source: gs2-extensions/parallel-building/trunk/src/perllib/inexport.pm@ 24626

Last change on this file since 24626 was 24626, checked in by jmt12, 13 years ago

An (almost) complete copy of the perllib directory from a (circa SEP2011) head checkout from Greenstone 2 trunk - in order to try and make merging in this extension a little easier later on (as there have been some major changes to buildcol.pl commited in the main trunk but not in the x64 branch)
File size: 38.4 KB
Line 
1###########################################################################
2#
3# inexport.pm -- useful class to support import.pl and export.pl
4# A component of the Greenstone digital library software
5# from the New Zealand Digital Library Project at the
6# University of Waikato, New Zealand.
7#
8# Copyright (C) 1999 New Zealand Digital Library Project
9#
10# This program is free software; you can redistribute it and/or modify
11# it under the terms of the GNU General Public License as published by
12# the Free Software Foundation; either version 2 of the License, or
13# (at your option) any later version.
14#
15# This program is distributed in the hope that it will be useful,
16# but WITHOUT ANY WARRANTY; without even the implied warranty of
17# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
18# GNU General Public License for more details.
19#
20# You should have received a copy of the GNU General Public License
21# along with this program; if not, write to the Free Software
22# Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
23#
24###########################################################################
25
26package inexport;
27
28use strict;
29
30no strict 'refs'; # allow filehandles to be variables and vice versa
31no strict 'subs'; # allow barewords (eg STDERR) as function arguments
32
33use arcinfo;
34use colcfg;
35use dbutil;
36use doc;
37use plugin;
38use plugout;
39use manifest;
40use inexport;
41use util;
42use scriptutil;
43use FileHandle;
44use gsprintf 'gsprintf';
45use printusage;
46use parse2;
47
48use File::Basename;
49
50sub new
51{
52 my $class = shift (@_);
53 my ($mode,$argv,$options,$opt_listall_options) = @_;
54
55 my $self = { 'xml' => 0, 'mode' => $mode };
56
57 # general options available to all plugins
58 my $arguments = $options->{'args'};
59 my $intArgLeftinAfterParsing = parse2::parse($argv,$arguments,$self,"allow_extra_options");
60 # Parse returns -1 if something has gone wrong
61 if ($intArgLeftinAfterParsing == -1)
62 {
63 &PrintUsage::print_txt_usage($options, "{import.params}");
64 die "\n";
65 }
66
67 my $language = $self->{'language'};
68 # If $language has been specified, load the appropriate resource bundle
69 # (Otherwise, the default resource bundle will be loaded automatically)
70 if ($language && $language =~ /\S/) {
71 &gsprintf::load_language_specific_resource_bundle($language);
72 }
73
74 if ($self->{'listall'}) {
75 if ($self->{'xml'}) {
76 &PrintUsage::print_xml_usage($opt_listall_options);
77 }
78 else
79 {
80 &PrintUsage::print_txt_usage($opt_listall_options,"{export.params}");
81 }
82 die "\n";
83 }
84
85
86 if ($self->{'xml'}) {
87 &PrintUsage::print_xml_usage($options);
88 print "\n";
89 return bless $self, $class;
90 }
91
92 if ($self->{'gli'}) { # the gli wants strings to be in UTF-8
93 &gsprintf::output_strings_in_UTF8;
94 }
95
96 # now check that we had exactly one leftover arg, which should be
97 # the collection name. We don't want to do this earlier, cos
98 # -xml arg doesn't need a collection name
99 # Or if the user specified -h, then we output the usage also
100
101 if ($intArgLeftinAfterParsing != 1 || (@$argv && $argv->[0] =~ /^\-+h/))
102 {
103 &PrintUsage::print_txt_usage($options, "{import.params}");
104 die "\n";
105 }
106
107 $self->{'close_out'} = 0;
108 my $out = $self->{'out'};
109 if ($out !~ /^(STDERR|STDOUT)$/i) {
110 open (OUT, ">$out") ||
111 (&gsprintf(STDERR, "{common.cannot_open_output_file}: $!\n", $out) && die);
112 $out = 'inexport::OUT';
113 $self->{'close_out'} = 1;
114 }
115 $out->autoflush(1);
116 $self->{'out'} = $out;
117
118 # @ARGV should be only one item, the name of the collection
119 $self->{'collection'} = shift @$argv;
120
121 if ((defined $self->{'jobs'}) && ($self->{'jobs'}>1)) {
122 require ParallelInexport;
123 }
124
125 return bless $self, $class;
126}
127
128# Simplified version of the contstructor for use with CGI scripts
129sub newCGI
130{
131 my $class = shift (@_);
132 my ($mode,$collect,$gsdl_cgi,$opt_site) = @_;
133
134 my $self = { 'xml' => 0, 'mode' => $mode };
135
136 $self->{'out'} = STDERR;
137
138 if (defined $gsdl_cgi) {
139 $self->{'site'} = $opt_site;
140 my $collect_dir = $gsdl_cgi->get_collection_dir($opt_site);
141 $self->{'collectdir'} = $collect_dir;
142 }
143 else {
144 $self->{'site'} = "";
145 $self->{'collectdir'} = &util::filename_cat($ENV{'GSDLHOME'},"collect");
146 }
147 $self->{'faillog'} = "";
148
149 $self->{'collection'} = $collect;
150
151 return bless $self, $class;
152}
153sub get_collection
154{
155 my $self = shift @_;
156
157 return $self->{'collection'};
158}
159
160
161sub read_collection_cfg
162{
163 my $self = shift @_;
164 my ($collection,$options) = @_;
165
166 my $collectdir = $self->{'collectdir'};
167 my $site = $self->{'site'};
168 my $out = $self->{'out'};
169
170 if (($collection = &colcfg::use_collection($site, $collection, $collectdir)) eq "") {
171 &PrintUsage::print_txt_usage($options, "{import.params}");
172 die "\n";
173 }
174
175 # add collection's perllib dir into include path in
176 # case we have collection specific modules
177 my $collection_perllib_path = $ENV{'GSDLCOLLECTDIR'} . '/perllib';
178 my $inc_paths = join(':', @INC);
179 if ($inc_paths !~ /$collection_perllib_path/)
180 {
181 unshift (@INC, $collection_perllib_path); # [jmt12]
182 }
183
184 # check that we can open the faillog
185 my $faillog = $self->{'faillog'};
186 if ($faillog eq "") {
187 $faillog = &util::filename_cat($ENV{'GSDLCOLLECTDIR'}, "etc", "fail.log");
188 }
189 open (FAILLOG, ">$faillog") ||
190 (&gsprintf(STDERR, "{import.cannot_open_fail_log}\n", $faillog) && die);
191
192
193 my $faillogname = $faillog;
194 $faillog = 'inexport::FAILLOG';
195 $faillog->autoflush(1);
196 $self->{'faillog'} = $faillog;
197 $self->{'faillogname'} = $faillogname;
198
199 # Read in the collection configuration file.
