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);
249 print( $fh SIG_FILE);
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 my $root = $self->_root;
383 my $is_dbm_deep = eval { local $SIG{'__DIE__'}; $value->isa( 'DBM::Deep' ) };
384 my $internal_ref = $is_dbm_deep && ($value->_root eq $root);
389 # Iterate through buckets, seeing if this is a new entry or a replace.
391 for (my $i=0; $i<$MAX_BUCKETS; $i++) {
392 my $subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
395 # Found empty bucket (end of list). Populate and exit loop.
399 $location = $internal_ref
400 ? $value->_base_offset
403 seek($fh, $tag->{offset} + ($i * $BUCKET_SIZE) + $root->{file_offset}, SEEK_SET);
404 print( $fh $md5 . pack($LONG_PACK, $location) );
408 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
411 # Found existing bucket with same key. Replace with new value.
416 $location = $value->_base_offset;
417 seek($fh, $tag->{offset} + ($i * $BUCKET_SIZE) + $root->{file_offset}, SEEK_SET);
418 print( $fh $md5 . pack($LONG_PACK, $location) );
422 seek($fh, $subloc + SIG_SIZE + $root->{file_offset}, SEEK_SET);
424 read( $fh, $size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
427 # If value is a hash, array, or raw value with equal or less size, we can
428 # reuse the same content area of the database. Otherwise, we have to create
429 # a new content area at the EOF.
432 my $r = Scalar::Util::reftype( $value ) || '';
433 if ( $r eq 'HASH' || $r eq 'ARRAY' ) {
434 $actual_length = $INDEX_SIZE;
436 # if autobless is enabled, must also take into consideration
437 # the class name, as it is stored along with key/value.
438 if ( $root->{autobless} ) {
439 my $value_class = Scalar::Util::blessed($value);
440 if ( defined $value_class && !$value->isa('DBM::Deep') ) {
441 $actual_length += length($value_class);
445 else { $actual_length = length($value); }
447 if ($actual_length <= $size) {
451 $location = $root->{end};
452 seek($fh, $tag->{offset} + ($i * $BUCKET_SIZE) + $HASH_SIZE + $root->{file_offset}, SEEK_SET);
453 print( $fh pack($LONG_PACK, $location) );
461 # If this is an internal reference, return now.
462 # No need to write value or plain key
469 # If bucket didn't fit into list, split into a new index level
472 seek($fh, $tag->{ref_loc} + $root->{file_offset}, SEEK_SET);
473 print( $fh pack($LONG_PACK, $root->{end}) );
475 my $index_tag = $self->_create_tag($root->{end}, SIG_INDEX, chr(0) x $INDEX_SIZE);
478 $keys .= $md5 . pack($LONG_PACK, 0);
480 for (my $i=0; $i<=$MAX_BUCKETS; $i++) {
481 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
483 my $old_subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
484 my $num = ord(substr($key, $tag->{ch} + 1, 1));
486 if ($offsets[$num]) {
487 my $offset = $offsets[$num] + SIG_SIZE + $DATA_LENGTH_SIZE;
488 seek($fh, $offset + $root->{file_offset}, SEEK_SET);
490 read( $fh, $subkeys, $BUCKET_LIST_SIZE);
492 for (my $k=0; $k<$MAX_BUCKETS; $k++) {
493 my $subloc = unpack($LONG_PACK, substr($subkeys, ($k * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
495 seek($fh, $offset + ($k * $BUCKET_SIZE) + $root->{file_offset}, SEEK_SET);
496 print( $fh $key . pack($LONG_PACK, $old_subloc || $root->{end}) );
502 $offsets[$num] = $root->{end};
503 seek($fh, $index_tag->{offset} + ($num * $LONG_SIZE) + $root->{file_offset}, SEEK_SET);
504 print( $fh pack($LONG_PACK, $root->{end}) );
506 my $blist_tag = $self->_create_tag($root->{end}, SIG_BLIST, chr(0) x $BUCKET_LIST_SIZE);
508 seek($fh, $blist_tag->{offset} + $root->{file_offset}, SEEK_SET);
509 print( $fh $key . pack($LONG_PACK, $old_subloc || $root->{end}) );
514 $location ||= $root->{end};
515 } # re-index bucket list
518 # Seek to content area and store signature, value and plaintext key
522 seek($fh, $location + $root->{file_offset}, SEEK_SET);
525 # Write signature based on content type, set content length and write actual value.
527 my $r = Scalar::Util::reftype($value) || '';
529 print( $fh TYPE_HASH );
530 print( $fh pack($DATA_LENGTH_PACK, $INDEX_SIZE) . chr(0) x $INDEX_SIZE );
531 $content_length = $INDEX_SIZE;
533 elsif ($r eq 'ARRAY') {
534 print( $fh TYPE_ARRAY );
535 print( $fh pack($DATA_LENGTH_PACK, $INDEX_SIZE) . chr(0) x $INDEX_SIZE );
536 $content_length = $INDEX_SIZE;
538 elsif (!defined($value)) {
539 print( $fh SIG_NULL );
540 print( $fh pack($DATA_LENGTH_PACK, 0) );
544 print( $fh SIG_DATA );
545 print( $fh pack($DATA_LENGTH_PACK, length($value)) . $value );
546 $content_length = length($value);
550 # Plain key is stored AFTER value, as keys are typically fetched less often.
552 print( $fh pack($DATA_LENGTH_PACK, length($plain_key)) . $plain_key );
555 # If value is blessed, preserve class name
557 if ( $root->{autobless} ) {
558 my $value_class = Scalar::Util::blessed($value);
559 if ( defined $value_class && $value_class ne 'DBM::Deep' ) {
561 # Blessed ref -- will restore later
564 print( $fh pack($DATA_LENGTH_PACK, length($value_class)) . $value_class );
565 $content_length += 1;
566 $content_length += $DATA_LENGTH_SIZE + length($value_class);
570 $content_length += 1;
575 # If this is a new content area, advance EOF counter
577 if ($location == $root->{end}) {
578 $root->{end} += SIG_SIZE;
579 $root->{end} += $DATA_LENGTH_SIZE + $content_length;
580 $root->{end} += $DATA_LENGTH_SIZE + length($plain_key);
584 # If content is a hash or array, create new child DBM::Deep object and
585 # pass each key or element to it.
588 my $branch = DBM::Deep->new(
590 base_offset => $location,
593 foreach my $key (keys %{$value}) {
594 $branch->STORE( $key, $value->{$key} );
597 elsif ($r eq 'ARRAY') {
598 my $branch = DBM::Deep->new(
600 base_offset => $location,
604 foreach my $element (@{$value}) {
605 $branch->STORE( $index, $element );
613 return $self->_throw_error("Fatal error: indexing failed -- possibly due to corruption in file");
616 sub _get_bucket_value {
618 # Fetch single value given tag and MD5 digested key.
621 my ($tag, $md5) = @_;
622 my $keys = $tag->{content};
627 # Iterate through buckets, looking for a key match
630 for (my $i=0; $i<$MAX_BUCKETS; $i++) {
631 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
632 my $subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
636 # Hit end of list, no match
641 if ( $md5 ne $key ) {
646 # Found match -- seek to offset and read signature
649 seek($fh, $subloc + $self->_root->{file_offset}, SEEK_SET);
650 read( $fh, $signature, SIG_SIZE);
653 # If value is a hash or array, return new DBM::Deep object with correct offset
655 if (($signature eq TYPE_HASH) || ($signature eq TYPE_ARRAY)) {
656 my $obj = DBM::Deep->new(
658 base_offset => $subloc,
662 if ($self->_root->{autobless}) {
664 # Skip over value and plain key to see if object needs
667 seek($fh, $DATA_LENGTH_SIZE + $INDEX_SIZE, SEEK_CUR);
670 read( $fh, $size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
671 if ($size) { seek($fh, $size, SEEK_CUR); }
674 read( $fh, $bless_bit, 1);
675 if (ord($bless_bit)) {
677 # Yes, object needs to be re-blessed
680 read( $fh, $size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
681 if ($size) { read( $fh, $class_name, $size); }
682 if ($class_name) { $obj = bless( $obj, $class_name ); }
690 # Otherwise return actual value
692 elsif ($signature eq SIG_DATA) {
695 read( $fh, $size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
696 if ($size) { read( $fh, $value, $size); }
701 # Key exists, but content is null
711 # Delete single key/value pair given tag and MD5 digested key.
