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 () { SIG_HASH }
104 sub TYPE_ARRAY () { SIG_ARRAY }
105 sub TYPE_SCALAR () { SIG_SCALAR }
111 if (scalar(@_) > 1) {
113 $proto->_throw_error( "Odd number of parameters to " . (caller(1))[2] );
117 elsif ( ref $_[0] ) {
118 unless ( eval { local $SIG{'__DIE__'}; %{$_[0]} || 1 } ) {
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 # locking implicitly enables autoflush
176 if ($args->{locking}) { $args->{autoflush} = 1; }
178 $self->{root} = exists $args->{root}
180 : DBM::Deep::_::Root->new( $args );
182 if (!defined($self->_fh)) { $self->_open(); }
189 require DBM::Deep::Hash;
190 return DBM::Deep::Hash->TIEHASH( @_ );
195 require DBM::Deep::Array;
196 return DBM::Deep::Array->TIEARRAY( @_ );
199 #XXX Unneeded now ...
205 # Open a fh to the database, create if nonexistent.
206 # Make sure file signature matches DBM::Deep spec.
208 my $self = $_[0]->_get_self;
210 if (defined($self->_fh)) { $self->_close(); }
213 local $SIG{'__DIE__'};
214 # Theoretically, adding O_BINARY should remove the need for the binmode
215 # Of course, testing it is going to be ... interesting.
216 my $flags = O_RDWR | O_CREAT | O_BINARY;
219 sysopen( $fh, $self->_root->{file}, $flags )
221 $self->_root->{fh} = $fh;
222 }; if ($@ ) { $self->_throw_error( "Received error: $@\n" ); }
223 if (! defined($self->_fh)) {
224 return $self->_throw_error("Cannot sysopen file: " . $self->_root->{file} . ": $!");
229 #XXX Can we remove this by using the right sysopen() flags?
230 # Maybe ... q.v. above
231 binmode $fh; # for win32
233 if ($self->_root->{autoflush}) {
234 my $old = select $fh;
239 seek($fh, 0 + $self->_root->{file_offset}, SEEK_SET);
242 my $bytes_read = read( $fh, $signature, length(SIG_FILE));
245 # File is empty -- write signature and master index
248 seek($fh, 0 + $self->_root->{file_offset}, SEEK_SET);
250 $self->_create_tag($self->_base_offset, $self->_type, chr(0) x $INDEX_SIZE);
252 my $plain_key = "[base]";
253 print($fh pack($DATA_LENGTH_PACK, length($plain_key)) . $plain_key );
255 # Flush the filehandle
256 my $old_fh = select $fh;
257 my $old_af = $|; $| = 1; $| = $old_af;
260 my @stats = stat($fh);
261 $self->_root->{inode} = $stats[1];
262 $self->_root->{end} = $stats[7];
268 # Check signature was valid
270 unless ($signature eq SIG_FILE) {
272 return $self->_throw_error("Signature not found -- file is not a Deep DB");
275 my @stats = stat($fh);
276 $self->_root->{inode} = $stats[1];
277 $self->_root->{end} = $stats[7];
280 # Get our type from master index signature
282 my $tag = $self->_load_tag($self->_base_offset);
284 #XXX We probably also want to store the hash algorithm name and not assume anything
285 #XXX The cool thing would be to allow a different hashing algorithm at every level
288 return $self->_throw_error("Corrupted file, no master index record");
290 if ($self->{type} ne $tag->{signature}) {
291 return $self->_throw_error("File type mismatch");
301 my $self = $_[0]->_get_self;
302 close $self->_root->{fh} if $self->_root->{fh};
303 $self->_root->{fh} = undef;
308 # Given offset, signature and content, create tag and write to disk
310 my ($self, $offset, $sig, $content) = @_;
311 my $size = length($content);
315 seek($fh, $offset + $self->_root->{file_offset}, SEEK_SET);
316 print($fh $sig . pack($DATA_LENGTH_PACK, $size) . $content );
318 if ($offset == $self->_root->{end}) {
319 $self->_root->{end} += SIG_SIZE + $DATA_LENGTH_SIZE + $size;
325 offset => $offset + SIG_SIZE + $DATA_LENGTH_SIZE,
332 # Given offset, load single tag and return signature, size and data
339 seek($fh, $offset + $self->_root->{file_offset}, SEEK_SET);
340 if (eof $fh) { return undef; }
343 read( $fh, $b, SIG_SIZE + $DATA_LENGTH_SIZE );
344 my ($sig, $size) = unpack( "A $DATA_LENGTH_PACK", $b );
347 read( $fh, $buffer, $size);
352 offset => $offset + SIG_SIZE + $DATA_LENGTH_SIZE,
359 # Given index tag, lookup single entry in index and return .
362 my ($tag, $index) = @_;
364 my $location = unpack($LONG_PACK, substr($tag->{content}, $index * $LONG_SIZE, $LONG_SIZE) );
365 if (!$location) { return; }
367 return $self->_load_tag( $location );
372 # Adds one key/value pair to bucket list, given offset, MD5 digest of key,
373 # plain (undigested) key and value.
376 my ($tag, $md5, $plain_key, $value) = @_;
377 my $keys = $tag->{content};
381 # added ref() check first to avoid eval and runtime exception for every
382 # scalar value being stored. performance tweak.
383 my $is_dbm_deep = eval { local $SIG{'__DIE__'}; $value->isa( 'DBM::Deep' ) };
385 my $internal_ref = $is_dbm_deep && ($value->_root eq $self->_root);
390 # Iterate through buckets, seeing if this is a new entry or a replace.
392 for (my $i=0; $i<$MAX_BUCKETS; $i++) {
393 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
394 my $subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
397 # Found empty bucket (end of list). Populate and exit loop.
401 $location = $internal_ref
402 ? $value->_base_offset
403 : $self->_root->{end};
405 seek($fh, $tag->{offset} + ($i * $BUCKET_SIZE) + $self->_root->{file_offset}, SEEK_SET);
406 print($fh $md5 . pack($LONG_PACK, $location) );
409 elsif ($md5 eq $key) {
411 # Found existing bucket with same key. Replace with new value.
416 $location = $value->_base_offset;
417 seek($fh, $tag->{offset} + ($i * $BUCKET_SIZE) + $self->_root->{file_offset}, SEEK_SET);
418 print($fh $md5 . pack($LONG_PACK, $location) );
421 seek($fh, $subloc + SIG_SIZE + $self->_root->{file_offset}, SEEK_SET);
423 read( $fh, $size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
426 # If value is a hash, array, or raw value with equal or less size, we can
427 # reuse the same content area of the database. Otherwise, we have to create
428 # a new content area at the EOF.
431 my $r = Scalar::Util::reftype( $value ) || '';
432 if ( $r eq 'HASH' || $r eq 'ARRAY' ) {
433 $actual_length = $INDEX_SIZE;
435 # if autobless is enabled, must also take into consideration
436 # the class name, as it is stored along with key/value.
437 if ( $self->_root->{autobless} ) {
438 my $value_class = Scalar::Util::blessed($value);
439 if ( defined $value_class && $value_class ne 'DBM::Deep' ) {
440 $actual_length += length($value_class);
444 else { $actual_length = length($value); }
446 if ($actual_length <= $size) {
450 $location = $self->_root->{end};
451 seek($fh, $tag->{offset} + ($i * $BUCKET_SIZE) + $HASH_SIZE + $self->_root->{file_offset}, SEEK_SET);
452 print($fh pack($LONG_PACK, $location) );
460 # If this is an internal reference, return now.
461 # No need to write value or plain key
468 # If bucket didn't fit into list, split into a new index level
471 seek($fh, $tag->{ref_loc} + $self->_root->{file_offset}, SEEK_SET);
472 print($fh pack($LONG_PACK, $self->_root->{end}) );
474 my $index_tag = $self->_create_tag($self->_root->{end}, SIG_INDEX, chr(0) x $INDEX_SIZE);
477 $keys .= $md5 . pack($LONG_PACK, 0);
479 for (my $i=0; $i<=$MAX_BUCKETS; $i++) {
480 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
482 my $old_subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
483 my $num = ord(substr($key, $tag->{ch} + 1, 1));
485 if ($offsets[$num]) {
486 my $offset = $offsets[$num] + SIG_SIZE + $DATA_LENGTH_SIZE;
487 seek($fh, $offset + $self->_root->{file_offset}, SEEK_SET);
489 read( $fh, $subkeys, $BUCKET_LIST_SIZE);
491 for (my $k=0; $k<$MAX_BUCKETS; $k++) {
492 my $subloc = unpack($LONG_PACK, substr($subkeys, ($k * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
494 seek($fh, $offset + ($k * $BUCKET_SIZE) + $self->_root->{file_offset}, SEEK_SET);
495 print($fh $key . pack($LONG_PACK, $old_subloc || $self->_root->{end}) );
501 $offsets[$num] = $self->_root->{end};
502 seek($fh, $index_tag->{offset} + ($num * $LONG_SIZE) + $self->_root->{file_offset}, SEEK_SET);
503 print($fh pack($LONG_PACK, $self->_root->{end}) );
505 my $blist_tag = $self->_create_tag($self->_root->{end}, SIG_BLIST, chr(0) x $BUCKET_LIST_SIZE);
507 seek($fh, $blist_tag->{offset} + $self->_root->{file_offset}, SEEK_SET);
508 print($fh $key . pack($LONG_PACK, $old_subloc || $self->_root->{end}) );
513 $location ||= $self->_root->{end};
514 } # re-index bucket list
517 # Seek to content area and store signature, value and plaintext key
521 seek($fh, $location + $self->_root->{file_offset}, SEEK_SET);
524 # Write signature based on content type, set content length and write actual value.
