config', 'format' ], forcecontent => 1); |
$config=XMLin("./import_xml/$type.xml", ForceArray => [ $type2tag{$type_base}, 'config', 'format' ], ForceContent => 1 ); |
669 |
|
|
670 |
|
# helper for progress bar |
671 |
|
sub fmt_time { |
672 |
|
my $t = shift || 0; |
673 |
|
my $out = ""; |
674 |
|
|
675 |
|
my ($ss,$mm,$hh) = gmtime($t); |
676 |
|
$out .= "${hh}h" if ($hh); |
677 |
|
$out .= sprintf("%02d:%02d", $mm,$ss); |
678 |
|
$out .= " " if ($hh == 0); |
679 |
|
return $out; |
680 |
|
} |
681 |
|
|
682 |
# output current progress indicator |
# output current progress indicator |
683 |
my $last_p = 0; |
my $last_p = 0; |
684 |
|
my $start_t = time(); |
685 |
sub progress { |
sub progress { |
686 |
return if (! $show_progress); |
return if (! $show_progress); |
687 |
my $current = shift; |
my $current = shift; |
688 |
my $total = shift || 1; |
my $total = shift || 1; |
689 |
my $p = int($current * 100 / $total); |
my $p = int($current * 100 / $total); |
690 |
if ($p != $last_p) { |
if ($p != $last_p) { |
691 |
printf STDERR ("%5d / %5d [%-51s] %-2d %% \r",$current,$total,"=" x ($p/2).">", $p ); |
my $rate = ($current / (time() - $start_t || 1)); |
692 |
|
my $eta = ($total-$current) / ($rate || 1); |
693 |
|
printf STDERR ("%5d [%-38s] %-5d %0.1f/s %s\r",$current,"=" x ($p/3)."$p%>", $total, $rate, fmt_time($eta)); |
694 |
$last_p = $p; |
$last_p = $p; |
695 |
} |
} |
696 |
} |
} |
712 |
# now read database |
# now read database |
713 |
print STDERR "using: $type...\n"; |
print STDERR "using: $type...\n"; |
714 |
|
|
715 |
|
# erase cache for tags by order in this database |
716 |
|
delete $cache->{tags_by_order}; |
717 |
|
|
718 |
if ($type_base eq "isis") { |
if ($type_base eq "isis") { |
719 |
|
|
720 |
my $isis_db = $cfg -> val($database, 'isis_db') || die "$database doesn't have 'isis_db' defined!"; |
my $isis_db = $cfg -> val($database, 'isis_db') || die "$database doesn't have 'isis_db' defined!"; |
800 |
print STDERR "\n"; |
print STDERR "\n"; |
801 |
|
|
802 |
} elsif ($type_base eq "excel") { |
} elsif ($type_base eq "excel") { |
803 |
use Spreadsheet::ParseExcel; |
require Spreadsheet::ParseExcel; |
804 |
use Spreadsheet::ParseExcel::Utility qw(int2col); |
require Spreadsheet::ParseExcel::Utility; |
805 |
|
import Spreadsheet::ParseExcel::Utility qw(int2col); |
806 |
|
|
807 |
$import2cp = Text::Iconv->new($config->{excel_codepage},$codepage); |
$import2cp = Text::Iconv->new($config->{excel_codepage},$codepage); |
808 |
my $excel_file = $cfg -> val($database, 'excel_file') || die "$database doesn't have 'excel_file' defined!"; |
my $excel_file = $cfg -> val($database, 'excel_file') || die "$database doesn't have 'excel_file' defined!"; |
830 |
for(my $iC = $oWorksheet->{MinCol} ; defined $oWorksheet->{MaxCol} && $iC <= $oWorksheet->{MaxCol} ; $iC++) { |
for(my $iC = $oWorksheet->{MinCol} ; defined $oWorksheet->{MaxCol} && $iC <= $oWorksheet->{MaxCol} ; $iC++) { |
831 |
my $cell = $oWorksheet->{Cells}[$iR][$iC]; |
my $cell = $oWorksheet->{Cells}[$iR][$iC]; |
832 |
if ($cell) { |
if ($cell) { |
833 |
$row->{int2col($iC)} = $cell->Value; |
# this conversion is a cludge. |
834 |
|
# Files from Excell could have |
835 |
|
# characters which don't fit into |
836 |
|
# destination encoding. |
837 |
|
$row->{int2col($iC)} = $utf2cp->convert($cell->Value) || $cell->Value; |
838 |
} |
} |
839 |
} |
} |
840 |
|
|
860 |
} |
} |
861 |
} elsif ($type_base eq "marc") { |
} elsif ($type_base eq "marc") { |
862 |
|
|
863 |
use MARC; |
require MARC::File::USMARC; |
864 |
|
|
865 |
