| 1 | ###########################################################################
|
|---|
| 2 | #
|
|---|
| 3 | # DirectoryPlugin.pm --
|
|---|
| 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 |
|
|---|
| 26 | # DirectoryPlugin is a plugin which recurses through directories processing
|
|---|
| 27 | # each file it finds - which basically means passing it down the plugin
|
|---|
| 28 | # pipeline
|
|---|
| 29 |
|
|---|
| 30 | package DirectoryPlugin;
|
|---|
| 31 |
|
|---|
| 32 | use extrametautil;
|
|---|
| 33 | use PrintInfo;
|
|---|
| 34 | use plugin;
|
|---|
| 35 | use util;
|
|---|
| 36 | use FileUtils;
|
|---|
| 37 | use metadatautil;
|
|---|
| 38 |
|
|---|
| 39 | use File::Basename;
|
|---|
| 40 | use strict;
|
|---|
| 41 | no strict 'refs';
|
|---|
| 42 | no strict 'subs';
|
|---|
| 43 |
|
|---|
| 44 | use Encode::Locale;
|
|---|
| 45 | use Encode;
|
|---|
| 46 | use Unicode::Normalize;
|
|---|
| 47 |
|
|---|
| 48 | BEGIN {
|
|---|
| 49 | @DirectoryPlugin::ISA = ('PrintInfo');
|
|---|
| 50 | }
|
|---|
| 51 |
|
|---|
| 52 | my $arguments =
|
|---|
| 53 | [ { 'name' => "block_exp",
|
|---|
| 54 | 'desc' => "{BasePlugin.block_exp}",
|
|---|
| 55 | 'type' => "regexp",
|
|---|
| 56 | 'deft' => &get_default_block_exp(),
|
|---|
| 57 | 'reqd' => "no" },
|
|---|
| 58 | # this option has been deprecated. leave it here for now so we can warn people not to use it
|
|---|
| 59 | { 'name' => "use_metadata_files",
|
|---|
| 60 | 'desc' => "{DirectoryPlugin.use_metadata_files}",
|
|---|
| 61 | 'type' => "flag",
|
|---|
| 62 | 'reqd' => "no",
|
|---|
| 63 | 'hiddengli' => "yes" },
|
|---|
| 64 | { 'name' => "recheck_directories",
|
|---|
| 65 | 'desc' => "{DirectoryPlugin.recheck_directories}",
|
|---|
| 66 | 'type' => "flag",
|
|---|
| 67 | 'reqd' => "no" } ];
|
|---|
| 68 |
|
|---|
| 69 | my $options = { 'name' => "DirectoryPlugin",
|
|---|
| 70 | 'desc' => "{DirectoryPlugin.desc}",
|
|---|
| 71 | 'abstract' => "no",
|
|---|
| 72 | 'inherits' => "yes",
|
|---|
| 73 | 'args' => $arguments };
|
|---|
| 74 |
|
|---|
| 75 | sub new {
|
|---|
| 76 | my ($class) = shift (@_);
|
|---|
| 77 | my ($pluginlist,$inputargs,$hashArgOptLists) = @_;
|
|---|
| 78 | push(@$pluginlist, $class);
|
|---|
| 79 |
|
|---|
| 80 | push(@{$hashArgOptLists->{"ArgList"}},@{$arguments});
|
|---|
| 81 | push(@{$hashArgOptLists->{"OptList"}},$options);
|
|---|
| 82 |
|
|---|
| 83 | my $self = new PrintInfo($pluginlist, $inputargs, $hashArgOptLists);
|
|---|
| 84 |
|
|---|
| 85 | print STDERR "INFO: This DirectoryPlugin supports version 2 manifest files\n";
|
|---|
| 86 |
|
|---|
| 87 | if ($self->{'info_only'}) {
|
|---|
| 88 | # don't worry about any options or initialisations etc
|
|---|
| 89 | return bless $self, $class;
|
|---|
| 90 | }
|
|---|
| 91 |
|
|---|
| 92 | # we have left this option in so we can warn people who are still using it
|
|---|
| 93 | if ($self->{'use_metadata_files'}) {
|
|---|
| 94 | die "ERROR: DirectoryPlugin -use_metadata_files option has been deprecated. Please remove the option and add MetadataXMLPlug to your plugin list instead!\n";
|
|---|
| 95 | }
|
|---|
| 96 |
|
|---|
| 97 | $self->{'num_processed'} = 0;
|
|---|
| 98 | $self->{'num_not_processed'} = 0;
|
|---|
| 99 | $self->{'num_blocked'} = 0;
|
|---|
| 100 | $self->{'num_archives'} = 0;
|
|---|
| 101 |
|
|---|
| 102 | $self->{'subdir_extrametakeys'} = {};
|
|---|
| 103 |
|
|---|
| 104 | return bless $self, $class;
|
|---|
| 105 | }
|
|---|
| 106 |
|
|---|
| 107 | # called once, at the start of processing
|
|---|
| 108 | sub init {
|
|---|
| 109 | my $self = shift (@_);
|
|---|
| 110 | my ($verbosity, $outhandle, $failhandle) = @_;
|
|---|
| 111 |
|
|---|
| 112 | # verbosity is passed through from the processor
|
|---|
| 113 | $self->{'verbosity'} = $verbosity;
|
|---|
| 114 |
|
|---|
