[31190] | 1 | # This class based on arcinfo.pm
|
---|
| 2 | package oaiinfo;
|
---|
| 3 |
|
---|
| 4 | use constant INFO_STATUS_INDEX => 0;
|
---|
| 5 | use constant INFO_TIMESTAMP_INDEX => 1;
|
---|
[31216] | 6 | use constant INFO_DATESTAMP_INDEX => 2;
|
---|
[31190] | 7 |
|
---|
[31904] | 8 | my $OID_EARLIEST_TIMESTAMP = "_earliesttimestamp";
|
---|
[31900] | 9 | # Declaring as my $OID_EARLIEST_TIMESTAMP rather than constant, because it's not straightforward
|
---|
| 10 | # to use string constant as hash key (need to concat with empty str).
|
---|
| 11 | # http://perldoc.perl.org/constant.html
|
---|
| 12 | # But beware of using perl 'constant' as hash key:
|
---|
| 13 | # https://stackoverflow.com/questions/96848/is-there-any-way-to-use-a-constant-as-hash-key-in-perl
|
---|
| 14 | # http://forums.devshed.com/perl-programming-6/massive-using-constants-hash-keys-603600.html
|
---|
| 15 | # https://perlmaven.com/constants-and-read-only-variables-in-perl
|
---|
| 16 | # http://neilb.org/reviews/constants.html - compares different ways to declare constants in perl
|
---|
| 17 |
|
---|
[31190] | 18 | use strict;
|
---|
| 19 |
|
---|
| 20 | use arcinfo;
|
---|
| 21 | use dbutil;
|
---|
| 22 |
|
---|
[31216] | 23 | # Store timestamp in 2 formats: internal and external (same as oailastmodified and oailastmodifieddate)
|
---|
| 24 | # These times indicate the last modified date for that document. In the case of the doc being deleted,
|
---|
| 25 | # it's the time the doc was deleted.
|
---|
[31190] | 26 |
|
---|
[31216] | 27 | # File format read in: OID <tab> (Deletion-)Status <tab> Timestamp <tab> Datestamp
|
---|
[31190] | 28 |
|
---|
[31900] | 29 | # A special record of the db contains the timestamp of the creation of the oai-inf.db for
|
---|
| 30 | # the collection, representing the collection's earliest datetimestamp.
|
---|
| 31 | # This record has $OID_EARLIEST_TIMESTAMP for OID.
|
---|
| 32 | # Its deletion status is maintained at NA, not applicable.
|
---|
[31904] | 33 | # In cases of older oai-inf.db files where there's no $OID_EARLIEST_TIMESTAMP record, this record
|
---|
[31900] | 34 | # is also created but with timestamp set to the oldest lastmodified date in oai-inf.db.
|
---|
| 35 |
|
---|
[31190] | 36 | # Deletion status can be:
|
---|
| 37 | # E = Doc with OID exists (has not been deleted from collection). Timestamp indicates last time of build
|
---|
| 38 | # D = Doc with OID has been deleted. Timestamp indicates time of deletion
|
---|
[31216] | 39 | # PD = Provisionally Deleted. The associated timestamps are momentarily unaltered.
|
---|
[31900] | 40 | # NA = Not Applicable. Only for the special record with $OID_EARLIEST_TIMESTAMP as OID.
|
---|
[31190] | 41 |
|
---|
| 42 | # oaidb is "always incremental": always reflects the I/B/R/D status of archive info db,
|
---|
| 43 | # before the indexing step of the build phase that alters the I/B/R/D contents of archive info db.
|
---|
| 44 | # (I=index, B=been indexed, R=reindex; D=delete)
|
---|
| 45 |
|
---|
| 46 | sub new {
|
---|
| 47 | my $class = shift(@_);
|
---|
[31409] | 48 | my ($config_filename, $infodbtype, $verbosity) = @_;
|
---|
[31190] | 49 |
|
---|
| 50 | my $self = {
|
---|
[31409] | 51 | 'verbosity' => $verbosity || 0,
|
---|
| 52 | 'verbosity_threshold' => 5, # start printing debugging info from verbosity >= threshold
|
---|
[31216] | 53 | 'info'=>{} # map of {OID, array[deletion-status,timestamp,datestamp]} pairs
|
---|
[31190] | 54 | };
|
---|
| 55 |
|
---|
| 56 | if(!defined $infodbtype) {
|
---|
| 57 | $infodbtype = &dbutil::get_default_infodb_type();
|
---|
| 58 | }
|
---|
| 59 | $infodbtype = "gdbm" if ($infodbtype eq "gdbm-txtgz");
|
---|
| 60 | $self->{'infodbtype'} = $infodbtype;
|
---|
| 61 |
|
---|
| 62 | # Create and store the db filenames we'll be working with (tmp and livedb)
|
---|
| 63 | my $etc_dir = &util::get_parent_folder($config_filename);
|
---|
| 64 |
|
---|
| 65 | my $perform_firsttime_init = 0;
|
---|
| 66 | $self->{'oaidb_live_filepath'} = &dbutil::get_infodb_file_path($infodbtype, "oai-inf", $etc_dir, $perform_firsttime_init);
|
---|
| 67 | $self->{'oaidb_tmp_filepath'} = &dbutil::get_infodb_file_path($infodbtype, "oai-inf-tmp", $etc_dir, $perform_firsttime_init);
|
---|
| 68 | $self->{'etc_dir'} = $etc_dir;
|
---|
[31409] | 69 | # print STDERR "############ LIVE DB: $self->{'oaidb_live_filepath'}\n" if $self->{'verbosity'} >= $self->{'verbosity_threshold'};
|
---|
| 70 | # print STDERR "############ TMP DB: $self->{'oaidb_tmp_filepath'}\n" if $self->{'verbosity'} >= $self->{'verbosity_threshold'};
|
---|
[31190] | 71 |
|
---|
| 72 | $self->{'oaidb_file_path'} = $self->{'oaidb_tmp_filepath'}; # db file we're working with
|
---|
| 73 |
|
---|
| 74 | return bless $self, $class;
|
---|
| 75 | }
|
---|
| 76 |
|
---|
| 77 | # this subroutine will work out the starting contents of the tmp-db (temporary oai db):
|
---|
| 78 | # whether it should start off empty, or with the contents of any existing live-db,
|
---|
| 79 | # or with the contents of any existing tmp-db.