714 my ($tag, $md5) = @_;
715 my $keys = $tag->{content};
720 # Iterate through buckets, looking for a key match
723 for (my $i=0; $i<$MAX_BUCKETS; $i++) {
724 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
725 my $subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
729 # Hit end of list, no match
734 if ( $md5 ne $key ) {
739 # Matched key -- delete bucket and return
741 seek($fh, $tag->{offset} + ($i * $BUCKET_SIZE) + $self->_root->{file_offset}, SEEK_SET);
742 print( $fh substr($keys, ($i+1) * $BUCKET_SIZE ) );
743 print( $fh chr(0) x $BUCKET_SIZE );
753 # Check existence of single key given tag and MD5 digested key.
756 my ($tag, $md5) = @_;
757 my $keys = $tag->{content};
760 # Iterate through buckets, looking for a key match
763 for (my $i=0; $i<$MAX_BUCKETS; $i++) {
764 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
765 my $subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
769 # Hit end of list, no match
774 if ( $md5 ne $key ) {
779 # Matched key -- return true
787 sub _find_bucket_list {
789 # Locate offset for bucket list, given digested key
795 # Locate offset for bucket list using digest index system
798 my $tag = $self->_load_tag($self->_base_offset);
799 if (!$tag) { return; }
801 while ($tag->{signature} ne SIG_BLIST) {
802 $tag = $self->_index_lookup($tag, ord(substr($md5, $ch, 1)));
803 if (!$tag) { return; }
810 sub _traverse_index {
812 # Scan index and recursively step into deeper levels, looking for next key.
814 my ($self, $offset, $ch, $force_return_next) = @_;
815 $force_return_next = undef unless $force_return_next;
817 my $tag = $self->_load_tag( $offset );
821 if ($tag->{signature} ne SIG_BLIST) {
822 my $content = $tag->{content};
824 if ($self->{return_next}) { $start = 0; }
825 else { $start = ord(substr($self->{prev_md5}, $ch, 1)); }
827 for (my $index = $start; $index < 256; $index++) {
828 my $subloc = unpack($LONG_PACK, substr($content, $index * $LONG_SIZE, $LONG_SIZE) );
830 my $result = $self->_traverse_index( $subloc, $ch + 1, $force_return_next );
831 if (defined($result)) { return $result; }
835 $self->{return_next} = 1;
838 elsif ($tag->{signature} eq SIG_BLIST) {
839 my $keys = $tag->{content};
840 if ($force_return_next) { $self->{return_next} = 1; }
843 # Iterate through buckets, looking for a key match
845 for (my $i=0; $i<$MAX_BUCKETS; $i++) {
846 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
847 my $subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
851 # End of bucket list -- return to outer loop
853 $self->{return_next} = 1;
856 elsif ($key eq $self->{prev_md5}) {
858 # Located previous key -- return next one found
860 $self->{return_next} = 1;
863 elsif ($self->{return_next}) {
865 # Seek to bucket location and skip over signature
867 seek($fh, $subloc + SIG_SIZE + $self->_root->{file_offset}, SEEK_SET);
870 # Skip over value to get to plain key
873 read( $fh, $size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
874 if ($size) { seek($fh, $size, SEEK_CUR); }
877 # Read in plain key and return as scalar
880 read( $fh, $size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
881 if ($size) { read( $fh, $plain_key, $size); }
887 $self->{return_next} = 1;
888 } # tag is a bucket list
895 # Locate next key, given digested previous one
897 my $self = $_[0]->_get_self;
899 $self->{prev_md5} = $_[1] ? $_[1] : undef;
900 $self->{return_next} = 0;
903 # If the previous key was not specifed, start at the top and
904 # return the first one found.
906 if (!$self->{prev_md5}) {
907 $self->{prev_md5} = chr(0) x $HASH_SIZE;
908 $self->{return_next} = 1;
911 return $self->_traverse_index( $self->_base_offset, 0 );
916 # If db locking is set, flock() the db file. If called multiple
917 # times before unlock(), then the same number of unlocks() must
918 # be called before the lock is released.
920 my $self = $_[0]->_get_self;
922 $type = LOCK_EX unless defined $type;
924 if (!defined($self->_fh)) { return; }
926 if ($self->_root->{locking}) {
927 if (!$self->_root->{locked}) {
928 flock($self->_fh, $type);
930 # refresh end counter in case file has changed size
931 my @stats = stat($self->_root->{file});
932 $self->_root->{end} = $stats[7];
934 # double-check file inode, in case another process
935 # has optimize()d our file while we were waiting.
936 if ($stats[1] != $self->_root->{inode}) {
937 $self->_open(); # re-open
938 flock($self->_fh, $type); # re-lock
939 $self->_root->{end} = (stat($self->_fh))[7]; # re-end
942 $self->_root->{locked}++;
952 # If db locking is set, unlock the db file. See note in lock()
953 # regarding calling lock() multiple times.
955 my $self = $_[0]->_get_self;
957 if (!defined($self->_fh)) { return; }
959 if ($self->_root->{locking} && $self->_root->{locked} > 0) {
960 $self->_root->{locked}--;
961 if (!$self->_root->{locked}) { flock($self->_fh, LOCK_UN); }
970 my $self = shift->_get_self;
971 my ($spot, $value) = @_;
976 elsif ( eval { local $SIG{__DIE__}; $value->isa( 'DBM::Deep' ) } ) {
977 my $type = $value->_type;
978 ${$spot} = $type eq TYPE_HASH ? {} : [];
979 $value->_copy_node( ${$spot} );
982 my $r = Scalar::Util::reftype( $value );
983 my $c = Scalar::Util::blessed( $value );
984 if ( $r eq 'ARRAY' ) {
985 ${$spot} = [ @{$value} ];
988 ${$spot} = { %{$value} };
990 ${$spot} = bless ${$spot}, $c
999 # Copy single level of keys or elements to new DB handle.
1000 # Recurse for nested structures
1002 my $self = shift->_get_self;
1005 if ($self->_type eq TYPE_HASH) {
1006 my $key = $self->first_key();
1008 my $value = $self->get($key);
1009 $self->_copy_value( \$db_temp->{$key}, $value );
1010 $key = $self->next_key($key);
1014 my $length = $self->length();
1015 for (my $index = 0; $index < $length; $index++) {
1016 my $value = $self->get($index);
1017 $self->_copy_value( \$db_temp->[$index], $value );
1026 # Recursively export into standard Perl hashes and arrays.