| 115 | # as are the outhandle and failhandle
|
|---|
| 116 | $self->{'outhandle'} = $outhandle if defined $outhandle;
|
|---|
| 117 | $self->{'failhandle'} = $failhandle;
|
|---|
| 118 |
|
|---|
| 119 | }
|
|---|
| 120 |
|
|---|
| 121 | # called once, after all passes have finished
|
|---|
| 122 | sub deinit {
|
|---|
| 123 | my ($self) = @_;
|
|---|
| 124 |
|
|---|
| 125 | }
|
|---|
| 126 |
|
|---|
| 127 | # called at the beginning of each plugin pass (import has one, building has many)
|
|---|
| 128 | sub begin {
|
|---|
| 129 | my $self = shift (@_);
|
|---|
| 130 | my ($pluginfo, $base_dir, $processor, $maxdocs) = @_;
|
|---|
| 131 |
|
|---|
| 132 | # Only lookup timestamp info for import.pl, and only if incremental is set
|
|---|
| 133 | my $proc_package_name = ref $processor;
|
|---|
| 134 | if ($proc_package_name !~ /buildproc$/ && $self->{'incremental'} == 1) {
|
|---|
| 135 | # Get the infodbtype value for this collection from the arcinfo object
|
|---|
| 136 | my $infodbtype = $processor->getoutputinfo()->{'infodbtype'};
|
|---|
| 137 | $infodbtype = "gdbm" if $infodbtype eq "gdbm-txtgz"; # in archives, cannot use txtgz version
|
|---|
| 138 | my $output_dir = $processor->getoutputdir();
|
|---|
| 139 | my $archives_inf = &dbutil::get_infodb_file_path($infodbtype, "archiveinf-doc", $output_dir);
|
|---|
| 140 |
|
|---|
| 141 | if ( -e $archives_inf ) {
|
|---|
| 142 | $self->{'inf_timestamp'} = -M $archives_inf;
|
|---|
| 143 | }
|
|---|
| 144 | }
|
|---|
| 145 | }
|
|---|
| 146 |
|
|---|
| 147 | sub remove_all {
|
|---|
| 148 | my $self = shift (@_);
|
|---|
| 149 | my ($pluginfo, $base_dir, $processor, $maxdocs) = @_;
|
|---|
| 150 | }
|
|---|
| 151 |
|
|---|
| 152 |
|
|---|
| 153 | sub remove_one {
|
|---|
| 154 | my $self = shift (@_);
|
|---|
| 155 | my ($file, $oids, $archivedir) = @_;
|
|---|
| 156 | return undef; # this will never be called for directories (will it??)
|
|---|
| 157 |
|
|---|
| 158 | }
|
|---|
| 159 |
|
|---|
| 160 |
|
|---|
| 161 | # called at the end of each plugin pass
|
|---|
| 162 | sub end {
|
|---|
| 163 | my ($self) = shift (@_);
|
|---|
| 164 |
|
|---|
| 165 | }
|
|---|
| 166 |
|
|---|
| 167 |
|
|---|
| 168 |
|
|---|
| 169 | # return 1 if this class might recurse using $pluginfo
|
|---|
| 170 | sub is_recursive {
|
|---|
| 171 | my $self = shift (@_);
|
|---|
| 172 |
|
|---|
| 173 | return 1;
|
|---|
| 174 | }
|
|---|
| 175 |
|
|---|
| 176 | sub get_default_block_exp {
|
|---|
| 177 | my $self = shift (@_);
|
|---|
| 178 |
|
|---|
| 179 | return '(?i)(CVS|\.svn|Thumbs\.db|OIDcount|\.DS_Store|~)$';
|
|---|
| 180 | }
|
|---|
| 181 |
|
|---|
| 182 | sub check_directory_path {
|
|---|
| 183 |
|
|---|
| 184 | my $self = shift(@_);
|
|---|
| 185 | my ($dirname) = @_;
|
|---|
| 186 |
|
|---|
| 187 | return undef unless (-d $dirname);
|
|---|
| 188 |
|
|---|
| 189 | return 0 if ($self->{'block_exp'} ne "" && $dirname =~ /$self->{'block_exp'}/);
|
|---|
| 190 |
|
|---|
| 191 | my $outhandle = $self->{'outhandle'};
|
|---|
| 192 |
|
|---|
| 193 | # check to make sure we're not reading the archives or index directory
|
|---|
| 194 | my $gsdlhome = quotemeta($ENV{'GSDLHOME'});
|
|---|
| 195 | if ($dirname =~ m/^$gsdlhome\/.*?\/import.*?\/(archives|index)$/) {
|
|---|
| 196 | print $outhandle "DirectoryPlugin: $dirname appears to be a reference to a Greenstone collection, skipping.\n";
|
|---|
| 197 | return 0;
|
|---|
| 198 | }
|
|---|
| 199 |
|
|---|
| 200 | # check to see we haven't got a cyclic path...
|
|---|
| 201 | if ($dirname =~ m%(/.*){,41}%) {
|
|---|
| 202 | print $outhandle "DirectoryPlugin: $dirname is 40 directories deep, is this a recursive path? if not increase constant in DirectoryPlugin.pm.\n";
|
|---|
| 203 | return 0;
|
|---|
| 204 | }
|
|---|
| 205 |
|
|---|
| 206 | # check to see we haven't got a cyclic path...