|
---|
| 80 | sub init_tmpdb {
|
---|
| 81 | my $self = shift(@_);
|
---|
| 82 | my ($removeold, $have_manifest) = @_;
|
---|
| 83 |
|
---|
| 84 | # if we have a manifest file, then we pretend we are fully incremental for oaiinfo db.
|
---|
| 85 | # removeold implies proper full-rebuild, whereas keepold or incremental means incremental
|
---|
| 86 | if($have_manifest) { # if we have a manifest file, we're not doing removeold/full-rebuild either
|
---|
| 87 | $removeold = 0;
|
---|
| 88 | }
|
---|
| 89 |
|
---|
[31192] | 90 | my $do_pd_step = ($removeold) ? 1 : 0;
|
---|
| 91 | # if $removeold, then proper full rebuild, will carry out step where all E will be marked as PD
|
---|
| 92 | # else some kind of incremental build, won't do the extra PD pass
|
---|
| 93 | # which is the step marking existing OIDs (E) as PD (provisionally deleted)
|
---|
[31190] | 94 |
|
---|
| 95 | my $oaidb_live_filepath = $self->{'oaidb_live_filepath'};
|
---|
| 96 | my $oaidb_tmp_filepath = $self->{'oaidb_tmp_filepath'};
|
---|
| 97 | my $infodbtype = $self->{'infodbtype'};
|
---|
| 98 | # Note: the live db can only exist if the collection has been activated at least once before
|
---|
| 99 | my $livedb_exists = &FileUtils::fileExists($oaidb_live_filepath);
|
---|
| 100 | my $tmpdb_exists = &FileUtils::fileExists($oaidb_tmp_filepath);
|
---|
| 101 |
|
---|
| 102 | my $initdb = 0;
|
---|
| 103 |
|
---|
| 104 | # work out what operation we need to do
|
---|
| 105 | # work with empty tmpdb
|
---|
| 106 | # copy_livedb_to_tmpdb
|
---|
| 107 | # work with existing tmpdb (so existing tmpdb will be topped up)
|
---|
| 108 |
|
---|
| 109 | # make_contents_of_tmpdb_empty
|
---|
| 110 | # make_contents_of_tmpdb_that_of_livedb
|
---|
| 111 | # continue_working_with_tmpdb ("contents_of_tmpdb_is_tmpdb")
|
---|
| 112 |
|
---|
| 113 | # We're going to prepare the starting state of tmpdb next.
|
---|
| 114 | # It can start off empty, start off with the contents of livedb, or it can start off with the contents
|
---|
| 115 | # of the existing tmp db. Which of these three it is depends on the 3 factors: whether livedb exists,
|
---|
| 116 | # whether tmpdb exists and whether or not removeold is true.
|
---|
| 117 | # i.o.w. which of the 3 outcomes it is depends on the truth table built on the following 3 variables:
|
---|
| 118 | # LDB = LiveDB exists
|
---|
| 119 | # TDB = TmpDB exists
|
---|
| 120 | # RO = Removeold
|
---|
| 121 | # OUTCOMES:
|
---|
| 122 | # clean slate (create an empty tmpdb/make tmpdb empty)
|
---|
| 123 | # top up tmpDB (work with existing tmpdb)
|
---|
| 124 | # copy LiveDB to TmpDB (liveDB's contents become the contents of TmpDB, and we'll work with that)
|
---|
| 125 | #
|
---|
| 126 | # TRUTH TABLE:
|
---|
| 127 | # ---------------------------------------
|
---|
| 128 | # LDB TDB RO | Outcome
|
---|
| 129 | # ---------------------------------------
|
---|
| 130 | # 0 0 0 | clean-slate
|
---|
| 131 | # 0 0 1 | clean-slate
|
---|
| 132 | # 0 1 0 | top-up-tmpdb
|
---|
| 133 | # 0 1 1 | erase tmpdb, clean-slate
|
---|
| 134 | # 1 0 0 | copy livedb to tmpdb
|
---|
| 135 | # 1 0 1 | copy livedb to tmpdb
|
---|
| 136 | # 1 1 0 | top-up-tmpdb
|
---|
| 137 | # 1 1 1 | copy livedb to tmpd
|
---|
| 138 | # ---------------------------------------
|
---|
| 139 | #
|
---|
| 140 | # Dr Bainbridge worked out using Karnaugh maps that, from the above truth table:
|
---|
| 141 | # => clean-slate/empty-tmpdb = !LDB && (RO || !TDB)
|
---|
| 142 | # => top-up-tmpdb/work-with-existing-tmpdb = !RO && TDB
|
---|
| 143 | # => copy-livedb-to-tmpdb = LDB && (!TDB || RO)
|
---|
| 144 | # I had most of these tests, except that I hadn't (yet) merged the two clean slate instances
|
---|
| 145 | # of first-build-ever and make-contents-of-tmpdb-empty
|
---|
| 146 |
|
---|
| 147 | #my $first_build_ever = (!$livedb_exists && !$tmpdb_exists);
|
---|
| 148 | #my $make_contents_of_tmpdb_empty = (!$livedb_exists && $tmpdb_exists && $removeold);
|
---|
| 149 | # Karnaugh map allows merging $first_build_ever and $make_contents_of_tmpdb_empty above
|
---|
| 150 | # into: my $work_with_empty_tmpdb = (!$livedb_exists && (!$tmpdb_exists || $removeold));
|
---|
| 151 | my $work_with_empty_tmpdb = (!$livedb_exists && (!$tmpdb_exists || $removeold));
|
---|
| 152 | my $make_contents_of_tmpdb_that_of_livedb = ($livedb_exists && (!$tmpdb_exists || $removeold));
|
---|
| 153 | my $work_with_existing_tmpdb = ($tmpdb_exists && !$removeold);
|
---|
| 154 |
|
---|
| 155 | if($work_with_empty_tmpdb) { # we'll use an empty tmpdb
|
---|
[31192] | 156 |
|
---|
[31190] | 157 | # If importing the collection for the very first time, neither db exists,
|
---|
| 158 | # so create an empty tmpdb.