200 my ($config_filename, $gs_mode) = &colcfg::get_collect_cfg_name($out);
201 my $collectcfg = &colcfg::read_collection_cfg ($config_filename, $gs_mode);
202
203 return ($config_filename,$collectcfg);
204}
205
206sub set_collection_options
207{
208 my $self = shift @_;
209 my ($collectcfg) = @_;
210
211 my $inexport_mode = $self->{'mode'};
212
213 my $verbosity = $self->{'verbosity'};
214 my $debug = $self->{'debug'};
215 my $importdir = $self->{'importdir'};
216 my $archivedir = $self->{'archivedir'} || $self->{'exportdir'} || "";
217 my $out = $self->{'out'};
218
219 # If the infodbtype value wasn't defined in the collect.cfg file, use the default
220 if (!defined($collectcfg->{'infodbtype'}))
221 {
222 $collectcfg->{'infodbtype'} = &dbutil::get_default_infodb_type();
223 }
224 if ($collectcfg->{'infodbtype'} eq "gdbm-txtgz") {
225 # we can't use the text version for archives dbs.
226 $collectcfg->{'infodbtype'} = "gdbm";
227 }
228 if (defined $collectcfg->{'importdir'} && $importdir eq "") {
229 $importdir = $collectcfg->{'importdir'};
230 }
231 if (defined $collectcfg->{'archivedir'} && $archivedir eq "") {
232 $archivedir = $collectcfg->{'archivedir'};
233 }
234 # fill in the default import and archives directories if none
235 # were supplied, turn all \ into / and remove trailing /
236 $importdir = &util::filename_cat ($ENV{'GSDLCOLLECTDIR'}, "import") if $importdir eq "";
237 $importdir =~ s/[\\\/]+/\//g;
238 $importdir =~ s/\/$//;
239 if (!-e $importdir) {
240 &gsprintf($out, "{import.no_import_dir}\n\n", $importdir);
241 die "\n";
242 }
243 $self->{'importdir'} = $importdir;
244
245 if ($archivedir eq "") {
246 if ($inexport_mode eq "import") {
247 $archivedir = &util::filename_cat ($ENV{'GSDLCOLLECTDIR'}, "archives");
248 }
249 elsif ($inexport_mode eq "export") {
250 $archivedir = &util::filename_cat ($ENV{'GSDLCOLLECTDIR'}, "export");
251 }
252 else {
253 print STDERR "Warning: Unrecognized import/export mode '$inexport_mode'\n";
254 print STDERR " Defaulting to 'archives' for file output\n";
255 $archivedir = &util::filename_cat ($ENV{'GSDLCOLLECTDIR'}, "archives");
256 }
257 }
258
259 $archivedir =~ s/[\\\/]+/\//g;
260 $archivedir =~ s/\/$//;
261 $self->{'archivedir'} = $archivedir;
262
263 if ($verbosity !~ /\d+/) {
264 if (defined $collectcfg->{'verbosity'} && $collectcfg->{'verbosity'} =~ /\d+/) {
265 $verbosity = $collectcfg->{'verbosity'};
266 } else {
267 $verbosity = 2; # the default
268 }
269 }
270 $self->{'verbosity'} = $verbosity;
271
272 if (defined $collectcfg->{'manifest'} && $self->{'manifest'} eq "") {
273 $self->{'manifest'} = $collectcfg->{'manifest'};
274 }
275
276 if (defined $collectcfg->{'gzip'} && !$self->{'gzip'}) {
277 if ($collectcfg->{'gzip'} =~ /^true$/i) {
278 $self->{'gzip'} = 1;
279 }
280 }
281
282 if ($self->{'maxdocs'} !~ /\-?\d+/) {
283 if (defined $collectcfg->{'maxdocs'} && $collectcfg->{'maxdocs'} =~ /\-?\d+/) {
284 $self->{'maxdocs'} = $collectcfg->{'maxdocs'};
285 } else {
286 $self->{'maxdocs'} = -1; # the default
287 }
288 }
289
290 if ((defined $self->{'groupsize'}) && ($self->{'groupsize'} == 1)) {
291 if (defined $collectcfg->{'groupsize'} && $collectcfg->{'groupsize'} =~ /\d+/) {
292 $self->{'groupsize'} = $collectcfg->{'groupsize'};
293 }
294 }
295
296 if (!defined $self->{'OIDtype'}
297 || ($self->{'OIDtype'} !~ /^(hash|incremental|assigned|dirname)$/ )) {
298 if (defined $collectcfg->{'OIDtype'}
299 && $collectcfg->{'OIDtype'} =~ /^(hash|incremental|assigned|dirname)$/) {
300 $self->{'OIDtype'} = $collectcfg->{'OIDtype'};
301 } else {
302 $self->{'OIDtype'} = "hash"; # the default
303 }
304 }
305
306 if ((!defined $self->{'OIDmetadata'}) || ($self->{'OIDmetadata'} eq "")) {
307 if (defined $collectcfg->{'OIDmetadata'}) {
308 $self->{'OIDmetadata'} = $collectcfg->{'OIDmetadata'};
309 } else {
310 $self->{'OIDmetadata'} = "dc.Identifier"; # the default
311 }
312 }
313
314 my $sortmeta = $self->{'sortmeta'};
315 if (defined $collectcfg->{'sortmeta'} && (!defined $sortmeta || $sortmeta eq "")) {
316 $sortmeta = $collectcfg->{'sortmeta'};
317 }
318 # sortmeta cannot be used with group size
319 $sortmeta = undef unless defined $sortmeta && $sortmeta =~ /\S/;
320 if (defined $sortmeta && $self->{'groupsize'} > 1) {
321 &gsprintf($out, "{import.cannot_sort}\n\n");
322 $sortmeta = undef;
323 }
324 $self->{'sortmeta'} = $sortmeta;
325
326 if (defined $collectcfg->{'removeprefix'} && $self->{'removeprefix'} eq "") {