|
|---|
| 207 | if ($dirname =~ m%.*?import/(.+?)/import/\1.*%) {
|
|---|
| 208 | print $outhandle "DirectoryPlugin: $dirname appears to be in a recursive loop...\n";
|
|---|
| 209 | return 0;
|
|---|
| 210 | }
|
|---|
| 211 |
|
|---|
| 212 | return 1;
|
|---|
| 213 | }
|
|---|
| 214 |
|
|---|
| 215 | # this may be called more than once
|
|---|
| 216 | sub sort_out_associated_files {
|
|---|
| 217 |
|
|---|
| 218 | my $self = shift (@_);
|
|---|
| 219 | my ($block_hash) = @_;
|
|---|
| 220 | if (!scalar (keys %{$block_hash->{'shared_fileroot'}})) {
|
|---|
| 221 | return;
|
|---|
| 222 | }
|
|---|
| 223 |
|
|---|
| 224 | $self->{'assocfile_info'} = {} unless defined $self->{'assocfile_info'};
|
|---|
| 225 | my $metadata = $self->{'assocfile_info'};
|
|---|
| 226 | foreach my $prefix (keys %{$block_hash->{'shared_fileroot'}}) {
|
|---|
| 227 | my $record = $block_hash->{'shared_fileroot'}->{$prefix};
|
|---|
| 228 |
|
|---|
| 229 | my $tie_to = $record->{'tie_to'};
|
|---|
| 230 | my $exts = $record->{'exts'};
|
|---|
| 231 |
|
|---|
| 232 | if ((defined $tie_to) && (scalar (keys %$exts) > 0)) {
|
|---|
| 233 | # set up fileblocks and assocfile_tobe
|
|---|
| 234 | my $base_file = "$prefix$tie_to";
|
|---|
| 235 | $metadata->{$base_file} = {} unless defined $metadata->{$base_file};
|
|---|
| 236 | my $base_file_metadata = $metadata->{$base_file};
|
|---|
| 237 |
|
|---|
| 238 | $base_file_metadata->{'gsdlassocfile_tobe'} = [] unless defined $base_file_metadata->{'gsdlassocfile_tobe'};
|
|---|
| 239 | my $assoc_tobe = $base_file_metadata->{'gsdlassocfile_tobe'};
|
|---|
| 240 | foreach my $e (keys %$exts) {
|
|---|
| 241 | # block the file
|
|---|
| 242 | &util::block_filename($block_hash,"$prefix$e");
|
|---|
| 243 | # set up as an associatd file
|
|---|
| 244 | print STDERR " $self->{'plugin_type'}: Associating $prefix$e with $tie_to version\n";
|
|---|
| 245 | my $mime_type = ""; # let system auto detect this
|
|---|
| 246 | push(@$assoc_tobe,"$prefix$e:$mime_type:");
|
|---|
| 247 |
|
|---|
| 248 | }
|
|---|
| 249 | }
|
|---|
| 250 | } # foreach record
|
|---|
| 251 |
|
|---|
| 252 | $block_hash->{'shared_fileroot'} = undef;
|
|---|
| 253 | $block_hash->{'shared_fileroot'} = {};
|
|---|
| 254 |
|
|---|
| 255 | }
|
|---|
| 256 |
|
|---|
| 257 |
|
|---|
| 258 | # do block exp OR special blocking ???
|
|---|
| 259 |
|
|---|
| 260 | sub file_is_blocked {
|
|---|
| 261 | my $self = shift (@_);
|
|---|
| 262 | my ($block_hash, $filename_full_path) = @_;
|
|---|
| 263 |
|
|---|
| 264 | $filename_full_path = &util::upgrade_if_dos_filename($filename_full_path);
|
|---|
| 265 |
|
|---|
| 266 | if (($ENV{'GSDLOS'} =~ m/^windows$/) && ($^O ne "cygwin")) {
|
|---|
| 267 | # on windows, all block paths are lowercased.
|
|---|
| 268 | my $lower_filename = lc ($filename_full_path);
|
|---|
| 269 | if (defined $block_hash->{'file_blocks'}->{$lower_filename}) {
|
|---|
| 270 | $self->{'num_blocked'} ++;
|
|---|
| 271 | return 1;
|
|---|
| 272 | }
|
|---|
| 273 | }
|
|---|
| 274 | else {
|
|---|
| 275 | if (defined $block_hash->{'file_blocks'}->{$filename_full_path}) {
|
|---|
| 276 | $self->{'num_blocked'} ++;
|
|---|
| 277 | return 1;
|
|---|
| 278 | }
|
|---|
| 279 | }
|
|---|
| 280 | # check Directory plugin's own block_exp
|
|---|
| 281 | if ($self->{'block_exp'} ne "" && $filename_full_path =~ /$self->{'block_exp'}/) {
|
|---|
| 282 | $self->{'num_blocked'} ++;
|
|---|
| 283 | return 1; # blocked
|
|---|
| 284 | }
|
|---|
| 285 | return 0;
|
|---|
| 286 | }
|
|---|
| 287 |
|
|---|
| 288 |
|
|---|
| 289 |
|
|---|
| 290 | sub file_block_read {
|
|---|
| 291 | my $self = shift (@_);
|
|---|
| 292 | my ($pluginfo, $base_dir, $file, $block_hash, $metadata, $gli) = @_;
|
|---|
| 293 |
|
|---|
| 294 | my $outhandle = $self->{'outhandle'};
|
|---|
| 295 | my $verbosity = $self->{'verbosity'};
|
|---|
| 296 |
|
|---|
| 297 | # Calculate the directory name and ensure it is a directory and
|
|---|
| 298 | # that it is not explicitly blocked.
|
|---|
| 299 | my $dirname = $file;
|
|---|
| 300 | $dirname = &FileUtils::filenameConcatenate($base_dir, $file) if $base_dir =~ /\w/;
|
|---|
| 301 |
|
|---|
| 302 | my $directory_ok = $self->check_directory_path($dirname);
|
|---|
| 303 | return $directory_ok unless (defined $directory_ok && $directory_ok == 1);
|
|---|
| 304 |
|
|---|
| 305 | print $outhandle "Global file scan checking directory: $dirname\n";
|
|---|
| 306 |
|
|---|
| 307 | $block_hash->{'all_files'} = {} unless defined $block_hash->{'all_files'};
|
|---|
| 308 | $block_hash->{'metadata_files'} = {} unless defined $block_hash->{'metadata_files'};
|
|---|
| 309 |
|
|---|
| 310 | $block_hash->{'file_blocks'} = {} unless defined $block_hash->{'file_blocks'};
|
|---|
| 311 | $block_hash->{'shared_fileroot'} = {} unless defined $block_hash->{'shared_fileroot'};
|
|---|
| 312 |
|
|---|
| 313 | # Recur over directory contents.