1028 my $self = $_[0]->_get_self;
1031 if ($self->_type eq TYPE_HASH) { $temp = {}; }
1032 elsif ($self->_type eq TYPE_ARRAY) { $temp = []; }
1035 $self->_copy_node( $temp );
1043 # Recursively import Perl hash/array structure
1045 #XXX This use of ref() seems to be ok
1046 if (!ref($_[0])) { return; } # Perl calls import() on use -- ignore
1048 my $self = $_[0]->_get_self;
1051 #XXX This use of ref() seems to be ok
1052 if (!ref($struct)) {
1054 # struct is not a reference, so just import based on our type
1058 if ($self->_type eq TYPE_HASH) { $struct = {@_}; }
1059 elsif ($self->_type eq TYPE_ARRAY) { $struct = [@_]; }
1062 my $r = Scalar::Util::reftype($struct) || '';
1063 if ($r eq "HASH" && $self->_type eq TYPE_HASH) {
1064 foreach my $key (keys %$struct) { $self->put($key, $struct->{$key}); }
1066 elsif ($r eq "ARRAY" && $self->_type eq TYPE_ARRAY) {
1067 $self->push( @$struct );
1070 return $self->_throw_error("Cannot import: type mismatch");
1078 # Rebuild entire database into new file, then move
1079 # it back on top of original.
1081 my $self = $_[0]->_get_self;
1083 #XXX Need to create a new test for this
1084 # if ($self->_root->{links} > 1) {
1085 # return $self->_throw_error("Cannot optimize: reference count is greater than 1");
1088 my $db_temp = DBM::Deep->new(
1089 file => $self->_root->{file} . '.tmp',
1090 type => $self->_type
1093 return $self->_throw_error("Cannot optimize: failed to open temp file: $!");
1097 $self->_copy_node( $db_temp );
1101 # Attempt to copy user, group and permissions over to new file
1103 my @stats = stat($self->_fh);
1104 my $perms = $stats[2] & 07777;
1105 my $uid = $stats[4];
1106 my $gid = $stats[5];
1107 chown( $uid, $gid, $self->_root->{file} . '.tmp' );
1108 chmod( $perms, $self->_root->{file} . '.tmp' );
1110 # q.v. perlport for more information on this variable
1111 if ( $^O eq 'MSWin32' || $^O eq 'cygwin' ) {
1113 # Potential race condition when optmizing on Win32 with locking.
1114 # The Windows filesystem requires that the filehandle be closed
1115 # before it is overwritten with rename(). This could be redone
1122 if (!rename $self->_root->{file} . '.tmp', $self->_root->{file}) {
1123 unlink $self->_root->{file} . '.tmp';
1125 return $self->_throw_error("Optimize failed: Cannot copy temp file over original: $!");
1137 # Make copy of object and return
1139 my $self = $_[0]->_get_self;
1141 return DBM::Deep->new(
1142 type => $self->_type,
1143 base_offset => $self->_base_offset,
1144 root => $self->_root
1149 my %is_legal_filter = map {
1152 store_key store_value
1153 fetch_key fetch_value
1158 # Setup filter function for storing or fetching the key or value
1160 my $self = $_[0]->_get_self;
1161 my $type = lc $_[1];
1162 my $func = $_[2] ? $_[2] : undef;
1164 if ( $is_legal_filter{$type} ) {
1165 $self->_root->{"filter_$type"} = $func;
1179 # Get access to the root structure
1181 my $self = $_[0]->_get_self;
1182 return $self->{root};
1187 # Get access to the raw fh
1189 #XXX It will be useful, though, when we split out HASH and ARRAY
1190 my $self = $_[0]->_get_self;
1191 return $self->_root->{fh};
1196 # Get type of current node (TYPE_HASH or TYPE_ARRAY)
1198 my $self = $_[0]->_get_self;
1199 return $self->{type};
1204 # Get base_offset of current node (TYPE_HASH or TYPE_ARRAY)
1206 my $self = $_[0]->_get_self;
1207 return $self->{base_offset};
1212 # Get last error string, or undef if no error
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));
1340 my $ref_loc = $tag->{offset} + ($num * $LONG_SIZE);
1341 my $new_tag = $self->_index_lookup($tag, $num);
1344 seek($fh, $ref_loc + $self->_root->{file_offset}, SEEK_SET);
1345 print( $fh pack($LONG_PACK, $self->_root->{end}) );
1347 $tag = $self->_create_tag($self->_root->{end}, SIG_BLIST, chr(0) x $BUCKET_LIST_SIZE);
1349 $tag->{ref_loc} = $ref_loc;
1357 $tag->{ref_loc} = $ref_loc;
1364 # Add key/value to bucket list
1366 my $result = $self->_add_bucket( $tag, $md5, $key, $value );
1375 # Fetch single value or element given plain key or array index
1377 my $self = shift->_get_self;
1381 # Make sure file is open
1383 if (!defined($self->_fh)) { $self->_open(); }
1385 my $md5 = $DIGEST_FUNC->($key);
1388 # Request shared lock for reading
1390 $self->lock( LOCK_SH );
1392 my $tag = $self->_find_bucket_list( $md5 );
1399 # Get value from bucket list
1401 my $result = $self->_get_bucket_value( $tag, $md5 );
1405 #XXX What is ref() checking here?
1406 #YYY Filters only apply on scalar values, so the ref check is making
1407 #YYY sure the fetched bucket is a scalar, not a child hash or array.
1408 return ($result && !ref($result) && $self->_root->{filter_fetch_value})
1409 ? $self->_root->{filter_fetch_value}->($result)
1415 # Delete single key/value pair or element given plain key or array index
1417 my $self = $_[0]->_get_self;
1420 my $md5 = $DIGEST_FUNC->($key);
1423 # Make sure file is open
1425 if (!defined($self->_fh)) { $self->_open(); }
1428 # Request exclusive lock for writing
1430 $self->lock( LOCK_EX );
1432 my $tag = $self->_find_bucket_list( $md5 );
1441 my $value = $self->_get_bucket_value( $tag, $md5 );
1442 if ($value && !ref($value) && $self->_root->{filter_fetch_value}) {
1443 $value = $self->_root->{filter_fetch_value}->($value);
1446 my $result = $self->_delete_bucket( $tag, $md5 );
1449 # If this object is an array and the key deleted was on the end of the stack,
1450 # decrement the length variable.
1460 # Check if a single key or element exists given plain key or array index
1462 my $self = $_[0]->_get_self;
1465 my $md5 = $DIGEST_FUNC->($key);
1468 # Make sure file is open
1470 if (!defined($self->_fh)) { $self->_open(); }
1473 # Request shared lock for reading
1475 $self->lock( LOCK_SH );
1477 my $tag = $self->_find_bucket_list( $md5 );
1480 # For some reason, the built-in exists() function returns '' for false
1488 # Check if bucket exists and return 1 or ''
1490 my $result = $self->_bucket_exists( $tag, $md5 ) || '';
1499 # Clear all keys from hash, or all elements from array.
1501 my $self = $_[0]->_get_self;
1504 # Make sure file is open
1506 if (!defined($self->_fh)) { $self->_open(); }
1509 # Request exclusive lock for writing
1511 $self->lock( LOCK_EX );
1513 my $fh = $self->_fh;
1515 seek($fh, $self->_base_offset + $self->_root->{file_offset}, SEEK_SET);
1521 $self->_create_tag($self->_base_offset, $self->_type, chr(0) x $INDEX_SIZE);
1529 # Public method aliases
1531 sub put { (shift)->STORE( @_ ) }
1532 sub store { (shift)->STORE( @_ ) }
1533 sub get { (shift)->FETCH( @_ ) }
1534 sub fetch { (shift)->FETCH( @_ ) }
1535 sub delete { (shift)->DELETE( @_ ) }
1536 sub exists { (shift)->EXISTS( @_ ) }
1537 sub clear { (shift)->CLEAR( @_ ) }
1539 package DBM::Deep::_::Root;
1553 filter_store_key => undef,
1554 filter_store_value => undef,
1555 filter_fetch_key => undef,
1556 filter_fetch_value => undef,
1562 if ( $self->{fh} && !$self->{file_offset} ) {
1563 $self->{file_offset} = tell( $self->{fh} );
1571 return unless $self;
1573 close $self->{fh} if $self->{fh};
1584 DBM::Deep - A pure perl multi-level hash/array DBM
1589 my $db = DBM::Deep->new( "foo.db" );
1591 $db->{key} = 'value'; # tie() style
1594 $db->put('key' => 'value'); # OO style
1595 print $db->get('key');