327 $self->{'removeprefix'} = $collectcfg->{'removeprefix'};
328 }
329
330 if (defined $collectcfg->{'removesuffix'} && $self->{'removesuffix'} eq "") {
331 $self->{'removesuffix'} = $collectcfg->{'removesuffix'};
332 }
333 if (defined $collectcfg->{'debug'} && $collectcfg->{'debug'} =~ /^true$/i) {
334 $self->{'debug'} = 1;
335 }
336 if (defined $collectcfg->{'gli'} && $collectcfg->{'gli'} =~ /^true$/i) {
337 $self->{'gli'} = 1;
338 }
339 $self->{'gli'} = 0 unless defined $self->{'gli'};
340
341 # check keepold and removeold
342 my $checkdir = ($inexport_mode eq "import") ? "archives" : "export";
343
344 my ($removeold, $keepold, $incremental, $incremental_mode)
345 = &scriptutil::check_removeold_and_keepold($self->{'removeold'}, $self->{'keepold'},
346 $self->{'incremental'}, $checkdir,
347 $collectcfg);
348
349 $self->{'removeold'} = $removeold;
350 $self->{'keepold'} = $keepold;
351 $self->{'incremental'} = $incremental;
352 $self->{'incremental_mode'} = $incremental_mode;
353}
354
355sub process_files
356{
357 my $self = shift @_;
358 my ($config_filename,$collectcfg) = @_;
359
360 my $inexport_mode = $self->{'mode'};
361
362 my $verbosity = $self->{'verbosity'};
363 my $debug = $self->{'debug'};
364
365 my $importdir = $self->{'importdir'};
366 my $archivedir = $self->{'archivedir'} || $self->{'exportdir'};
367
368 my $incremental = $self->{'incremental'};
369 my $incremental_mode = $self->{'incremental_mode'};
370
371 my $removeold = $self->{'removeold'};
372 my $keepold = $self->{'keepold'};
373
374 my $saveas = $self->{'saveas'};
375 my $OIDtype = $self->{'OIDtype'};
376 my $OIDmetadata = $self->{'OIDmetadata'};
377
378 my $out = $self->{'out'};
379 my $faillog = $self->{'faillog'};
380
381 my $maxdocs = $self->{'maxdocs'};
382 my $gzip = $self->{'gzip'};
383 my $groupsize = $self->{'groupsize'};
384 my $sortmeta = $self->{'sortmeta'};
385
386 my $removeprefix = $self->{'removeprefix'};
387 my $removesuffix = $self->{'removesuffix'};
388
389 my $gli = $self->{'gli'};
390
391 my $jobs = $self->{'jobs'};
392 my $epoch = $self->{'epoch'};
393
394 # related to export
395 my $xsltfile = $self->{'xsltfile'};
396 my $group_marc = $self->{'group_marc'};
397 my $mapping_file = $self->{'mapping_file'};
398 my $xslt_mets = $self->{'xslt_mets'};
399 my $xslt_txt = $self->{'xslt_txt'};
400 my $fedora_namespace = $self->{'fedora_namespace'};
401 my $metadata_prefix = $self->{'metadata_prefix'};
402
403 if ($inexport_mode eq "import") {
404 print STDERR "<Import>\n" if $gli;
405 }
406 else {
407 print STDERR "<export>\n" if $gli;
408 }
409
410 my $manifest_lookup = new manifest($collectcfg->{'infodbtype'},$archivedir);
411 if ($self->{'manifest'} ne "") {
412 my $manifest_filename = $self->{'manifest'};
413
414 if (!&util::filename_is_absolute($manifest_filename)) {
415 $manifest_filename = &util::filename_cat ($ENV{'GSDLCOLLECTDIR'}, $manifest_filename);
416 }
417
418 $self->{'manifest'} =~ s/[\\\/]+/\//g;
419 $self->{'manifest'} =~ s/\/$//;
420
421 $manifest_lookup->parse($manifest_filename);
422 }
423
424 my $manifest = $self->{'manifest'};
425
426 # load all the plugins
427 my $plugins = [];
428 if (defined $collectcfg->{'plugin'}) {
429 $plugins = $collectcfg->{'plugin'};
430 }
431
432 my $plugin_incr_mode = $incremental_mode;
433 if ($manifest ne "") {
434 # if we have a manifest file, then we pretend we are fully incremental for plugins
435 $plugin_incr_mode = "all";
436 }
437 #some global options for the plugins
438 my @global_opts = ();
439
440 my $pluginfo = &plugin::load_plugins ($plugins, $verbosity, $out, $faillog, \@global_opts, $plugin_incr_mode);
441 if (scalar(@$pluginfo) == 0) {
442 &gsprintf($out, "{import.no_plugins_loaded}\n");
443 die "\n";
444 }
445
446 # remove the old contents of the archives directory (and tmp
447 # directory) if needed
448
449 if ($removeold) {
450 if (-e $archivedir) {
451 &gsprintf($out, "{import.removing_archives}\n");
452 &util::rm_r ($archivedir);
453 }
454 my $tmpdir = &util::filename_cat ($ENV{'GSDLCOLLECTDIR'}, "tmp");
455 $tmpdir =~ s/[\\\/]+/\//g;
456 $tmpdir =~ s/\/$//;
457 if (-e $tmpdir) {
458 &gsprintf($out, "{import.removing_tmpdir}\n");
459 &util::rm_r ($tmpdir);
460 }
461 }
462
463 # create the archives dir if needed
464 &util::mk_all_dir($archivedir);
465
466 # read the archive information file
467
468 # BACKWARDS COMPATIBILITY: Just in case there are old .ldb/.bdb files (won't do anything for other infodbtypes)