526 my $r = Scalar::Util::reftype($value) || '';
528 print($fh TYPE_HASH );
529 print($fh pack($DATA_LENGTH_PACK, $INDEX_SIZE) . chr(0) x $INDEX_SIZE );
530 $content_length = $INDEX_SIZE;
532 elsif ($r eq 'ARRAY') {
533 print($fh TYPE_ARRAY );
534 print($fh pack($DATA_LENGTH_PACK, $INDEX_SIZE) . chr(0) x $INDEX_SIZE );
535 $content_length = $INDEX_SIZE;
537 elsif (!defined($value)) {
538 print($fh SIG_NULL );
539 print($fh pack($DATA_LENGTH_PACK, 0) );
543 print($fh SIG_DATA );
544 print($fh pack($DATA_LENGTH_PACK, length($value)) . $value );
545 $content_length = length($value);
549 # Plain key is stored AFTER value, as keys are typically fetched less often.
551 print($fh pack($DATA_LENGTH_PACK, length($plain_key)) . $plain_key );
554 # If value is blessed, preserve class name
556 if ( $self->_root->{autobless} ) {
557 my $value_class = Scalar::Util::blessed($value);
558 if ( defined $value_class && $value_class ne 'DBM::Deep' ) {
560 # Blessed ref -- will restore later
563 print($fh pack($DATA_LENGTH_PACK, length($value_class)) . $value_class );
564 $content_length += 1;
565 $content_length += $DATA_LENGTH_SIZE + length($value_class);
569 $content_length += 1;
574 # If this is a new content area, advance EOF counter
576 if ($location == $self->_root->{end}) {
577 $self->_root->{end} += SIG_SIZE;
578 $self->_root->{end} += $DATA_LENGTH_SIZE + $content_length;
579 $self->_root->{end} += $DATA_LENGTH_SIZE + length($plain_key);
583 # If content is a hash or array, create new child DBM::Deep object and
584 # pass each key or element to it.
587 my $branch = DBM::Deep->new(
589 base_offset => $location,
590 root => $self->_root,
592 foreach my $key (keys %{$value}) {
593 $branch->STORE( $key, $value->{$key} );
596 elsif ($r eq 'ARRAY') {
597 my $branch = DBM::Deep->new(
599 base_offset => $location,
600 root => $self->_root,
603 foreach my $element (@{$value}) {
604 $branch->STORE( $index, $element );
612 return $self->_throw_error("Fatal error: indexing failed -- possibly due to corruption in file");
615 sub _get_bucket_value {
617 # Fetch single value given tag and MD5 digested key.
620 my ($tag, $md5) = @_;
621 my $keys = $tag->{content};
626 # Iterate through buckets, looking for a key match
629 for (my $i=0; $i<$MAX_BUCKETS; $i++) {
630 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
631 my $subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
635 # Hit end of list, no match
640 if ( $md5 ne $key ) {
645 # Found match -- seek to offset and read signature
648 seek($fh, $subloc + $self->_root->{file_offset}, SEEK_SET);
649 read( $fh, $signature, SIG_SIZE);
652 # If value is a hash or array, return new DBM::Deep object with correct offset
654 if (($signature eq TYPE_HASH) || ($signature eq TYPE_ARRAY)) {
655 my $obj = DBM::Deep->new(
657 base_offset => $subloc,
661 if ($self->_root->{autobless}) {
663 # Skip over value and plain key to see if object needs
666 seek($fh, $DATA_LENGTH_SIZE + $INDEX_SIZE, SEEK_CUR);
669 read( $fh, $size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
670 if ($size) { seek($fh, $size, SEEK_CUR); }
673 read( $fh, $bless_bit, 1);
674 if (ord($bless_bit)) {
676 # Yes, object needs to be re-blessed
679 read( $fh, $size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
680 if ($size) { read( $fh, $class_name, $size); }
681 if ($class_name) { $obj = bless( $obj, $class_name ); }
689 # Otherwise return actual value
691 elsif ($signature eq SIG_DATA) {
694 read( $fh, $size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
695 if ($size) { read( $fh, $value, $size); }
700 # Key exists, but content is null
710 # Delete single key/value pair given tag and MD5 digested key.
713 my ($tag, $md5) = @_;
714 my $keys = $tag->{content};
719 # Iterate through buckets, looking for a key match
722 for (my $i=0; $i<$MAX_BUCKETS; $i++) {
723 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
724 my $subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
728 # Hit end of list, no match
733 if ( $md5 ne $key ) {
738 # Matched key -- delete bucket and return
740 seek($fh, $tag->{offset} + ($i * $BUCKET_SIZE) + $self->_root->{file_offset}, SEEK_SET);
741 print($fh substr($keys, ($i+1) * $BUCKET_SIZE ) );
742 print($fh chr(0) x $BUCKET_SIZE );
752 # Check existence of single key given tag and MD5 digested key.
755 my ($tag, $md5) = @_;
756 my $keys = $tag->{content};
759 # Iterate through buckets, looking for a key match
762 for (my $i=0; $i<$MAX_BUCKETS; $i++) {
763 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
764 my $subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
768 # Hit end of list, no match
773 if ( $md5 ne $key ) {
778 # Matched key -- return true
786 sub _find_bucket_list {
788 # Locate offset for bucket list, given digested key
794 # Locate offset for bucket list using digest index system
797 my $tag = $self->_load_tag($self->_base_offset);
798 if (!$tag) { return; }
800 while ($tag->{signature} ne SIG_BLIST) {
801 $tag = $self->_index_lookup($tag, ord(substr($md5, $ch, 1)));
802 if (!$tag) { return; }
809 sub _traverse_index {
811 # Scan index and recursively step into deeper levels, looking for next key.
813 my ($self, $offset, $ch, $force_return_next) = @_;
814 $force_return_next = undef unless $force_return_next;
816 my $tag = $self->_load_tag( $offset );
820 if ($tag->{signature} ne SIG_BLIST) {
821 my $content = $tag->{content};
823 if ($self->{return_next}) { $start = 0; }
824 else { $start = ord(substr($self->{prev_md5}, $ch, 1)); }
826 for (my $index = $start; $index < 256; $index++) {
827 my $subloc = unpack($LONG_PACK, substr($content, $index * $LONG_SIZE, $LONG_SIZE) );
829 my $result = $self->_traverse_index( $subloc, $ch + 1, $force_return_next );
830 if (defined($result)) { return $result; }
834 $self->{return_next} = 1;
837 elsif ($tag->{signature} eq SIG_BLIST) {
838 my $keys = $tag->{content};
839 if ($force_return_next) { $self->{return_next} = 1; }
842 # Iterate through buckets, looking for a key match
844 for (my $i=0; $i<$MAX_BUCKETS; $i++) {
845 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
846 my $subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
850 # End of bucket list -- return to outer loop
852 $self->{return_next} = 1;
855 elsif ($key eq $self->{prev_md5}) {
857 # Located previous key -- return next one found
859 $self->{return_next} = 1;
862 elsif ($self->{return_next}) {
864 # Seek to bucket location and skip over signature
866 seek($fh, $subloc + SIG_SIZE + $self->_root->{file_offset}, SEEK_SET);
869 # Skip over value to get to plain key
872 read( $fh, $size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
873 if ($size) { seek($fh, $size, SEEK_CUR); }
876 # Read in plain key and return as scalar
879 read( $fh, $size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
880 if ($size) { read( $fh, $plain_key, $size); }
886 $self->{return_next} = 1;
887 } # tag is a bucket list
894 # Locate next key, given digested previous one
896 my $self = $_[0]->_get_self;
898 $self->{prev_md5} = $_[1] ? $_[1] : undef;
899 $self->{return_next} = 0;
902 # If the previous key was not specifed, start at the top and
903 # return the first one found.
905 if (!$self->{prev_md5}) {
906 $self->{prev_md5} = chr(0) x $HASH_SIZE;
907 $self->{return_next} = 1;
910 return $self->_traverse_index( $self->_base_offset, 0 );
915 # If db locking is set, flock() the db file. If called multiple
916 # times before unlock(), then the same number of unlocks() must
917 # be called before the lock is released.
919 my $self = $_[0]->_get_self;
921 $type = LOCK_EX unless defined $type;
923 if (!defined($self->_fh)) { return; }
925 if ($self->_root->{locking}) {
926 if (!$self->_root->{locked}) {
927 flock($self->_fh, $type);
929 # refresh end counter in case file has changed size
930 my @stats = stat($self->_root->{file});
931 $self->_root->{end} = $stats[7];
933 # double-check file inode, in case another process
934 # has optimize()d our file while we were waiting.
935 if ($stats[1] != $self->_root->{inode}) {
936 $self->_open(); # re-open
937 flock($self->_fh, $type); # re-lock
938 $self->_root->{end} = (stat($self->_fh))[7]; # re-end
941 $self->_root->{locked}++;
951 # If db locking is set, unlock the db file. See note in lock()
952 # regarding calling lock() multiple times.