1597 # true multi-level support
1598 $db->{my_complex} = [
1599 'hello', { perl => 'rules' },
1605 A unique flat-file database module, written in pure perl. True
1606 multi-level hash/array support (unlike MLDBM, which is faked), hybrid
1607 OO / tie() interface, cross-platform FTPable files, and quite fast. Can
1608 handle millions of keys and unlimited hash levels without significant
1609 slow-down. Written from the ground-up in pure perl -- this is NOT a
1610 wrapper around a C-based DBM. Out-of-the-box compatibility with Unix,
1611 Mac OS X and Windows.
1615 Hopefully you are using Perl's excellent CPAN module, which will download
1616 and install the module for you. If not, get the tarball, and run these
1628 Construction can be done OO-style (which is the recommended way), or using
1629 Perl's tie() function. Both are examined here.
1631 =head2 OO CONSTRUCTION
1633 The recommended way to construct a DBM::Deep object is to use the new()
1634 method, which gets you a blessed, tied hash or array reference.
1636 my $db = DBM::Deep->new( "foo.db" );
1638 This opens a new database handle, mapped to the file "foo.db". If this
1639 file does not exist, it will automatically be created. DB files are
1640 opened in "r+" (read/write) mode, and the type of object returned is a
1641 hash, unless otherwise specified (see L<OPTIONS> below).
1643 You can pass a number of options to the constructor to specify things like
1644 locking, autoflush, etc. This is done by passing an inline hash:
1646 my $db = DBM::Deep->new(
1652 Notice that the filename is now specified I<inside> the hash with
1653 the "file" parameter, as opposed to being the sole argument to the
1654 constructor. This is required if any options are specified.
1655 See L<OPTIONS> below for the complete list.
1659 You can also start with an array instead of a hash. For this, you must
1660 specify the C<type> parameter:
1662 my $db = DBM::Deep->new(
1664 type => DBM::Deep->TYPE_ARRAY
1667 B<Note:> Specifing the C<type> parameter only takes effect when beginning
1668 a new DB file. If you create a DBM::Deep object with an existing file, the
1669 C<type> will be loaded from the file header, and an error will be thrown if
1670 the wrong type is passed in.
1672 =head2 TIE CONSTRUCTION
1674 Alternately, you can create a DBM::Deep handle by using Perl's built-in
1675 tie() function. The object returned from tie() can be used to call methods,
1676 such as lock() and unlock(), but cannot be used to assign to the DBM::Deep
1677 file (as expected with most tie'd objects).
1680 my $db = tie %hash, "DBM::Deep", "foo.db";
1683 my $db = tie @array, "DBM::Deep", "bar.db";
1685 As with the OO constructor, you can replace the DB filename parameter with
1686 a hash containing one or more options (see L<OPTIONS> just below for the
1689 tie %hash, "DBM::Deep", {
1697 There are a number of options that can be passed in when constructing your
1698 DBM::Deep objects. These apply to both the OO- and tie- based approaches.
1704 Filename of the DB file to link the handle to. You can pass a full absolute
1705 filesystem path, partial path, or a plain filename if the file is in the
1706 current working directory. This is a required parameter (though q.v. fh).
1710 If you want, you can pass in the fh instead of the file. This is most useful for doing
1713 my $db = DBM::Deep->new( { fh => \*DATA } );
1715 You are responsible for making sure that the fh has been opened appropriately for your
1716 needs. If you open it read-only and attempt to write, an exception will be thrown. If you
1717 open it write-only or append-only, an exception will be thrown immediately as DBM::Deep
1718 needs to read from the fh.
1722 This is the offset within the file that the DBM::Deep db starts. Most of the time, you will
1723 not need to set this. However, it's there if you want it.
1725 If you pass in fh and do not set this, it will be set appropriately.
1729 This parameter specifies what type of object to create, a hash or array. Use
1730 one of these two constants: C<DBM::Deep-E<gt>TYPE_HASH> or C<DBM::Deep-E<gt>TYPE_ARRAY>.
1731 This only takes effect when beginning a new file. This is an optional
1732 parameter, and defaults to C<DBM::Deep-E<gt>TYPE_HASH>.
1736 Specifies whether locking is to be enabled. DBM::Deep uses Perl's Fnctl flock()
1737 function to lock the database in exclusive mode for writes, and shared mode for
1738 reads. Pass any true value to enable. This affects the base DB handle I<and
1739 any child hashes or arrays> that use the same DB file. This is an optional
1740 parameter, and defaults to 0 (disabled). See L<LOCKING> below for more.
1744 Specifies whether autoflush is to be enabled on the underlying filehandle.
1745 This obviously slows down write operations, but is required if you may have
1746 multiple processes accessing the same DB file (also consider enable I<locking>).
1747 Pass any true value to enable. This is an optional parameter, and defaults to 0
1752 If I<autobless> mode is enabled, DBM::Deep will preserve blessed hashes, and
1753 restore them when fetched. This is an B<experimental> feature, and does have
1754 side-effects. Basically, when hashes are re-blessed into their original
1755 classes, they are no longer blessed into the DBM::Deep class! So you won't be
1756 able to call any DBM::Deep methods on them. You have been warned.
1757 This is an optional parameter, and defaults to 0 (disabled).
1761 See L<FILTERS> below.
1765 Setting I<debug> mode will make all errors non-fatal, dump them out to
1766 STDERR, and continue on. This is for debugging purposes only, and probably
1767 not what you want. This is an optional parameter, and defaults to 0 (disabled).
1769 B<NOTE>: This parameter is considered deprecated and should not be used anymore.
1773 =head1 TIE INTERFACE
1775 With DBM::Deep you can access your databases using Perl's standard hash/array
1776 syntax. Because all DBM::Deep objects are I<tied> to hashes or arrays, you can
1777 treat them as such. DBM::Deep will intercept all reads/writes and direct them
1778 to the right place -- the DB file. This has nothing to do with the
1779 L<TIE CONSTRUCTION> section above. This simply tells you how to use DBM::Deep
1780 using regular hashes and arrays, rather than calling functions like C<get()>
1781 and C<put()> (although those work too). It is entirely up to you how to want
1782 to access your databases.
1786 You can treat any DBM::Deep object like a normal Perl hash reference. Add keys,
1787 or even nested hashes (or arrays) using standard Perl syntax:
1789 my $db = DBM::Deep->new( "foo.db" );
1791 $db->{mykey} = "myvalue";
1793 $db->{myhash}->{subkey} = "subvalue";
1795 print $db->{myhash}->{subkey} . "\n";
1797 You can even step through hash keys using the normal Perl C<keys()> function:
1799 foreach my $key (keys %$db) {
1800 print "$key: " . $db->{$key} . "\n";
1803 Remember that Perl's C<keys()> function extracts I<every> key from the hash and
1804 pushes them onto an array, all before the loop even begins. If you have an
1805 extra large hash, this may exhaust Perl's memory. Instead, consider using
1806 Perl's C<each()> function, which pulls keys/values one at a time, using very
1809 while (my ($key, $value) = each %$db) {
1810 print "$key: $value\n";
1813 Please note that when using C<each()>, you should always pass a direct
1814 hash reference, not a lookup. Meaning, you should B<never> do this:
1817 while (my ($key, $value) = each %{$db->{foo}}) { # BAD
1819 This causes an infinite loop, because for each iteration, Perl is calling
1820 FETCH() on the $db handle, resulting in a "new" hash for foo every time, so
1821 it effectively keeps returning the first key over and over again. Instead,
1822 assign a temporary variable to C<$db->{foo}>, then pass that to each().
