7 # Multi-level database module for storing hash trees, arrays and simple
8 # key/value pairs into FTP-able, cross-platform binary database files.
10 # Type `perldoc DBM::Deep` for complete documentation.
14 # tie %db, 'DBM::Deep', 'my_database.db'; # standard tie() method
16 # my $db = new DBM::Deep( 'my_database.db' ); # preferred OO method
18 # $db->{my_scalar} = 'hello world';
19 # $db->{my_hash} = { larry => 'genius', hashes => 'fast' };
20 # $db->{my_array} = [ 1, 2, 3, time() ];
21 # $db->{my_complex} = [ 'hello', { perl => 'rules' }, 42, 99 ];
22 # push @{$db->{my_array}}, 'another value';
23 # my @key_list = keys %{$db->{my_hash}};
24 # print "This module " . $db->{my_complex}->[1]->{perl} . "!\n";
27 # (c) 2002-2005 Joseph Huckaby. All Rights Reserved.
28 # This program is free software; you can redistribute it and/or
29 # modify it under the same terms as Perl itself.
34 use Fcntl qw( :DEFAULT :flock :seek );
38 use vars qw( $VERSION );
42 # Set to 4 and 'N' for 32-bit offset tags (default). Theoretical limit of 4 GB per file.
43 # (Perl must be compiled with largefile support for files > 2 GB)
45 # Set to 8 and 'Q' for 64-bit offsets. Theoretical limit of 16 XB per file.
46 # (Perl must be compiled with largefile and 64-bit long support)
52 # Set to 4 and 'N' for 32-bit data length prefixes. Limit of 4 GB for each key/value.
53 # Upgrading this is possible (see above) but probably not necessary. If you need
54 # more than 4 GB for a single key or value, this module is really not for you :-)
56 #my $DATA_LENGTH_SIZE = 4;
57 #my $DATA_LENGTH_PACK = 'N';
58 our ($LONG_SIZE, $LONG_PACK, $DATA_LENGTH_SIZE, $DATA_LENGTH_PACK);
61 # Maximum number of buckets per list before another level of indexing is done.
62 # Increase this value for slightly greater speed, but larger database files.
63 # DO NOT decrease this value below 16, due to risk of recursive reindex overrun.
68 # Better not adjust anything below here, unless you're me :-)
72 # Setup digest function for keys
74 our ($DIGEST_FUNC, $HASH_SIZE);
75 #my $DIGEST_FUNC = \&Digest::MD5::md5;
78 # Precalculate index and bucket sizes based on values above.
81 my ($INDEX_SIZE, $BUCKET_SIZE, $BUCKET_LIST_SIZE);
88 # Setup file and tag signatures. These should never change.
90 sub SIG_FILE () { 'DPDB' }
91 sub SIG_HASH () { 'H' }
92 sub SIG_ARRAY () { 'A' }
93 sub SIG_SCALAR () { 'S' }
94 sub SIG_NULL () { 'N' }
95 sub SIG_DATA () { 'D' }
96 sub SIG_INDEX () { 'I' }
97 sub SIG_BLIST () { 'B' }
101 # Setup constants for users to pass to new()
103 sub TYPE_HASH () { return SIG_HASH; }
104 sub TYPE_ARRAY () { return SIG_ARRAY; }
105 sub TYPE_SCALAR () { return SIG_SCALAR; }
111 if (scalar(@_) > 1) {
113 $proto->_throw_error( "Odd number of parameters to " . (caller(1))[2] );
117 elsif ( my $type = Scalar::Util::reftype($_[0]) ) {
118 if ( $type ne 'HASH' ) {
119 $proto->_throw_error( "Not a hashref in args to " . (caller(1))[2] );
124 $args = { file => shift };
132 # Class constructor method for Perl OO interface.
133 # Calls tie() and returns blessed reference to tied hash or array,
134 # providing a hybrid OO/tie interface.
137 my $args = $class->_get_args( @_ );
140 # Check if we want a tied hash or array.
143 if (defined($args->{type}) && $args->{type} eq TYPE_ARRAY) {
144 $class = 'DBM::Deep::Array';
145 require DBM::Deep::Array;
146 tie @$self, $class, %$args;
149 $class = 'DBM::Deep::Hash';
150 require DBM::Deep::Hash;
151 tie %$self, $class, %$args;
154 return bless $self, $class;
159 # Setup $self and bless into this class.
164 # These are the defaults to be optionally overridden below
167 base_offset => length(SIG_FILE),
170 foreach my $param ( keys %$self ) {
171 next unless exists $args->{$param};
172 $self->{$param} = delete $args->{$param}
175 $self->{root} = exists $args->{root}
177 : DBM::Deep::_::Root->new( $args );
179 if (!defined($self->fh)) { $self->_open(); }
186 require DBM::Deep::Hash;
187 return DBM::Deep::Hash->TIEHASH( @_ );
192 require DBM::Deep::Array;
193 return DBM::Deep::Array->TIEARRAY( @_ );
196 #XXX Unneeded now ...
202 # Open a fh to the database, create if nonexistent.
203 # Make sure file signature matches DBM::Deep spec.
205 my $self = $_[0]->_get_self;
207 if (defined($self->fh)) { $self->_close(); }
210 # Theoretically, adding O_BINARY should remove the need for the binmode
211 # Of course, testing it is going to be ... interesting.
212 my $flags = O_RDWR | O_CREAT | O_BINARY;
215 sysopen( $fh, $self->root->{file}, $flags )
217 $self->root->{fh} = $fh;
218 }; if ($@ ) { $self->_throw_error( "Received error: $@\n" ); }
219 if (! defined($self->fh)) {
220 return $self->_throw_error("Cannot sysopen file: " . $self->root->{file} . ": $!");
225 #XXX Can we remove this by using the right sysopen() flags?
226 # Maybe ... q.v. above
227 binmode $fh; # for win32
229 if ($self->root->{autoflush}) {
230 my $old = select $fh;
236 seek($fh, 0, SEEK_SET);
239 my $bytes_read = read( $fh, $signature, length(SIG_FILE));
242 # File is empty -- write signature and master index
245 seek($fh, 0, SEEK_SET);
247 $self->_create_tag($self->base_offset, $self->type, chr(0) x $INDEX_SIZE);
249 my $plain_key = "[base]";
250 print($fh pack($DATA_LENGTH_PACK, length($plain_key)) . $plain_key );
252 # Flush the filehandle
253 my $old_fh = select $fh;
259 my @stats = stat($fh);
260 $self->root->{inode} = $stats[1];
261 $self->root->{end} = $stats[7];
267 # Check signature was valid
269 unless ($signature eq SIG_FILE) {
271 return $self->_throw_error("Signature not found -- file is not a Deep DB");
274 my @stats = stat($fh);
275 $self->root->{inode} = $stats[1];
276 $self->root->{end} = $stats[7];
279 # Get our type from master index signature
281 my $tag = $self->_load_tag($self->base_offset);
283 #XXX We probably also want to store the hash algorithm name and not assume anything
284 #XXX The cool thing would be to allow a different hashing algorithm at every level
287 return $self->_throw_error("Corrupted file, no master index record");
289 if ($self->{type} ne $tag->{signature}) {
290 return $self->_throw_error("File type mismatch");
300 my $self = $_[0]->_get_self;
301 close $self->root->{fh} if $self->root->{fh};
302 $self->root->{fh} = undef;
307 # Given offset, signature and content, create tag and write to disk
309 my ($self, $offset, $sig, $content) = @_;
310 my $size = length($content);
314 seek($fh, $offset, SEEK_SET);
315 print($fh $sig . pack($DATA_LENGTH_PACK, $size) . $content );
317 if ($offset == $self->root->{end}) {
318 $self->root->{end} += SIG_SIZE + $DATA_LENGTH_SIZE + $size;
324 offset => $offset + SIG_SIZE + $DATA_LENGTH_SIZE,
331 # Given offset, load single tag and return signature, size and data
338 seek($fh, $offset, SEEK_SET);
339 if (eof $fh) { return undef; }
342 read( $fh, $sig, SIG_SIZE);
345 read( $fh, $size, $DATA_LENGTH_SIZE);
346 $size = unpack($DATA_LENGTH_PACK, $size);
349 read( $fh, $buffer, $size);
354 offset => $offset + SIG_SIZE + $DATA_LENGTH_SIZE,
361 # Given index tag, lookup single entry in index and return .
364 my ($tag, $index) = @_;
366 my $location = unpack($LONG_PACK, substr($tag->{content}, $index * $LONG_SIZE, $LONG_SIZE) );
367 if (!$location) { return; }
369 return $self->_load_tag( $location );
374 # Adds one key/value pair to bucket list, given offset, MD5 digest of key,
375 # plain (undigested) key and value.
378 my ($tag, $md5, $plain_key, $value) = @_;
379 my $keys = $tag->{content};
383 # added ref() check first to avoid eval and runtime exception for every
384 # scalar value being stored. performance tweak.
385 my $is_dbm_deep = ref($value) && eval { $value->isa( 'DBM::Deep' ) };
387 my $internal_ref = $is_dbm_deep && ($value->root eq $self->root);
392 # Iterate through buckets, seeing if this is a new entry or a replace.
394 for (my $i=0; $i<$MAX_BUCKETS; $i++) {
395 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
396 my $subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
399 # Found empty bucket (end of list). Populate and exit loop.
403 $location = $internal_ref
404 ? $value->base_offset
405 : $self->root->{end};
407 seek($fh, $tag->{offset} + ($i * $BUCKET_SIZE), SEEK_SET);
408 print($fh $md5 . pack($LONG_PACK, $location) );
411 elsif ($md5 eq $key) {
413 # Found existing bucket with same key. Replace with new value.