|
---|
| 159 | #
|
---|
| 160 | # We also create an empty tmpdb when livedb doesn't exist and $removeold is true.
|
---|
| 161 | # This can happen if we've never run activate (so no livedb),
|
---|
| 162 | # yet had done some import (and perhaps building) followed by a full re-import now.
|
---|
| 163 | # Since there was no activate and we're doing a removeold/full-rebuild now, can just
|
---|
[32512] | 164 | # work with a new tmpdb: even though one already existed, its contents can be wiped out.
|
---|
| 165 | # In such a scenario, we'll be deleting tmpdb. Then there will be no livedb nor any tmpdb
|
---|
[31190] | 166 | # any more, so same situation as if importing the very first time when no oaidb exists either.
|
---|
| 167 |
|
---|
| 168 | &dbutil::remove_db_file($self->{'infodbtype'}, $oaidb_tmp_filepath) if $tmpdb_exists; # remove the db file and any assoc files
|
---|
| 169 | $initdb = 1; # new tmpdb
|
---|
| 170 |
|
---|
| 171 | # if the oai db is created the first time, it's like incremental and
|
---|
| 172 | # "keepold" (keepold means "only add, don't reprocess existing"). So
|
---|
| 173 | # no need to do the special passes dealing with "provisional deletes".
|
---|
[31192] | 174 | $do_pd_step = 0;
|
---|
[31190] | 175 |
|
---|
| 176 | } elsif ($make_contents_of_tmpdb_that_of_livedb) {
|
---|
| 177 |
|
---|
| 178 | # If the livedb exists and we're doing a full rebuild ($removeold is true),
|
---|
| 179 | # copy livedb to tmp regardless of if tmpdb already exists.
|
---|
| 180 | # Or if the livedb exists and tmpdb doesn't exist, it doesn't matter
|
---|
| 181 | # if we're incremental or not: also copy live to tmp and work with tmp.
|
---|
| 182 |
|
---|
| 183 | # copy livedb to tmpdb
|
---|
| 184 | &dbutil::remove_db_file($self->{'infodbtype'}, $oaidb_tmp_filepath) if $tmpdb_exists; # remove the db file and any assoc files
|
---|
| 185 | &FileUtils::copyFiles($oaidb_live_filepath, $oaidb_tmp_filepath);
|
---|
| 186 |
|
---|
[31723] | 187 | $initdb = 0; # tmpdb exists, since we just copied livedb to tmpdb, so no need to init existing tmpdb
|
---|
[31190] | 188 |
|
---|
| 189 | } else { # $work_with_existing_tmpdb, so we'll build on top of what's presently already in tmpdb
|
---|
| 190 | # (we'll be topping up the current tmpdb)
|
---|
| 191 |
|
---|
| 192 | # !$removeold, meaning incremental
|
---|
| 193 | # If incremental and have a tmpdb already, regardless of whether livedb exists,
|
---|
| 194 | # then work with the existing tmpdb file, as this means we've been
|
---|
| 195 | # importing (perhaps followed by building) repeatedly without activating the
|
---|
| 196 | # last time but want to maintain the (incremental) changes in tmpdb.
|
---|
| 197 |
|
---|
| 198 | $initdb = 0;
|
---|
| 199 |
|
---|
| 200 | } # Dr Bainbridge drew up Karnaugh maps on the truth table, which proved that all cases
|
---|
| 201 | # are indeed covered above, so don't need any other catch-all else here
|
---|
| 202 |
|
---|
| 203 | $self->{'oaidb_file_path'} = &dbutil::get_infodb_file_path($infodbtype, "oai-inf-tmp", $self->{'etc_dir'}, $initdb);
|
---|
| 204 | # final param follows jmt's $perform_firsttime_init in inexport.pm
|
---|
| 205 |
|
---|
[31409] | 206 | # print STDERR "@@@@@ oaidb: $self->{'oaidb_file_path'}\n" if $self->{'verbosity'} >= $self->{'verbosity_threshold'};
|
---|
[31900] | 207 |
|
---|
| 208 | return ($do_pd_step, $initdb);
|
---|
[31190] | 209 | }
|
---|
| 210 |
|
---|
| 211 | sub get_filepath {
|
---|
| 212 | my $self = shift (@_);
|
---|
| 213 | return $self->{'oaidb_file_path'};
|
---|
| 214 | }
|
---|
| 215 |
|
---|
| 216 | sub import_stage {
|
---|
| 217 | my $self = shift (@_);
|
---|
| 218 | my ($removeold, $have_manifest) = @_;
|
---|
[31409] | 219 |
|
---|
[31900] | 220 | my ($do_pd_step, $is_new_db) = $self->init_tmpdb($removeold, $have_manifest);
|
---|
| 221 | # returns 1 for $do_pd_step if the step to mark oaidb entries as PD is required
|
---|
[31192] | 222 | # if we're doing full rebuilding and it's NOT the first time creating the oai_inf db,
|
---|
| 223 | # then the tasks to do with PD (provisionally deleted) OAI OIDs should be carried out
|
---|
[31900] | 224 | # Returns 1 for is_new_db to allow further one time initialisation of the new oai-inf.db
|
---|
[31190] | 225 |
|
---|
| 226 | $self->load_info();
|
---|
| 227 | $self->print_info(); # DEBUGGING
|
---|
| 228 |
|
---|
[31900] | 229 | # A special record of the oai-inf.db will contain the timestamp when the oai-inf.db was created.
|
---|
| 230 | # This represents the collection's "earliest datetimestamp". It should remain unaltered
|
---|
| 231 | # for as long as oai-inf db exists. This record has the special OID of $OID_EARLIEST_TIMESTAMP.
|
---|
| 232 | # This record should not be marked as PD, but remain as E, as it can't ever be deleted.
|
---|
| 233 | # Although the status field for the $OID_EARLIEST_TIMESTAMP record is actually meaningless.