1826 As with hashes, you can treat any DBM::Deep object like a normal Perl array
1827 reference. This includes inserting, removing and manipulating elements,
1828 and the C<push()>, C<pop()>, C<shift()>, C<unshift()> and C<splice()> functions.
1829 The object must have first been created using type C<DBM::Deep-E<gt>TYPE_ARRAY>,
1830 or simply be a nested array reference inside a hash. Example:
1832 my $db = DBM::Deep->new(
1833 file => "foo-array.db",
1834 type => DBM::Deep->TYPE_ARRAY
1838 push @$db, "bar", "baz";
1839 unshift @$db, "bah";
1841 my $last_elem = pop @$db; # baz
1842 my $first_elem = shift @$db; # bah
1843 my $second_elem = $db->[1]; # bar
1845 my $num_elements = scalar @$db;
1849 In addition to the I<tie()> interface, you can also use a standard OO interface
1850 to manipulate all aspects of DBM::Deep databases. Each type of object (hash or
1851 array) has its own methods, but both types share the following common methods:
1852 C<put()>, C<get()>, C<exists()>, C<delete()> and C<clear()>.
1856 =item * new() / clone()
1858 These are the constructor and copy-functions.
1860 =item * put() / store()
1862 Stores a new hash key/value pair, or sets an array element value. Takes two
1863 arguments, the hash key or array index, and the new value. The value can be
1864 a scalar, hash ref or array ref. Returns true on success, false on failure.
1866 $db->put("foo", "bar"); # for hashes
1867 $db->put(1, "bar"); # for arrays
1869 =item * get() / fetch()
1871 Fetches the value of a hash key or array element. Takes one argument: the hash
1872 key or array index. Returns a scalar, hash ref or array ref, depending on the
1875 my $value = $db->get("foo"); # for hashes
1876 my $value = $db->get(1); # for arrays
1880 Checks if a hash key or array index exists. Takes one argument: the hash key
1881 or array index. Returns true if it exists, false if not.
1883 if ($db->exists("foo")) { print "yay!\n"; } # for hashes
1884 if ($db->exists(1)) { print "yay!\n"; } # for arrays
1888 Deletes one hash key/value pair or array element. Takes one argument: the hash
1889 key or array index. Returns true on success, false if not found. For arrays,
1890 the remaining elements located after the deleted element are NOT moved over.
1891 The deleted element is essentially just undefined, which is exactly how Perl's
1892 internal arrays work. Please note that the space occupied by the deleted
1893 key/value or element is B<not> reused again -- see L<UNUSED SPACE RECOVERY>
1894 below for details and workarounds.
1896 $db->delete("foo"); # for hashes
1897 $db->delete(1); # for arrays
1901 Deletes B<all> hash keys or array elements. Takes no arguments. No return
1902 value. Please note that the space occupied by the deleted keys/values or
1903 elements is B<not> reused again -- see L<UNUSED SPACE RECOVERY> below for
1904 details and workarounds.
1906 $db->clear(); # hashes or arrays
1908 =item * lock() / unlock()
1914 Recover lost disk space.
1916 =item * import() / export()
1918 Data going in and out.
1920 =item * set_digest() / set_pack() / set_filter()
1922 q.v. adjusting the interal parameters.
1924 =item * error() / clear_error()
1926 Error handling methods. These are deprecated and will be removed in 1.00.
1932 For hashes, DBM::Deep supports all the common methods described above, and the
1933 following additional methods: C<first_key()> and C<next_key()>.
1939 Returns the "first" key in the hash. As with built-in Perl hashes, keys are
1940 fetched in an undefined order (which appears random). Takes no arguments,
1941 returns the key as a scalar value.
1943 my $key = $db->first_key();
1947 Returns the "next" key in the hash, given the previous one as the sole argument.
1948 Returns undef if there are no more keys to be fetched.
1950 $key = $db->next_key($key);
1954 Here are some examples of using hashes:
1956 my $db = DBM::Deep->new( "foo.db" );
1958 $db->put("foo", "bar");
1959 print "foo: " . $db->get("foo") . "\n";
1961 $db->put("baz", {}); # new child hash ref
1962 $db->get("baz")->put("buz", "biz");
1963 print "buz: " . $db->get("baz")->get("buz") . "\n";
1965 my $key = $db->first_key();
1967 print "$key: " . $db->get($key) . "\n";
1968 $key = $db->next_key($key);
1971 if ($db->exists("foo")) { $db->delete("foo"); }
1975 For arrays, DBM::Deep supports all the common methods described above, and the
1976 following additional methods: C<length()>, C<push()>, C<pop()>, C<shift()>,
1977 C<unshift()> and C<splice()>.
1983 Returns the number of elements in the array. Takes no arguments.
1985 my $len = $db->length();
1989 Adds one or more elements onto the end of the array. Accepts scalars, hash
1990 refs or array refs. No return value.
1992 $db->push("foo", "bar", {});
1996 Fetches the last element in the array, and deletes it. Takes no arguments.
1997 Returns undef if array is empty. Returns the element value.
1999 my $elem = $db->pop();
2003 Fetches the first element in the array, deletes it, then shifts all the
2004 remaining elements over to take up the space. Returns the element value. This
2005 method is not recommended with large arrays -- see L<LARGE ARRAYS> below for
2008 my $elem = $db->shift();
2012 Inserts one or more elements onto the beginning of the array, shifting all
2013 existing elements over to make room. Accepts scalars, hash refs or array refs.
2014 No return value. This method is not recommended with large arrays -- see
2015 <LARGE ARRAYS> below for details.
2017 $db->unshift("foo", "bar", {});
2021 Performs exactly like Perl's built-in function of the same name. See L<perldoc
2022 -f splice> for usage -- it is too complicated to document here. This method is
2023 not recommended with large arrays -- see L<LARGE ARRAYS> below for details.
2027 Here are some examples of using arrays:
2029 my $db = DBM::Deep->new(
2031 type => DBM::Deep->TYPE_ARRAY
2034 $db->push("bar", "baz");
2035 $db->unshift("foo");
2038 my $len = $db->length();
2039 print "length: $len\n"; # 4
2041 for (my $k=0; $k<$len; $k++) {
2042 print "$k: " . $db->get($k) . "\n";
2045 $db->splice(1, 2, "biz", "baf");
2047 while (my $elem = shift @$db) {
2048 print "shifted: $elem\n";
2053 Enable automatic file locking by passing a true value to the C<locking>
2054 parameter when constructing your DBM::Deep object (see L<SETUP> above).
2056 my $db = DBM::Deep->new(
2061 This causes DBM::Deep to C<flock()> the underlying filehandle with exclusive
2062 mode for writes, and shared mode for reads. This is required if you have
2063 multiple processes accessing the same database file, to avoid file corruption.
2064 Please note that C<flock()> does NOT work for files over NFS. See L<DB OVER
2065 NFS> below for more.
2067 =head2 EXPLICIT LOCKING
2069 You can explicitly lock a database, so it remains locked for multiple
2070 transactions. This is done by calling the C<lock()> method, and passing an
2071 optional lock mode argument (defaults to exclusive mode). This is particularly
2072 useful for things like counters, where the current value needs to be fetched,
2073 then incremented, then stored again.
2076 my $counter = $db->get("counter");
2078 $db->put("counter", $counter);
2087 You can pass C<lock()> an optional argument, which specifies which mode to use
2088 (exclusive or shared). Use one of these two constants: C<DBM::Deep-E<gt>LOCK_EX>
2089 or C<DBM::Deep-E<gt>LOCK_SH>. These are passed directly to C<flock()>, and are the
2090 same as the constants defined in Perl's C<Fcntl> module.