418 $location = $value->base_offset;
419 seek($fh, $tag->{offset} + ($i * $BUCKET_SIZE), SEEK_SET);
420 print($fh $md5 . pack($LONG_PACK, $location) );
423 seek($fh, $subloc + SIG_SIZE, SEEK_SET);
425 read( $fh, $size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
428 # If value is a hash, array, or raw value with equal or less size, we can
429 # reuse the same content area of the database. Otherwise, we have to create
430 # a new content area at the EOF.
433 my $r = Scalar::Util::reftype( $value ) || '';
434 if ( $r eq 'HASH' || $r eq 'ARRAY' ) {
435 $actual_length = $INDEX_SIZE;
437 # if autobless is enabled, must also take into consideration
438 # the class name, as it is stored along with key/value.
439 if ( $self->root->{autobless} ) {
440 my $value_class = Scalar::Util::blessed($value);
441 if ( defined $value_class && $value_class ne 'DBM::Deep' ) {
442 $actual_length += length($value_class);
446 else { $actual_length = length($value); }
448 if ($actual_length <= $size) {
452 $location = $self->root->{end};
453 seek($fh, $tag->{offset} + ($i * $BUCKET_SIZE) + $HASH_SIZE, SEEK_SET);
454 print($fh pack($LONG_PACK, $location) );
462 # If this is an internal reference, return now.
463 # No need to write value or plain key
470 # If bucket didn't fit into list, split into a new index level
473 seek($fh, $tag->{ref_loc}, SEEK_SET);
474 print($fh pack($LONG_PACK, $self->root->{end}) );
476 my $index_tag = $self->_create_tag($self->root->{end}, SIG_INDEX, chr(0) x $INDEX_SIZE);
479 $keys .= $md5 . pack($LONG_PACK, 0);
481 for (my $i=0; $i<=$MAX_BUCKETS; $i++) {
482 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
484 my $old_subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
485 my $num = ord(substr($key, $tag->{ch} + 1, 1));
487 if ($offsets[$num]) {
488 my $offset = $offsets[$num] + SIG_SIZE + $DATA_LENGTH_SIZE;
489 seek($fh, $offset, SEEK_SET);
491 read( $fh, $subkeys, $BUCKET_LIST_SIZE);
493 for (my $k=0; $k<$MAX_BUCKETS; $k++) {
494 my $subloc = unpack($LONG_PACK, substr($subkeys, ($k * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
496 seek($fh, $offset + ($k * $BUCKET_SIZE), SEEK_SET);
497 print($fh $key . pack($LONG_PACK, $old_subloc || $self->root->{end}) );
503 $offsets[$num] = $self->root->{end};
504 seek($fh, $index_tag->{offset} + ($num * $LONG_SIZE), SEEK_SET);
505 print($fh pack($LONG_PACK, $self->root->{end}) );
507 my $blist_tag = $self->_create_tag($self->root->{end}, SIG_BLIST, chr(0) x $BUCKET_LIST_SIZE);
509 seek($fh, $blist_tag->{offset}, SEEK_SET);
510 print($fh $key . pack($LONG_PACK, $old_subloc || $self->root->{end}) );
515 $location ||= $self->root->{end};
516 } # re-index bucket list
519 # Seek to content area and store signature, value and plaintext key
523 seek($fh, $location, SEEK_SET);
526 # Write signature based on content type, set content length and write actual value.
528 my $r = Scalar::Util::reftype($value) || '';
530 print($fh TYPE_HASH );
531 print($fh pack($DATA_LENGTH_PACK, $INDEX_SIZE) . chr(0) x $INDEX_SIZE );
532 $content_length = $INDEX_SIZE;
534 elsif ($r eq 'ARRAY') {
535 print($fh TYPE_ARRAY );
536 print($fh pack($DATA_LENGTH_PACK, $INDEX_SIZE) . chr(0) x $INDEX_SIZE );
537 $content_length = $INDEX_SIZE;
539 elsif (!defined($value)) {
540 print($fh SIG_NULL );
541 print($fh pack($DATA_LENGTH_PACK, 0) );
545 print($fh SIG_DATA );
546 print($fh pack($DATA_LENGTH_PACK, length($value)) . $value );
547 $content_length = length($value);
551 # Plain key is stored AFTER value, as keys are typically fetched less often.
553 print($fh pack($DATA_LENGTH_PACK, length($plain_key)) . $plain_key );
556 # If value is blessed, preserve class name
558 if ( $self->root->{autobless} ) {
559 my $value_class = Scalar::Util::blessed($value);
560 if ( defined $value_class && $value_class ne 'DBM::Deep' ) {
562 # Blessed ref -- will restore later
565 print($fh pack($DATA_LENGTH_PACK, length($value_class)) . $value_class );
566 $content_length += 1;
567 $content_length += $DATA_LENGTH_SIZE + length($value_class);
571 $content_length += 1;
576 # If this is a new content area, advance EOF counter
578 if ($location == $self->root->{end}) {
579 $self->root->{end} += SIG_SIZE;
580 $self->root->{end} += $DATA_LENGTH_SIZE + $content_length;
581 $self->root->{end} += $DATA_LENGTH_SIZE + length($plain_key);
585 # If content is a hash or array, create new child DBM::Deep object and
586 # pass each key or element to it.
589 my $branch = DBM::Deep->new(
591 base_offset => $location,
594 foreach my $key (keys %{$value}) {
595 $branch->STORE( $key, $value->{$key} );
598 elsif ($r eq 'ARRAY') {
599 my $branch = DBM::Deep->new(
601 base_offset => $location,
605 foreach my $element (@{$value}) {
606 $branch->STORE( $index, $element );
614 return $self->_throw_error("Fatal error: indexing failed -- possibly due to corruption in file");
617 sub _get_bucket_value {
619 # Fetch single value given tag and MD5 digested key.
622 my ($tag, $md5) = @_;
623 my $keys = $tag->{content};
628 # Iterate through buckets, looking for a key match
631 for (my $i=0; $i<$MAX_BUCKETS; $i++) {
632 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
633 my $subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
637 # Hit end of list, no match
642 if ( $md5 ne $key ) {
647 # Found match -- seek to offset and read signature
650 seek($fh, $subloc, SEEK_SET);
651 read( $fh, $signature, SIG_SIZE);
654 # If value is a hash or array, return new DBM::Deep object with correct offset
656 if (($signature eq TYPE_HASH) || ($signature eq TYPE_ARRAY)) {
657 my $obj = DBM::Deep->new(
659 base_offset => $subloc,
663 if ($self->root->{autobless}) {
665 # Skip over value and plain key to see if object needs
668 seek($fh, $DATA_LENGTH_SIZE + $INDEX_SIZE, SEEK_CUR);
671 read( $fh, $size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
672 if ($size) { seek($fh, $size, SEEK_CUR); }
675 read( $fh, $bless_bit, 1);
676 if (ord($bless_bit)) {
678 # Yes, object needs to be re-blessed
681 read( $fh, $size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
682 if ($size) { read( $fh, $class_name, $size); }
683 if ($class_name) { $obj = bless( $obj, $class_name ); }
691 # Otherwise return actual value
693 elsif ($signature eq SIG_DATA) {
696 read( $fh, $size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
697 if ($size) { read( $fh, $value, $size); }
702 # Key exists, but content is null
712 # Delete single key/value pair given tag and MD5 digested key.
715 my ($tag, $md5) = @_;
716 my $keys = $tag->{content};
721 # Iterate through buckets, looking for a key match
724 for (my $i=0; $i<$MAX_BUCKETS; $i++) {
725 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
726 my $subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
730 # Hit end of list, no match
735 if ( $md5 ne $key ) {
740 # Matched key -- delete bucket and return
742 seek($fh, $tag->{offset} + ($i * $BUCKET_SIZE), SEEK_SET);
743 print($fh substr($keys, ($i+1) * $BUCKET_SIZE ) );
744 print($fh chr(0) x $BUCKET_SIZE );
754 # Check existence of single key given tag and MD5 digested key.
757 my ($tag, $md5) = @_;
758 my $keys = $tag->{content};
761 # Iterate through buckets, looking for a key match
764 for (my $i=0; $i<$MAX_BUCKETS; $i++) {
765 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
766 my $subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
770 # Hit end of list, no match
775 if ( $md5 ne $key ) {
780 # Matched key -- return true
788 sub _find_bucket_list {
790 # Locate offset for bucket list, given digested key
796 # Locate offset for bucket list using digest index system
799 my $tag = $self->_load_tag($self->base_offset);
800 if (!$tag) { return; }
802 while ($tag->{signature} ne SIG_BLIST) {
803 $tag = $self->_index_lookup($tag, ord(substr($md5, $ch, 1)));
804 if (!$tag) { return; }
811 sub _traverse_index {
813 # Scan index and recursively step into deeper levels, looking for next key.