|
---|
| 234 | my $save_to_db = $self->insert_coll_earliest_timestamp($is_new_db);
|
---|
| 235 |
|
---|
[31190] | 236 | if ($do_pd_step) {
|
---|
| 237 | $self->mark_all_existing_as_provisionallydeleted();
|
---|
| 238 | $self->print_info(); # DEBUGGING
|
---|
[31900] | 239 |
|
---|
| 240 | $save_to_db = 1;
|
---|
| 241 | }
|
---|
[31190] | 242 |
|
---|
[31900] | 243 | if($save_to_db) {
|
---|
| 244 | # save changes to $self->{'info'} out to db file, now that we're done
|
---|
| 245 | $self->save_info();
|
---|
[31190] | 246 | }
|
---|
| 247 |
|
---|
| 248 | }
|
---|
| 249 |
|
---|
| 250 | sub building_stage_before_indexing() {
|
---|
| 251 | my $self = shift (@_);
|
---|
| 252 | my ($archivedir) = @_;
|
---|
| 253 |
|
---|
| 254 | # load archive info db into memory
|
---|
| 255 | my $arcinfo_doc_filename = &dbutil::get_infodb_file_path($self->{'infodbtype'}, "archiveinf-doc", $archivedir);
|
---|
[32097] | 256 | #my $arcinfo_src_filename = &dbutil::get_infodb_file_path($self->{'infodbtype'}, "archiveinf-src", $archivedir); # unused
|
---|
[31190] | 257 | my $archive_info = new arcinfo ($self->{'infodbtype'});
|
---|
| 258 | $archive_info->load_info ($arcinfo_doc_filename);
|
---|
| 259 |
|
---|
| 260 | #my $started_from_scratch = &FileUtils::fileTest($self->{'oaidb_tmp_filepath'}, '-z'); # 1 if tmpdb is empty
|
---|
| 261 | # -z test for file is empty http://www.perlmonks.org/?node_id=927447
|
---|
| 262 |
|
---|
| 263 | # load the oaidb file's contents into memory.
|
---|
| 264 | $self->load_info();
|
---|
| 265 | $self->print_info(); # DEBUGGING
|
---|
| 266 |
|
---|
| 267 | # process all the index, reindex and delete operations as indicated in arcinfo,
|
---|
| 268 | # all the while ensuring all PDs are changed back to E for OIDs that exist in both arcinfo and oaiinfo db.
|
---|
| 269 |
|
---|
[31900] | 270 | my $arcinfo_map = $archive_info->{'info'};
|
---|
[31218] | 271 |
|
---|
| 272 | foreach my $OID (keys %$arcinfo_map) {
|
---|
[31190] | 273 | my $arcinf_tuple = $archive_info->{'info'}->{$OID};
|
---|
| 274 | my $indexing_status = $arcinf_tuple->[arcinfo::INFO_STATUS_INDEX];
|
---|
| 275 | # use packageName::constant to refer to constants declared in another package,
|
---|
| 276 | # see http://perldoc.perl.org/constant.html
|
---|
| 277 |
|
---|
[31409] | 278 | print STDERR "######## OID: $OID - status: $indexing_status\n" if $self->{'verbosity'} >= $self->{'verbosity_threshold'};
|
---|
[31190] | 279 |
|
---|
| 280 | if($indexing_status eq "I") {
|
---|
| 281 | $self->index($OID); # add new as E with current timestamp/or set existing as E with orig timestamp
|
---|
| 282 | } elsif($indexing_status eq "R") {
|
---|
| 283 | $self->reindex($OID); # update timestamp and ensure marked as E (if oid doesn't exist, add new)
|
---|
| 284 | } elsif($indexing_status eq "D") {
|
---|
| 285 | $self->delete($OID); # set as D with current timestamp
|
---|
| 286 | } elsif($indexing_status eq "B") { # B for "been indexed"
|
---|
| 287 | $self->been_indexed($OID); # will flip any PD to E if oid exists, else will add new entry for oid
|
---|
| 288 | # A new entry may be required if the collection had been built prior to turning this into
|
---|
| 289 | # an oaicollection. But what if we always maintain an oaidb? Still call $self->index() here.
|
---|
| 290 | } else {
|
---|
[31900] | 291 | if ($self->{'verbosity'} >= $self->{'verbosity_threshold'}) {
|
---|
| 292 | print STDERR "### oaiinfo::building_stage_before_indexing(): Unrecognised indexing status $indexing_status\n";
|
---|
| 293 | }
|
---|
[31190] | 294 | }
|
---|
| 295 | }
|
---|
| 296 |
|
---|
| 297 | # once all docs processed, go through oaiiinfo db changing any PDs to D along with current timestamp
|
---|
| 298 | # to indicate that they're deleted
|
---|
| 299 | $self->mark_all_provisionallydeleted_as_deleted();
|
---|
| 300 | $self->print_info();
|
---|
| 301 |
|
---|
| 302 | # let's save to db file now that we're done
|
---|
| 303 | $self->save_info();
|
---|
| 304 |
|
---|
| 305 | }
|
---|
| 306 |
|
---|
| 307 | sub activate_collection { # move tmp db to live db
|
---|
| 308 | my $self = shift (@_);
|
---|
| 309 |
|
---|
| 310 | my $oaidb_live_filepath = $self->{'oaidb_live_filepath'};
|
---|
| 311 | my $oaidb_tmp_filepath = $self->{'oaidb_tmp_filepath'};
|
---|
| 312 |
|
---|
| 313 | my $livedb_exists = &FileUtils::fileExists($oaidb_live_filepath);
|
---|
| 314 | my $tmpdb_exists = &FileUtils::fileExists($oaidb_tmp_filepath);
|
---|
| 315 |
|
---|
| 316 | if($tmpdb_exists) {
|
---|
| 317 | if($livedb_exists) {
|
---|
[31723] | 318 | #&dbutil::remove_db_file($self->{'infodbtype'}, $oaidb_live_filepath); # remove the db file and any assoc files
|
---|