2092 $db->lock( DBM::Deep->LOCK_SH );
2096 =head1 IMPORTING/EXPORTING
2098 You can import existing complex structures by calling the C<import()> method,
2099 and export an entire database into an in-memory structure using the C<export()>
2100 method. Both are examined here.
2104 Say you have an existing hash with nested hashes/arrays inside it. Instead of
2105 walking the structure and adding keys/elements to the database as you go,
2106 simply pass a reference to the C<import()> method. This recursively adds
2107 everything to an existing DBM::Deep object for you. Here is an example:
2112 array1 => [ "elem0", "elem1", "elem2" ],
2114 subkey1 => "subvalue1",
2115 subkey2 => "subvalue2"
2119 my $db = DBM::Deep->new( "foo.db" );
2120 $db->import( $struct );
2122 print $db->{key1} . "\n"; # prints "value1"
2124 This recursively imports the entire C<$struct> object into C<$db>, including
2125 all nested hashes and arrays. If the DBM::Deep object contains exsiting data,
2126 keys are merged with the existing ones, replacing if they already exist.
2127 The C<import()> method can be called on any database level (not just the base
2128 level), and works with both hash and array DB types.
2130 B<Note:> Make sure your existing structure has no circular references in it.
2131 These will cause an infinite loop when importing.
2135 Calling the C<export()> method on an existing DBM::Deep object will return
2136 a reference to a new in-memory copy of the database. The export is done
2137 recursively, so all nested hashes/arrays are all exported to standard Perl
2138 objects. Here is an example:
2140 my $db = DBM::Deep->new( "foo.db" );
2142 $db->{key1} = "value1";
2143 $db->{key2} = "value2";
2145 $db->{hash1}->{subkey1} = "subvalue1";
2146 $db->{hash1}->{subkey2} = "subvalue2";
2148 my $struct = $db->export();
2150 print $struct->{key1} . "\n"; # prints "value1"
2152 This makes a complete copy of the database in memory, and returns a reference
2153 to it. The C<export()> method can be called on any database level (not just
2154 the base level), and works with both hash and array DB types. Be careful of
2155 large databases -- you can store a lot more data in a DBM::Deep object than an
2156 in-memory Perl structure.
2158 B<Note:> Make sure your database has no circular references in it.
2159 These will cause an infinite loop when exporting.
2163 DBM::Deep has a number of hooks where you can specify your own Perl function
2164 to perform filtering on incoming or outgoing data. This is a perfect
2165 way to extend the engine, and implement things like real-time compression or
2166 encryption. Filtering applies to the base DB level, and all child hashes /
2167 arrays. Filter hooks can be specified when your DBM::Deep object is first
2168 constructed, or by calling the C<set_filter()> method at any time. There are
2169 four available filter hooks, described below:
2173 =item * filter_store_key
2175 This filter is called whenever a hash key is stored. It
2176 is passed the incoming key, and expected to return a transformed key.
2178 =item * filter_store_value
2180 This filter is called whenever a hash key or array element is stored. It
2181 is passed the incoming value, and expected to return a transformed value.
2183 =item * filter_fetch_key
2185 This filter is called whenever a hash key is fetched (i.e. via
2186 C<first_key()> or C<next_key()>). It is passed the transformed key,
2187 and expected to return the plain key.
2189 =item * filter_fetch_value
2191 This filter is called whenever a hash key or array element is fetched.
2192 It is passed the transformed value, and expected to return the plain value.
2196 Here are the two ways to setup a filter hook:
2198 my $db = DBM::Deep->new(
2200 filter_store_value => \&my_filter_store,
2201 filter_fetch_value => \&my_filter_fetch
2206 $db->set_filter( "filter_store_value", \&my_filter_store );
2207 $db->set_filter( "filter_fetch_value", \&my_filter_fetch );
2209 Your filter function will be called only when dealing with SCALAR keys or
2210 values. When nested hashes and arrays are being stored/fetched, filtering
2211 is bypassed. Filters are called as static functions, passed a single SCALAR
2212 argument, and expected to return a single SCALAR value. If you want to
2213 remove a filter, set the function reference to C<undef>:
2215 $db->set_filter( "filter_store_value", undef );
2217 =head2 REAL-TIME ENCRYPTION EXAMPLE
2219 Here is a working example that uses the I<Crypt::Blowfish> module to
2220 do real-time encryption / decryption of keys & values with DBM::Deep Filters.
2221 Please visit L<http://search.cpan.org/search?module=Crypt::Blowfish> for more
2222 on I<Crypt::Blowfish>. You'll also need the I<Crypt::CBC> module.
2225 use Crypt::Blowfish;
2228 my $cipher = Crypt::CBC->new({
2229 'key' => 'my secret key',
2230 'cipher' => 'Blowfish',
2232 'regenerate_key' => 0,
2233 'padding' => 'space',
2237 my $db = DBM::Deep->new(
2238 file => "foo-encrypt.db",
2239 filter_store_key => \&my_encrypt,
2240 filter_store_value => \&my_encrypt,
2241 filter_fetch_key => \&my_decrypt,
2242 filter_fetch_value => \&my_decrypt,
2245 $db->{key1} = "value1";
2246 $db->{key2} = "value2";
2247 print "key1: " . $db->{key1} . "\n";
2248 print "key2: " . $db->{key2} . "\n";
2254 return $cipher->encrypt( $_[0] );
2257 return $cipher->decrypt( $_[0] );
2260 =head2 REAL-TIME COMPRESSION EXAMPLE
2262 Here is a working example that uses the I<Compress::Zlib> module to do real-time
2263 compression / decompression of keys & values with DBM::Deep Filters.
2264 Please visit L<http://search.cpan.org/search?module=Compress::Zlib> for
2265 more on I<Compress::Zlib>.
2270 my $db = DBM::Deep->new(
2271 file => "foo-compress.db",
2272 filter_store_key => \&my_compress,
2273 filter_store_value => \&my_compress,
2274 filter_fetch_key => \&my_decompress,
2275 filter_fetch_value => \&my_decompress,
2278 $db->{key1} = "value1";
2279 $db->{key2} = "value2";
2280 print "key1: " . $db->{key1} . "\n";
2281 print "key2: " . $db->{key2} . "\n";
2287 return Compress::Zlib::memGzip( $_[0] ) ;
2290 return Compress::Zlib::memGunzip( $_[0] ) ;
2293 B<Note:> Filtering of keys only applies to hashes. Array "keys" are
2294 actually numerical index numbers, and are not filtered.
2296 =head1 ERROR HANDLING
2298 Most DBM::Deep methods return a true value for success, and call die() on
2299 failure. You can wrap calls in an eval block to catch the die. Also, the
2300 actual error message is stored in an internal scalar, which can be fetched by
2301 calling the C<error()> method.
2303 my $db = DBM::Deep->new( "foo.db" ); # create hash
2304 eval { $db->push("foo"); }; # ILLEGAL -- push is array-only call
2306 print $@; # prints error message
2307 print $db->error(); # prints error message
2309 You can then call C<clear_error()> to clear the current error state.
2313 If you set the C<debug> option to true when creating your DBM::Deep object,
2314 all errors are considered NON-FATAL, and dumped to STDERR. This should only
2315 be used for debugging purposes and not production work. DBM::Deep expects errors
2316 to be thrown, not propagated back up the stack.
2318 B<NOTE>: error() and clear_error() are considered deprecated and I<will> be removed
2319 in 1.00. Please don't use them. Instead, wrap all your functions with in eval-blocks.
2321 =head1 LARGEFILE SUPPORT
2323 If you have a 64-bit system, and your Perl is compiled with both LARGEFILE
2324 and 64-bit support, you I<may> be able to create databases larger than 2 GB.