815 my ($self, $offset, $ch, $force_return_next) = @_;
816 $force_return_next = undef unless $force_return_next;
818 my $tag = $self->_load_tag( $offset );
822 if ($tag->{signature} ne SIG_BLIST) {
823 my $content = $tag->{content};
825 if ($self->{return_next}) { $start = 0; }
826 else { $start = ord(substr($self->{prev_md5}, $ch, 1)); }
828 for (my $index = $start; $index < 256; $index++) {
829 my $subloc = unpack($LONG_PACK, substr($content, $index * $LONG_SIZE, $LONG_SIZE) );
831 my $result = $self->_traverse_index( $subloc, $ch + 1, $force_return_next );
832 if (defined($result)) { return $result; }
836 $self->{return_next} = 1;
839 elsif ($tag->{signature} eq SIG_BLIST) {
840 my $keys = $tag->{content};
841 if ($force_return_next) { $self->{return_next} = 1; }
844 # Iterate through buckets, looking for a key match
846 for (my $i=0; $i<$MAX_BUCKETS; $i++) {
847 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
848 my $subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
852 # End of bucket list -- return to outer loop
854 $self->{return_next} = 1;
857 elsif ($key eq $self->{prev_md5}) {
859 # Located previous key -- return next one found
861 $self->{return_next} = 1;
864 elsif ($self->{return_next}) {
866 # Seek to bucket location and skip over signature
868 seek($fh, $subloc + SIG_SIZE, SEEK_SET);
871 # Skip over value to get to plain key
874 read( $fh, $size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
875 if ($size) { seek($fh, $size, SEEK_CUR); }
878 # Read in plain key and return as scalar
881 read( $fh, $size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
882 if ($size) { read( $fh, $plain_key, $size); }
888 $self->{return_next} = 1;
889 } # tag is a bucket list
896 # Locate next key, given digested previous one
898 my $self = $_[0]->_get_self;
900 $self->{prev_md5} = $_[1] ? $_[1] : undef;
901 $self->{return_next} = 0;
904 # If the previous key was not specifed, start at the top and
905 # return the first one found.
907 if (!$self->{prev_md5}) {
908 $self->{prev_md5} = chr(0) x $HASH_SIZE;
909 $self->{return_next} = 1;
912 return $self->_traverse_index( $self->base_offset, 0 );
917 # If db locking is set, flock() the db file. If called multiple
918 # times before unlock(), then the same number of unlocks() must
919 # be called before the lock is released.
921 my $self = $_[0]->_get_self;
923 $type = LOCK_EX unless defined $type;
925 if (!defined($self->fh)) { return; }
927 if ($self->root->{locking}) {
928 if (!$self->root->{locked}) {
929 flock($self->fh, $type);
931 # double-check file inode, in case another process
932 # has optimize()d our file while we were waiting.
933 if ((stat($self->root->{file}))[1] != $self->root->{inode}) {
934 $self->_open(); # re-open
935 flock($self->fh, $type); # re-lock
938 $self->root->{locked}++;
948 # If db locking is set, unlock the db file. See note in lock()
949 # regarding calling lock() multiple times.
951 my $self = $_[0]->_get_self;
953 if (!defined($self->fh)) { return; }
955 if ($self->root->{locking} && $self->root->{locked} > 0) {
956 $self->root->{locked}--;
957 if (!$self->root->{locked}) { flock($self->fh, LOCK_UN); }
965 #XXX These uses of ref() need verified
968 # Copy single level of keys or elements to new DB handle.
969 # Recurse for nested structures
971 my $self = $_[0]->_get_self;
974 if ($self->type eq TYPE_HASH) {
975 my $key = $self->first_key();
977 my $value = $self->get($key);
978 #XXX This doesn't work with autobless
979 if (!ref($value)) { $db_temp->{$key} = $value; }
981 my $type = $value->type;
982 if ($type eq TYPE_HASH) { $db_temp->{$key} = {}; }
983 else { $db_temp->{$key} = []; }
984 $value->_copy_node( $db_temp->{$key} );
986 $key = $self->next_key($key);
990 my $length = $self->length();
991 for (my $index = 0; $index < $length; $index++) {
992 my $value = $self->get($index);
993 if (!ref($value)) { $db_temp->[$index] = $value; }
994 #XXX NO tests for this code
996 my $type = $value->type;
997 if ($type eq TYPE_HASH) { $db_temp->[$index] = {}; }
998 else { $db_temp->[$index] = []; }
999 $value->_copy_node( $db_temp->[$index] );
1007 # Recursively export into standard Perl hashes and arrays.
1009 my $self = $_[0]->_get_self;
1012 if ($self->type eq TYPE_HASH) { $temp = {}; }
1013 elsif ($self->type eq TYPE_ARRAY) { $temp = []; }
1016 $self->_copy_node( $temp );
1024 # Recursively import Perl hash/array structure
1026 #XXX This use of ref() seems to be ok
1027 if (!ref($_[0])) { return; } # Perl calls import() on use -- ignore
1029 my $self = $_[0]->_get_self;
1032 #XXX This use of ref() seems to be ok
1033 if (!ref($struct)) {
1035 # struct is not a reference, so just import based on our type
1039 if ($self->type eq TYPE_HASH) { $struct = {@_}; }
1040 elsif ($self->type eq TYPE_ARRAY) { $struct = [@_]; }
1043 my $r = Scalar::Util::reftype($struct) || '';
1044 if ($r eq "HASH" && $self->type eq TYPE_HASH) {
1045 foreach my $key (keys %$struct) { $self->put($key, $struct->{$key}); }
1047 elsif ($r eq "ARRAY" && $self->type eq TYPE_ARRAY) {
1048 $self->push( @$struct );
1051 return $self->_throw_error("Cannot import: type mismatch");
1059 # Rebuild entire database into new file, then move
1060 # it back on top of original.
1062 my $self = $_[0]->_get_self;
1064 #XXX Need to create a new test for this
1065 # if ($self->root->{links} > 1) {
1066 # return $self->_throw_error("Cannot optimize: reference count is greater than 1");
1069 my $db_temp = DBM::Deep->new(
1070 file => $self->root->{file} . '.tmp',
1074 return $self->_throw_error("Cannot optimize: failed to open temp file: $!");
1078 $self->_copy_node( $db_temp );
1082 # Attempt to copy user, group and permissions over to new file
1084 my @stats = stat($self->fh);
1085 my $perms = $stats[2] & 07777;
1086 my $uid = $stats[4];
1087 my $gid = $stats[5];
1088 chown( $uid, $gid, $self->root->{file} . '.tmp' );
1089 chmod( $perms, $self->root->{file} . '.tmp' );
1091 # q.v. perlport for more information on this variable
1092 if ( $^O eq 'MSWin32' || $^O eq 'cygwin' ) {
1094 # Potential race condition when optmizing on Win32 with locking.
1095 # The Windows filesystem requires that the filehandle be closed
1096 # before it is overwritten with rename(). This could be redone
1103 if (!rename $self->root->{file} . '.tmp', $self->root->{file}) {
1104 unlink $self->root->{file} . '.tmp';
1106 return $self->_throw_error("Optimize failed: Cannot copy temp file over original: $!");
1118 # Make copy of object and return
1120 my $self = $_[0]->_get_self;
1122 return DBM::Deep->new(
1123 type => $self->type,
1124 base_offset => $self->base_offset,
1130 my %is_legal_filter = map {
1133 store_key store_value
1134 fetch_key fetch_value
1139 # Setup filter function for storing or fetching the key or value
1141 my $self = $_[0]->_get_self;
1142 my $type = lc $_[1];
1143 my $func = $_[2] ? $_[2] : undef;
1145 if ( $is_legal_filter{$type} ) {
1146 $self->root->{"filter_$type"} = $func;
1160 # Get access to the root structure
1162 my $self = $_[0]->_get_self;
1163 return $self->{root};
1168 # Get access to the raw fh
1170 #XXX It will be useful, though, when we split out HASH and ARRAY
1171 my $self = $_[0]->_get_self;
1172 return $self->root->{fh};
1177 # Get type of current node (TYPE_HASH or TYPE_ARRAY)
1179 my $self = $_[0]->_get_self;
1180 return $self->{type};
1185 # Get base_offset of current node (TYPE_HASH or TYPE_ARRAY)
1187 my $self = $_[0]->_get_self;
1188 return $self->{base_offset};
1193 # Get last error string, or undef if no error
1196 #? ( _get_self($_[0])->{root}->{error} or undef )
1197 ? ( $_[0]->_get_self->{root}->{error} or undef )
1207 # Store error string in self
1209 my $self = $_[0]->_get_self;
1210 my $error_text = $_[1];
1212 if ( Scalar::Util::blessed $self ) {
1213 $self->root->{error} = $error_text;
1215 unless ($self->root->{debug}) {
1216 die "DBM::Deep: $error_text\n";
1219 warn "DBM::Deep: $error_text\n";
1223 die "DBM::Deep: $error_text\n";
1231 my $self = $_[0]->_get_self;
1233 undef $self->root->{error};
1238 # Precalculate index, bucket and bucket list sizes
1241 #XXX I don't like this ...
1242 set_pack() unless defined $LONG_SIZE;
1244 $INDEX_SIZE = 256 * $LONG_SIZE;
1245 $BUCKET_SIZE = $HASH_SIZE + $LONG_SIZE;
1246 $BUCKET_LIST_SIZE = $MAX_BUCKETS * $BUCKET_SIZE;
1251 # Set pack/unpack modes (see file header for more)
1253 my ($long_s, $long_p, $data_s, $data_p) = @_;
1255 $LONG_SIZE = $long_s ? $long_s : 4;
1256 $LONG_PACK = $long_p ? $long_p : 'N';
1258 $DATA_LENGTH_SIZE = $data_s ? $data_s : 4;
1259 $DATA_LENGTH_PACK = $data_p ? $data_p : 'N';
1266 # Set key digest function (default is MD5)
1268 my ($digest_func, $hash_size) = @_;
1270 $DIGEST_FUNC = $digest_func ? $digest_func : \&Digest::MD5::md5;
1271 $HASH_SIZE = $hash_size ? $hash_size : 16;
1277 # tie() methods (hashes and arrays)
1282 # Store single hash key/value or array element in database.