|
|---|
| 314 | my (@dir, $subfile);
|
|---|
| 315 | #my $count = 0;
|
|---|
| 316 |
|
|---|
| 317 | print $outhandle "DirectoryPlugin block: getting directory $dirname\n" if ($verbosity > 2);
|
|---|
| 318 |
|
|---|
| 319 | # find all the files in the directory
|
|---|
| 320 | if (!opendir (DIR, $dirname)) {
|
|---|
| 321 | if ($gli) {
|
|---|
| 322 | print STDERR "<ProcessingError n='$file' r='Could not read directory $dirname'>\n";
|
|---|
| 323 | }
|
|---|
| 324 | print $outhandle "DirectoryPlugin: WARNING - couldn't read directory $dirname\n";
|
|---|
| 325 | return -1; # error in processing
|
|---|
| 326 | }
|
|---|
| 327 | @dir = sort readdir (DIR);
|
|---|
| 328 | closedir (DIR);
|
|---|
| 329 |
|
|---|
| 330 | for (my $i = 0; $i < scalar(@dir); $i++) {
|
|---|
| 331 | my $raw_subfile = $dir[$i];
|
|---|
| 332 | next if ($raw_subfile =~ m/^\.\.?$/);
|
|---|
| 333 |
|
|---|
| 334 | my $this_file_base_dir = $base_dir;
|
|---|
| 335 | my $raw_file_subfile = &FileUtils::filenameConcatenate($file, $raw_subfile);
|
|---|
| 336 |
|
|---|
| 337 | # Recursively read each $raw_subfile
|
|---|
| 338 | print $outhandle "DirectoryPlugin block recurring: ". Encode::decode("utf8", $raw_file_subfile) ."\n" if ($verbosity > 2);
|
|---|
| 339 | print $outhandle "DirectoryPlugin block recurring: ". Encode::decode(locale =>$raw_file_subfile) ."\n" if ($verbosity > 2);
|
|---|
| 340 |
|
|---|
| 341 | #$count += &plugin::file_block_read ($pluginfo, $this_file_base_dir,
|
|---|
| 342 |
|
|---|
| 343 | &plugin::file_block_read ($pluginfo, $this_file_base_dir,
|
|---|
| 344 | $raw_file_subfile,
|
|---|
| 345 | $block_hash, $metadata, $gli);
|
|---|
| 346 |
|
|---|
| 347 | }
|
|---|
| 348 | $self->sort_out_associated_files($block_hash);
|
|---|
| 349 | #return $count;
|
|---|
| 350 | return 1;
|
|---|
| 351 |
|
|---|
| 352 | }
|
|---|
| 353 |
|
|---|
| 354 | # We don't do metadata_read
|
|---|
| 355 | sub metadata_read {
|
|---|
| 356 | my $self = shift (@_);
|
|---|
| 357 | my ($pluginfo, $base_dir, $file, $block_hash,
|
|---|
| 358 | $extrametakeys, $extrametadata, $extrametafile,
|
|---|
| 359 | $processor, $gli, $aux) = @_;
|
|---|
| 360 |
|
|---|
| 361 | return undef;
|
|---|
| 362 | }
|
|---|
| 363 |
|
|---|
| 364 |
|
|---|
| 365 | # return number of files processed, undef if can't process
|
|---|
| 366 | # Note that $base_dir might be "" and that $file might
|
|---|
| 367 | # include directories
|
|---|
| 368 |
|
|---|
| 369 | # This function passes around metadata hash structures. Metadata hash
|
|---|
| 370 | # structures are hashes that map from a (scalar) key (the metadata element
|
|---|
| 371 | # name) to either a scalar metadata value or a reference to an array of
|
|---|
| 372 | # such values.
|
|---|
| 373 |
|
|---|
| 374 | sub read {
|
|---|
| 375 | my $self = shift (@_);
|
|---|
| 376 | my ($pluginfo, $base_dir, $file, $block_hash, $in_metadata, $processor, $maxdocs, $total_count, $gli) = @_;
|
|---|
| 377 | my $outhandle = $self->{'outhandle'};
|
|---|
| 378 | my $verbosity = $self->{'verbosity'};
|
|---|
| 379 |
|
|---|
| 380 | # Calculate the directory name and ensure it is a directory and
|
|---|
| 381 | # that it is not explicitly blocked.
|
|---|
| 382 | my $dirname;
|
|---|
| 383 | if ($file eq "") {
|
|---|
| 384 | $dirname = $base_dir;
|
|---|
| 385 | } else {
|
|---|
| 386 | $dirname = $file;
|
|---|
| 387 | $dirname = &FileUtils::filenameConcatenate($base_dir, $file) if $base_dir =~ /\w/;
|
|---|
| 388 | }
|
|---|
| 389 |
|
|---|
| 390 | my $directory_ok = $self->check_directory_path($dirname);
|
|---|
| 391 | return $directory_ok unless (defined $directory_ok && $directory_ok == 1);
|
|---|
| 392 |
|
|---|
| 393 | if (($verbosity > 2) && ((scalar keys %$in_metadata) > 0)) {
|
|---|
| 394 | print $outhandle "DirectoryPlugin: metadata passed in: ",
|
|---|
| 395 | join(", ", keys %$in_metadata), "\n";
|
|---|
| 396 | }
|
|---|
| 397 |
|
|---|
| 398 |
|
|---|
| 399 | # Recur over directory contents.