[31190] | 319 | &dbutil::rename_db_file_to($self->{'infodbtype'}, $oaidb_live_filepath, $oaidb_live_filepath.".bak"); # rename the db file and any assoc files
|
---|
| 320 | }
|
---|
| 321 | #&FileUtils::moveFiles($oaidb_tmp_filepath, $oaidb_live_filepath);
|
---|
| 322 | &dbutil::rename_db_file_to($self->{'infodbtype'}, $oaidb_tmp_filepath, $oaidb_live_filepath); # rename the db file and any assoc files
|
---|
| 323 |
|
---|
[31409] | 324 | if ($self->{'verbosity'} >= $self->{'verbosity_threshold'}) {
|
---|
| 325 | print STDERR "#### Should now have MOVED $self->{'oaidb_tmp_filepath'} to $self->{'oaidb_live_filepath'}\n";
|
---|
| 326 | }
|
---|
[31190] | 327 |
|
---|
| 328 | } else {
|
---|
[31900] | 329 | if ($self->{'verbosity'} >= $self->{'verbosity_threshold'}) {
|
---|
| 330 | print STDERR "@@@@@ In oaiinfo::activate_collection():\n";
|
---|
| 331 | print STDERR "@@@@@ No tmpdb at $self->{'oaidb_tmp_filepath'}\n";
|
---|
| 332 | print STDERR "@@@@@ to make 'live' by moving to $self->{'oaidb_live_filepath'}.\n";
|
---|
| 333 | }
|
---|
[31190] | 334 | }
|
---|
| 335 | }
|
---|
| 336 |
|
---|
| 337 | ##################### SPECIFIC TO PD-STEP ####################
|
---|
| 338 |
|
---|
| 339 |
|
---|
| 340 | # mark all existing, E (non-deleted) OIDs as Provisionally Deleted (PD)
|
---|
| 341 | # this subroutine doesn't save to oai-inf.DB
|
---|
| 342 | # the caller should call save_info when they want to save to the db
|
---|
| 343 | sub mark_all_existing_as_provisionallydeleted {
|
---|
| 344 | my $self = shift (@_);
|
---|
| 345 |
|
---|
[31409] | 346 | print STDERR "@@@@@ oaiinfo::mark_all_E_as_PD(): Marking the E entries as PD\n" if $self->{'verbosity'} >= $self->{'verbosity_threshold'};
|
---|
[31190] | 347 |
|
---|
[31900] | 348 | my $infomap = $self->{'info'};
|
---|
[31218] | 349 |
|
---|
| 350 | foreach my $OID (keys %$infomap) { # Mac Mountain Lion wants %$map, won't accept %$self->{'info'}
|
---|
[31192] | 351 | my $OID_info = $self->{'info'}->{$OID};
|
---|
| 352 | my $curr_status = $OID_info->[INFO_STATUS_INDEX];
|
---|
| 353 | if($curr_status eq "E") {
|
---|
| 354 | $OID_info->[INFO_STATUS_INDEX] = "PD";
|
---|
[31190] | 355 | }
|
---|
| 356 | }
|
---|
| 357 | }
|
---|
| 358 |
|
---|
| 359 | # mark all OIDs that are Provisionally Deleted (PD) as deleted, and set to current timestamp
|
---|
| 360 | # To be called at end of build. Again, the caller should save to DB by calling save_info.
|
---|
| 361 | sub mark_all_provisionallydeleted_as_deleted {
|
---|
| 362 | my $self = shift (@_);
|
---|
| 363 |
|
---|
[31409] | 364 | print STDERR "@@@@@ oaiinfo::mark_all_PD_as_D(): Marking the PD entries as D\n" if $self->{'verbosity'} >= $self->{'verbosity_threshold'};
|
---|
[31190] | 365 |
|
---|
[31900] | 366 | my $infomap = $self->{'info'};
|
---|
[31218] | 367 |
|
---|
| 368 | foreach my $OID (keys %$infomap) {
|
---|
[31192] | 369 | my $OID_info = $self->{'info'}->{$OID};
|
---|
| 370 | my $curr_status = $OID_info->[INFO_STATUS_INDEX];
|
---|
| 371 | if($curr_status eq "PD") {
|
---|
[31190] | 372 | $self->set_info($OID, "D", $self->get_current_time());
|
---|
| 373 | }
|
---|
| 374 | }
|
---|
| 375 | }
|
---|
| 376 |
|
---|
| 377 |
|
---|
| 378 | ##################### GENERAL, NOT SPECIFIC TO PD-STEP ####################
|
---|
| 379 |
|
---|
| 380 | sub print_info {
|
---|
| 381 | my $self = shift (@_);
|
---|
| 382 |
|
---|
[31900] | 383 | if ($self->{'verbosity'} < $self->{'verbosity_threshold'}) {
|
---|
| 384 | return;
|
---|
| 385 | }
|
---|
[31409] | 386 |
|
---|
[31190] | 387 | print STDERR "###########################################################\n";
|
---|
| 388 | print STDERR "@@@@@ oaiinfo::print_info(): oaidb in memory contains: \n";
|
---|
| 389 |
|
---|
[31900] | 390 | my $infomap = $self->{'info'};
|
---|
[31218] | 391 |
|
---|
| 392 | foreach my $OID (keys %$infomap) {
|
---|
[31190] | 393 | print STDERR "OID: $OID";
|
---|
| 394 | print STDERR " status: " . $self->{'info'}->{$OID}->[INFO_STATUS_INDEX];
|
---|
| 395 | print STDERR " time: " . $self->{'info'}->{$OID}->[INFO_TIMESTAMP_INDEX];
|
---|
[31216] | 396 | print STDERR " date: " . $self->{'info'}->{$OID}->[INFO_DATESTAMP_INDEX];
|
---|
[31190] | 397 | print STDERR "\n";
|
---|
| 398 | }
|
---|
| 399 |
|
---|
| 400 | print STDERR "###########################################################\n";
|
---|
| 401 | }
|
---|
| 402 |
|
---|
| 403 |
|
---|
[31900] | 404 | # When a fresh oai-inf.db is created, this method is called to add the db's special
|
---|
| 405 | # record representing the collection's earliest timestamp.
|
---|
| 406 | # OID=$OID_EARLIEST_TIMESTAMP, deletion_status=NA for not applicable, and current timestamp/date.