954 my $self = $_[0]->_get_self;
956 if (!defined($self->_fh)) { return; }
958 if ($self->_root->{locking} && $self->_root->{locked} > 0) {
959 $self->_root->{locked}--;
960 if (!$self->_root->{locked}) { flock($self->_fh, LOCK_UN); }
969 my $self = shift->_get_self;
970 my ($spot, $value) = @_;
975 elsif ( eval { local $SIG{__DIE__}; $value->isa( 'DBM::Deep' ) } ) {
976 my $type = $value->_type;
977 ${$spot} = $type eq TYPE_HASH ? {} : [];
978 $value->_copy_node( ${$spot} );
981 my $r = Scalar::Util::reftype( $value );
982 my $c = Scalar::Util::blessed( $value );
983 if ( $r eq 'ARRAY' ) {
984 ${$spot} = [ @{$value} ];
987 ${$spot} = { %{$value} };
989 ${$spot} = bless ${$spot}, $c
998 # Copy single level of keys or elements to new DB handle.
999 # Recurse for nested structures
1001 my $self = shift->_get_self;
1004 if ($self->_type eq TYPE_HASH) {
1005 my $key = $self->first_key();
1007 my $value = $self->get($key);
1008 $self->_copy_value( \$db_temp->{$key}, $value );
1009 $key = $self->next_key($key);
1013 my $length = $self->length();
1014 for (my $index = 0; $index < $length; $index++) {
1015 my $value = $self->get($index);
1016 $self->_copy_value( \$db_temp->[$index], $value );
1025 # Recursively export into standard Perl hashes and arrays.
1027 my $self = $_[0]->_get_self;
1030 if ($self->_type eq TYPE_HASH) { $temp = {}; }
1031 elsif ($self->_type eq TYPE_ARRAY) { $temp = []; }
1034 $self->_copy_node( $temp );
1042 # Recursively import Perl hash/array structure
1044 #XXX This use of ref() seems to be ok
1045 if (!ref($_[0])) { return; } # Perl calls import() on use -- ignore
1047 my $self = $_[0]->_get_self;
1050 #XXX This use of ref() seems to be ok
1051 if (!ref($struct)) {
1053 # struct is not a reference, so just import based on our type
1057 if ($self->_type eq TYPE_HASH) { $struct = {@_}; }
1058 elsif ($self->_type eq TYPE_ARRAY) { $struct = [@_]; }
1061 my $r = Scalar::Util::reftype($struct) || '';
1062 if ($r eq "HASH" && $self->_type eq TYPE_HASH) {
1063 foreach my $key (keys %$struct) { $self->put($key, $struct->{$key}); }
1065 elsif ($r eq "ARRAY" && $self->_type eq TYPE_ARRAY) {
1066 $self->push( @$struct );
1069 return $self->_throw_error("Cannot import: type mismatch");
1077 # Rebuild entire database into new file, then move
1078 # it back on top of original.
1080 my $self = $_[0]->_get_self;
1082 #XXX Need to create a new test for this
1083 # if ($self->_root->{links} > 1) {
1084 # return $self->_throw_error("Cannot optimize: reference count is greater than 1");
1087 my $db_temp = DBM::Deep->new(
1088 file => $self->_root->{file} . '.tmp',
1089 type => $self->_type
1092 return $self->_throw_error("Cannot optimize: failed to open temp file: $!");
1096 $self->_copy_node( $db_temp );
1100 # Attempt to copy user, group and permissions over to new file
1102 my @stats = stat($self->_fh);
1103 my $perms = $stats[2] & 07777;
1104 my $uid = $stats[4];
1105 my $gid = $stats[5];
1106 chown( $uid, $gid, $self->_root->{file} . '.tmp' );
1107 chmod( $perms, $self->_root->{file} . '.tmp' );
1109 # q.v. perlport for more information on this variable
1110 if ( $^O eq 'MSWin32' || $^O eq 'cygwin' ) {
1112 # Potential race condition when optmizing on Win32 with locking.
1113 # The Windows filesystem requires that the filehandle be closed
1114 # before it is overwritten with rename(). This could be redone
1121 if (!rename $self->_root->{file} . '.tmp', $self->_root->{file}) {
1122 unlink $self->_root->{file} . '.tmp';
1124 return $self->_throw_error("Optimize failed: Cannot copy temp file over original: $!");
1136 # Make copy of object and return
1138 my $self = $_[0]->_get_self;
1140 return DBM::Deep->new(
1141 type => $self->_type,
1142 base_offset => $self->_base_offset,
1143 root => $self->_root
1148 my %is_legal_filter = map {
1151 store_key store_value
1152 fetch_key fetch_value
1157 # Setup filter function for storing or fetching the key or value
1159 my $self = $_[0]->_get_self;
1160 my $type = lc $_[1];
1161 my $func = $_[2] ? $_[2] : undef;
1163 if ( $is_legal_filter{$type} ) {
1164 $self->_root->{"filter_$type"} = $func;
1178 # Get access to the root structure
1180 my $self = $_[0]->_get_self;
1181 return $self->{root};
1186 # Get access to the raw fh
1188 #XXX It will be useful, though, when we split out HASH and ARRAY
1189 my $self = $_[0]->_get_self;
1190 return $self->_root->{fh};
1195 # Get type of current node (TYPE_HASH or TYPE_ARRAY)
1197 my $self = $_[0]->_get_self;
1198 return $self->{type};
1203 # Get base_offset of current node (TYPE_HASH or TYPE_ARRAY)
1205 my $self = $_[0]->_get_self;
1206 return $self->{base_offset};
1211 # Get last error string, or undef if no error
1214 #? ( _get_self($_[0])->{root}->{error} or undef )
1215 ? ( $_[0]->_get_self->{root}->{error} or undef )
1225 # Store error string in self
1227 my $error_text = $_[1];
1229 if ( Scalar::Util::blessed $_[0] ) {
1230 my $self = $_[0]->_get_self;
1231 $self->_root->{error} = $error_text;
1233 unless ($self->_root->{debug}) {
1234 die "DBM::Deep: $error_text\n";
1237 warn "DBM::Deep: $error_text\n";
1241 die "DBM::Deep: $error_text\n";
1249 my $self = $_[0]->_get_self;
1251 undef $self->_root->{error};
1254 sub _precalc_sizes {
1256 # Precalculate index, bucket and bucket list sizes
1259 #XXX I don't like this ...
1260 set_pack() unless defined $LONG_SIZE;
1262 $INDEX_SIZE = 256 * $LONG_SIZE;
1263 $BUCKET_SIZE = $HASH_SIZE + $LONG_SIZE;
1264 $BUCKET_LIST_SIZE = $MAX_BUCKETS * $BUCKET_SIZE;
1269 # Set pack/unpack modes (see file header for more)
1271 my ($long_s, $long_p, $data_s, $data_p) = @_;
1273 $LONG_SIZE = $long_s ? $long_s : 4;
1274 $LONG_PACK = $long_p ? $long_p : 'N';
1276 $DATA_LENGTH_SIZE = $data_s ? $data_s : 4;
1277 $DATA_LENGTH_PACK = $data_p ? $data_p : 'N';
1284 # Set key digest function (default is MD5)
1286 my ($digest_func, $hash_size) = @_;
1288 $DIGEST_FUNC = $digest_func ? $digest_func : \&Digest::MD5::md5;
1289 $HASH_SIZE = $hash_size ? $hash_size : 16;
1295 # tie() methods (hashes and arrays)
1300 # Store single hash key/value or array element in database.
1302 my $self = $_[0]->_get_self;
1305 # User may be storing a hash, in which case we do not want it run
1306 # through the filtering system
1307 my $value = ($self->_root->{filter_store_value} && !ref($_[2]))
1308 ? $self->_root->{filter_store_value}->($_[2])
1311 my $md5 = $DIGEST_FUNC->($key);
1314 # Make sure file is open
1316 if (!defined($self->_fh) && !$self->_open()) {
1322 # Request exclusive lock for writing
1324 $self->lock( LOCK_EX );
1326 my $fh = $self->_fh;
1329 # Locate offset for bucket list using digest index system
1331 my $tag = $self->_load_tag($self->_base_offset);
1333 $tag = $self->_create_tag($self->_base_offset, SIG_INDEX, chr(0) x $INDEX_SIZE);
1337 while ($tag->{signature} ne SIG_BLIST) {
1338 my $num = ord(substr($md5, $ch, 1));
1339 my $new_tag = $self->_index_lookup($tag, $num);
1341 my $ref_loc = $tag->{offset} + ($num * $LONG_SIZE);
1342 seek($fh, $ref_loc + $self->_root->{file_offset}, SEEK_SET);
1343 print($fh pack($LONG_PACK, $self->_root->{end}) );
1345 $tag = $self->_create_tag($self->_root->{end}, SIG_BLIST, chr(0) x $BUCKET_LIST_SIZE);
1346 $tag->{ref_loc} = $ref_loc;
1351 my $ref_loc = $tag->{offset} + ($num * $LONG_SIZE);
1353 $tag->{ref_loc} = $ref_loc;
1360 # Add key/value to bucket list
1362 my $result = $self->_add_bucket( $tag, $md5, $key, $value );
1371 # Fetch single value or element given plain key or array index
1373 my $self = shift->_get_self;
1377 # Make sure file is open
1379 if (!defined($self->_fh)) { $self->_open(); }
1381 my $md5 = $DIGEST_FUNC->($key);
1384 # Request shared lock for reading
1386 $self->lock( LOCK_SH );
1388 my $tag = $self->_find_bucket_list( $md5 );
1395 # Get value from bucket list
1397 my $result = $self->_get_bucket_value( $tag, $md5 );
1401 #XXX What is ref() checking here?