|
|---|
| 400 | my @dir;
|
|---|
| 401 |
|
|---|
| 402 | print $outhandle "DirectoryPlugin read: getting directory $dirname\n" if ($verbosity > 2);
|
|---|
| 403 |
|
|---|
| 404 | # find all the files in the directory
|
|---|
| 405 | if (!opendir (DIR, $dirname)) {
|
|---|
| 406 | if ($gli) {
|
|---|
| 407 | print STDERR "<ProcessingError n='$file' r='Could not read directory $dirname'>\n";
|
|---|
| 408 | }
|
|---|
| 409 | print $outhandle "DirectoryPlugin: WARNING - couldn't read directory $dirname\n";
|
|---|
| 410 | return -1; # error in processing
|
|---|
| 411 | }
|
|---|
| 412 | @dir = sort readdir (DIR);
|
|---|
| 413 | map { $_ = &unicode::raw_filename_to_url_encoded($_); } @dir;
|
|---|
| 414 | closedir (DIR);
|
|---|
| 415 | # Re-order the files in the list so any directories ending with .all are moved to the end
|
|---|
| 416 | for (my $i = scalar(@dir) - 1; $i >= 0; $i--) {
|
|---|
| 417 | if (-d &FileUtils::filenameConcatenate($dirname, $dir[$i]) && $dir[$i] =~ /\.all$/) {
|
|---|
| 418 | push(@dir, splice(@dir, $i, 1));
|
|---|
| 419 | }
|
|---|
| 420 | }
|
|---|
| 421 |
|
|---|
| 422 | # Chain through to the rest of the read function (now split off and named
|
|---|
| 423 | # read_phase2)
|
|---|
| 424 | my $count = $self->read_phase2($pluginfo, $dirname, \@dir, $base_dir, $file, $block_hash, $in_metadata, $processor, $maxdocs, $total_count, $gli);
|
|---|
| 425 |
|
|---|
| 426 | return $count;
|
|---|
| 427 | }
|
|---|
| 428 |
|
|---|
| 429 | sub read_phase2
|
|---|
| 430 | {
|
|---|
| 431 | my $self = shift (@_);
|
|---|
| 432 | my ($pluginfo, $dirname, $dir_ref, $base_dir, $file, $block_hash, $in_metadata, $processor, $maxdocs, $total_count, $gli) = @_;
|
|---|
| 433 | # These were defined in read (phase 1)
|
|---|
| 434 | my @dir = @{$dir_ref};
|
|---|
| 435 | my $subfile;
|
|---|
| 436 |
|
|---|
| 437 | my $outhandle = $self->{'outhandle'};
|
|---|
| 438 | my $verbosity = $self->{'verbosity'};
|
|---|
| 439 |
|
|---|
| 440 | # setup the metadata structures. we do a metadata_read pass to see if there is any additional metadata, then pass it to read
|
|---|
| 441 |
|
|---|
| 442 | my $additionalmetadata = 0; # is there extra metadata available?
|
|---|
| 443 | my %extrametadata; # maps from filespec to extra metadata keys
|
|---|
| 444 | my %extrametafile; # maps from filespec to the metadata.xml (or similar) file it came from
|
|---|
| 445 | my @extrametakeys; # keys of %extrametadata in order read
|
|---|
| 446 |
|
|---|
| 447 |
|
|---|
| 448 | my $os_dirsep = &util::get_os_dirsep();
|
|---|
| 449 | my $dirsep = &util::get_dirsep();
|
|---|
| 450 | my $base_dir_regexp = $base_dir;
|
|---|
| 451 | $base_dir_regexp =~ s/\//$os_dirsep/g;
|
|---|
| 452 |
|
|---|
| 453 | # Want to get relative path of local_dirname within the base_directory
|
|---|
| 454 | # but with URL style slashes.
|
|---|
| 455 | my $local_dirname = &util::filename_within_directory_url_format($dirname, $base_dir);
|
|---|
| 456 |
|
|---|
| 457 | # if we are in import folder, then local_dirname will be empty
|
|---|
| 458 | if ($local_dirname ne "") {
|
|---|
| 459 | # look for extra metadata passed down from higher folders
|
|---|
| 460 | $local_dirname .= "/"; # closing slash must be URL type slash also and not $dirsep;
|
|---|
| 461 | if (defined $self->{'subdir_extrametakeys'}->{$local_dirname}) {
|
|---|
| 462 | my $extrakeys = $self->{'subdir_extrametakeys'}->{$local_dirname};
|
|---|
| 463 | foreach my $ek (@$extrakeys) {
|
|---|
| 464 | my $extrakeys_re = $ek->{'re'};
|
|---|
| 465 | my $extrakeys_md = $ek->{'md'};
|
|---|
| 466 | my $extrakeys_mf = $ek->{'mf'};
|
|---|
| 467 | &extrametautil::addmetakey(\@extrametakeys, $extrakeys_re);
|
|---|
| 468 | &extrametautil::setmetadata(\%extrametadata, $extrakeys_re, $extrakeys_md);
|
|---|
| 469 | &extrametautil::setmetafile(\%extrametafile, $extrakeys_re, $extrakeys_mf);
|
|---|
| 470 | }
|
|---|
| 471 | delete($self->{'subdir_extrametakeys'}->{$local_dirname});
|
|---|
| 472 | }
|
|---|
| 473 | }
|
|---|
| 474 | # apply metadata pass for each of the files in the directory -- ignore
|
|---|
| 475 | # maxdocs here
|
|---|
| 476 | my $num_files = scalar(@dir);
|
|---|
| 477 | for (my $i = 0; $i < scalar(@dir); $i++) {
|
|---|
| 478 | my $subfile = $dir[$i];
|
|---|
| 479 | next if ($subfile =~ m/^\.\.?$/);
|
|---|
| 480 |
|
|---|
| 481 | my $this_file_base_dir = $base_dir;
|
|---|
| 482 | my $raw_subfile = &unicode::url_encoded_to_raw_filename($subfile);