469 &util::rename_ldb_or_bdb_file(&util::filename_cat($archivedir, "archiveinf-doc"));
470 &util::rename_ldb_or_bdb_file(&util::filename_cat($archivedir, "archiveinf-src"));
471
472 my $arcinfo_doc_filename = &dbutil::get_infodb_file_path($collectcfg->{'infodbtype'}, "archiveinf-doc", $archivedir);
473 my $arcinfo_src_filename = &dbutil::get_infodb_file_path($collectcfg->{'infodbtype'}, "archiveinf-src", $archivedir);
474
475 my $archive_info = new arcinfo ($collectcfg->{'infodbtype'});
476 $archive_info->load_info ($arcinfo_doc_filename);
477
478 if ($manifest eq "") {
479 # Load in list of files in import folder from last import (if present)
480 $archive_info->load_prev_import_filelist ($arcinfo_src_filename);
481 }
482
483 ####Use Plugout####
484 my $plugout;
485
486 if ($inexport_mode eq "import") {
487 if (defined $collectcfg->{'plugout'}) {
488 # If a plugout was specified in the collect.cfg file, assume it is sensible
489 # We can't check the name because it could be anything, if it is a custom plugout
490 $plugout = $collectcfg->{'plugout'};
491 }
492 else{
493 if ($saveas !~ /^(GreenstoneXML|GreenstoneMETS)$/) {
494 push @$plugout,"GreenstoneXMLPlugout";
495 }
496 else{
497 push @$plugout,$saveas."Plugout";
498 }
499 }
500 }
501 else {
502 if (defined $collectcfg->{'plugout'} && $collectcfg->{'plugout'} =~ /^(.*METS|DSpace|MARCXML)Plugout/) {
503 $plugout = $collectcfg->{'plugout'};
504 }
505 else{
506 if ($saveas !~ /^(GreenstoneMETS|FedoraMETS|DSpace|MARCXML)$/) {
507 push @$plugout,"GreenstoneMETSPlugout";
508 }
509 else{
510 push @$plugout,$saveas."Plugout";
511 }
512 }
513 }
514
515 my $plugout_name = $plugout->[0];
516
517 push @$plugout,("-output_info",$archive_info) if (defined $archive_info);
518 push @$plugout,("-verbosity",$verbosity) if (defined $verbosity);
519 push @$plugout,("-debug") if ($debug);
520 push @$plugout,("-group_size",$groupsize) if (defined $groupsize);
521 push @$plugout,("-gzip_output") if ($gzip);
522 push @$plugout,("-output_handle",$out) if (defined $out);
523
524 push @$plugout,("-xslt_file",$xsltfile) if (defined $xsltfile && $xsltfile ne "");
525
526 if ($plugout_name =~ m/^MARCXMLPlugout$/) {
527 push @$plugout,("-group") if ($group_marc);
528 push @$plugout,("-mapping_file",$mapping_file) if (defined $mapping_file && $mapping_file ne "");
529 }
530 if ($plugout_name =~ m/^.*METSPlugout$/) {
531 push @$plugout,("-xslt_mets",$xslt_mets) if (defined $xslt_mets && $xslt_mets ne "");
532 push @$plugout,("-xslt_txt",$xslt_txt) if (defined $xslt_txt && $xslt_txt ne "");
533 }
534
535 if ($plugout_name eq "FedoraMETSPlugout") {
536 push @$plugout,("-fedora_namespace",$fedora_namespace) if (defined $fedora_namespace && $fedora_namespace ne "");
537 }
538
539 if ($plugout_name eq "DSpacePlugout") {
540 push @$plugout,("-metadata_prefix",$metadata_prefix) if (defined $metadata_prefix && $metadata_prefix ne "");
541 }
542
543 my $processor = &plugout::load_plugout($plugout);
544 $processor->setoutputdir ($archivedir);
545 $processor->set_sortmeta ($sortmeta, $removeprefix, $removesuffix) if defined $sortmeta;
546 $processor->set_OIDtype ($OIDtype, $OIDmetadata);
547
548 &plugin::begin($pluginfo, $importdir, $processor, $maxdocs, $gli);
549
550 if ($removeold) {
551 # occasionally, plugins may want to do something on remove
552 # old, eg pharos image indexing
553 &plugin::remove_all($pluginfo, $importdir, $processor, $maxdocs, $gli);
554 }
555
556 # process the import directory
557 my $block_hash = {};
558 $block_hash->{'new_files'} = {};
559 $block_hash->{'reindex_files'} = {};
560 my $metadata = {};
561
562 # global blocking pass may set up some metadata
563 &plugin::file_block_read($pluginfo, $importdir, "", $block_hash, $metadata, $gli);
564
565 if ($manifest ne "") {
566 #
567 # 1. Process delete files first
568 #
569 my @deleted_files = keys %{$manifest_lookup->{'delete'}};
570 my @full_deleted_files = ();
571
572 # ensure all filenames are absolute
573 foreach my $df (@deleted_files) {
574 my $full_df =
575 (&util::filename_is_absolute($df))
576 ? $df
577 : &util::filename_cat($importdir,$df);
578
579 if (-d $full_df) {
580 &add_dir_contents_to_list($full_df, \@full_deleted_files);
581 } else {
582 push(@full_deleted_files,$full_df);
583 }
584 }
585
586 &plugin::remove_some($pluginfo, $collectcfg->{'infodbtype'}, $archivedir, \@full_deleted_files);
587 mark_docs_for_deletion($archive_info,{},
588 \@full_deleted_files,