2325 DBM::Deep by default uses 32-bit file offset tags, but these can be changed
2326 by calling the static C<set_pack()> method before you do anything else.
2328 DBM::Deep::set_pack(8, 'Q');
2330 This tells DBM::Deep to pack all file offsets with 8-byte (64-bit) quad words
2331 instead of 32-bit longs. After setting these values your DB files have a
2332 theoretical maximum size of 16 XB (exabytes).
2334 B<Note:> Changing these values will B<NOT> work for existing database files.
2335 Only change this for new files, and make sure it stays set consistently
2336 throughout the file's life. If you do set these values, you can no longer
2337 access 32-bit DB files. You can, however, call C<set_pack(4, 'N')> to change
2338 back to 32-bit mode.
2340 B<Note:> I have not personally tested files > 2 GB -- all my systems have
2341 only a 32-bit Perl. However, I have received user reports that this does
2344 =head1 LOW-LEVEL ACCESS
2346 If you require low-level access to the underlying filehandle that DBM::Deep uses,
2347 you can call the C<_fh()> method, which returns the handle:
2349 my $fh = $db->_fh();
2351 This method can be called on the root level of the datbase, or any child
2352 hashes or arrays. All levels share a I<root> structure, which contains things
2353 like the filehandle, a reference counter, and all the options specified
2354 when you created the object. You can get access to this root structure by
2355 calling the C<root()> method.
2357 my $root = $db->_root();
2359 This is useful for changing options after the object has already been created,
2360 such as enabling/disabling locking, or debug modes. You can also
2361 store your own temporary user data in this structure (be wary of name
2362 collision), which is then accessible from any child hash or array.
2364 =head1 CUSTOM DIGEST ALGORITHM
2366 DBM::Deep by default uses the I<Message Digest 5> (MD5) algorithm for hashing
2367 keys. However you can override this, and use another algorithm (such as SHA-256)
2368 or even write your own. But please note that DBM::Deep currently expects zero
2369 collisions, so your algorithm has to be I<perfect>, so to speak.
2370 Collision detection may be introduced in a later version.
2374 You can specify a custom digest algorithm by calling the static C<set_digest()>
2375 function, passing a reference to a subroutine, and the length of the algorithm's
2376 hashes (in bytes). This is a global static function, which affects ALL DBM::Deep
2377 objects. Here is a working example that uses a 256-bit hash from the
2378 I<Digest::SHA256> module. Please see
2379 L<http://search.cpan.org/search?module=Digest::SHA256> for more.
2384 my $context = Digest::SHA256::new(256);
2386 DBM::Deep::set_digest( \&my_digest, 32 );
2388 my $db = DBM::Deep->new( "foo-sha.db" );
2390 $db->{key1} = "value1";
2391 $db->{key2} = "value2";
2392 print "key1: " . $db->{key1} . "\n";
2393 print "key2: " . $db->{key2} . "\n";
2399 return substr( $context->hash($_[0]), 0, 32 );
2402 B<Note:> Your returned digest strings must be B<EXACTLY> the number
2403 of bytes you specify in the C<set_digest()> function (in this case 32).
2405 =head1 CIRCULAR REFERENCES
2407 DBM::Deep has B<experimental> support for circular references. Meaning you
2408 can have a nested hash key or array element that points to a parent object.
2409 This relationship is stored in the DB file, and is preserved between sessions.
2412 my $db = DBM::Deep->new( "foo.db" );
2415 $db->{circle} = $db; # ref to self
2417 print $db->{foo} . "\n"; # prints "foo"
2418 print $db->{circle}->{foo} . "\n"; # prints "foo" again
2420 One catch is, passing the object to a function that recursively walks the
2421 object tree (such as I<Data::Dumper> or even the built-in C<optimize()> or
2422 C<export()> methods) will result in an infinite loop. The other catch is,
2423 if you fetch the I<key> of a circular reference (i.e. using the C<first_key()>
2424 or C<next_key()> methods), you will get the I<target object's key>, not the
2425 ref's key. This gets even more interesting with the above example, where
2426 the I<circle> key points to the base DB object, which technically doesn't
2427 have a key. So I made DBM::Deep return "[base]" as the key name in that
2430 =head1 CAVEATS / ISSUES / BUGS
2432 This section describes all the known issues with DBM::Deep. It you have found
2433 something that is not listed here, please send e-mail to L<jhuckaby@cpan.org>.
2435 =head2 UNUSED SPACE RECOVERY
2437 One major caveat with DBM::Deep is that space occupied by existing keys and
2438 values is not recovered when they are deleted. Meaning if you keep deleting
2439 and adding new keys, your file will continuously grow. I am working on this,
2440 but in the meantime you can call the built-in C<optimize()> method from time to
2441 time (perhaps in a crontab or something) to recover all your unused space.
2443 $db->optimize(); # returns true on success
2445 This rebuilds the ENTIRE database into a new file, then moves it on top of
2446 the original. The new file will have no unused space, thus it will take up as
2447 little disk space as possible. Please note that this operation can take
2448 a long time for large files, and you need enough disk space to temporarily hold
2449 2 copies of your DB file. The temporary file is created in the same directory
2450 as the original, named with a ".tmp" extension, and is deleted when the
2451 operation completes. Oh, and if locking is enabled, the DB is automatically
2452 locked for the entire duration of the copy.
2454 B<WARNING:> Only call optimize() on the top-level node of the database, and
2455 make sure there are no child references lying around. DBM::Deep keeps a reference
2456 counter, and if it is greater than 1, optimize() will abort and return undef.
2458 =head2 AUTOVIVIFICATION
2460 Unfortunately, autovivification doesn't work with tied hashes. This appears to
2461 be a bug in Perl's tie() system, as I<Jakob Schmidt> encountered the very same
2462 issue with his I<DWH_FIle> module (see L<http://search.cpan.org/search?module=DWH_File>),
2463 and it is also mentioned in the BUGS section for the I<MLDBM> module <see
2464 L<http://search.cpan.org/search?module=MLDBM>). Basically, on a new db file,
2467 $db->{foo}->{bar} = "hello";
2469 Since "foo" doesn't exist, you cannot add "bar" to it. You end up with "foo"
2470 being an empty hash. Try this instead, which works fine:
2472 $db->{foo} = { bar => "hello" };
2474 As of Perl 5.8.7, this bug still exists. I have walked very carefully through
2475 the execution path, and Perl indeed passes an empty hash to the STORE() method.
2476 Probably a bug in Perl.
2478 =head2 FILE CORRUPTION
2480 The current level of error handling in DBM::Deep is minimal. Files I<are> checked
2481 for a 32-bit signature when opened, but other corruption in files can cause
2482 segmentation faults. DBM::Deep may try to seek() past the end of a file, or get
2483 stuck in an infinite loop depending on the level of corruption. File write
2484 operations are not checked for failure (for speed), so if you happen to run
2485 out of disk space, DBM::Deep will probably fail in a bad way. These things will
2486 be addressed in a later version of DBM::Deep.
2490 Beware of using DB files over NFS. DBM::Deep uses flock(), which works well on local
2491 filesystems, but will NOT protect you from file corruption over NFS. I've heard
2492 about setting up your NFS server with a locking daemon, then using lockf() to
2493 lock your files, but your mileage may vary there as well. From what I
2494 understand, there is no real way to do it. However, if you need access to the
2495 underlying filehandle in DBM::Deep for using some other kind of locking scheme like
2496 lockf(), see the L<LOW-LEVEL ACCESS> section above.
2498 =head2 COPYING OBJECTS
2500 Beware of copying tied objects in Perl. Very strange things can happen.
2501 Instead, use DBM::Deep's C<clone()> method which safely copies the object and
2502 returns a new, blessed, tied hash or array to the same level in the DB.