|
|---|
| 483 |
|
|---|
| 484 | my $raw_file_subfile = &FileUtils::filenameConcatenate($file, $raw_subfile);
|
|---|
| 485 | my $raw_full_filename = &FileUtils::filenameConcatenate($this_file_base_dir, $raw_file_subfile);
|
|---|
| 486 |
|
|---|
| 487 | if ($self->file_is_blocked($block_hash,$raw_full_filename)) {
|
|---|
| 488 | print STDERR "DirectoryPlugin: file $raw_full_filename was blocked for metadata_read\n" if ($verbosity > 2);
|
|---|
| 489 | next;
|
|---|
| 490 | }
|
|---|
| 491 |
|
|---|
| 492 | # Recursively read each $raw_subfile
|
|---|
| 493 | print $outhandle "DirectoryPlugin metadata recurring: $raw_subfile\n" if ($verbosity > 2);
|
|---|
| 494 |
|
|---|
| 495 | &plugin::metadata_read ($pluginfo, $this_file_base_dir,
|
|---|
| 496 | $raw_file_subfile,$block_hash,
|
|---|
| 497 | \@extrametakeys, \%extrametadata,
|
|---|
| 498 | \%extrametafile,
|
|---|
| 499 | $processor, $gli);
|
|---|
| 500 | $additionalmetadata = 1;
|
|---|
| 501 | }
|
|---|
| 502 |
|
|---|
| 503 | # filter out any extrametakeys that mention subdirectories and store
|
|---|
| 504 | # for later use (i.e. when that sub-directory is being processed)
|
|---|
| 505 | foreach my $ek (@extrametakeys) { # where each Extrametakey (which is a filename) is stored as a url-style regex
|
|---|
| 506 |
|
|---|
| 507 | my ($subdir_re,$extrakey_dir) = &util::url_fileparse($ek);
|
|---|
| 508 |
|
|---|
| 509 | if ($extrakey_dir ne "") {
|
|---|
| 510 | # a subdir was specified
|
|---|
| 511 | my $md = &extrametautil::getmetadata(\%extrametadata, $ek);
|
|---|
| 512 | my $mf = &extrametautil::getmetafile(\%extrametafile, $ek);
|
|---|
| 513 |
|
|---|
| 514 | my $subdir_extrametakeys = $self->{'subdir_extrametakeys'};
|
|---|
| 515 | my $subdir_rec = { 're' => $subdir_re, 'md' => $md, 'mf' => $mf };
|
|---|
| 516 |
|
|---|
| 517 | # when it's looked up, it must be relative to the base dir
|
|---|
| 518 | push(@{$subdir_extrametakeys->{"$local_dirname$extrakey_dir"}},$subdir_rec);
|
|---|
| 519 | }
|
|---|
| 520 | }
|
|---|
| 521 |
|
|---|
| 522 | # import each of the files in the directory
|
|---|
| 523 | my $count=0;
|
|---|
| 524 | for (my $i = 0; $i <= scalar(@dir); $i++) {
|
|---|
| 525 | # When every file in the directory has been done, pause for a moment (figuratively!)
|
|---|
| 526 | # If the -recheck_directories argument hasn't been provided, stop now (default)
|
|---|
| 527 | # Otherwise, re-read the contents of the directory to check for new files
|
|---|
| 528 | # Any new files are added to the @dir list and are processed as normal
|
|---|
| 529 | # This is necessary when documents to be indexed are specified in bibliographic DBs
|
|---|
| 530 | # These files are copied/downloaded and stored in a new folder at import time
|
|---|
| 531 | if ($i == $num_files) {
|
|---|
| 532 | last unless $self->{'recheck_directories'};
|
|---|
| 533 |
|
|---|
| 534 | # Re-read the files in the directory to see if there are any new files
|
|---|
| 535 | last if (!opendir (DIR, $dirname));
|
|---|
| 536 | my @dirnow = sort readdir (DIR);
|
|---|
| 537 | map { $_ = &unicode::raw_filename_to_url_encoded($_) } @dirnow;
|
|---|
| 538 | closedir (DIR);
|
|---|
| 539 |
|
|---|
| 540 | # We're only interested if there are more files than there were before
|
|---|
| 541 | last if (scalar(@dirnow) <= scalar(@dir));
|
|---|
| 542 |
|
|---|
| 543 | # Any new files are added to the end of @dir to get processed by the loop
|
|---|
| 544 | my $j;
|
|---|
| 545 | foreach my $subfilenow (@dirnow) {
|
|---|
| 546 | for ($j = 0; $j < $num_files; $j++) {
|
|---|
| 547 | last if ($subfilenow eq $dir[$j]);
|
|---|
| 548 | }
|
|---|
| 549 | if ($j == $num_files) {
|
|---|
| 550 | # New file
|
|---|
| 551 | push(@dir, $subfilenow);
|
|---|
| 552 | }
|
|---|
| 553 | }
|
|---|
| 554 | # When the new files have been processed, check again
|
|---|
| 555 | $num_files = scalar(@dir);
|
|---|
| 556 | }
|
|---|
| 557 |
|
|---|
| 558 | my $subfile = $dir[$i];
|
|---|
| 559 | last if ($maxdocs != -1 && ($count + $total_count) >= $maxdocs);
|
|---|
| 560 | next if ($subfile =~ /^\.\.?$/);
|
|---|
| 561 |
|
|---|
| 562 | my $this_file_base_dir = $base_dir;
|
|---|
| 563 | my $raw_subfile = &unicode::url_encoded_to_raw_filename($subfile);
|
|---|
| 564 | # get the canonical unicode version of the filename. This may not match
|
|---|
| 565 | # the filename on the file system. We will use it to compare to regex
|
|---|
| 566 | # in the metadata table.