1402 #YYY Filters only apply on scalar values, so the ref check is making
1403 #YYY sure the fetched bucket is a scalar, not a child hash or array.
1404 return ($result && !ref($result) && $self->_root->{filter_fetch_value})
1405 ? $self->_root->{filter_fetch_value}->($result)
1411 # Delete single key/value pair or element given plain key or array index
1413 my $self = $_[0]->_get_self;
1416 my $md5 = $DIGEST_FUNC->($key);
1419 # Make sure file is open
1421 if (!defined($self->_fh)) { $self->_open(); }
1424 # Request exclusive lock for writing
1426 $self->lock( LOCK_EX );
1428 my $tag = $self->_find_bucket_list( $md5 );
1437 my $value = $self->_get_bucket_value( $tag, $md5 );
1438 if ($value && !ref($value) && $self->_root->{filter_fetch_value}) {
1439 $value = $self->_root->{filter_fetch_value}->($value);
1442 my $result = $self->_delete_bucket( $tag, $md5 );
1445 # If this object is an array and the key deleted was on the end of the stack,
1446 # decrement the length variable.
1456 # Check if a single key or element exists given plain key or array index
1458 my $self = $_[0]->_get_self;
1461 my $md5 = $DIGEST_FUNC->($key);
1464 # Make sure file is open
1466 if (!defined($self->_fh)) { $self->_open(); }
1469 # Request shared lock for reading
1471 $self->lock( LOCK_SH );
1473 my $tag = $self->_find_bucket_list( $md5 );
1476 # For some reason, the built-in exists() function returns '' for false
1484 # Check if bucket exists and return 1 or ''
1486 my $result = $self->_bucket_exists( $tag, $md5 ) || '';
1495 # Clear all keys from hash, or all elements from array.
1497 my $self = $_[0]->_get_self;
1500 # Make sure file is open
1502 if (!defined($self->_fh)) { $self->_open(); }
1505 # Request exclusive lock for writing
1507 $self->lock( LOCK_EX );
1509 my $fh = $self->_fh;
1511 seek($fh, $self->_base_offset + $self->_root->{file_offset}, SEEK_SET);
1517 $self->_create_tag($self->_base_offset, $self->_type, chr(0) x $INDEX_SIZE);
1525 # Public method aliases
1527 sub put { (shift)->STORE( @_ ) }
1528 sub store { (shift)->STORE( @_ ) }
1529 sub get { (shift)->FETCH( @_ ) }
1530 sub fetch { (shift)->FETCH( @_ ) }
1531 sub delete { (shift)->DELETE( @_ ) }
1532 sub exists { (shift)->EXISTS( @_ ) }
1533 sub clear { (shift)->CLEAR( @_ ) }
1535 package DBM::Deep::_::Root;
1549 filter_store_key => undef,
1550 filter_store_value => undef,
1551 filter_fetch_key => undef,
1552 filter_fetch_value => undef,
1558 if ( $self->{fh} && !$self->{file_offset} ) {
1559 $self->{file_offset} = tell( $self->{fh} );
1567 return unless $self;
1569 close $self->{fh} if $self->{fh};
1580 DBM::Deep - A pure perl multi-level hash/array DBM
1585 my $db = DBM::Deep->new( "foo.db" );
1587 $db->{key} = 'value'; # tie() style
1590 $db->put('key' => 'value'); # OO style
1591 print $db->get('key');
1593 # true multi-level support
1594 $db->{my_complex} = [
1595 'hello', { perl => 'rules' },
1601 A unique flat-file database module, written in pure perl. True
1602 multi-level hash/array support (unlike MLDBM, which is faked), hybrid
1603 OO / tie() interface, cross-platform FTPable files, and quite fast. Can
1604 handle millions of keys and unlimited hash levels without significant
1605 slow-down. Written from the ground-up in pure perl -- this is NOT a
1606 wrapper around a C-based DBM. Out-of-the-box compatibility with Unix,
1607 Mac OS X and Windows.
1611 Hopefully you are using Perl's excellent CPAN module, which will download
1612 and install the module for you. If not, get the tarball, and run these
1624 Construction can be done OO-style (which is the recommended way), or using
1625 Perl's tie() function. Both are examined here.
1627 =head2 OO CONSTRUCTION
1629 The recommended way to construct a DBM::Deep object is to use the new()
1630 method, which gets you a blessed, tied hash or array reference.
1632 my $db = DBM::Deep->new( "foo.db" );
1634 This opens a new database handle, mapped to the file "foo.db". If this
1635 file does not exist, it will automatically be created. DB files are
1636 opened in "r+" (read/write) mode, and the type of object returned is a
1637 hash, unless otherwise specified (see L<OPTIONS> below).
1639 You can pass a number of options to the constructor to specify things like
1640 locking, autoflush, etc. This is done by passing an inline hash:
1642 my $db = DBM::Deep->new(
1648 Notice that the filename is now specified I<inside> the hash with
1649 the "file" parameter, as opposed to being the sole argument to the
1650 constructor. This is required if any options are specified.
1651 See L<OPTIONS> below for the complete list.
1655 You can also start with an array instead of a hash. For this, you must
1656 specify the C<type> parameter:
1658 my $db = DBM::Deep->new(
1660 type => DBM::Deep->TYPE_ARRAY
1663 B<Note:> Specifing the C<type> parameter only takes effect when beginning
1664 a new DB file. If you create a DBM::Deep object with an existing file, the
1665 C<type> will be loaded from the file header, and an error will be thrown if
1666 the wrong type is passed in.
1668 =head2 TIE CONSTRUCTION
1670 Alternately, you can create a DBM::Deep handle by using Perl's built-in
1671 tie() function. The object returned from tie() can be used to call methods,
1672 such as lock() and unlock(), but cannot be used to assign to the DBM::Deep
1673 file (as expected with most tie'd objects).
1676 my $db = tie %hash, "DBM::Deep", "foo.db";
1679 my $db = tie @array, "DBM::Deep", "bar.db";
1681 As with the OO constructor, you can replace the DB filename parameter with
1682 a hash containing one or more options (see L<OPTIONS> just below for the
1685 tie %hash, "DBM::Deep", {
1693 There are a number of options that can be passed in when constructing your
1694 DBM::Deep objects. These apply to both the OO- and tie- based approaches.
1700 Filename of the DB file to link the handle to. You can pass a full absolute
1701 filesystem path, partial path, or a plain filename if the file is in the
1702 current working directory. This is a required parameter (though q.v. fh).
1706 If you want, you can pass in the fh instead of the file. This is most useful for doing
1709 my $db = DBM::Deep->new( { fh => \*DATA } );
1711 You are responsible for making sure that the fh has been opened appropriately for your
1712 needs. If you open it read-only and attempt to write, an exception will be thrown. If you
1713 open it write-only or append-only, an exception will be thrown immediately as DBM::Deep
1714 needs to read from the fh.
1718 This is the offset within the file that the DBM::Deep db starts. Most of the time, you will
1719 not need to set this. However, it's there if you want it.
1721 If you pass in fh and do not set this, it will be set appropriately.
1725 This parameter specifies what type of object to create, a hash or array. Use
1726 one of these two constants: C<DBM::Deep-E<gt>TYPE_HASH> or C<DBM::Deep-E<gt>TYPE_ARRAY>.
1727 This only takes effect when beginning a new file. This is an optional
1728 parameter, and defaults to C<DBM::Deep-E<gt>TYPE_HASH>.
1732 Specifies whether locking is to be enabled. DBM::Deep uses Perl's Fnctl flock()
1733 function to lock the database in exclusive mode for writes, and shared mode for
1734 reads. Pass any true value to enable. This affects the base DB handle I<and
1735 any child hashes or arrays> that use the same DB file. This is an optional
1736 parameter, and defaults to 0 (disabled). See L<LOCKING> below for more.
1740 Specifies whether autoflush is to be enabled on the underlying filehandle.
1741 This obviously slows down write operations, but is required if you may have
1742 multiple processes accessing the same DB file (also consider enable I<locking>).
1743 Pass any true value to enable. This is an optional parameter, and defaults to 0
1748 If I<autobless> mode is enabled, DBM::Deep will preserve blessed hashes, and
1749 restore them when fetched. This is an B<experimental> feature, and does have
1750 side-effects. Basically, when hashes are re-blessed into their original
1751 classes, they are no longer blessed into the DBM::Deep class! So you won't be
1752 able to call any DBM::Deep methods on them. You have been warned.
1753 This is an optional parameter, and defaults to 0 (disabled).
1757 See L<FILTERS> below.
1761 Setting I<debug> mode will make all errors non-fatal, dump them out to
1762 STDERR, and continue on. This is for debugging purposes only, and probably
1763 not what you want. This is an optional parameter, and defaults to 0 (disabled).
1767 Instead of passing a file path, you can instead pass a handle to an pre-opened
1768 filehandle. Note: Beware of using the magick *DATA handle, as this actually
1769 contains your entire Perl script, as well as the data following the __DATA__
1770 marker. This will not work, because DBM::Deep uses absolute seek()s into the
1771 file. Instead, consider reading *DATA into an IO::Scalar handle, then passing
1772 in that. Also please note optimize() will NOT work when passing in only a
1773 handle. Pass in a real filename in order to use optimize().