1284 my $self = $_[0]->_get_self;
1287 # User may be storing a hash, in which case we do not want it run
1288 # through the filtering system
1289 my $value = ($self->root->{filter_store_value} && !ref($_[2]))
1290 ? $self->root->{filter_store_value}->($_[2])
1293 my $md5 = $DIGEST_FUNC->($key);
1296 # Make sure file is open
1298 if (!defined($self->fh) && !$self->_open()) {
1304 # Request exclusive lock for writing
1306 $self->lock( LOCK_EX );
1311 # If locking is enabled, set 'end' parameter again, in case another
1312 # DB instance appended to our file while we were unlocked.
1314 if ($self->root->{locking} || $self->root->{volatile}) {
1315 $self->root->{end} = (stat($fh))[7];
1319 # Locate offset for bucket list using digest index system
1321 my $tag = $self->_load_tag($self->base_offset);
1323 $tag = $self->_create_tag($self->base_offset, SIG_INDEX, chr(0) x $INDEX_SIZE);
1327 while ($tag->{signature} ne SIG_BLIST) {
1328 my $num = ord(substr($md5, $ch, 1));
1329 my $new_tag = $self->_index_lookup($tag, $num);
1331 my $ref_loc = $tag->{offset} + ($num * $LONG_SIZE);
1332 seek($fh, $ref_loc, SEEK_SET);
1333 print($fh pack($LONG_PACK, $self->root->{end}) );
1335 $tag = $self->_create_tag($self->root->{end}, SIG_BLIST, chr(0) x $BUCKET_LIST_SIZE);
1336 $tag->{ref_loc} = $ref_loc;
1341 my $ref_loc = $tag->{offset} + ($num * $LONG_SIZE);
1343 $tag->{ref_loc} = $ref_loc;
1350 # Add key/value to bucket list
1352 my $result = $self->_add_bucket( $tag, $md5, $key, $value );
1361 # Fetch single value or element given plain key or array index
1363 my $self = shift->_get_self;
1367 # Make sure file is open
1369 if (!defined($self->fh)) { $self->_open(); }
1371 my $md5 = $DIGEST_FUNC->($key);
1374 # Request shared lock for reading
1376 $self->lock( LOCK_SH );
1378 my $tag = $self->_find_bucket_list( $md5 );
1385 # Get value from bucket list
1387 my $result = $self->_get_bucket_value( $tag, $md5 );
1391 #XXX What is ref() checking here?
1392 #YYY Filters only apply on scalar values, so the ref check is making
1393 #YYY sure the fetched bucket is a scalar, not a child hash or array.
1394 return ($result && !ref($result) && $self->root->{filter_fetch_value})
1395 ? $self->root->{filter_fetch_value}->($result)
1401 # Delete single key/value pair or element given plain key or array index
1403 my $self = $_[0]->_get_self;
1406 my $md5 = $DIGEST_FUNC->($key);
1409 # Make sure file is open
1411 if (!defined($self->fh)) { $self->_open(); }
1414 # Request exclusive lock for writing
1416 $self->lock( LOCK_EX );
1418 my $tag = $self->_find_bucket_list( $md5 );
1427 my $value = $self->_get_bucket_value( $tag, $md5 );
1428 if ($value && !ref($value) && $self->root->{filter_fetch_value}) {
1429 $value = $self->root->{filter_fetch_value}->($value);
1432 my $result = $self->_delete_bucket( $tag, $md5 );
1435 # If this object is an array and the key deleted was on the end of the stack,
1436 # decrement the length variable.
1446 # Check if a single key or element exists given plain key or array index
1448 my $self = $_[0]->_get_self;
1451 my $md5 = $DIGEST_FUNC->($key);
1454 # Make sure file is open
1456 if (!defined($self->fh)) { $self->_open(); }
1459 # Request shared lock for reading
1461 $self->lock( LOCK_SH );
1463 my $tag = $self->_find_bucket_list( $md5 );
1466 # For some reason, the built-in exists() function returns '' for false
1474 # Check if bucket exists and return 1 or ''
1476 my $result = $self->_bucket_exists( $tag, $md5 ) || '';
1485 # Clear all keys from hash, or all elements from array.
1487 my $self = $_[0]->_get_self;
1490 # Make sure file is open
1492 if (!defined($self->fh)) { $self->_open(); }
1495 # Request exclusive lock for writing
1497 $self->lock( LOCK_EX );
1501 seek($fh, $self->base_offset, SEEK_SET);
1507 $self->_create_tag($self->base_offset, $self->type, chr(0) x $INDEX_SIZE);
1515 # Public method aliases
1517 sub put { (shift)->STORE( @_ ) }
1518 sub store { (shift)->STORE( @_ ) }
1519 sub get { (shift)->FETCH( @_ ) }
1520 sub fetch { (shift)->FETCH( @_ ) }
1521 sub delete { (shift)->DELETE( @_ ) }
1522 sub exists { (shift)->EXISTS( @_ ) }
1523 sub clear { (shift)->CLEAR( @_ ) }
1525 package DBM::Deep::_::Root;
1539 filter_store_key => undef,
1540 filter_store_value => undef,
1541 filter_fetch_key => undef,
1542 filter_fetch_value => undef,
1553 return unless $self;
1555 close $self->{fh} if $self->{fh};
1566 DBM::Deep - A pure perl multi-level hash/array DBM
1571 my $db = DBM::Deep->new( "foo.db" );
1573 $db->{key} = 'value'; # tie() style
1576 $db->put('key' => 'value'); # OO style
1577 print $db->get('key');
1579 # true multi-level support
1580 $db->{my_complex} = [
1581 'hello', { perl => 'rules' },
1587 A unique flat-file database module, written in pure perl. True
1588 multi-level hash/array support (unlike MLDBM, which is faked), hybrid
1589 OO / tie() interface, cross-platform FTPable files, and quite fast. Can
1590 handle millions of keys and unlimited hash levels without significant
1591 slow-down. Written from the ground-up in pure perl -- this is NOT a
1592 wrapper around a C-based DBM. Out-of-the-box compatibility with Unix,
1593 Mac OS X and Windows.
1597 Hopefully you are using Perl's excellent CPAN module, which will download
1598 and install the module for you. If not, get the tarball, and run these
1610 Construction can be done OO-style (which is the recommended way), or using
1611 Perl's tie() function. Both are examined here.
1613 =head2 OO CONSTRUCTION
1615 The recommended way to construct a DBM::Deep object is to use the new()
1616 method, which gets you a blessed, tied hash or array reference.
1618 my $db = DBM::Deep->new( "foo.db" );
1620 This opens a new database handle, mapped to the file "foo.db". If this
1621 file does not exist, it will automatically be created. DB files are
1622 opened in "r+" (read/write) mode, and the type of object returned is a
1623 hash, unless otherwise specified (see L<OPTIONS> below).
1625 You can pass a number of options to the constructor to specify things like
1626 locking, autoflush, etc. This is done by passing an inline hash:
1628 my $db = DBM::Deep->new(
1634 Notice that the filename is now specified I<inside> the hash with
1635 the "file" parameter, as opposed to being the sole argument to the
1636 constructor. This is required if any options are specified.
1637 See L<OPTIONS> below for the complete list.
1641 You can also start with an array instead of a hash. For this, you must
1642 specify the C<type> parameter:
1644 my $db = DBM::Deep->new(
1646 type => DBM::Deep->TYPE_ARRAY
1649 B<Note:> Specifing the C<type> parameter only takes effect when beginning
1650 a new DB file. If you create a DBM::Deep object with an existing file, the
1651 C<type> will be loaded from the file header, and an error will be thrown if
1652 the wrong type is passed in.
1654 =head2 TIE CONSTRUCTION
1656 Alternately, you can create a DBM::Deep handle by using Perl's built-in
1657 tie() function. The object returned from tie() can be used to call methods,
1658 such as lock() and unlock(), but cannot be used to assign to the DBM::Deep
1659 file (as expected with most tie'd objects).
1662 my $db = tie %hash, "DBM::Deep", "foo.db";
1665 my $db = tie @array, "DBM::Deep", "bar.db";
1667 As with the OO constructor, you can replace the DB filename parameter with
1668 a hash containing one or more options (see L<OPTIONS> just below for the
1671 tie %hash, "DBM::Deep", {
1679 There are a number of options that can be passed in when constructing your
1680 DBM::Deep objects. These apply to both the OO- and tie- based approaches.
1686 Filename of the DB file to link the handle to. You can pass a full absolute
1687 filesystem path, partial path, or a plain filename if the file is in the
1688 current working directory. This is a required parameter.
1692 This parameter specifies what type of object to create, a hash or array. Use
1693 one of these two constants: C<DBM::Deep-E<gt>TYPE_HASH> or C<DBM::Deep-E<gt>TYPE_ARRAY>.
1694 This only takes effect when beginning a new file. This is an optional
1695 parameter, and defaults to C<DBM::Deep-E<gt>TYPE_HASH>.
1699 Specifies whether locking is to be enabled. DBM::Deep uses Perl's Fnctl flock()
1700 function to lock the database in exclusive mode for writes, and shared mode for
1701 reads. Pass any true value to enable. This affects the base DB handle I<and
1702 any child hashes or arrays> that use the same DB file. This is an optional
1703 parameter, and defaults to 0 (disabled). See L<LOCKING> below for more.
1707 Specifies whether autoflush is to be enabled on the underlying filehandle.
1708 This obviously slows down write operations, but is required if you may have
1709 multiple processes accessing the same DB file (also consider enable I<locking>
1710 or at least I<volatile>). Pass any true value to enable. This is an optional
1711 parameter, and defaults to 0 (disabled).