|
|---|
| 567 | my $unicode_subfile = &util::raw_filename_to_unicode($dirname, $raw_subfile);
|
|---|
| 568 | my $raw_file_subfile = &FileUtils::filenameConcatenate($file, $raw_subfile);
|
|---|
| 569 | my $raw_full_filename
|
|---|
| 570 | = &FileUtils::filenameConcatenate($this_file_base_dir,$raw_file_subfile);
|
|---|
| 571 |
|
|---|
| 572 | if ($self->file_is_blocked($block_hash,$raw_full_filename)) {
|
|---|
| 573 | print STDERR "DirectoryPlugin: file $raw_full_filename was blocked for read\n" if ($verbosity > 2);
|
|---|
| 574 | next;
|
|---|
| 575 | }
|
|---|
| 576 | print STDERR "** DirectoryPlugin processing $raw_full_filename\n";
|
|---|
| 577 | # Follow Windows shortcuts
|
|---|
| 578 | if ($raw_subfile =~ m/(?i)\.lnk$/ && (($ENV{'GSDLOS'} =~ m/^windows$/i) && ($^O ne "cygwin"))) {
|
|---|
| 579 | require Win32::Shortcut;
|
|---|
| 580 | my $shortcut = new Win32::Shortcut(&FileUtils::filenameConcatenate($dirname, $raw_subfile));
|
|---|
| 581 | if ($shortcut) {
|
|---|
| 582 | # The file to be processed is now the target of the shortcut
|
|---|
| 583 | $this_file_base_dir = "";
|
|---|
| 584 | $file = "";
|
|---|
| 585 | $raw_subfile = $shortcut->Path;
|
|---|
| 586 | }
|
|---|
| 587 | }
|
|---|
| 588 |
|
|---|
| 589 | # check for a symlink pointing back to a leading directory
|
|---|
| 590 | if (-d "$dirname/$raw_subfile" && -l "$dirname/$raw_subfile") {
|
|---|
| 591 | # readlink gives a "fatal error" on systems that don't implement
|
|---|
| 592 | # symlinks. This assumes the the -l test above would fail on those.
|
|---|
| 593 | my $linkdest=readlink "$dirname/$raw_subfile";
|
|---|
| 594 | if (!defined ($linkdest)) {
|
|---|
| 595 | # system error - file not found?
|
|---|
| 596 | warn "DirectoryPlugin: symlink problem - $!";
|
|---|
| 597 | } else {
|
|---|
| 598 | # see if link points to current or a parent directory
|
|---|
| 599 | if ($linkdest =~ m@^[\./\\]+$@ ||
|
|---|
| 600 | index($dirname, $linkdest) != -1) {
|
|---|
| 601 | warn "DirectoryPlugin: Ignoring recursive symlink ($dirname/$raw_subfile -> $linkdest)\n";
|
|---|
| 602 | next;
|
|---|
| 603 | ;
|
|---|
| 604 | }
|
|---|
| 605 | }
|
|---|
| 606 | }
|
|---|
| 607 |
|
|---|
| 608 | print $outhandle "DirectoryPlugin: preparing metadata for $raw_subfile\n" if ($verbosity > 2);
|
|---|
| 609 |
|
|---|
| 610 | # Make a copy of $in_metadata to pass to $raw_subfile
|
|---|
| 611 | my $out_metadata = {};
|
|---|
| 612 | &metadatautil::combine_metadata_structures($out_metadata, $in_metadata);
|
|---|
| 613 |
|
|---|
| 614 | # check the assocfile_info
|
|---|
| 615 | if (defined $self->{'assocfile_info'}->{$raw_full_filename}) {
|
|---|
| 616 | &metadatautil::combine_metadata_structures($out_metadata, $self->{'assocfile_info'}->{$raw_full_filename});
|
|---|
| 617 | }
|
|---|
| 618 |
|
|---|
| 619 | ### Now we need to look up the metadata table to see if there is any
|
|---|
| 620 | # extra metadata for us. We need the canonical unicode version here.
|
|---|
| 621 | if ($additionalmetadata == 1) {
|
|---|
| 622 | foreach my $filespec (@extrametakeys) {
|
|---|
| 623 | if ($unicode_subfile =~ /^$filespec$/) {
|
|---|
| 624 | print $outhandle "File \"$unicode_subfile\" matches filespec \"$filespec\"\n"
|
|---|
| 625 | if ($verbosity > 2);
|
|---|
| 626 | my $mdref = &extrametautil::getmetadata(\%extrametadata, $filespec);
|
|---|
| 627 | my $mfref = &extrametautil::getmetafile(\%extrametafile, $filespec);
|
|---|
| 628 |
|
|---|
| 629 | # Add the list files where the metadata came from
|
|---|
| 630 | # into the metadata table so we can track this
|
|---|
| 631 | # This mechanism is similar to how gsdlassocfile works
|
|---|
| 632 |
|
|---|
| 633 | my @metafile_pair = ();
|
|---|
| 634 | foreach my $l (keys %$mfref) {
|
|---|
| 635 | my $f = $mfref->{$l};
|
|---|
| 636 | push (@metafile_pair, "$f : $l");
|
|---|
| 637 | }
|
|---|
| 638 |
|
|---|
| 639 | $mdref->{'gsdlmetafile'} = \@metafile_pair;
|
|---|
| 640 |
|
|---|
| 641 | &metadatautil::combine_metadata_structures($out_metadata, $mdref);
|
|---|
| 642 | }
|
|---|
| 643 | }
|
|---|
| 644 | }
|
|---|
| 645 |
|
|---|
| 646 | if (defined $self->{'inf_timestamp'}) {
|
|---|
| 647 | # Look to see if it's a completely new file
|
|---|
| 648 |
|
|---|
| 649 | if (!$block_hash->{'new_files'}->{$raw_full_filename}) {
|
|---|
| 650 | # Not a new file, must be an existing file
|
|---|
| 651 | # Let' see if it's newer than the last import.pl
|
|---|
| 652 |
|
|---|
| 653 |
|
|---|
| 654 | if (! -d $raw_full_filename) {
|
|---|
| 655 | if (!$block_hash->{'reindex_files'}->{$raw_full_filename}) {
|
|---|
| 656 | # filename has been around for longer than inf_timestamp
|
|---|
| 657 | print $outhandle "**** Skipping $unicode_subfile\n" if ($verbosity >3);
|
|---|
| 658 | next;
|
|---|
| 659 | }
|
|---|
| 660 | else {
|
|---|
| 661 | # Remove old folder in archives (might hash to something different)
|
|---|
| 662 | # *** should be doing this on a Del one as well
|
|---|
| 663 | # but leave folder name?? and ensure hashs to
|
|---|
| 664 | # same again??