$import2cp = Text::Iconv->new($config->{marc_codepage},$codepage); |
$import2cp = Text::Iconv->new($config->{marc_codepage},$codepage); |
866 |
my $marc_file = $cfg -> val($database, 'marc_file') || die "$database doesn't have 'marc_file' defined!"; |
my $marc_file = $cfg -> val($database, 'marc_file') || die "$database doesn't have 'marc_file' defined!"; |
867 |
|
|
868 |
# optional argument is format |
# optional argument is format |
869 |
my $format = x($config->{format}) || 'usmarc'; |
warn "marc_format is no longer used!" if ($config->{marc_format}); |
|
|
|
870 |
print STDERR "Reading MARC file '$marc_file'\n"; |
print STDERR "Reading MARC file '$marc_file'\n"; |
871 |
|
|
872 |
my $marc = new MARC; |
my $marc = MARC::File::USMARC->in( $marc_file ) |
873 |
my $nr = $marc->openmarc({ |
|| die "Can't open MARC file '$marc_file': ".$MARC::File::ERROR; |
874 |
file=>$marc_file, format=>$format |
|
875 |
}) || die "Can't open MARC file '$marc_file'"; |
# count records in MARC file |
876 |
|
sub marc_count { |
877 |
|
my $filename = shift || die; |
878 |
|
my $file = MARC::File::USMARC->in($filename) || die $MARC::File::ERROR; |
879 |
|
my $count = 0; |
880 |
|
while ($file->skip()) { |
881 |
|
$count++; |
882 |
|
} |
883 |
|
return $count; |
884 |
|
} |
885 |
|
|
886 |
my $i=0; # record nr. |
my $count = marc_count($marc_file) || warn "no records in '$marc_file'?"; |
887 |
|
|
888 |
my $rec; |
my $i = 0; |
889 |
|
|
890 |
while ($marc->nextmarc(1)) { |
while( my $rec = $marc->next() ) { |
891 |
|
|
892 |
# XXX |
progress($i++,$count); |
|
fakeprogress($i++); |
|
893 |
|
|
894 |
my $swishpath = $database."#".$i; |
my $swishpath = $database."#".$i; |
895 |
|
|
896 |
if (my $xml = data2xml($type_base,$marc,$add_xml,$cfg,$database)) { |
if (my $xml = data2xml($type_base,$rec,$add_xml,$cfg,$database)) { |
897 |
$xml = $cp2utf->convert($xml); |
$xml = $cp2utf->convert($xml); |
898 |
use bytes; # as opposed to chars |
use bytes; # as opposed to chars |
899 |
print "Path-Name: $swishpath\n"; |
print "Path-Name: $swishpath\n"; |
901 |
print "Document-Type: XML\n\n$xml\n"; |
print "Document-Type: XML\n\n$xml\n"; |
902 |
} |
} |
903 |
} |
} |
904 |
|
|
905 |
|
print STDERR "\n"; |
906 |
|
|
907 |
} elsif ($type_base eq "feed") { |
} elsif ($type_base eq "feed") { |
908 |
|
|
909 |
$import2cp = Text::Iconv->new($config->{feed_codepage},$codepage); |
$import2cp = Text::Iconv->new($config->{feed_codepage},$codepage); |
958 |
|
|
959 |
all2xml.pl - read various file formats and dump XML for SWISH-E |
all2xml.pl - read various file formats and dump XML for SWISH-E |
960 |
|
|
961 |
|
=head1 SYNOPSYS |
962 |
|
|
963 |
|
$ all2xml.pl [test.conf] |
964 |
|
|
965 |
=head1 DESCRIPTION |
=head1 DESCRIPTION |
966 |
|
|
967 |
This command will read ISIS data file using OpenIsis perl module, MARC |
This command will read ISIS data file using OpenIsis perl module, MARC |
970 |
this script B<isn't general xml generator> from isis files (isis allready |
this script B<isn't general xml generator> from isis files (isis allready |
971 |
has something like that). Output of this script is tailor-made for SWISH-E. |
has something like that). Output of this script is tailor-made for SWISH-E. |
972 |
|
|
973 |
|
If no configuration file is specified, it will use default one called |
974 |
|
C<all2xml.conf>. |
975 |
|
|
976 |
=head1 BUGS |
=head1 BUGS |
977 |
|
|
978 |
Documentation is really lacking. However, in true Open Source spirit, source |
Documentation is really lacking. However, in true Open Source spirit, source |