589 $archivedir, $verbosity, "delete");
590
591
592 #
593 # 2. Now files for reindexing
594 #
595
596 my @reindex_files = keys %{$manifest_lookup->{'reindex'}};
597 my @full_reindex_files = ();
598 # ensure all filenames are absolute
599 foreach my $rf (@reindex_files) {
600 my $full_rf =
601 (&util::filename_is_absolute($rf))
602 ? $rf
603 : &util::filename_cat($importdir,$rf);
604
605 if (-d $full_rf) {
606 &add_dir_contents_to_list($full_rf, \@full_reindex_files);
607 } else {
608 push(@full_reindex_files,$full_rf);
609 }
610 }
611
612 &plugin::remove_some($pluginfo, $collectcfg->{'infodbtype'}, $archivedir, \@full_reindex_files);
613 mark_docs_for_deletion($archive_info,{},\@full_reindex_files, $archivedir,$verbosity, "reindex");
614
615 # And now to ensure the new version of the file processed by
616 # appropriate plugin, we need to add it to block_hash reindex list
617 foreach my $full_rf (@full_reindex_files) {
618 $block_hash->{'reindex_files'}->{$full_rf} = 1;
619 }
620
621
622 #
623 # 3. Now finally any new files - add to block_hash new_files list
624 #
625
626 my @new_files = keys %{$manifest_lookup->{'index'}};
627 my @full_new_files = ();
628
629 foreach my $nf (@new_files) {
630 # ensure filename is absolute
631 my $full_nf =
632 (&util::filename_is_absolute($nf))
633 ? $nf
634 : &util::filename_cat($importdir,$nf);
635
636 if (-d $full_nf) {
637 &add_dir_contents_to_list($full_nf, \@full_new_files);
638 } else {
639 push(@full_new_files,$full_nf);
640 }
641 }
642
643 my $arcinfo_src_filename = &dbutil::get_infodb_file_path($collectcfg->{'infodbtype'}, "archiveinf-src", $archivedir);
644 my $arcinfodb_map = {};
645 &dbutil::read_infodb_file($collectcfg->{'infodbtype'}, $arcinfo_src_filename, $arcinfodb_map);
646 foreach my $f (@full_new_files) {
647 # check that we haven't seen it already
648 if (defined $arcinfodb_map->{$f}) {
649 # TODO make better warning
650 print STDERR "Warning: $f already in src archive, \n";
651 } else {
652 $block_hash->{'new_files'}->{$f} = 1;
653 }
654 }
655
656 undef $arcinfodb_map;
657 }
658 else {
659 # if incremental, we read through the import folder to see whats changed.
660
661 if ($incremental || $incremental_mode eq "onlyadd") {
662 prime_doc_oid_count($archivedir);
663
664 # Can now work out which files were new, already existed, and have
665 # been deleted
666
667 new_vs_old_import_diff($archive_info,$block_hash,$importdir,
668 $archivedir,$verbosity,$incremental_mode);
669
670 my @new_files = sort keys %{$block_hash->{'new_files'}};
671 if (scalar(@new_files>0)) {
672 print STDERR "New files and modified metadata files since last import:\n ";
673 print STDERR join("\n ",@new_files), "\n";
674 }
675
676 if ($incremental) {
677 # only look for deletions if we are truely incremental
678 my @deleted_files = sort keys %{$block_hash->{'deleted_files'}};
679 # Filter out any in gsdl/tmp area
680 my @filtered_deleted_files = ();
681 my $gsdl_tmp_area = &util::filename_cat($ENV{'GSDLHOME'}, "tmp");
682 my $collect_tmp_area = &util::filename_cat($ENV{'GSDLCOLLECTDIR'}, "tmp");
683 $gsdl_tmp_area = &util::filename_to_regex($gsdl_tmp_area);
684 $collect_tmp_area = &util::filename_to_regex($collect_tmp_area);
685
686 foreach my $df (@deleted_files) {
687 next if ($df =~ m/^$gsdl_tmp_area/);
688 next if ($df =~ m/^$collect_tmp_area/);
689
690 push(@filtered_deleted_files,$df);
691 }
692
693
694 @deleted_files = @filtered_deleted_files;
695
696 if (scalar(@deleted_files)>0) {
697 print STDERR "Files deleted since last import:\n ";
698 print STDERR join("\n ",@deleted_files), "\n";
699
700
701 &plugin::remove_some($pluginfo, $collectcfg->{'infodbtype'}, $archivedir, \@deleted_files);
702
703 mark_docs_for_deletion($archive_info,$block_hash,\@deleted_files, $archivedir,$verbosity, "delete");
704 }
705
706 my @reindex_files = sort keys %{$block_hash->{'reindex_files'}};
707
708 if (scalar(@reindex_files)>0) {
709 print STDERR "Files to reindex since last import:\n ";
710 print STDERR join("\n ",@reindex_files), "\n";
711 &plugin::remove_some($pluginfo, $collectcfg->{'infodbtype'}, $archivedir, \@reindex_files);
712 mark_docs_for_deletion($archive_info,$block_hash,\@reindex_files, $archivedir,$verbosity, "reindex");
713 }
714
715 }
716 }
717 }
718
719 # Check for existence of the file that's to contain earliestDateStamp in archivesdir
720 # Do nothing if the file already exists (file exists on incremental build).