|
|---|
| 665 |
|
|---|
| 666 | # Then let through as new doc??
|
|---|
| 667 |
|
|---|
| 668 | # mark to doc-oids that rely on it for re-indexing
|
|---|
| 669 | }
|
|---|
| 670 | }
|
|---|
| 671 | }
|
|---|
| 672 | }
|
|---|
| 673 |
|
|---|
| 674 | # Recursively read each $subfile
|
|---|
| 675 | print $outhandle "DirectoryPlugin recurring: $unicode_subfile\n" if ($verbosity > 2);
|
|---|
| 676 |
|
|---|
| 677 | $count += &plugin::read ($pluginfo, $this_file_base_dir,
|
|---|
| 678 | $raw_file_subfile, $block_hash,
|
|---|
| 679 | $out_metadata, $processor, $maxdocs, ($total_count + $count), $gli);
|
|---|
| 680 | }
|
|---|
| 681 |
|
|---|
| 682 | return $count;
|
|---|
| 683 | }
|
|---|
| 684 |
|
|---|
| 685 | sub compile_stats {
|
|---|
| 686 | my $self = shift(@_);
|
|---|
| 687 | my ($stats) = @_;
|
|---|
| 688 | }
|
|---|
| 689 |
|
|---|
| 690 | # Manifest files, version 2, provide an explicit listing of the documents to be
|
|---|
| 691 | # processed by Greenstone. This allows a user to avoid expensive file tree
|
|---|
| 692 | # searches - a crucial requirement for very-large scale collections and
|
|---|
| 693 | # parallel processing. However, we still want to leverage the metadata parsing
|
|---|
| 694 | # functionality found here in DirectoryPlugin. Thus we have this special call
|
|---|
| 695 | # to read that expects a single file. The normal read function starts by
|
|---|
| 696 | # listing the files in a given directory and then performs a number of actions
|
|---|
| 697 | # over them (including recursing down into any further directories found). We
|
|---|
| 698 | # circumvent that behaviour by 'pretending' to already have a directory listing
|
|---|
| 699 | # containing at most two file - the file passed in, and an accompanying
|
|---|
| 700 | # metadata.xml file if one exists.
|
|---|
| 701 | sub read_for_manifest_v2
|
|---|
| 702 | {
|
|---|
| 703 | my $self = shift (@_);
|
|---|
| 704 | my ($pluginfo, $file, $block_hash, $processor, $gli) = @_;
|
|---|
| 705 | my $base_dir = '';
|
|---|
| 706 | my $in_metadata = {};
|
|---|
| 707 | my $maxdocs = -1;
|
|---|
| 708 | my $total_count = 0;
|
|---|
| 709 | # Ensure we have the full path of the file to process
|
|---|
| 710 | my $full_path = $file;
|
|---|
| 711 | if ($base_dir =~ /\w/)
|
|---|
| 712 | {
|
|---|
| 713 | $full_path = &FileUtils::filenameConcatenate($base_dir, $file);
|
|---|
| 714 | }
|
|---|
| 715 | # Unlike the vanilla read(), directories are unacceptable
|
|---|
| 716 | if (!-f $full_path)
|
|---|
| 717 | {
|
|---|
| 718 | return 0;
|
|---|
| 719 | }
|
|---|
| 720 | # Now split the full path into a directory and a filename
|
|---|
| 721 | my ($dirname, $the_file) = $full_path =~ /^(.*)\/([^\/]+)$/;
|
|---|
| 722 | # We will prepopulate a 'directory listing' with this file
|
|---|
| 723 | my @dir = ($the_file);
|
|---|
| 724 | # See if there is an accompanying
|
|---|
| 725 | my $metadata_xml_path = $dirname . '/metadata.xml';
|
|---|
| 726 | if (-f $metadata_xml_path)
|
|---|
| 727 | {
|
|---|
| 728 | unshift(@dir, 'metadata.xml');
|
|---|
| 729 | }
|
|---|
| 730 | # Chain through to the normal read process, but with out 'forged' directory
|
|---|
| 731 | # listing so as to avoid all the costs of actually listing / recursing.
|
|---|
| 732 | my $count = $self->read_phase2($pluginfo, $dirname, \@dir, $base_dir, $dirname, $block_hash, $in_metadata, $processor, $maxdocs, $total_count, $gli);
|
|---|
| 733 | # We don't return count, but test that it is 1 exactly.
|
|---|
| 734 | if ($count != 1)
|
|---|
| 735 | {
|
|---|
| 736 | print STDERR "ERROR! The count of documents processed from a single call to DirectoryPlugin::read_for_manifest_v2() is not 1.\n";
|
|---|
| 737 | }
|
|---|
| 738 | }
|
|---|
| 739 |
|
|---|
| 740 | 1;
|
|---|