|
---|
| 407 | # For older oai-inf.db's that don't yet have this record, a record will be added too,
|
---|
| 408 | # but with the timestamp set to the oldest last modified date for the collection's docs.
|
---|
| 409 | sub insert_coll_earliest_timestamp {
|
---|
| 410 | my $self = shift (@_);
|
---|
| 411 | my ($is_new_db) = @_;
|
---|
| 412 |
|
---|
| 413 | my $current_time = $self->get_current_time();
|
---|
| 414 | my $save_to_db = 0;
|
---|
| 415 |
|
---|
| 416 |
|
---|
| 417 | print STDERR "@@@@@ oaiinfo::insert_coll_earliest_timestamp(): " if $self->{'verbosity'} >= $self->{'verbosity_threshold'};
|
---|
| 418 |
|
---|
| 419 | if($is_new_db) {
|
---|
| 420 |
|
---|
| 421 | print STDERR "New db. Setting timestamp of oai-inf.db creation.\n" if $self->{'verbosity'} >= $self->{'verbosity_threshold'};
|
---|
| 422 |
|
---|
| 423 | $self->set_info($OID_EARLIEST_TIMESTAMP, "NA", $current_time);
|
---|
| 424 | $save_to_db = 1;
|
---|
| 425 | }
|
---|
| 426 |
|
---|
| 427 | else { # oai-inf.db already exists, ensure it has an [$OID_EARLIEST_TIMESTAMP] set
|
---|
| 428 |
|
---|
| 429 | my $earliesttimestamp_record = $self->{'info'}->{$OID_EARLIEST_TIMESTAMP};
|
---|
| 430 |
|
---|
| 431 | if (!defined $earliesttimestamp_record) {
|
---|
| 432 | # oai-inf.db exists, but doesn't contain an [$OID_EARLIEST_TIMESTAMP] record yet.
|
---|
| 433 | # Let's create one for it:
|
---|
| 434 | # Work out the earliest lastmodified datetime in the collection, by inspecting
|
---|
| 435 | # the last modified timestamp for each doc in the collection
|
---|
| 436 |
|
---|
| 437 | my $earliest_timestamp = $current_time;
|
---|
| 438 |
|
---|
| 439 | my $infomap = $self->{'info'}; # Mac Mountain Lion wants %$map, won't accept %$self->{'info'}
|
---|
| 440 | foreach my $OID (keys %$infomap) {
|
---|
| 441 | my $OID_info = $self->{'info'}->{$OID};
|
---|
| 442 | my $lastmodified = $OID_info->[INFO_TIMESTAMP_INDEX];
|
---|
| 443 | if($lastmodified < $earliest_timestamp) {
|
---|
| 444 | $earliest_timestamp = $lastmodified;
|
---|
| 445 | }
|
---|
| 446 | }
|
---|
| 447 |
|
---|
| 448 | print STDERR "Collection timestamp not yet set for $OID_EARLIEST_TIMESTAMP. Setting to earliest found: $earliest_timestamp\n" if $self->{'verbosity'} >= $self->{'verbosity_threshold'};
|
---|
| 449 |
|
---|
| 450 | $self->set_info($OID_EARLIEST_TIMESTAMP, "NA", $earliest_timestamp);
|
---|
| 451 | $save_to_db = 1;
|
---|
| 452 | } else {
|
---|
| 453 | print STDERR "Collection timestamp was already set\n" if $self->{'verbosity'} >= $self->{'verbosity_threshold'};
|
---|
| 454 | }
|
---|
| 455 | }
|
---|
| 456 |
|
---|
| 457 | return $save_to_db;
|
---|
| 458 | }
|
---|
| 459 |
|
---|
| 460 |
|
---|
[31190] | 461 | # Find the OID, if it exists, make its status=E for existing. Leave its timestamp alone.
|
---|
| 462 | # If the OID doesn't yet exist, add it as a new entry with status=E and with current timestamp.
|
---|
| 463 | sub index { # Add a new oid with current time and E. If the oid was already present, mark as E
|
---|
| 464 | my $self = shift (@_);
|
---|
| 465 | my ($OID) = @_;
|
---|
| 466 |
|
---|
| 467 | my $OID_info = $self->{'info'}->{$OID};
|
---|
| 468 |
|
---|
| 469 | if (defined $OID_info) { # if OID is present, this will change status back to E, timestamp unchanged
|
---|
| 470 | $OID_info->[INFO_STATUS_INDEX] = "E";
|
---|
| 471 |
|
---|
| 472 | } else { # if OID is not present, then it's now added as existing from current time on
|
---|
| 473 | $self->set_info($OID, "E", $self->get_current_time());
|
---|
| 474 | }
|
---|
| 475 | }
|
---|
| 476 |
|
---|
| 477 | # Upon reindexing a document with identifier OID, change its timestamp to current time
|
---|
| 478 | # if a new OID, then add as new entry with status=E and current timestamp
|
---|
| 479 | sub reindex { # update timestamp if oid is already present, if not (unlikely), add as new
|
---|
| 480 | my $self = shift (@_);
|
---|
| 481 | my ($OID) = @_;
|
---|
| 482 |
|
---|
| 483 | my $OID_info = $self->{'info'}->{$OID};
|
---|
| 484 | $self->set_info($OID, "E", $self->get_current_time()); # Takes care of 3 things:
|
---|
| 485 | # if OID exists, updates modified time to indicate the doc has been reindexed
|
---|
| 486 | # if OID exists, ensures any status=PD is flipped back to E for this OID doc (as we know it exists);
|
---|
| 487 | # if the OID doesn't yet exist, adds a new OID entry with status=E and current timestamp.
|
---|
| 488 |
|
---|
| 489 | }
|
---|
| 490 |
|
---|
| 491 | # Does the same as index():
|
---|
| 492 | # OIDs that have been indexed upon rebuild may still be new to the oaidb: GS2 collections
|
---|
| 493 | # are not OAI collections by default, unlike GS3 collections. Imagine rebuilding a (GS2) collection
|
---|
| 494 | # 5 times and then setting them to be an OAI collection. In that case, the doc OIDs in the collection
|
---|
| 495 | # may not be in the oaidb yet. Unless, we decide (as is the present case) to always maintain an oaidb
|
---|
| 496 | # (always creating an oaidb regardless of whether the collection has OAI support turned on or not).