721 # If the file doesn't exist, as happens on full build, create it and write out the current datestamp into it
722 # In buildcol, read the file's contents and set the earliestdateStamp in GS2's build.cfg / GS3's buildconfig.xml
723 # In doc.pm have set_oaiLastModified similar to set_lastmodified, and create the doc fields
724 # oailastmodified and oailastmodifieddate
725 my $earliestDatestampFile = &util::filename_cat($archivedir, "earliestDatestamp");
726 if (!-f $earliestDatestampFile && -d $archivedir) {
727 my $current_time_in_seconds = time; # in seconds
728
729 if(open(FOUT, ">$earliestDatestampFile")) {
730 # || (&gsprintf(STDERR, "{common.cannot_open}: $!\n", $earliestDatestampFile) && die);
731 print FOUT $current_time_in_seconds;
732 close(FOUT);
733 }
734 else {
735 &gsprintf(STDERR, "{import.cannot_write_earliestdatestamp}\n", $earliestDatestampFile);
736 }
737
738 }
739
740 # now, whichever mode we are in, we can process the entire import folder
741 if ((defined $jobs) && ($jobs > 1))
742 {
743 # if jobs are set to >1, run in parallel using MPI helper
744 # [hs, 1 july 2010]
745 &ParallelInexport::farm_out_processes($jobs, $epoch, $importdir, $block_hash,
746 $self->{'collection'}, $self->{'site'});
747 }
748 else
749 {
750 &plugin::read ($pluginfo, $importdir, "", $block_hash, $metadata, $processor, $maxdocs, 0, $gli);
751 }
752
753
754 if ($saveas eq "FedoraMETS") {
755 # create collection "doc obj" for Fedora that contains
756 # collection-level metadata
757
758 my $doc_obj = new doc($config_filename,"nonindexed_doc","none");
759 $doc_obj->set_OID("collection");
760
761 my $col_name = undef;
762 my $col_meta = $collectcfg->{'collectionmeta'};
763
764 if (defined $col_meta) {
765 store_collectionmeta($col_meta,"collectionname",$doc_obj); # in GS3 this is a collection's name
766 store_collectionmeta($col_meta,"collectionextra",$doc_obj); # in GS3 this is a collection's description
767 }
768 $processor->process($doc_obj);
769 }
770
771 &plugin::end($pluginfo, $processor);
772
773 &plugin::deinit($pluginfo, $processor);
774
775 # Store the value of OIDCount (used in doc.pm) so it can be
776 # restored correctly to this value on an incremental build
777 store_doc_oid_count($archivedir);
778
779 # write out the archive information file
780 $processor->close_file_output() if (defined $groupsize) && ($groupsize > 1);
781 $processor->close_group_output() if $processor->is_group();
782
783 # for backwards compatability with archvies.inf file
784 if ($arcinfo_doc_filename =~ m/(contents)|(\.inf)$/) {
785 $archive_info->save_info($arcinfo_doc_filename);
786 }
787 else {
788 $archive_info->save_revinfo_db($arcinfo_src_filename);
789 }
790
791 return $pluginfo;
792}
793
794
795sub generate_statistics
796{
797 my $self = shift @_;
798 my ($pluginfo) = @_;
799
800 my $inexport_mode = $self->{'mode'};
801
802 my $statsfile = $self->{'statsfile'};
803 my $out = $self->{'out'};
804 my $faillogname = $self->{'faillogname'};
805 my $gli = $self->{'gli'};
806 my $jobs = $self->{'jobs'};
807
808 # write out import stats
809
810 if ((!defined $jobs) || ($jobs == 1))
811 {
812 # only output statistics if there are multiple jobs
813 # [hs, 1 july 2010]
814
815 my $close_stats = 0;
816 if ($statsfile !~ /^(STDERR|STDOUT)$/i) {
817 if (open (STATS, ">$statsfile")) {
818 $statsfile = 'inexport::STATS';
819 $close_stats = 1;
820 } else {
821 &gsprintf($out, "{import.cannot_open_stats_file}", $statsfile);
822 &gsprintf($out, "{import.stats_backup}\n");
823 $statsfile = 'STDERR';
824 }
825 }
826
827 &gsprintf($out, "\n");
828 &gsprintf($out, "*********************************************\n");
829 &gsprintf($out, "{$inexport_mode.complete}\n");
830 &gsprintf($out, "*********************************************\n");
831
832 &plugin::write_stats($pluginfo, $statsfile, $faillogname, $gli);
833 if ($close_stats) {
834 close STATS;
835 }
836 }
837
838 close OUT if $self->{'close_out'};
839 close FAILLOG;
840}
841
842
843sub store_collectionmeta
844{
845 my ($collectionmeta,$field,$doc_obj) = @_;
846
847 my $section = $doc_obj->get_top_section();
848
849 my $field_hash = $collectionmeta->{$field};
850
851 foreach my $k (keys %$field_hash)
852 {
853 my $val = $field_hash->{$k};
854
855 ### print STDERR "*** $k = $field_hash->{$k}\n";
856
857 my $md_label = "ex.$field";
858
859
860 if ($k =~ m/^\[l=(.*?)\]$/)
861 {
862
863 my $md_suffix = $1;
864 $md_label .= "^$md_suffix";
865 }
866
867
868 $doc_obj->add_utf8_metadata($section,$md_label, $val);
869
870 # see collConfigxml.pm: GS2's "collectionextra" is called "description" in GS3,
871 # while "collectionname" in GS2 is called "name" in GS3.
872 # Variable $nameMap variable in collConfigxml.pm maps between GS2 and GS3
873 if (($md_label eq "ex.collectionname^en") || ($md_label eq "ex.collectionname"))
874 {
875 $doc_obj->add_utf8_metadata($section,"dc.Title", $val);
876 }
877
878 }
879}
880
881
882sub oid_count_file {
883 my ($archivedir) = @_;
884 return &util::filename_cat ($archivedir, "OIDcount");
885}
886
887
888sub prime_doc_oid_count
889{
890 my ($archivedir) = @_;
891 my $oid_count_filename = &oid_count_file($archivedir);
892
893 if (-e $oid_count_filename) {
894 if (open(OIDIN,"<$oid_count_filename")) {
895 my $OIDcount = <OIDIN>;
896 chomp $OIDcount;
897 close(OIDIN);
898
899 $doc::OIDcount = $OIDcount;
900 }
901 else {
902 &gsprintf(STDERR, "{import.cannot_read_OIDcount}\n", $oid_count_filename);
903 }
904 }
905
906}
907
908sub store_doc_oid_count
909{
910 # Use the file "OIDcount" in the archives directory to record
911 # what value doc.pm got up to
912
913 my ($archivedir) = @_;
914 my $oid_count_filename = &oid_count_file($archivedir);
915
916
917 if (open(OIDOUT,">$oid_count_filename")) {
918 print OIDOUT $doc::OIDcount, "\n";
919
920 close(OIDOUT);
921 }
922 else {
923 &gsprintf(STDERR, "{import.cannot_write_OIDcount}\n", $oid_count_filename);
924 }
925}
926
927
928
929sub new_vs_old_import_diff
930{
931 my ($archive_info,$block_hash,$importdir,$archivedir,$verbosity,$incremental_mode) = @_;
932
933 # Get the infodbtype value for this collection from the arcinfo object
934 my $infodbtype = $archive_info->{'infodbtype'};
935
936 # in this method, we want to know if metadata files are modified or not.