1777 =head1 TIE INTERFACE
1779 With DBM::Deep you can access your databases using Perl's standard hash/array
1780 syntax. Because all DBM::Deep objects are I<tied> to hashes or arrays, you can
1781 treat them as such. DBM::Deep will intercept all reads/writes and direct them
1782 to the right place -- the DB file. This has nothing to do with the
1783 L<TIE CONSTRUCTION> section above. This simply tells you how to use DBM::Deep
1784 using regular hashes and arrays, rather than calling functions like C<get()>
1785 and C<put()> (although those work too). It is entirely up to you how to want
1786 to access your databases.
1790 You can treat any DBM::Deep object like a normal Perl hash reference. Add keys,
1791 or even nested hashes (or arrays) using standard Perl syntax:
1793 my $db = DBM::Deep->new( "foo.db" );
1795 $db->{mykey} = "myvalue";
1797 $db->{myhash}->{subkey} = "subvalue";
1799 print $db->{myhash}->{subkey} . "\n";
1801 You can even step through hash keys using the normal Perl C<keys()> function:
1803 foreach my $key (keys %$db) {
1804 print "$key: " . $db->{$key} . "\n";
1807 Remember that Perl's C<keys()> function extracts I<every> key from the hash and
1808 pushes them onto an array, all before the loop even begins. If you have an
1809 extra large hash, this may exhaust Perl's memory. Instead, consider using
1810 Perl's C<each()> function, which pulls keys/values one at a time, using very
1813 while (my ($key, $value) = each %$db) {
1814 print "$key: $value\n";
1817 Please note that when using C<each()>, you should always pass a direct
1818 hash reference, not a lookup. Meaning, you should B<never> do this:
1821 while (my ($key, $value) = each %{$db->{foo}}) { # BAD
1823 This causes an infinite loop, because for each iteration, Perl is calling
1824 FETCH() on the $db handle, resulting in a "new" hash for foo every time, so
1825 it effectively keeps returning the first key over and over again. Instead,
1826 assign a temporary variable to C<$db->{foo}>, then pass that to each().
1830 As with hashes, you can treat any DBM::Deep object like a normal Perl array
1831 reference. This includes inserting, removing and manipulating elements,
1832 and the C<push()>, C<pop()>, C<shift()>, C<unshift()> and C<splice()> functions.
1833 The object must have first been created using type C<DBM::Deep-E<gt>TYPE_ARRAY>,
1834 or simply be a nested array reference inside a hash. Example:
1836 my $db = DBM::Deep->new(
1837 file => "foo-array.db",
1838 type => DBM::Deep->TYPE_ARRAY
1842 push @$db, "bar", "baz";
1843 unshift @$db, "bah";
1845 my $last_elem = pop @$db; # baz
1846 my $first_elem = shift @$db; # bah
1847 my $second_elem = $db->[1]; # bar
1849 my $num_elements = scalar @$db;
1853 In addition to the I<tie()> interface, you can also use a standard OO interface
1854 to manipulate all aspects of DBM::Deep databases. Each type of object (hash or
1855 array) has its own methods, but both types share the following common methods:
1856 C<put()>, C<get()>, C<exists()>, C<delete()> and C<clear()>.
1860 =item * new() / clone()
1862 These are the constructor and copy-functions.
1864 =item * put() / store()
1866 Stores a new hash key/value pair, or sets an array element value. Takes two
1867 arguments, the hash key or array index, and the new value. The value can be
1868 a scalar, hash ref or array ref. Returns true on success, false on failure.
1870 $db->put("foo", "bar"); # for hashes
1871 $db->put(1, "bar"); # for arrays
1873 =item * get() / fetch()
1875 Fetches the value of a hash key or array element. Takes one argument: the hash
1876 key or array index. Returns a scalar, hash ref or array ref, depending on the
1879 my $value = $db->get("foo"); # for hashes
1880 my $value = $db->get(1); # for arrays
1884 Checks if a hash key or array index exists. Takes one argument: the hash key
1885 or array index. Returns true if it exists, false if not.
1887 if ($db->exists("foo")) { print "yay!\n"; } # for hashes
1888 if ($db->exists(1)) { print "yay!\n"; } # for arrays
1892 Deletes one hash key/value pair or array element. Takes one argument: the hash
1893 key or array index. Returns true on success, false if not found. For arrays,
1894 the remaining elements located after the deleted element are NOT moved over.
1895 The deleted element is essentially just undefined, which is exactly how Perl's
1896 internal arrays work. Please note that the space occupied by the deleted
1897 key/value or element is B<not> reused again -- see L<UNUSED SPACE RECOVERY>
1898 below for details and workarounds.
1900 $db->delete("foo"); # for hashes
1901 $db->delete(1); # for arrays
1905 Deletes B<all> hash keys or array elements. Takes no arguments. No return
1906 value. Please note that the space occupied by the deleted keys/values or
1907 elements is B<not> reused again -- see L<UNUSED SPACE RECOVERY> below for
1908 details and workarounds.
1910 $db->clear(); # hashes or arrays
1912 =item * lock() / unlock()
1918 Recover lost disk space.
1920 =item * import() / export()
1922 Data going in and out.
1924 =item * set_digest() / set_pack() / set_filter()
1926 q.v. adjusting the interal parameters.
1928 =item * error() / clear_error()
1930 Error handling methods (may be deprecated).
1936 For hashes, DBM::Deep supports all the common methods described above, and the
1937 following additional methods: C<first_key()> and C<next_key()>.
1943 Returns the "first" key in the hash. As with built-in Perl hashes, keys are
1944 fetched in an undefined order (which appears random). Takes no arguments,
1945 returns the key as a scalar value.
1947 my $key = $db->first_key();
1951 Returns the "next" key in the hash, given the previous one as the sole argument.
1952 Returns undef if there are no more keys to be fetched.
1954 $key = $db->next_key($key);
1958 Here are some examples of using hashes:
1960 my $db = DBM::Deep->new( "foo.db" );
1962 $db->put("foo", "bar");
1963 print "foo: " . $db->get("foo") . "\n";
1965 $db->put("baz", {}); # new child hash ref
1966 $db->get("baz")->put("buz", "biz");
1967 print "buz: " . $db->get("baz")->get("buz") . "\n";
1969 my $key = $db->first_key();
1971 print "$key: " . $db->get($key) . "\n";
1972 $key = $db->next_key($key);
1975 if ($db->exists("foo")) { $db->delete("foo"); }
1979 For arrays, DBM::Deep supports all the common methods described above, and the
1980 following additional methods: C<length()>, C<push()>, C<pop()>, C<shift()>,
1981 C<unshift()> and C<splice()>.
1987 Returns the number of elements in the array. Takes no arguments.
1989 my $len = $db->length();
1993 Adds one or more elements onto the end of the array. Accepts scalars, hash
1994 refs or array refs. No return value.
1996 $db->push("foo", "bar", {});
2000 Fetches the last element in the array, and deletes it. Takes no arguments.
2001 Returns undef if array is empty. Returns the element value.
2003 my $elem = $db->pop();
2007 Fetches the first element in the array, deletes it, then shifts all the
2008 remaining elements over to take up the space. Returns the element value. This
2009 method is not recommended with large arrays -- see L<LARGE ARRAYS> below for
2012 my $elem = $db->shift();
2016 Inserts one or more elements onto the beginning of the array, shifting all
2017 existing elements over to make room. Accepts scalars, hash refs or array refs.
2018 No return value. This method is not recommended with large arrays -- see
2019 <LARGE ARRAYS> below for details.
2021 $db->unshift("foo", "bar", {});
2025 Performs exactly like Perl's built-in function of the same name. See L<perldoc
2026 -f splice> for usage -- it is too complicated to document here. This method is
2027 not recommended with large arrays -- see L<LARGE ARRAYS> below for details.
2031 Here are some examples of using arrays:
2033 my $db = DBM::Deep->new(
2035 type => DBM::Deep->TYPE_ARRAY
2038 $db->push("bar", "baz");
2039 $db->unshift("foo");
2042 my $len = $db->length();
2043 print "length: $len\n"; # 4
2045 for (my $k=0; $k<$len; $k++) {
2046 print "$k: " . $db->get($k) . "\n";
2049 $db->splice(1, 2, "biz", "baf");
2051 while (my $elem = shift @$db) {
2052 print "shifted: $elem\n";
2057 Enable automatic file locking by passing a true value to the C<locking>
2058 parameter when constructing your DBM::Deep object (see L<SETUP> above).
2060 my $db = DBM::Deep->new(
2065 This causes DBM::Deep to C<flock()> the underlying filehandle with exclusive
2066 mode for writes, and shared mode for reads. This is required if you have
2067 multiple processes accessing the same database file, to avoid file corruption.
2068 Please note that C<flock()> does NOT work for files over NFS. See L<DB OVER
2069 NFS> below for more.
2071 =head2 EXPLICIT LOCKING
2073 You can explicitly lock a database, so it remains locked for multiple
2074 transactions. This is done by calling the C<lock()> method, and passing an
2075 optional lock mode argument (defaults to exclusive mode). This is particularly
2076 useful for things like counters, where the current value needs to be fetched,
2077 then incremented, then stored again.