|
---|
| 497 | sub been_indexed {
|
---|
| 498 | my $self = shift (@_);
|
---|
| 499 | my ($OID) = @_;
|
---|
| 500 |
|
---|
| 501 | $self->index($OID);
|
---|
| 502 | }
|
---|
| 503 |
|
---|
| 504 | # Upon deleting a document with identifier OID,
|
---|
| 505 | # set status to deleted and change its timestamp to current time
|
---|
| 506 | sub delete {
|
---|
| 507 | my $self = shift (@_);
|
---|
| 508 | my ($OID) = @_;
|
---|
| 509 |
|
---|
| 510 | # the following method will set to current time if no timestamp provided,
|
---|
[31900] | 511 | # But by being explicit here, the code is easier to follow
|
---|
[31216] | 512 | $self->set_info($OID, "D", $self->get_current_time());
|
---|
[31190] | 513 |
|
---|
| 514 | }
|
---|
| 515 |
|
---|
| 516 | #############################################################
|
---|
| 517 | sub get_current_time {
|
---|
| 518 | my $self = shift (@_);
|
---|
| 519 | return time; # current time
|
---|
| 520 |
|
---|
| 521 | # localtime(time) returns an array of values (day, month, year, hour, min, seconds) or singular string
|
---|
| 522 | # return localtime; # same as localtime(time); # http://perldoc.perl.org/functions/localtime.html
|
---|
| 523 |
|
---|
| 524 | }
|
---|
| 525 |
|
---|
[31216] | 526 | sub get_datestamp {
|
---|
| 527 | my $self = shift (@_);
|
---|
| 528 | my ($timestamp) = @_;
|
---|
| 529 |
|
---|
| 530 | my ($seconds, $minutes, $hours, $day_of_month, $month, $year,
|
---|
| 531 | $wday, $yday, $isdst) = localtime($timestamp);
|
---|
| 532 |
|
---|
| 533 | my $datestamp = sprintf("%d%02d%02d",1900+$year,$month+1,$day_of_month);
|
---|
| 534 |
|
---|
| 535 | return $datestamp;
|
---|
| 536 | }
|
---|
| 537 |
|
---|
[31190] | 538 | sub _load_info_txt
|
---|
| 539 | {
|
---|
| 540 | my $self = shift (@_);
|
---|
| 541 | my ($filename) = @_;
|
---|
| 542 |
|
---|
| 543 | if (defined $filename && &FileUtils::fileExists($filename)) {
|
---|
| 544 | open (INFILE, $filename) ||
|
---|
| 545 | die "oaiinfo::load_info couldn't read $filename\n";
|
---|
| 546 |
|
---|
| 547 | my ($line, @lineparts);
|
---|
| 548 | while (defined ($line = <INFILE>)) {
|
---|
| 549 | $line =~ s/\cM|\cJ//g; # remove end-of-line characters
|
---|
| 550 | @lineparts = split ("\t", $line);
|
---|
| 551 | if (scalar(@lineparts) >= 2) {
|
---|
[31192] | 552 | $self->set_info (@lineparts);
|
---|
[31190] | 553 | }
|
---|
| 554 | }
|
---|
| 555 | close (INFILE);
|
---|
| 556 | }
|
---|
| 557 |
|
---|
| 558 | }
|
---|
| 559 |
|
---|
| 560 | sub _load_info_db
|
---|
| 561 | {
|
---|
| 562 | my $self = shift (@_);
|
---|
| 563 | my ($filename) = @_;
|
---|
| 564 |
|
---|
| 565 | my $infodb_map = {};
|
---|
| 566 |
|
---|
| 567 | &dbutil::read_infodb_file($self->{'infodbtype'}, $filename, $infodb_map);
|
---|
| 568 |
|
---|
[31218] | 569 | foreach my $oid ( keys %$infodb_map ) {
|
---|
[31190] | 570 | my $vals = $infodb_map->{$oid};
|
---|
| 571 | # interested in oid, timestamp, deletion status
|
---|
| 572 |
|
---|
| 573 | my ($deletion_status) = ($vals=~/^<status>(.*)$/m);
|
---|
| 574 | my ($timestamp) = ($vals=~/^<timestamp>(.*)$/m);
|
---|
[31216] | 575 | my ($datestamp) = ($vals=~/^<datestamp>(.*)$/m);
|
---|
[31190] | 576 |
|
---|
[31216] | 577 | $self->add_info ($oid, $deletion_status, $timestamp, $datestamp);
|
---|
[31190] | 578 | }
|
---|
| 579 | }
|
---|
| 580 |
|
---|
| 581 | # if no filename is passed in (and you don't generally want to), then
|
---|
| 582 | # it tries to load in <collection>/etc/oai-inf.<db> if it exists
|
---|
| 583 | sub load_info {
|
---|
| 584 | my $self = shift (@_);
|
---|
| 585 | my ($filename) = @_;
|
---|
| 586 |
|
---|
| 587 | $self->{'info'} = {};
|
---|
| 588 |
|
---|
| 589 | $filename = $self->{'oaidb_file_path'} unless defined $filename;
|
---|
| 590 |
|
---|
| 591 | if (&FileUtils::fileExists($filename)) {
|
---|
| 592 | if ($filename =~ m/\.inf$/) {
|
---|
| 593 | $self->_load_info_txt($filename);
|
---|
| 594 | }
|
---|
| 595 | else {
|
---|
| 596 | $self->_load_info_db($filename);
|
---|
| 597 | }
|
---|
| 598 | }
|
---|
| 599 |
|
---|
| 600 | }
|
---|
| 601 |
|
---|
| 602 | sub _save_info_txt {
|
---|
| 603 | my $self = shift (@_);
|
---|
| 604 | my ($filename) = @_;
|
---|
| 605 |
|
---|
| 606 | my ($OID, $info);
|
---|
| 607 |
|
---|
| 608 | open (OUTFILE, ">$filename") ||
|
---|
| 609 | die "oaiinfo::save_info couldn't write $filename\n";
|
---|
| 610 |
|
---|
| 611 | foreach $info (@{$self->get_OID_list()}) {
|
---|
| 612 | if (defined $info) {
|
---|
| 613 | print OUTFILE join("\t", @$info), "\n";
|
---|
| 614 | }
|
---|
| 615 | }
|
---|
| 616 | close (OUTFILE);
|
---|
| 617 | }
|
---|
| 618 |
|
---|
| 619 | # if no filename is passed in (and you don't generally want to), then
|
---|
| 620 | # this subroutine tries to write to <collection>/etc/oai-inf.<db>.