937 my $arcinfo_doc_filename = &dbutil::get_infodb_file_path($infodbtype, "archiveinf-doc", $archivedir);
938
939 my $archiveinf_timestamp = -M $arcinfo_doc_filename;
940
941 # First convert all files to absolute form
942 # This is to support the situation where the import folder is not
943 # the default
944
945 my $prev_all_files = $archive_info->{'prev_import_filelist'};
946 my $full_prev_all_files = {};
947
948 foreach my $prev_file (keys %$prev_all_files) {
949
950 if (!&util::filename_is_absolute($prev_file)) {
951 my $full_prev_file = &util::filename_cat($ENV{'GSDLCOLLECTDIR'},$prev_file);
952 $full_prev_all_files->{$full_prev_file} = $prev_file;
953 }
954 else {
955 $full_prev_all_files->{$prev_file} = $prev_file;
956 }
957 }
958
959
960 # Figure out which are the new files, existing files and so
961 # by implication the files from the previous import that are not
962 # there any more => mark them for deletion
963 foreach my $curr_file (keys %{$block_hash->{'all_files'}}) {
964
965 my $full_curr_file = $curr_file;
966
967 # entry in 'all_files' is moved to either 'existing_files',
968 # 'deleted_files', 'new_files', or 'new_or_modified_metadata_files'
969
970 if (!&util::filename_is_absolute($curr_file)) {
971 # add in import dir to make absolute
972 $full_curr_file = &util::filename_cat($importdir,$curr_file);
973 }
974
975 # figure out if new file or not
976 if (defined $full_prev_all_files->{$full_curr_file}) {
977 # delete it so that only files that need deleting are left
978 delete $full_prev_all_files->{$full_curr_file};
979
980 # had it before. is it a metadata file?
981 if ($block_hash->{'metadata_files'}->{$full_curr_file}) {
982
983 # is it modified??
984 if (-M $full_curr_file < $archiveinf_timestamp) {
985 print STDERR "*** Detected a *modified metadata* file: $full_curr_file\n" if $verbosity >= 2;
986 # its newer than last build
987 $block_hash->{'new_or_modified_metadata_files'}->{$full_curr_file} = 1;
988 }
989 }
990 else {
991 if ($incremental_mode eq "all") {
992
993 # had it before
994 $block_hash->{'existing_files'}->{$full_curr_file} = 1;
995
996 }
997 else {
998 # Warning in "onlyadd" mode, but had it before!
999 print STDERR "Warning: File $full_curr_file previously imported.\n";
1000 print STDERR " Treating as new file\n";
1001
1002 $block_hash->{'new_files'}->{$full_curr_file} = 1;
1003
1004 }
1005 }
1006 }
1007 else {
1008 if ($block_hash->{'metadata_files'}->{$full_curr_file}) {
1009 # the new file is the special sort of file greenstone uses
1010 # to attach metadata to src documents
1011 # i.e metadata.xml
1012 # (but note, the filename used is not constrained in
1013 # Greenstone to always be this)
1014
1015 print STDERR "*** Detected *new* metadata file: $full_curr_file\n" if $verbosity >= 2;
1016 $block_hash->{'new_or_modified_metadata_files'}->{$full_curr_file} = 1;
1017 }
1018 else {
1019 $block_hash->{'new_files'}->{$full_curr_file} = 1;
1020 }
1021 }
1022
1023
1024 delete $block_hash->{'all_files'}->{$curr_file};
1025 }
1026
1027
1028
1029
1030 # Deal with complication of new or modified metadata files by forcing
1031 # everything from this point down in the file hierarchy to
1032 # be freshly imported.
1033 #
1034 # This may mean files that have not changed are reindexed, but does
1035 # guarantee by the end of processing all new metadata is correctly
1036 # associated with the relevant document(s).
1037
1038 foreach my $new_mdf (keys %{$block_hash->{'new_or_modified_metadata_files'}}) {
1039 my ($fileroot,$situated_dir,$ext) = fileparse($new_mdf, "\\.[^\\.]+\$");
1040
1041 $situated_dir =~ s/[\\\/]+$//; # remove tailing slashes
1042 $situated_dir =~ s/\\/\\\\/g; # need to protect windows slash \ in regular expression
1043
1044 # Go through existing_files, and mark anything that is contained
1045 # within 'situated_dir' to be reindexed (in case some of the metadata
1046 # attaches to one of these files)
1047
1048 my $reindex_files = [];
1049
1050 foreach my $existing_f (keys %{$block_hash->{'existing_files'}}) {
1051
1052 if ($existing_f =~ m/^$situated_dir/) {
1053
1054 print STDERR "**** Existing file $existing_f\nis located within\n$situated_dir\n";
1055
1056 push(@$reindex_files,$existing_f);
1057 $block_hash->{'reindex_files'}->{$existing_f} = 1;
1058 delete $block_hash->{'existing_files'}->{$existing_f};
1059
1060 }
1061 }
1062
1063 # metadata file needs to be in new_files list so parsed by MetadataXMLPlug
1064 # (or equivalent)
1065 $block_hash->{'new_files'}->{$new_mdf} = 1;
1066
1067 }
1068
1069 # go through remaining existing files and work out what has changed and needs to be reindexed.
1070 my @existing_files = sort keys %{$block_hash->{'existing_files'}};
1071
1072 my $reindex_files = [];
1073
1074 foreach my $existing_filename (@existing_files) {
1075 if (-M $existing_filename < $archiveinf_timestamp) {
1076 # file is newer than last build
1077
1078 my $existing_file = $existing_filename;
1079 #my $collectdir = &util::filename_cat($ENV{'GSDLCOLLECTDIR'});
1080
1081 #my $collectdir_resafe = &util::filename_to_regex($collectdir);
1082 #$existing_file =~ s/^$collectdir_resafe(\\|\/)?//;
1083
1084 print STDERR "**** Reindexing existing file: $existing_file\n";
1085
1086 push(@$reindex_files,$existing_file);
1087 $block_hash->{'reindex_files'}->{$existing_filename} = 1;
1088 }
1089
1090 }
1091
1092
1093 # By this point full_prev_all_files contains the files
1094 # mentioned in archiveinf-src.db but are not in the 'import'
1095 # folder (or whatever was specified through -importdir ...)