2080 my $counter = $db->get("counter");
2082 $db->put("counter", $counter);
2091 You can pass C<lock()> an optional argument, which specifies which mode to use
2092 (exclusive or shared). Use one of these two constants: C<DBM::Deep-E<gt>LOCK_EX>
2093 or C<DBM::Deep-E<gt>LOCK_SH>. These are passed directly to C<flock()>, and are the
2094 same as the constants defined in Perl's C<Fcntl> module.
2096 $db->lock( DBM::Deep->LOCK_SH );
2100 =head1 IMPORTING/EXPORTING
2102 You can import existing complex structures by calling the C<import()> method,
2103 and export an entire database into an in-memory structure using the C<export()>
2104 method. Both are examined here.
2108 Say you have an existing hash with nested hashes/arrays inside it. Instead of
2109 walking the structure and adding keys/elements to the database as you go,
2110 simply pass a reference to the C<import()> method. This recursively adds
2111 everything to an existing DBM::Deep object for you. Here is an example:
2116 array1 => [ "elem0", "elem1", "elem2" ],
2118 subkey1 => "subvalue1",
2119 subkey2 => "subvalue2"
2123 my $db = DBM::Deep->new( "foo.db" );
2124 $db->import( $struct );
2126 print $db->{key1} . "\n"; # prints "value1"
2128 This recursively imports the entire C<$struct> object into C<$db>, including
2129 all nested hashes and arrays. If the DBM::Deep object contains exsiting data,
2130 keys are merged with the existing ones, replacing if they already exist.
2131 The C<import()> method can be called on any database level (not just the base
2132 level), and works with both hash and array DB types.
2134 B<Note:> Make sure your existing structure has no circular references in it.
2135 These will cause an infinite loop when importing.
2139 Calling the C<export()> method on an existing DBM::Deep object will return
2140 a reference to a new in-memory copy of the database. The export is done
2141 recursively, so all nested hashes/arrays are all exported to standard Perl
2142 objects. Here is an example:
2144 my $db = DBM::Deep->new( "foo.db" );
2146 $db->{key1} = "value1";
2147 $db->{key2} = "value2";
2149 $db->{hash1}->{subkey1} = "subvalue1";
2150 $db->{hash1}->{subkey2} = "subvalue2";
2152 my $struct = $db->export();
2154 print $struct->{key1} . "\n"; # prints "value1"
2156 This makes a complete copy of the database in memory, and returns a reference
2157 to it. The C<export()> method can be called on any database level (not just
2158 the base level), and works with both hash and array DB types. Be careful of
2159 large databases -- you can store a lot more data in a DBM::Deep object than an
2160 in-memory Perl structure.
2162 B<Note:> Make sure your database has no circular references in it.
2163 These will cause an infinite loop when exporting.
2167 DBM::Deep has a number of hooks where you can specify your own Perl function
2168 to perform filtering on incoming or outgoing data. This is a perfect
2169 way to extend the engine, and implement things like real-time compression or
2170 encryption. Filtering applies to the base DB level, and all child hashes /
2171 arrays. Filter hooks can be specified when your DBM::Deep object is first
2172 constructed, or by calling the C<set_filter()> method at any time. There are
2173 four available filter hooks, described below:
2177 =item * filter_store_key
2179 This filter is called whenever a hash key is stored. It
2180 is passed the incoming key, and expected to return a transformed key.
2182 =item * filter_store_value
2184 This filter is called whenever a hash key or array element is stored. It
2185 is passed the incoming value, and expected to return a transformed value.
2187 =item * filter_fetch_key
2189 This filter is called whenever a hash key is fetched (i.e. via
2190 C<first_key()> or C<next_key()>). It is passed the transformed key,
2191 and expected to return the plain key.
2193 =item * filter_fetch_value
2195 This filter is called whenever a hash key or array element is fetched.
2196 It is passed the transformed value, and expected to return the plain value.
2200 Here are the two ways to setup a filter hook:
2202 my $db = DBM::Deep->new(
2204 filter_store_value => \&my_filter_store,
2205 filter_fetch_value => \&my_filter_fetch
2210 $db->set_filter( "filter_store_value", \&my_filter_store );
2211 $db->set_filter( "filter_fetch_value", \&my_filter_fetch );
2213 Your filter function will be called only when dealing with SCALAR keys or
2214 values. When nested hashes and arrays are being stored/fetched, filtering
2215 is bypassed. Filters are called as static functions, passed a single SCALAR
2216 argument, and expected to return a single SCALAR value. If you want to
2217 remove a filter, set the function reference to C<undef>:
2219 $db->set_filter( "filter_store_value", undef );
2221 =head2 REAL-TIME ENCRYPTION EXAMPLE
2223 Here is a working example that uses the I<Crypt::Blowfish> module to
2224 do real-time encryption / decryption of keys & values with DBM::Deep Filters.
2225 Please visit L<http://search.cpan.org/search?module=Crypt::Blowfish> for more
2226 on I<Crypt::Blowfish>. You'll also need the I<Crypt::CBC> module.
2229 use Crypt::Blowfish;
2232 my $cipher = Crypt::CBC->new({
2233 'key' => 'my secret key',
2234 'cipher' => 'Blowfish',
2236 'regenerate_key' => 0,
2237 'padding' => 'space',
2241 my $db = DBM::Deep->new(
2242 file => "foo-encrypt.db",
2243 filter_store_key => \&my_encrypt,
2244 filter_store_value => \&my_encrypt,
2245 filter_fetch_key => \&my_decrypt,
2246 filter_fetch_value => \&my_decrypt,
2249 $db->{key1} = "value1";
2250 $db->{key2} = "value2";
2251 print "key1: " . $db->{key1} . "\n";
2252 print "key2: " . $db->{key2} . "\n";
2258 return $cipher->encrypt( $_[0] );
2261 return $cipher->decrypt( $_[0] );
2264 =head2 REAL-TIME COMPRESSION EXAMPLE
2266 Here is a working example that uses the I<Compress::Zlib> module to do real-time
2267 compression / decompression of keys & values with DBM::Deep Filters.
2268 Please visit L<http://search.cpan.org/search?module=Compress::Zlib> for
2269 more on I<Compress::Zlib>.
2274 my $db = DBM::Deep->new(
2275 file => "foo-compress.db",
2276 filter_store_key => \&my_compress,
2277 filter_store_value => \&my_compress,
2278 filter_fetch_key => \&my_decompress,
2279 filter_fetch_value => \&my_decompress,
2282 $db->{key1} = "value1";
2283 $db->{key2} = "value2";
2284 print "key1: " . $db->{key1} . "\n";
2285 print "key2: " . $db->{key2} . "\n";
2291 return Compress::Zlib::memGzip( $_[0] ) ;
2294 return Compress::Zlib::memGunzip( $_[0] ) ;
2297 B<Note:> Filtering of keys only applies to hashes. Array "keys" are
2298 actually numerical index numbers, and are not filtered.
2300 =head1 ERROR HANDLING
2302 Most DBM::Deep methods return a true value for success, and call die() on
2303 failure. You can wrap calls in an eval block to catch the die. Also, the
2304 actual error message is stored in an internal scalar, which can be fetched by
2305 calling the C<error()> method.
2307 my $db = DBM::Deep->new( "foo.db" ); # create hash
2308 eval { $db->push("foo"); }; # ILLEGAL -- push is array-only call
2310 print $@; # prints error message
2311 print $db->error(); # prints error message
2313 You can then call C<clear_error()> to clear the current error state.
2317 If you set the C<debug> option to true when creating your DBM::Deep object,
2318 all errors are considered NON-FATAL, and dumped to STDERR. This should only
2319 be used for debugging purposes and not production work. DBM::Deep expects errors
2320 to be thrown, not propagated back up the stack.
2322 =head1 LARGEFILE SUPPORT
2324 If you have a 64-bit system, and your Perl is compiled with both LARGEFILE
2325 and 64-bit support, you I<may> be able to create databases larger than 2 GB.
2326 DBM::Deep by default uses 32-bit file offset tags, but these can be changed
2327 by calling the static C<set_pack()> method before you do anything else.
2329 DBM::Deep::set_pack(8, 'Q');
2331 This tells DBM::Deep to pack all file offsets with 8-byte (64-bit) quad words
2332 instead of 32-bit longs. After setting these values your DB files have a
2333 theoretical maximum size of 16 XB (exabytes).
2335 B<Note:> Changing these values will B<NOT> work for existing database files.
2336 Only change this for new files, and make sure it stays set consistently
2337 throughout the file's life. If you do set these values, you can no longer
2338 access 32-bit DB files. You can, however, call C<set_pack(4, 'N')> to change
2339 back to 32-bit mode.
2341 B<Note:> I have not personally tested files > 2 GB -- all my systems have
2342 only a 32-bit Perl. However, I have received user reports that this does
2345 =head1 LOW-LEVEL ACCESS
2347 If you require low-level access to the underlying filehandle that DBM::Deep uses,
2348 you can call the C<_fh()> method, which returns the handle:
2350 my $fh = $db->_fh();
2352 This method can be called on the root level of the datbase, or any child
2353 hashes or arrays. All levels share a I<root> structure, which contains things
2354 like the filehandle, a reference counter, and all the options specified
2355 when you created the object. You can get access to this root structure by
2356 calling the C<root()> method.
2358 my $root = $db->_root();
2360 This is useful for changing options after the object has already been created,
2361 such as enabling/disabling locking, or debug modes. You can also
2362 store your own temporary user data in this structure (be wary of name
2363 collision), which is then accessible from any child hash or array.