|
---|
| 621 | sub _save_info_db {
|
---|
| 622 | my $self = shift (@_);
|
---|
| 623 | my ($filename) = @_;
|
---|
| 624 |
|
---|
| 625 | $filename = $self->{'oaidb_file_path'} unless defined $filename;
|
---|
| 626 | my $infodbtype = $self->{'infodbtype'};
|
---|
| 627 |
|
---|
| 628 | # write out again. Open file for overwriting, not appending.
|
---|
[31900] | 629 | # Then write out data structure $self->{'info'} that has been maintaining the data in-memory.
|
---|
[31190] | 630 | my $infodb_handle = &dbutil::open_infodb_write_handle($infodbtype, $filename);
|
---|
[31218] | 631 |
|
---|
[31900] | 632 | my $infomap = $self->{'info'};
|
---|
[31218] | 633 | foreach my $oid ( keys %$infomap ) {
|
---|
[31190] | 634 | my $OID_info = $self->{'info'}->{$oid};
|
---|
[31216] | 635 | my $val = "<status>".$OID_info->[INFO_STATUS_INDEX];
|
---|
| 636 | $val .= "\n<timestamp>".$OID_info->[INFO_TIMESTAMP_INDEX];
|
---|
| 637 | $val .= "\n<datestamp>".$OID_info->[INFO_DATESTAMP_INDEX]."\n";
|
---|
[31190] | 638 | &dbutil::write_infodb_rawentry($infodbtype,$infodb_handle,$oid,$val);
|
---|
| 639 | }
|
---|
| 640 | &dbutil::close_infodb_write_handle($infodbtype, $infodb_handle);
|
---|
| 641 | }
|
---|
| 642 |
|
---|
| 643 | sub save_info {
|
---|
| 644 | my $self = shift (@_);
|
---|
| 645 | my ($filename) = @_;
|
---|
| 646 |
|
---|
| 647 | if(defined $filename) {
|
---|
| 648 | if ($filename =~ m/(contents)|(\.inf)$/) {
|
---|
| 649 | $self->_save_info_txt($filename);
|
---|
| 650 | }
|
---|
| 651 | else {
|
---|
| 652 | $self->_save_info_db($filename);
|
---|
| 653 | }
|
---|
| 654 | } else {
|
---|
| 655 | $self->_save_info_db();
|
---|
| 656 | }
|
---|
| 657 | }
|
---|
| 658 |
|
---|
| 659 |
|
---|
| 660 | sub set_info { # sets existing or appends
|
---|
| 661 | my $self = shift (@_);
|
---|
| 662 | my ($OID, $del_status, $timestamp) = @_;
|
---|
[31216] | 663 |
|
---|
[31190] | 664 | if(!defined $timestamp) { # get current date timestamp
|
---|
| 665 | $timestamp = $self->get_current_time();
|
---|
| 666 | }
|
---|
[31216] | 667 | my $datestamp = $self->get_datestamp($timestamp);
|
---|
[31190] | 668 |
|
---|
[31216] | 669 | $self->{'info'}->{$OID} = [$del_status, $timestamp, $datestamp];
|
---|
| 670 |
|
---|
[31190] | 671 | }
|
---|
| 672 |
|
---|
[31216] | 673 | sub add_info { # called to load a single record from file into memory, so it should be provided all 4 fields
|
---|
| 674 | my $self = shift (@_);
|
---|
| 675 | my ($OID, $del_status, $timestamp, $datestamp) = @_;
|
---|
[31190] | 676 |
|
---|
[31216] | 677 | $self->{'info'}->{$OID} = [$del_status, $timestamp, $datestamp];
|
---|
| 678 | }
|
---|
| 679 |
|
---|
| 680 |
|
---|
| 681 | # returns a list of the form [[OID, deletion_status, timestamp, datestamp], ...]
|
---|
[31190] | 682 | sub get_OID_list
|
---|
| 683 | {
|
---|
| 684 | my $self = shift (@_);
|
---|
| 685 |
|
---|
| 686 | my @list = ();
|
---|
[31218] | 687 |
|
---|
[31900] | 688 | my $infomap = $self->{'info'};
|
---|
[31218] | 689 | foreach my $OID (keys %$infomap) {
|
---|
[31190] | 690 | my $OID_info = $self->{'info'}->{$OID};
|
---|
| 691 |
|
---|
| 692 | push (@list, [$OID, $OID_info->[INFO_STATUS_INDEX],
|
---|
[31216] | 693 | $OID_info->[INFO_TIMESTAMP_INDEX],
|
---|
| 694 | $OID_info->[INFO_DATESTAMP_INDEX]
|
---|
| 695 | ]);
|
---|
[31190] | 696 | }
|
---|
| 697 |
|
---|
| 698 | return \@list;
|
---|
| 699 | }
|
---|
| 700 |
|
---|
| 701 |
|
---|
| 702 | # returns the number of entries so far, including deleted ones
|
---|
| 703 | # http://stackoverflow.com/questions/1109095/how-can-i-find-the-number-of-keys-in-a-hash-in-perl
|
---|
| 704 | sub size {
|
---|
| 705 | my $self = shift (@_);
|
---|
[31900] | 706 | my $infomap = $self->{'info'};
|
---|
[31218] | 707 | return (scalar keys %$infomap);
|
---|
[31190] | 708 | }
|
---|
| 709 |
|
---|
| 710 | 1;
|
---|