1715 If I<volatile> mode is enabled, DBM::Deep will stat() the DB file before each
1716 STORE() operation. This is required if an outside force may change the size of
1717 the file between transactions. Locking also implicitly enables volatile. This
1718 is useful if you want to use a different locking system or write your own. Pass
1719 any true value to enable. This is an optional parameter, and defaults to 0
1724 If I<autobless> mode is enabled, DBM::Deep will preserve blessed hashes, and
1725 restore them when fetched. This is an B<experimental> feature, and does have
1726 side-effects. Basically, when hashes are re-blessed into their original
1727 classes, they are no longer blessed into the DBM::Deep class! So you won't be
1728 able to call any DBM::Deep methods on them. You have been warned.
1729 This is an optional parameter, and defaults to 0 (disabled).
1733 See L<FILTERS> below.
1737 Setting I<debug> mode will make all errors non-fatal, dump them out to
1738 STDERR, and continue on. This is for debugging purposes only, and probably
1739 not what you want. This is an optional parameter, and defaults to 0 (disabled).
1743 Instead of passing a file path, you can instead pass a handle to an pre-opened
1744 filehandle. Note: Beware of using the magick *DATA handle, as this actually
1745 contains your entire Perl script, as well as the data following the __DATA__
1746 marker. This will not work, because DBM::Deep uses absolute seek()s into the
1747 file. Instead, consider reading *DATA into an IO::Scalar handle, then passing
1748 in that. Also please note optimize() will NOT work when passing in only a
1749 handle. Pass in a real filename in order to use optimize().
1753 =head1 TIE INTERFACE
1755 With DBM::Deep you can access your databases using Perl's standard hash/array
1756 syntax. Because all DBM::Deep objects are I<tied> to hashes or arrays, you can
1757 treat them as such. DBM::Deep will intercept all reads/writes and direct them
1758 to the right place -- the DB file. This has nothing to do with the
1759 L<TIE CONSTRUCTION> section above. This simply tells you how to use DBM::Deep
1760 using regular hashes and arrays, rather than calling functions like C<get()>
1761 and C<put()> (although those work too). It is entirely up to you how to want
1762 to access your databases.
1766 You can treat any DBM::Deep object like a normal Perl hash reference. Add keys,
1767 or even nested hashes (or arrays) using standard Perl syntax:
1769 my $db = DBM::Deep->new( "foo.db" );
1771 $db->{mykey} = "myvalue";
1773 $db->{myhash}->{subkey} = "subvalue";
1775 print $db->{myhash}->{subkey} . "\n";
1777 You can even step through hash keys using the normal Perl C<keys()> function:
1779 foreach my $key (keys %$db) {
1780 print "$key: " . $db->{$key} . "\n";
1783 Remember that Perl's C<keys()> function extracts I<every> key from the hash and
1784 pushes them onto an array, all before the loop even begins. If you have an
1785 extra large hash, this may exhaust Perl's memory. Instead, consider using
1786 Perl's C<each()> function, which pulls keys/values one at a time, using very
1789 while (my ($key, $value) = each %$db) {
1790 print "$key: $value\n";
1793 Please note that when using C<each()>, you should always pass a direct
1794 hash reference, not a lookup. Meaning, you should B<never> do this:
1797 while (my ($key, $value) = each %{$db->{foo}}) { # BAD
1799 This causes an infinite loop, because for each iteration, Perl is calling
1800 FETCH() on the $db handle, resulting in a "new" hash for foo every time, so
1801 it effectively keeps returning the first key over and over again. Instead,
1802 assign a temporary variable to C<$db->{foo}>, then pass that to each().
1806 As with hashes, you can treat any DBM::Deep object like a normal Perl array
1807 reference. This includes inserting, removing and manipulating elements,
1808 and the C<push()>, C<pop()>, C<shift()>, C<unshift()> and C<splice()> functions.
1809 The object must have first been created using type C<DBM::Deep-E<gt>TYPE_ARRAY>,
1810 or simply be a nested array reference inside a hash. Example:
1812 my $db = DBM::Deep->new(
1813 file => "foo-array.db",
1814 type => DBM::Deep->TYPE_ARRAY
1818 push @$db, "bar", "baz";
1819 unshift @$db, "bah";
1821 my $last_elem = pop @$db; # baz
1822 my $first_elem = shift @$db; # bah
1823 my $second_elem = $db->[1]; # bar
1825 my $num_elements = scalar @$db;
1829 In addition to the I<tie()> interface, you can also use a standard OO interface
1830 to manipulate all aspects of DBM::Deep databases. Each type of object (hash or
1831 array) has its own methods, but both types share the following common methods:
1832 C<put()>, C<get()>, C<exists()>, C<delete()> and C<clear()>.
1836 =item * put() / store()
1838 Stores a new hash key/value pair, or sets an array element value. Takes two
1839 arguments, the hash key or array index, and the new value. The value can be
1840 a scalar, hash ref or array ref. Returns true on success, false on failure.
1842 $db->put("foo", "bar"); # for hashes
1843 $db->put(1, "bar"); # for arrays
1845 =item * get() / fetch()
1847 Fetches the value of a hash key or array element. Takes one argument: the hash
1848 key or array index. Returns a scalar, hash ref or array ref, depending on the
1851 my $value = $db->get("foo"); # for hashes
1852 my $value = $db->get(1); # for arrays
1856 Checks if a hash key or array index exists. Takes one argument: the hash key
1857 or array index. Returns true if it exists, false if not.
1859 if ($db->exists("foo")) { print "yay!\n"; } # for hashes
1860 if ($db->exists(1)) { print "yay!\n"; } # for arrays
1864 Deletes one hash key/value pair or array element. Takes one argument: the hash
1865 key or array index. Returns true on success, false if not found. For arrays,
1866 the remaining elements located after the deleted element are NOT moved over.
1867 The deleted element is essentially just undefined, which is exactly how Perl's
1868 internal arrays work. Please note that the space occupied by the deleted
1869 key/value or element is B<not> reused again -- see L<UNUSED SPACE RECOVERY>
1870 below for details and workarounds.
1872 $db->delete("foo"); # for hashes
1873 $db->delete(1); # for arrays
1877 Deletes B<all> hash keys or array elements. Takes no arguments. No return
1878 value. Please note that the space occupied by the deleted keys/values or
1879 elements is B<not> reused again -- see L<UNUSED SPACE RECOVERY> below for
1880 details and workarounds.
1882 $db->clear(); # hashes or arrays
1888 For hashes, DBM::Deep supports all the common methods described above, and the
1889 following additional methods: C<first_key()> and C<next_key()>.
1895 Returns the "first" key in the hash. As with built-in Perl hashes, keys are
1896 fetched in an undefined order (which appears random). Takes no arguments,
1897 returns the key as a scalar value.
1899 my $key = $db->first_key();
1903 Returns the "next" key in the hash, given the previous one as the sole argument.
1904 Returns undef if there are no more keys to be fetched.
1906 $key = $db->next_key($key);
1910 Here are some examples of using hashes:
1912 my $db = DBM::Deep->new( "foo.db" );
1914 $db->put("foo", "bar");
1915 print "foo: " . $db->get("foo") . "\n";
1917 $db->put("baz", {}); # new child hash ref
1918 $db->get("baz")->put("buz", "biz");
1919 print "buz: " . $db->get("baz")->get("buz") . "\n";
1921 my $key = $db->first_key();
1923 print "$key: " . $db->get($key) . "\n";
1924 $key = $db->next_key($key);
1927 if ($db->exists("foo")) { $db->delete("foo"); }
1931 For arrays, DBM::Deep supports all the common methods described above, and the
1932 following additional methods: C<length()>, C<push()>, C<pop()>, C<shift()>,
1933 C<unshift()> and C<splice()>.
1939 Returns the number of elements in the array. Takes no arguments.
1941 my $len = $db->length();
1945 Adds one or more elements onto the end of the array. Accepts scalars, hash
1946 refs or array refs. No return value.
1948 $db->push("foo", "bar", {});
1952 Fetches the last element in the array, and deletes it. Takes no arguments.
1953 Returns undef if array is empty. Returns the element value.
1955 my $elem = $db->pop();
1959 Fetches the first element in the array, deletes it, then shifts all the
1960 remaining elements over to take up the space. Returns the element value. This
1961 method is not recommended with large arrays -- see L<LARGE ARRAYS> below for
1964 my $elem = $db->shift();
1968 Inserts one or more elements onto the beginning of the array, shifting all
1969 existing elements over to make room. Accepts scalars, hash refs or array refs.
1970 No return value. This method is not recommended with large arrays -- see
1971 <LARGE ARRAYS> below for details.
1973 $db->unshift("foo", "bar", {});
1977 Performs exactly like Perl's built-in function of the same name. See L<perldoc
1978 -f splice> for usage -- it is too complicated to document here. This method is
1979 not recommended with large arrays -- see L<LARGE ARRAYS> below for details.
1983 Here are some examples of using arrays:
1985 my $db = DBM::Deep->new(
1987 type => DBM::Deep->TYPE_ARRAY
1990 $db->push("bar", "baz");
1991 $db->unshift("foo");
1994 my $len = $db->length();
1995 print "length: $len\n"; # 4
1997 for (my $k=0; $k<$len; $k++) {
1998 print "$k: " . $db->get($k) . "\n";
2001 $db->splice(1, 2, "biz", "baf");
2003 while (my $elem = shift @$db) {
2004 print "shifted: $elem\n";
2009 Enable automatic file locking by passing a true value to the C<locking>
2010 parameter when constructing your DBM::Deep object (see L<SETUP> above).