2365 =head1 CUSTOM DIGEST ALGORITHM
2367 DBM::Deep by default uses the I<Message Digest 5> (MD5) algorithm for hashing
2368 keys. However you can override this, and use another algorithm (such as SHA-256)
2369 or even write your own. But please note that DBM::Deep currently expects zero
2370 collisions, so your algorithm has to be I<perfect>, so to speak.
2371 Collision detection may be introduced in a later version.
2375 You can specify a custom digest algorithm by calling the static C<set_digest()>
2376 function, passing a reference to a subroutine, and the length of the algorithm's
2377 hashes (in bytes). This is a global static function, which affects ALL DBM::Deep
2378 objects. Here is a working example that uses a 256-bit hash from the
2379 I<Digest::SHA256> module. Please see
2380 L<http://search.cpan.org/search?module=Digest::SHA256> for more.
2385 my $context = Digest::SHA256::new(256);
2387 DBM::Deep::set_digest( \&my_digest, 32 );
2389 my $db = DBM::Deep->new( "foo-sha.db" );
2391 $db->{key1} = "value1";
2392 $db->{key2} = "value2";
2393 print "key1: " . $db->{key1} . "\n";
2394 print "key2: " . $db->{key2} . "\n";
2400 return substr( $context->hash($_[0]), 0, 32 );
2403 B<Note:> Your returned digest strings must be B<EXACTLY> the number
2404 of bytes you specify in the C<set_digest()> function (in this case 32).
2406 =head1 CIRCULAR REFERENCES
2408 DBM::Deep has B<experimental> support for circular references. Meaning you
2409 can have a nested hash key or array element that points to a parent object.
2410 This relationship is stored in the DB file, and is preserved between sessions.
2413 my $db = DBM::Deep->new( "foo.db" );
2416 $db->{circle} = $db; # ref to self
2418 print $db->{foo} . "\n"; # prints "foo"
2419 print $db->{circle}->{foo} . "\n"; # prints "foo" again
2421 One catch is, passing the object to a function that recursively walks the
2422 object tree (such as I<Data::Dumper> or even the built-in C<optimize()> or
2423 C<export()> methods) will result in an infinite loop. The other catch is,
2424 if you fetch the I<key> of a circular reference (i.e. using the C<first_key()>
2425 or C<next_key()> methods), you will get the I<target object's key>, not the
2426 ref's key. This gets even more interesting with the above example, where
2427 the I<circle> key points to the base DB object, which technically doesn't
2428 have a key. So I made DBM::Deep return "[base]" as the key name in that
2431 =head1 CAVEATS / ISSUES / BUGS
2433 This section describes all the known issues with DBM::Deep. It you have found
2434 something that is not listed here, please send e-mail to L<jhuckaby@cpan.org>.
2436 =head2 UNUSED SPACE RECOVERY
2438 One major caveat with DBM::Deep is that space occupied by existing keys and
2439 values is not recovered when they are deleted. Meaning if you keep deleting
2440 and adding new keys, your file will continuously grow. I am working on this,
2441 but in the meantime you can call the built-in C<optimize()> method from time to
2442 time (perhaps in a crontab or something) to recover all your unused space.
2444 $db->optimize(); # returns true on success
2446 This rebuilds the ENTIRE database into a new file, then moves it on top of
2447 the original. The new file will have no unused space, thus it will take up as
2448 little disk space as possible. Please note that this operation can take
2449 a long time for large files, and you need enough disk space to temporarily hold
2450 2 copies of your DB file. The temporary file is created in the same directory
2451 as the original, named with a ".tmp" extension, and is deleted when the
2452 operation completes. Oh, and if locking is enabled, the DB is automatically
2453 locked for the entire duration of the copy.
2455 B<WARNING:> Only call optimize() on the top-level node of the database, and
2456 make sure there are no child references lying around. DBM::Deep keeps a reference
2457 counter, and if it is greater than 1, optimize() will abort and return undef.
2459 =head2 AUTOVIVIFICATION
2461 Unfortunately, autovivification doesn't work with tied hashes. This appears to
2462 be a bug in Perl's tie() system, as I<Jakob Schmidt> encountered the very same
2463 issue with his I<DWH_FIle> module (see L<http://search.cpan.org/search?module=DWH_File>),
2464 and it is also mentioned in the BUGS section for the I<MLDBM> module <see
2465 L<http://search.cpan.org/search?module=MLDBM>). Basically, on a new db file,
2468 $db->{foo}->{bar} = "hello";
2470 Since "foo" doesn't exist, you cannot add "bar" to it. You end up with "foo"
2471 being an empty hash. Try this instead, which works fine:
2473 $db->{foo} = { bar => "hello" };
2475 As of Perl 5.8.7, this bug still exists. I have walked very carefully through
2476 the execution path, and Perl indeed passes an empty hash to the STORE() method.
2477 Probably a bug in Perl.
2479 =head2 FILE CORRUPTION
2481 The current level of error handling in DBM::Deep is minimal. Files I<are> checked
2482 for a 32-bit signature when opened, but other corruption in files can cause
2483 segmentation faults. DBM::Deep may try to seek() past the end of a file, or get
2484 stuck in an infinite loop depending on the level of corruption. File write
2485 operations are not checked for failure (for speed), so if you happen to run
2486 out of disk space, DBM::Deep will probably fail in a bad way. These things will
2487 be addressed in a later version of DBM::Deep.
2491 Beware of using DB files over NFS. DBM::Deep uses flock(), which works well on local
2492 filesystems, but will NOT protect you from file corruption over NFS. I've heard
2493 about setting up your NFS server with a locking daemon, then using lockf() to
2494 lock your files, but your mileage may vary there as well. From what I
2495 understand, there is no real way to do it. However, if you need access to the
2496 underlying filehandle in DBM::Deep for using some other kind of locking scheme like
2497 lockf(), see the L<LOW-LEVEL ACCESS> section above.
2499 =head2 COPYING OBJECTS
2501 Beware of copying tied objects in Perl. Very strange things can happen.
2502 Instead, use DBM::Deep's C<clone()> method which safely copies the object and
2503 returns a new, blessed, tied hash or array to the same level in the DB.
2505 my $copy = $db->clone();
2507 B<Note>: Since clone() here is cloning the object, not the database location, any
2508 modifications to either $db or $copy will be visible in both.
2512 Beware of using C<shift()>, C<unshift()> or C<splice()> with large arrays.
2513 These functions cause every element in the array to move, which can be murder
2514 on DBM::Deep, as every element has to be fetched from disk, then stored again in
2515 a different location. This will be addressed in the forthcoming version 1.00.
2519 This section discusses DBM::Deep's speed and memory usage.
2523 Obviously, DBM::Deep isn't going to be as fast as some C-based DBMs, such as
2524 the almighty I<BerkeleyDB>. But it makes up for it in features like true
2525 multi-level hash/array support, and cross-platform FTPable files. Even so,
2526 DBM::Deep is still pretty fast, and the speed stays fairly consistent, even
2527 with huge databases. Here is some test data:
2529 Adding 1,000,000 keys to new DB file...
2531 At 100 keys, avg. speed is 2,703 keys/sec
2532 At 200 keys, avg. speed is 2,642 keys/sec
2533 At 300 keys, avg. speed is 2,598 keys/sec
2534 At 400 keys, avg. speed is 2,578 keys/sec
2535 At 500 keys, avg. speed is 2,722 keys/sec
2536 At 600 keys, avg. speed is 2,628 keys/sec
2537 At 700 keys, avg. speed is 2,700 keys/sec
2538 At 800 keys, avg. speed is 2,607 keys/sec
2539 At 900 keys, avg. speed is 2,190 keys/sec
2540 At 1,000 keys, avg. speed is 2,570 keys/sec
2541 At 2,000 keys, avg. speed is 2,417 keys/sec
2542 At 3,000 keys, avg. speed is 1,982 keys/sec
2543 At 4,000 keys, avg. speed is 1,568 keys/sec
2544 At 5,000 keys, avg. speed is 1,533 keys/sec
2545 At 6,000 keys, avg. speed is 1,787 keys/sec
2546 At 7,000 keys, avg. speed is 1,977 keys/sec
2547 At 8,000 keys, avg. speed is 2,028 keys/sec
2548 At 9,000 keys, avg. speed is 2,077 keys/sec
2549 At 10,000 keys, avg. speed is 2,031 keys/sec
2550 At 20,000 keys, avg. speed is 1,970 keys/sec
2551 At 30,000 keys, avg. speed is 2,050 keys/sec
2552 At 40,000 keys, avg. speed is 2,073 keys/sec
2553 At 50,000 keys, avg. speed is 1,973 keys/sec
2554 At 60,000 keys, avg. speed is 1,914 keys/sec
2555 At 70,000 keys, avg. speed is 2,091 keys/sec
2556 At 80,000 keys, avg. speed is 2,103 keys/sec
2557 At 90,000 keys, avg. speed is 1,886 keys/sec
2558 At 100,000 keys, avg. speed is 1,970 keys/sec
2559 At 200,000 keys, avg. speed is 2,053 keys/sec
2560 At 300,000 keys, avg. speed is 1,697 keys/sec
2561 At 400,000 keys, avg. speed is 1,838 keys/sec
2562 At 500,000 keys, avg. speed is 1,941 keys/sec
2563 At 600,000 keys, avg. speed is 1,930 keys/sec
2564 At 700,000 keys, avg. speed is 1,735 keys/sec
2565 At 800,000 keys, avg. speed is 1,795 keys/sec
2566 At 900,000 keys, avg. speed is 1,221 keys/sec
2567 At 1,000,000 keys, avg. speed is 1,077 keys/sec
2569 This test was performed on a PowerMac G4 1gHz running Mac OS X 10.3.2 & Perl
2570 5.8.1, with an 80GB Ultra ATA/100 HD spinning at 7200RPM. The hash keys and
2571 values were between 6 - 12 chars in length. The DB file ended up at 210MB.