1096
1097 # This list can contain files that were created in the 'tmp' or
1098 # 'cache' areas (such as screen-size and thumbnail images).
1099 #
1100 # In building the final list of files to delete, we test to see if
1101 # it exists on the filesystem and if it does (unusual for a "normal"
1102 # file in import, but possible in the case of 'tmp' files),
1103 # supress it from going into the final list
1104
1105 my $collectdir = $ENV{'GSDLCOLLECTDIR'};
1106
1107 my @deleted_files = values %$full_prev_all_files;
1108 map { my $curr_file = $_;
1109 my $full_curr_file = $curr_file;
1110
1111 if (!&util::filename_is_absolute($curr_file)) {
1112 # add in import dir to make absolute
1113
1114 $full_curr_file = &util::filename_cat($collectdir,$curr_file);
1115 }
1116
1117
1118 if (!-e $full_curr_file) {
1119 $block_hash->{'deleted_files'}->{$curr_file} = 1;
1120 }
1121 } @deleted_files;
1122
1123
1124
1125}
1126
1127
1128# this is used to delete "deleted" docs, and to remove old versions of "changed" docs
1129# $mode is 'delete' or 'reindex'
1130sub mark_docs_for_deletion
1131{
1132 my ($archive_info,$block_hash,$deleted_files,$archivedir,$verbosity,$mode) = @_;
1133
1134 my $mode_text = "deleted from index";
1135 if ($mode eq "reindex") {
1136 $mode_text = "reindexed";
1137 }
1138
1139 # Get the infodbtype value for this collection from the arcinfo object
1140 my $infodbtype = $archive_info->{'infodbtype'};
1141
1142 my $arcinfo_doc_filename = &dbutil::get_infodb_file_path($infodbtype, "archiveinf-doc", $archivedir);
1143 my $arcinfo_src_filename = &dbutil::get_infodb_file_path($infodbtype, "archiveinf-src", $archivedir);
1144
1145
1146 # record files marked for deletion in arcinfo
1147 foreach my $file (@$deleted_files) {
1148 # use 'archiveinf-src' info database file to look up all the OIDs
1149 # that this file is used in (note in most cases, it's just one OID)
1150
1151 my $src_rec = &dbutil::read_infodb_entry($infodbtype, $arcinfo_src_filename, $file);
1152 my $oids = $src_rec->{'oid'};
1153 my $file_record_deleted = 0;
1154
1155 # delete the src record
1156 my $src_infodb_file_handle = &dbutil::open_infodb_write_handle($infodbtype, $arcinfo_src_filename, "append");
1157 &dbutil::delete_infodb_entry($infodbtype, $src_infodb_file_handle, $file);
1158 &dbutil::close_infodb_write_handle($infodbtype, $src_infodb_file_handle);
1159
1160
1161 foreach my $oid (@$oids) {
1162
1163 # find the source doc (the primary file that becomes this oid)
1164 my $doc_rec = &dbutil::read_infodb_entry($infodbtype, $arcinfo_doc_filename, $oid);
1165 my $doc_source_file = $doc_rec->{'src-file'}->[0];
1166 if (!&util::filename_is_absolute($doc_source_file)) {
1167 $doc_source_file = &util::filename_cat($ENV{'GSDLCOLLECTDIR'},$doc_source_file);
1168 }
1169
1170 if ($doc_source_file ne $file) {
1171 # its an associated or metadata file
1172
1173 # mark source doc for reimport as one of its assoc files has changed or deleted
1174 $block_hash->{'reindex_files'}->{$doc_source_file} = 1;
1175
1176 }
1177 my $curr_status = $archive_info->get_status_info($oid);
1178 if (defined($curr_status) && (($curr_status ne "D"))) {
1179 if ($verbosity>1) {
1180 print STDERR "$oid ($doc_source_file) marked to be $mode_text on next buildcol.pl\n";
1181 }
1182 # mark oid for deletion (it will be deleted or reimported)
1183 $archive_info->set_status_info($oid,"D");
1184 my $val = &dbutil::read_infodb_rawentry($infodbtype, $arcinfo_doc_filename, $oid);
1185 $val =~ s/^<index-status>(.*)$/<index-status>D/m;
1186
1187 my $val_rec = &dbutil::convert_infodb_string_to_hash($val);
1188 my $doc_infodb_file_handle = &dbutil::open_infodb_write_handle($infodbtype, $arcinfo_doc_filename, "append");
1189
1190 &dbutil::write_infodb_entry($infodbtype, $doc_infodb_file_handle, $oid, $val_rec);
1191 &dbutil::close_infodb_write_handle($infodbtype, $doc_infodb_file_handle);
1192 }
1193 }
1194
1195 }
1196
1197 # now go through and check that we haven't marked any primary
1198 # files for reindex (because their associated files have
1199 # changed/deleted) when they have been deleted themselves. only in
1200 # delete mode.
1201
1202 if ($mode eq "delete") {
1203 foreach my $file (@$deleted_files) {
1204 if (defined $block_hash->{'reindex_files'}->{$file}) {
1205 delete $block_hash->{'reindex_files'}->{$file};
1206 }
1207 }
1208 }
1209
1210
1211}
1212
1213sub add_dir_contents_to_list {
1214
1215 my ($dirname, $list) = @_;
1216
1217 # Recur over directory contents.
1218 my (@dir, $subfile);
1219
1220 # find all the files in the directory
1221 if (!opendir (DIR, $dirname)) {
1222 print STDERR "inexport: WARNING - couldn't read directory $dirname\n";
1223 return -1; # error in processing
1224 }
1225 @dir = readdir (DIR);
1226 closedir (DIR);
1227
1228 for (my $i = 0; $i < scalar(@dir); $i++) {
1229 my $subfile = $dir[$i];
1230 next if ($subfile =~ m/^\.\.?$/);
1231 next if ($subfile =~ /^\.svn$/);
1232 my $full_file = &util::filename_cat($dirname, $subfile);
1233 if (-d $full_file) {
1234 &add_dir_contents_to_list($full_file, $list);
1235 } else {
1236 push (@$list, $full_file);
1237 }
1238 }
1239
1240}
1241
1242
12431;
Note: See TracBrowser for help on using the repository browser.