2012 my $db = DBM::Deep->new(
2017 This causes DBM::Deep to C<flock()> the underlying filehandle with exclusive
2018 mode for writes, and shared mode for reads. This is required if you have
2019 multiple processes accessing the same database file, to avoid file corruption.
2020 Please note that C<flock()> does NOT work for files over NFS. See L<DB OVER
2021 NFS> below for more.
2023 =head2 EXPLICIT LOCKING
2025 You can explicitly lock a database, so it remains locked for multiple
2026 transactions. This is done by calling the C<lock()> method, and passing an
2027 optional lock mode argument (defaults to exclusive mode). This is particularly
2028 useful for things like counters, where the current value needs to be fetched,
2029 then incremented, then stored again.
2032 my $counter = $db->get("counter");
2034 $db->put("counter", $counter);
2043 You can pass C<lock()> an optional argument, which specifies which mode to use
2044 (exclusive or shared). Use one of these two constants: C<DBM::Deep-E<gt>LOCK_EX>
2045 or C<DBM::Deep-E<gt>LOCK_SH>. These are passed directly to C<flock()>, and are the
2046 same as the constants defined in Perl's C<Fcntl> module.
2048 $db->lock( DBM::Deep->LOCK_SH );
2052 If you want to implement your own file locking scheme, be sure to create your
2053 DBM::Deep objects setting the C<volatile> option to true. This hints to DBM::Deep
2054 that the DB file may change between transactions. See L<LOW-LEVEL ACCESS>
2057 =head1 IMPORTING/EXPORTING
2059 You can import existing complex structures by calling the C<import()> method,
2060 and export an entire database into an in-memory structure using the C<export()>
2061 method. Both are examined here.
2065 Say you have an existing hash with nested hashes/arrays inside it. Instead of
2066 walking the structure and adding keys/elements to the database as you go,
2067 simply pass a reference to the C<import()> method. This recursively adds
2068 everything to an existing DBM::Deep object for you. Here is an example:
2073 array1 => [ "elem0", "elem1", "elem2" ],
2075 subkey1 => "subvalue1",
2076 subkey2 => "subvalue2"
2080 my $db = DBM::Deep->new( "foo.db" );
2081 $db->import( $struct );
2083 print $db->{key1} . "\n"; # prints "value1"
2085 This recursively imports the entire C<$struct> object into C<$db>, including
2086 all nested hashes and arrays. If the DBM::Deep object contains exsiting data,
2087 keys are merged with the existing ones, replacing if they already exist.
2088 The C<import()> method can be called on any database level (not just the base
2089 level), and works with both hash and array DB types.
2091 B<Note:> Make sure your existing structure has no circular references in it.
2092 These will cause an infinite loop when importing.
2096 Calling the C<export()> method on an existing DBM::Deep object will return
2097 a reference to a new in-memory copy of the database. The export is done
2098 recursively, so all nested hashes/arrays are all exported to standard Perl
2099 objects. Here is an example:
2101 my $db = DBM::Deep->new( "foo.db" );
2103 $db->{key1} = "value1";
2104 $db->{key2} = "value2";
2106 $db->{hash1}->{subkey1} = "subvalue1";
2107 $db->{hash1}->{subkey2} = "subvalue2";
2109 my $struct = $db->export();
2111 print $struct->{key1} . "\n"; # prints "value1"
2113 This makes a complete copy of the database in memory, and returns a reference
2114 to it. The C<export()> method can be called on any database level (not just
2115 the base level), and works with both hash and array DB types. Be careful of
2116 large databases -- you can store a lot more data in a DBM::Deep object than an
2117 in-memory Perl structure.
2119 B<Note:> Make sure your database has no circular references in it.
2120 These will cause an infinite loop when exporting.
2124 DBM::Deep has a number of hooks where you can specify your own Perl function
2125 to perform filtering on incoming or outgoing data. This is a perfect
2126 way to extend the engine, and implement things like real-time compression or
2127 encryption. Filtering applies to the base DB level, and all child hashes /
2128 arrays. Filter hooks can be specified when your DBM::Deep object is first
2129 constructed, or by calling the C<set_filter()> method at any time. There are
2130 four available filter hooks, described below:
2134 =item * filter_store_key
2136 This filter is called whenever a hash key is stored. It
2137 is passed the incoming key, and expected to return a transformed key.
2139 =item * filter_store_value
2141 This filter is called whenever a hash key or array element is stored. It
2142 is passed the incoming value, and expected to return a transformed value.
2144 =item * filter_fetch_key
2146 This filter is called whenever a hash key is fetched (i.e. via
2147 C<first_key()> or C<next_key()>). It is passed the transformed key,
2148 and expected to return the plain key.
2150 =item * filter_fetch_value
2152 This filter is called whenever a hash key or array element is fetched.
2153 It is passed the transformed value, and expected to return the plain value.
2157 Here are the two ways to setup a filter hook:
2159 my $db = DBM::Deep->new(
2161 filter_store_value => \&my_filter_store,
2162 filter_fetch_value => \&my_filter_fetch
2167 $db->set_filter( "filter_store_value", \&my_filter_store );
2168 $db->set_filter( "filter_fetch_value", \&my_filter_fetch );
2170 Your filter function will be called only when dealing with SCALAR keys or
2171 values. When nested hashes and arrays are being stored/fetched, filtering
2172 is bypassed. Filters are called as static functions, passed a single SCALAR
2173 argument, and expected to return a single SCALAR value. If you want to
2174 remove a filter, set the function reference to C<undef>:
2176 $db->set_filter( "filter_store_value", undef );
2178 =head2 REAL-TIME ENCRYPTION EXAMPLE
2180 Here is a working example that uses the I<Crypt::Blowfish> module to
2181 do real-time encryption / decryption of keys & values with DBM::Deep Filters.
2182 Please visit L<http://search.cpan.org/search?module=Crypt::Blowfish> for more
2183 on I<Crypt::Blowfish>. You'll also need the I<Crypt::CBC> module.
2186 use Crypt::Blowfish;
2189 my $cipher = Crypt::CBC->new({
2190 'key' => 'my secret key',
2191 'cipher' => 'Blowfish',
2193 'regenerate_key' => 0,
2194 'padding' => 'space',
2198 my $db = DBM::Deep->new(
2199 file => "foo-encrypt.db",
2200 filter_store_key => \&my_encrypt,
2201 filter_store_value => \&my_encrypt,
2202 filter_fetch_key => \&my_decrypt,
2203 filter_fetch_value => \&my_decrypt,
2206 $db->{key1} = "value1";
2207 $db->{key2} = "value2";
2208 print "key1: " . $db->{key1} . "\n";
2209 print "key2: " . $db->{key2} . "\n";
2215 return $cipher->encrypt( $_[0] );
2218 return $cipher->decrypt( $_[0] );
2221 =head2 REAL-TIME COMPRESSION EXAMPLE
2223 Here is a working example that uses the I<Compress::Zlib> module to do real-time
2224 compression / decompression of keys & values with DBM::Deep Filters.
2225 Please visit L<http://search.cpan.org/search?module=Compress::Zlib> for
2226 more on I<Compress::Zlib>.
2231 my $db = DBM::Deep->new(
2232 file => "foo-compress.db",
2233 filter_store_key => \&my_compress,
2234 filter_store_value => \&my_compress,
2235 filter_fetch_key => \&my_decompress,
2236 filter_fetch_value => \&my_decompress,
2239 $db->{key1} = "value1";
2240 $db->{key2} = "value2";
2241 print "key1: " . $db->{key1} . "\n";
2242 print "key2: " . $db->{key2} . "\n";
2248 return Compress::Zlib::memGzip( $_[0] ) ;
2251 return Compress::Zlib::memGunzip( $_[0] ) ;
2254 B<Note:> Filtering of keys only applies to hashes. Array "keys" are
2255 actually numerical index numbers, and are not filtered.
2257 =head1 ERROR HANDLING
2259 Most DBM::Deep methods return a true value for success, and call die() on
2260 failure. You can wrap calls in an eval block to catch the die. Also, the
2261 actual error message is stored in an internal scalar, which can be fetched by
2262 calling the C<error()> method.
2264 my $db = DBM::Deep->new( "foo.db" ); # create hash
2265 eval { $db->push("foo"); }; # ILLEGAL -- push is array-only call
2267 print $@; # prints error message
2268 print $db->error(); # prints error message
2270 You can then call C<clear_error()> to clear the current error state.
2274 If you set the C<debug> option to true when creating your DBM::Deep object,
2275 all errors are considered NON-FATAL, and dumped to STDERR. This should only
2276 be used for debugging purposes and not production work. DBM::Deep expects errors
2277 to be thrown, not propagated back up the stack.
2279 =head1 LARGEFILE SUPPORT
2281 If you have a 64-bit system, and your Perl is compiled with both LARGEFILE
2282 and 64-bit support, you I<may> be able to create databases larger than 2 GB.
2283 DBM::Deep by default uses 32-bit file offset tags, but these can be changed
2284 by calling the static C<set_pack()> method before you do anything else.
2286 DBM::Deep::set_pack(8, 'Q');
2288 This tells DBM::Deep to pack all file offsets with 8-byte (64-bit) quad words
2289 instead of 32-bit longs. After setting these values your DB files have a
2290 theoretical maximum size of 16 XB (exabytes).
2292 B<Note:> Changing these values will B<NOT> work for existing database files.
2293 Only change this for new files, and make sure it stays set consistently
2294 throughout the file's life. If you do set these values, you can no longer
2295 access 32-bit DB files. You can, however, call C<set_pack(4, 'N')> to change
2296 back to 32-bit mode.