2572 Run time was 12 min 3 sec.
2576 One of the great things about DBM::Deep is that it uses very little memory.
2577 Even with huge databases (1,000,000+ keys) you will not see much increased
2578 memory on your process. DBM::Deep relies solely on the filesystem for storing
2579 and fetching data. Here is output from I</usr/bin/top> before even opening a
2582 PID USER PRI NI SIZE RSS SHARE STAT %CPU %MEM TIME COMMAND
2583 22831 root 11 0 2716 2716 1296 R 0.0 0.2 0:07 perl
2585 Basically the process is taking 2,716K of memory. And here is the same
2586 process after storing and fetching 1,000,000 keys:
2588 PID USER PRI NI SIZE RSS SHARE STAT %CPU %MEM TIME COMMAND
2589 22831 root 14 0 2772 2772 1328 R 0.0 0.2 13:32 perl
2591 Notice the memory usage increased by only 56K. Test was performed on a 700mHz
2592 x86 box running Linux RedHat 7.2 & Perl 5.6.1.
2594 =head1 DB FILE FORMAT
2596 In case you were interested in the underlying DB file format, it is documented
2597 here in this section. You don't need to know this to use the module, it's just
2598 included for reference.
2602 DBM::Deep files always start with a 32-bit signature to identify the file type.
2603 This is at offset 0. The signature is "DPDB" in network byte order. This is
2604 checked for when the file is opened and an error will be thrown if it's not found.
2608 The DBM::Deep file is in a I<tagged format>, meaning each section of the file
2609 has a standard header containing the type of data, the length of data, and then
2610 the data itself. The type is a single character (1 byte), the length is a
2611 32-bit unsigned long in network byte order, and the data is, well, the data.
2612 Here is how it unfolds:
2616 Immediately after the 32-bit file signature is the I<Master Index> record.
2617 This is a standard tag header followed by 1024 bytes (in 32-bit mode) or 2048
2618 bytes (in 64-bit mode) of data. The type is I<H> for hash or I<A> for array,
2619 depending on how the DBM::Deep object was constructed.
2621 The index works by looking at a I<MD5 Hash> of the hash key (or array index
2622 number). The first 8-bit char of the MD5 signature is the offset into the
2623 index, multipled by 4 in 32-bit mode, or 8 in 64-bit mode. The value of the
2624 index element is a file offset of the next tag for the key/element in question,
2625 which is usually a I<Bucket List> tag (see below).
2627 The next tag I<could> be another index, depending on how many keys/elements
2628 exist. See L<RE-INDEXING> below for details.
2632 A I<Bucket List> is a collection of 16 MD5 hashes for keys/elements, plus
2633 file offsets to where the actual data is stored. It starts with a standard
2634 tag header, with type I<B>, and a data size of 320 bytes in 32-bit mode, or
2635 384 bytes in 64-bit mode. Each MD5 hash is stored in full (16 bytes), plus
2636 the 32-bit or 64-bit file offset for the I<Bucket> containing the actual data.
2637 When the list fills up, a I<Re-Index> operation is performed (See
2638 L<RE-INDEXING> below).
2642 A I<Bucket> is a tag containing a key/value pair (in hash mode), or a
2643 index/value pair (in array mode). It starts with a standard tag header with
2644 type I<D> for scalar data (string, binary, etc.), or it could be a nested
2645 hash (type I<H>) or array (type I<A>). The value comes just after the tag
2646 header. The size reported in the tag header is only for the value, but then,
2647 just after the value is another size (32-bit unsigned long) and then the plain
2648 key itself. Since the value is likely to be fetched more often than the plain
2649 key, I figured it would be I<slightly> faster to store the value first.
2651 If the type is I<H> (hash) or I<A> (array), the value is another I<Master Index>
2652 record for the nested structure, where the process begins all over again.
2656 After a I<Bucket List> grows to 16 records, its allocated space in the file is
2657 exhausted. Then, when another key/element comes in, the list is converted to a
2658 new index record. However, this index will look at the next char in the MD5
2659 hash, and arrange new Bucket List pointers accordingly. This process is called
2660 I<Re-Indexing>. Basically, a new index tag is created at the file EOF, and all
2661 17 (16 + new one) keys/elements are removed from the old Bucket List and
2662 inserted into the new index. Several new Bucket Lists are created in the
2663 process, as a new MD5 char from the key is being examined (it is unlikely that
2664 the keys will all share the same next char of their MD5s).
2666 Because of the way the I<MD5> algorithm works, it is impossible to tell exactly
2667 when the Bucket Lists will turn into indexes, but the first round tends to
2668 happen right around 4,000 keys. You will see a I<slight> decrease in
2669 performance here, but it picks back up pretty quick (see L<SPEED> above). Then
2670 it takes B<a lot> more keys to exhaust the next level of Bucket Lists. It's
2671 right around 900,000 keys. This process can continue nearly indefinitely --
2672 right up until the point the I<MD5> signatures start colliding with each other,
2673 and this is B<EXTREMELY> rare -- like winning the lottery 5 times in a row AND
2674 getting struck by lightning while you are walking to cash in your tickets.
2675 Theoretically, since I<MD5> hashes are 128-bit values, you I<could> have up to
2676 340,282,366,921,000,000,000,000,000,000,000,000,000 keys/elements (I believe
2677 this is 340 unodecillion, but don't quote me).
2681 When a new key/element is stored, the key (or index number) is first run through
2682 I<Digest::MD5> to get a 128-bit signature (example, in hex:
2683 b05783b0773d894396d475ced9d2f4f6). Then, the I<Master Index> record is checked
2684 for the first char of the signature (in this case I<b0>). If it does not exist,
2685 a new I<Bucket List> is created for our key (and the next 15 future keys that
2686 happen to also have I<b> as their first MD5 char). The entire MD5 is written
2687 to the I<Bucket List> along with the offset of the new I<Bucket> record (EOF at
2688 this point, unless we are replacing an existing I<Bucket>), where the actual
2689 data will be stored.
2693 Fetching an existing key/element involves getting a I<Digest::MD5> of the key
2694 (or index number), then walking along the indexes. If there are enough
2695 keys/elements in this DB level, there might be nested indexes, each linked to
2696 a particular char of the MD5. Finally, a I<Bucket List> is pointed to, which
2697 contains up to 16 full MD5 hashes. Each is checked for equality to the key in
2698 question. If we found a match, the I<Bucket> tag is loaded, where the value and
2699 plain key are stored.
2701 Fetching the plain key occurs when calling the I<first_key()> and I<next_key()>
2702 methods. In this process the indexes are walked systematically, and each key
2703 fetched in increasing MD5 order (which is why it appears random). Once the
2704 I<Bucket> is found, the value is skipped and the plain key returned instead.
2705 B<Note:> Do not count on keys being fetched as if the MD5 hashes were
2706 alphabetically sorted. This only happens on an index-level -- as soon as the
2707 I<Bucket Lists> are hit, the keys will come out in the order they went in --
2708 so it's pretty much undefined how the keys will come out -- just like Perl's
2711 =head1 CODE COVERAGE
2713 We use B<Devel::Cover> to test the code coverage of our tests, below is the
2714 B<Devel::Cover> report on this module's test suite.
2716 ---------------------------- ------ ------ ------ ------ ------ ------ ------
2717 File stmt bran cond sub pod time total
2718 ---------------------------- ------ ------ ------ ------ ------ ------ ------
2719 blib/lib/DBM/Deep.pm 95.0 83.2 68.7 98.2 100.0 57.8 90.7
2720 blib/lib/DBM/Deep/Array.pm 98.9 88.9 87.5 100.0 n/a 27.4 96.4
2721 blib/lib/DBM/Deep/Hash.pm 95.3 80.0 100.0 100.0 n/a 14.8 92.4
2722 Total 95.8 83.9 72.8 98.8 100.0 100.0 91.8
2723 ---------------------------- ------ ------ ------ ------ ------ ------ ------
2725 =head1 MORE INFORMATION
2727 Check out the DBM::Deep Google Group at L<http://groups.google.com/group/DBM-Deep>
2728 or send email to L<DBM-Deep@googlegroups.com>.
2732 Joseph Huckaby, L<jhuckaby@cpan.org>
2734 Rob Kinyon, L<rkinyon@cpan.org>
2736 Special thanks to Adam Sah and Rich Gaushell! You know why :-)
2740 perltie(1), Tie::Hash(3), Digest::MD5(3), Fcntl(3), flock(2), lockf(3), nfs(5),
2741 Digest::SHA256(3), Crypt::Blowfish(3), Compress::Zlib(3)
2745 Copyright (c) 2002-2006 Joseph Huckaby. All Rights Reserved.
2746 This is free software, you may use it and distribute it under the
2747 same terms as Perl itself.