2298 B<Note:> I have not personally tested files > 2 GB -- all my systems have
2299 only a 32-bit Perl. However, I have received user reports that this does
2302 =head1 LOW-LEVEL ACCESS
2304 If you require low-level access to the underlying filehandle that DBM::Deep uses,
2305 you can call the C<fh()> method, which returns the handle:
2309 This method can be called on the root level of the datbase, or any child
2310 hashes or arrays. All levels share a I<root> structure, which contains things
2311 like the filehandle, a reference counter, and all the options specified
2312 when you created the object. You can get access to this root structure by
2313 calling the C<root()> method.
2315 my $root = $db->root();
2317 This is useful for changing options after the object has already been created,
2318 such as enabling/disabling locking, volatile or debug modes. You can also
2319 store your own temporary user data in this structure (be wary of name
2320 collision), which is then accessible from any child hash or array.
2322 =head1 CUSTOM DIGEST ALGORITHM
2324 DBM::Deep by default uses the I<Message Digest 5> (MD5) algorithm for hashing
2325 keys. However you can override this, and use another algorithm (such as SHA-256)
2326 or even write your own. But please note that DBM::Deep currently expects zero
2327 collisions, so your algorithm has to be I<perfect>, so to speak.
2328 Collision detection may be introduced in a later version.
2332 You can specify a custom digest algorithm by calling the static C<set_digest()>
2333 function, passing a reference to a subroutine, and the length of the algorithm's
2334 hashes (in bytes). This is a global static function, which affects ALL DBM::Deep
2335 objects. Here is a working example that uses a 256-bit hash from the
2336 I<Digest::SHA256> module. Please see
2337 L<http://search.cpan.org/search?module=Digest::SHA256> for more.
2342 my $context = Digest::SHA256::new(256);
2344 DBM::Deep::set_digest( \&my_digest, 32 );
2346 my $db = DBM::Deep->new( "foo-sha.db" );
2348 $db->{key1} = "value1";
2349 $db->{key2} = "value2";
2350 print "key1: " . $db->{key1} . "\n";
2351 print "key2: " . $db->{key2} . "\n";
2357 return substr( $context->hash($_[0]), 0, 32 );
2360 B<Note:> Your returned digest strings must be B<EXACTLY> the number
2361 of bytes you specify in the C<set_digest()> function (in this case 32).
2363 =head1 CIRCULAR REFERENCES
2365 DBM::Deep has B<experimental> support for circular references. Meaning you
2366 can have a nested hash key or array element that points to a parent object.
2367 This relationship is stored in the DB file, and is preserved between sessions.
2370 my $db = DBM::Deep->new( "foo.db" );
2373 $db->{circle} = $db; # ref to self
2375 print $db->{foo} . "\n"; # prints "foo"
2376 print $db->{circle}->{foo} . "\n"; # prints "foo" again
2378 One catch is, passing the object to a function that recursively walks the
2379 object tree (such as I<Data::Dumper> or even the built-in C<optimize()> or
2380 C<export()> methods) will result in an infinite loop. The other catch is,
2381 if you fetch the I<key> of a circular reference (i.e. using the C<first_key()>
2382 or C<next_key()> methods), you will get the I<target object's key>, not the
2383 ref's key. This gets even more interesting with the above example, where
2384 the I<circle> key points to the base DB object, which technically doesn't
2385 have a key. So I made DBM::Deep return "[base]" as the key name in that
2388 =head1 CAVEATS / ISSUES / BUGS
2390 This section describes all the known issues with DBM::Deep. It you have found
2391 something that is not listed here, please send e-mail to L<jhuckaby@cpan.org>.
2393 =head2 UNUSED SPACE RECOVERY
2395 One major caveat with DBM::Deep is that space occupied by existing keys and
2396 values is not recovered when they are deleted. Meaning if you keep deleting
2397 and adding new keys, your file will continuously grow. I am working on this,
2398 but in the meantime you can call the built-in C<optimize()> method from time to
2399 time (perhaps in a crontab or something) to recover all your unused space.
2401 $db->optimize(); # returns true on success
2403 This rebuilds the ENTIRE database into a new file, then moves it on top of
2404 the original. The new file will have no unused space, thus it will take up as
2405 little disk space as possible. Please note that this operation can take
2406 a long time for large files, and you need enough disk space to temporarily hold
2407 2 copies of your DB file. The temporary file is created in the same directory
2408 as the original, named with a ".tmp" extension, and is deleted when the
2409 operation completes. Oh, and if locking is enabled, the DB is automatically
2410 locked for the entire duration of the copy.
2412 B<WARNING:> Only call optimize() on the top-level node of the database, and
2413 make sure there are no child references lying around. DBM::Deep keeps a reference
2414 counter, and if it is greater than 1, optimize() will abort and return undef.
2416 =head2 AUTOVIVIFICATION
2418 Unfortunately, autovivification doesn't work with tied hashes. This appears to
2419 be a bug in Perl's tie() system, as I<Jakob Schmidt> encountered the very same
2420 issue with his I<DWH_FIle> module (see L<http://search.cpan.org/search?module=DWH_File>),
2421 and it is also mentioned in the BUGS section for the I<MLDBM> module <see
2422 L<http://search.cpan.org/search?module=MLDBM>). Basically, on a new db file,
2425 $db->{foo}->{bar} = "hello";
2427 Since "foo" doesn't exist, you cannot add "bar" to it. You end up with "foo"
2428 being an empty hash. Try this instead, which works fine:
2430 $db->{foo} = { bar => "hello" };
2432 As of Perl 5.8.7, this bug still exists. I have walked very carefully through
2433 the execution path, and Perl indeed passes an empty hash to the STORE() method.
2434 Probably a bug in Perl.
2436 =head2 FILE CORRUPTION
2438 The current level of error handling in DBM::Deep is minimal. Files I<are> checked
2439 for a 32-bit signature when opened, but other corruption in files can cause
2440 segmentation faults. DBM::Deep may try to seek() past the end of a file, or get
2441 stuck in an infinite loop depending on the level of corruption. File write
2442 operations are not checked for failure (for speed), so if you happen to run
2443 out of disk space, DBM::Deep will probably fail in a bad way. These things will
2444 be addressed in a later version of DBM::Deep.
2448 Beware of using DB files over NFS. DBM::Deep uses flock(), which works well on local
2449 filesystems, but will NOT protect you from file corruption over NFS. I've heard
2450 about setting up your NFS server with a locking daemon, then using lockf() to
2451 lock your files, but your mileage may vary there as well. From what I
2452 understand, there is no real way to do it. However, if you need access to the
2453 underlying filehandle in DBM::Deep for using some other kind of locking scheme like
2454 lockf(), see the L<LOW-LEVEL ACCESS> section above.
2456 =head2 COPYING OBJECTS
2458 Beware of copying tied objects in Perl. Very strange things can happen.
2459 Instead, use DBM::Deep's C<clone()> method which safely copies the object and
2460 returns a new, blessed, tied hash or array to the same level in the DB.
2462 my $copy = $db->clone();
2464 B<Note>: Since clone() here is cloning the object, not the database location, any
2465 modifications to either $db or $copy will be visible in both.
2469 Beware of using C<shift()>, C<unshift()> or C<splice()> with large arrays.
2470 These functions cause every element in the array to move, which can be murder
2471 on DBM::Deep, as every element has to be fetched from disk, then stored again in
2472 a different location. This will be addressed in the forthcoming version 1.00.
2476 This section discusses DBM::Deep's speed and memory usage.
2480 Obviously, DBM::Deep isn't going to be as fast as some C-based DBMs, such as
2481 the almighty I<BerkeleyDB>. But it makes up for it in features like true
2482 multi-level hash/array support, and cross-platform FTPable files. Even so,
2483 DBM::Deep is still pretty fast, and the speed stays fairly consistent, even
2484 with huge databases. Here is some test data:
2486 Adding 1,000,000 keys to new DB file...
2488 At 100 keys, avg. speed is 2,703 keys/sec
2489 At 200 keys, avg. speed is 2,642 keys/sec
2490 At 300 keys, avg. speed is 2,598 keys/sec
2491 At 400 keys, avg. speed is 2,578 keys/sec
2492 At 500 keys, avg. speed is 2,722 keys/sec
2493 At 600 keys, avg. speed is 2,628 keys/sec
2494 At 700 keys, avg. speed is 2,700 keys/sec
2495 At 800 keys, avg. speed is 2,607 keys/sec
2496 At 900 keys, avg. speed is 2,190 keys/sec
2497 At 1,000 keys, avg. speed is 2,570 keys/sec
2498 At 2,000 keys, avg. speed is 2,417 keys/sec
2499 At 3,000 keys, avg. speed is 1,982 keys/sec
2500 At 4,000 keys, avg. speed is 1,568 keys/sec
2501 At 5,000 keys, avg. speed is 1,533 keys/sec
2502 At 6,000 keys, avg. speed is 1,787 keys/sec
2503 At 7,000 keys, avg. speed is 1,977 keys/sec
2504 At 8,000 keys, avg. speed is 2,028 keys/sec
2505 At 9,000 keys, avg. speed is 2,077 keys/sec
2506 At 10,000 keys, avg. speed is 2,031 keys/sec
2507 At 20,000 keys, avg. speed is 1,970 keys/sec
2508 At 30,000 keys, avg. speed is 2,050 keys/sec
2509 At 40,000 keys, avg. speed is 2,073 keys/sec
2510 At 50,000 keys, avg. speed is 1,973 keys/sec
2511 At 60,000 keys, avg. speed is 1,914 keys/sec
2512 At 70,000 keys, avg. speed is 2,091 keys/sec
2513 At 80,000 keys, avg. speed is 2,103 keys/sec
2514 At 90,000 keys, avg. speed is 1,886 keys/sec
2515 At 100,000 keys, avg. speed is 1,970 keys/sec
2516 At 200,000 keys, avg. speed is 2,053 keys/sec
2517 At 300,000 keys, avg. speed is 1,697 keys/sec
2518 At 400,000 keys, avg. speed is 1,838 keys/sec
2519 At 500,000 keys, avg. speed is 1,941 keys/sec
2520 At 600,000 keys, avg. speed is 1,930 keys/sec
2521 At 700,000 keys, avg. speed is 1,735 keys/sec
2522 At 800,000 keys, avg. speed is 1,795 keys/sec
2523 At 900,000 keys, avg. speed is 1,221 keys/sec
2524 At 1,000,000 keys, avg. speed is 1,077 keys/sec
2526 This test was performed on a PowerMac G4 1gHz running Mac OS X 10.3.2 & Perl
2527 5.8.1, with an 80GB Ultra ATA/100 HD spinning at 7200RPM. The hash keys and
2528 values were between 6 - 12 chars in length. The DB file ended up at 210MB.
2529 Run time was 12 min 3 sec.
2533 One of the great things about DBM::Deep is that it uses very little memory.
2534 Even with huge databases (1,000,000+ keys) you will not see much increased
2535 memory on your process. DBM::Deep relies solely on the filesystem for storing
2536 and fetching data. Here is output from I</usr/bin/top> before even opening a
2539 PID USER PRI NI SIZE RSS SHARE STAT %CPU %MEM TIME COMMAND
2540 22831 root 11 0 2716 2716 1296 R 0.0 0.2 0:07 perl
2542 Basically the process is taking 2,716K of memory. And here is the same
2543 process after storing and fetching 1,000,000 keys:
2545 PID USER PRI NI SIZE RSS SHARE STAT %CPU %MEM TIME COMMAND
2546 22831 root 14 0 2772 2772 1328 R 0.0 0.2 13:32 perl
2548 Notice the memory usage increased by only 56K. Test was performed on a 700mHz
2549 x86 box running Linux RedHat 7.2 & Perl 5.6.1.
2551 =head1 DB FILE FORMAT
2553 In case you were interested in the underlying DB file format, it is documented
2554 here in this section. You don't need to know this to use the module, it's just
2555 included for reference.
2559 DBM::Deep files always start with a 32-bit signature to identify the file type.
2560 This is at offset 0. The signature is "DPDB" in network byte order. This is
2561 checked for when the file is opened and an error will be thrown if it's not found.
2565 The DBM::Deep file is in a I<tagged format>, meaning each section of the file
2566 has a standard header containing the type of data, the length of data, and then
2567 the data itself. The type is a single character (1 byte), the length is a
2568 32-bit unsigned long in network byte order, and the data is, well, the data.
2569 Here is how it unfolds:
2573 Immediately after the 32-bit file signature is the I<Master Index> record.
2574 This is a standard tag header followed by 1024 bytes (in 32-bit mode) or 2048
2575 bytes (in 64-bit mode) of data. The type is I<H> for hash or I<A> for array,
2576 depending on how the DBM::Deep object was constructed.
2578 The index works by looking at a I<MD5 Hash> of the hash key (or array index
2579 number). The first 8-bit char of the MD5 signature is the offset into the
2580 index, multipled by 4 in 32-bit mode, or 8 in 64-bit mode. The value of the
2581 index element is a file offset of the next tag for the key/element in question,
2582 which is usually a I<Bucket List> tag (see below).
2584 The next tag I<could> be another index, depending on how many keys/elements
2585 exist. See L<RE-INDEXING> below for details.
2589 A I<Bucket List> is a collection of 16 MD5 hashes for keys/elements, plus
2590 file offsets to where the actual data is stored. It starts with a standard
2591 tag header, with type I<B>, and a data size of 320 bytes in 32-bit mode, or
2592 384 bytes in 64-bit mode. Each MD5 hash is stored in full (16 bytes), plus
2593 the 32-bit or 64-bit file offset for the I<Bucket> containing the actual data.
2594 When the list fills up, a I<Re-Index> operation is performed (See
2595 L<RE-INDEXING> below).
2599 A I<Bucket> is a tag containing a key/value pair (in hash mode), or a
2600 index/value pair (in array mode). It starts with a standard tag header with
2601 type I<D> for scalar data (string, binary, etc.), or it could be a nested
2602 hash (type I<H>) or array (type I<A>). The value comes just after the tag
2603 header. The size reported in the tag header is only for the value, but then,
2604 just after the value is another size (32-bit unsigned long) and then the plain
2605 key itself. Since the value is likely to be fetched more often than the plain
2606 key, I figured it would be I<slightly> faster to store the value first.
2608 If the type is I<H> (hash) or I<A> (array), the value is another I<Master Index>
2609 record for the nested structure, where the process begins all over again.
2613 After a I<Bucket List> grows to 16 records, its allocated space in the file is
2614 exhausted. Then, when another key/element comes in, the list is converted to a
2615 new index record. However, this index will look at the next char in the MD5
2616 hash, and arrange new Bucket List pointers accordingly. This process is called
2617 I<Re-Indexing>. Basically, a new index tag is created at the file EOF, and all
2618 17 (16 + new one) keys/elements are removed from the old Bucket List and
2619 inserted into the new index. Several new Bucket Lists are created in the
2620 process, as a new MD5 char from the key is being examined (it is unlikely that
2621 the keys will all share the same next char of their MD5s).
2623 Because of the way the I<MD5> algorithm works, it is impossible to tell exactly
2624 when the Bucket Lists will turn into indexes, but the first round tends to
2625 happen right around 4,000 keys. You will see a I<slight> decrease in
2626 performance here, but it picks back up pretty quick (see L<SPEED> above). Then
2627 it takes B<a lot> more keys to exhaust the next level of Bucket Lists. It's
2628 right around 900,000 keys. This process can continue nearly indefinitely --
2629 right up until the point the I<MD5> signatures start colliding with each other,
2630 and this is B<EXTREMELY> rare -- like winning the lottery 5 times in a row AND
2631 getting struck by lightning while you are walking to cash in your tickets.
2632 Theoretically, since I<MD5> hashes are 128-bit values, you I<could> have up to
2633 340,282,366,921,000,000,000,000,000,000,000,000,000 keys/elements (I believe
2634 this is 340 unodecillion, but don't quote me).
2638 When a new key/element is stored, the key (or index number) is first run through
2639 I<Digest::MD5> to get a 128-bit signature (example, in hex:
2640 b05783b0773d894396d475ced9d2f4f6). Then, the I<Master Index> record is checked
2641 for the first char of the signature (in this case I<b>). If it does not exist,
2642 a new I<Bucket List> is created for our key (and the next 15 future keys that
2643 happen to also have I<b> as their first MD5 char). The entire MD5 is written
2644 to the I<Bucket List> along with the offset of the new I<Bucket> record (EOF at
2645 this point, unless we are replacing an existing I<Bucket>), where the actual
2646 data will be stored.
2650 Fetching an existing key/element involves getting a I<Digest::MD5> of the key
2651 (or index number), then walking along the indexes. If there are enough
2652 keys/elements in this DB level, there might be nested indexes, each linked to
2653 a particular char of the MD5. Finally, a I<Bucket List> is pointed to, which
2654 contains up to 16 full MD5 hashes. Each is checked for equality to the key in
2655 question. If we found a match, the I<Bucket> tag is loaded, where the value and
2656 plain key are stored.
2658 Fetching the plain key occurs when calling the I<first_key()> and I<next_key()>
2659 methods. In this process the indexes are walked systematically, and each key
2660 fetched in increasing MD5 order (which is why it appears random). Once the
2661 I<Bucket> is found, the value is skipped the plain key returned instead.
2662 B<Note:> Do not count on keys being fetched as if the MD5 hashes were
2663 alphabetically sorted. This only happens on an index-level -- as soon as the
2664 I<Bucket Lists> are hit, the keys will come out in the order they went in --
2665 so it's pretty much undefined how the keys will come out -- just like Perl's
2668 =head1 CODE COVERAGE
2670 We use B<Devel::Cover> to test the code coverage of my tests, below is the
2671 B<Devel::Cover> report on this module's test suite.
2673 ---------------------------- ------ ------ ------ ------ ------ ------ ------
2674 File stmt bran cond sub pod time total
2675 ---------------------------- ------ ------ ------ ------ ------ ------ ------
2676 blib/lib/DBM/Deep.pm 93.7 82.5 71.9 96.5 25.9 82.8 87.9
2677 blib/lib/DBM/Deep/Array.pm 98.8 88.0 90.9 100.0 n/a 12.8 96.3
2678 blib/lib/DBM/Deep/Hash.pm 95.2 80.0 100.0 100.0 n/a 4.4 92.3
2679 Total 94.8 83.2 76.5 97.6 25.9 100.0 89.7
2680 ---------------------------- ------ ------ ------ ------ ------ ------ ------
2684 Joseph Huckaby, L<jhuckaby@cpan.org>
2685 Rob Kinyon, L<rkinyon@cpan.org>
2687 Special thanks to Adam Sah and Rich Gaushell! You know why :-)
2691 perltie(1), Tie::Hash(3), Digest::MD5(3), Fcntl(3), flock(2), lockf(3), nfs(5),
2692 Digest::SHA256(3), Crypt::Blowfish(3), Compress::Zlib(3)
2696 Copyright (c) 2002-2006 Joseph Huckaby. All Rights Reserved.
2697 This is free software, you may use it and distribute it under the
2698 same terms as Perl itself.