2504 my $copy = $db->clone();
2506 B<Note>: Since clone() here is cloning the object, not the database location, any
2507 modifications to either $db or $copy will be visible in both.
2511 Beware of using C<shift()>, C<unshift()> or C<splice()> with large arrays.
2512 These functions cause every element in the array to move, which can be murder
2513 on DBM::Deep, as every element has to be fetched from disk, then stored again in
2514 a different location. This will be addressed in the forthcoming version 1.00.
2518 This section discusses DBM::Deep's speed and memory usage.
2522 Obviously, DBM::Deep isn't going to be as fast as some C-based DBMs, such as
2523 the almighty I<BerkeleyDB>. But it makes up for it in features like true
2524 multi-level hash/array support, and cross-platform FTPable files. Even so,
2525 DBM::Deep is still pretty fast, and the speed stays fairly consistent, even
2526 with huge databases. Here is some test data:
2528 Adding 1,000,000 keys to new DB file...
2530 At 100 keys, avg. speed is 2,703 keys/sec
2531 At 200 keys, avg. speed is 2,642 keys/sec
2532 At 300 keys, avg. speed is 2,598 keys/sec
2533 At 400 keys, avg. speed is 2,578 keys/sec
2534 At 500 keys, avg. speed is 2,722 keys/sec
2535 At 600 keys, avg. speed is 2,628 keys/sec
2536 At 700 keys, avg. speed is 2,700 keys/sec
2537 At 800 keys, avg. speed is 2,607 keys/sec
2538 At 900 keys, avg. speed is 2,190 keys/sec
2539 At 1,000 keys, avg. speed is 2,570 keys/sec
2540 At 2,000 keys, avg. speed is 2,417 keys/sec
2541 At 3,000 keys, avg. speed is 1,982 keys/sec
2542 At 4,000 keys, avg. speed is 1,568 keys/sec
2543 At 5,000 keys, avg. speed is 1,533 keys/sec
2544 At 6,000 keys, avg. speed is 1,787 keys/sec
2545 At 7,000 keys, avg. speed is 1,977 keys/sec
2546 At 8,000 keys, avg. speed is 2,028 keys/sec
2547 At 9,000 keys, avg. speed is 2,077 keys/sec
2548 At 10,000 keys, avg. speed is 2,031 keys/sec
2549 At 20,000 keys, avg. speed is 1,970 keys/sec
2550 At 30,000 keys, avg. speed is 2,050 keys/sec
2551 At 40,000 keys, avg. speed is 2,073 keys/sec
2552 At 50,000 keys, avg. speed is 1,973 keys/sec
2553 At 60,000 keys, avg. speed is 1,914 keys/sec
2554 At 70,000 keys, avg. speed is 2,091 keys/sec
2555 At 80,000 keys, avg. speed is 2,103 keys/sec
2556 At 90,000 keys, avg. speed is 1,886 keys/sec
2557 At 100,000 keys, avg. speed is 1,970 keys/sec
2558 At 200,000 keys, avg. speed is 2,053 keys/sec
2559 At 300,000 keys, avg. speed is 1,697 keys/sec
2560 At 400,000 keys, avg. speed is 1,838 keys/sec
2561 At 500,000 keys, avg. speed is 1,941 keys/sec
2562 At 600,000 keys, avg. speed is 1,930 keys/sec
2563 At 700,000 keys, avg. speed is 1,735 keys/sec
2564 At 800,000 keys, avg. speed is 1,795 keys/sec
2565 At 900,000 keys, avg. speed is 1,221 keys/sec
2566 At 1,000,000 keys, avg. speed is 1,077 keys/sec
2568 This test was performed on a PowerMac G4 1gHz running Mac OS X 10.3.2 & Perl
2569 5.8.1, with an 80GB Ultra ATA/100 HD spinning at 7200RPM. The hash keys and
2570 values were between 6 - 12 chars in length. The DB file ended up at 210MB.
2571 Run time was 12 min 3 sec.
2575 One of the great things about DBM::Deep is that it uses very little memory.
2576 Even with huge databases (1,000,000+ keys) you will not see much increased
2577 memory on your process. DBM::Deep relies solely on the filesystem for storing
2578 and fetching data. Here is output from I</usr/bin/top> before even opening a
2581 PID USER PRI NI SIZE RSS SHARE STAT %CPU %MEM TIME COMMAND
2582 22831 root 11 0 2716 2716 1296 R 0.0 0.2 0:07 perl
2584 Basically the process is taking 2,716K of memory. And here is the same
2585 process after storing and fetching 1,000,000 keys:
2587 PID USER PRI NI SIZE RSS SHARE STAT %CPU %MEM TIME COMMAND
2588 22831 root 14 0 2772 2772 1328 R 0.0 0.2 13:32 perl
2590 Notice the memory usage increased by only 56K. Test was performed on a 700mHz
2591 x86 box running Linux RedHat 7.2 & Perl 5.6.1.
2593 =head1 DB FILE FORMAT
2595 In case you were interested in the underlying DB file format, it is documented
2596 here in this section. You don't need to know this to use the module, it's just
2597 included for reference.
2601 DBM::Deep files always start with a 32-bit signature to identify the file type.
2602 This is at offset 0. The signature is "DPDB" in network byte order. This is
2603 checked for when the file is opened and an error will be thrown if it's not found.
2607 The DBM::Deep file is in a I<tagged format>, meaning each section of the file
2608 has a standard header containing the type of data, the length of data, and then
2609 the data itself. The type is a single character (1 byte), the length is a
2610 32-bit unsigned long in network byte order, and the data is, well, the data.
2611 Here is how it unfolds:
2615 Immediately after the 32-bit file signature is the I<Master Index> record.
2616 This is a standard tag header followed by 1024 bytes (in 32-bit mode) or 2048
2617 bytes (in 64-bit mode) of data. The type is I<H> for hash or I<A> for array,
2618 depending on how the DBM::Deep object was constructed.
2620 The index works by looking at a I<MD5 Hash> of the hash key (or array index
2621 number). The first 8-bit char of the MD5 signature is the offset into the
2622 index, multipled by 4 in 32-bit mode, or 8 in 64-bit mode. The value of the
2623 index element is a file offset of the next tag for the key/element in question,
2624 which is usually a I<Bucket List> tag (see below).
2626 The next tag I<could> be another index, depending on how many keys/elements
2627 exist. See L<RE-INDEXING> below for details.
2631 A I<Bucket List> is a collection of 16 MD5 hashes for keys/elements, plus
2632 file offsets to where the actual data is stored. It starts with a standard
2633 tag header, with type I<B>, and a data size of 320 bytes in 32-bit mode, or
2634 384 bytes in 64-bit mode. Each MD5 hash is stored in full (16 bytes), plus
2635 the 32-bit or 64-bit file offset for the I<Bucket> containing the actual data.
2636 When the list fills up, a I<Re-Index> operation is performed (See
2637 L<RE-INDEXING> below).
2641 A I<Bucket> is a tag containing a key/value pair (in hash mode), or a
2642 index/value pair (in array mode). It starts with a standard tag header with
2643 type I<D> for scalar data (string, binary, etc.), or it could be a nested
2644 hash (type I<H>) or array (type I<A>). The value comes just after the tag
2645 header. The size reported in the tag header is only for the value, but then,
2646 just after the value is another size (32-bit unsigned long) and then the plain
2647 key itself. Since the value is likely to be fetched more often than the plain
2648 key, I figured it would be I<slightly> faster to store the value first.
2650 If the type is I<H> (hash) or I<A> (array), the value is another I<Master Index>
2651 record for the nested structure, where the process begins all over again.
2655 After a I<Bucket List> grows to 16 records, its allocated space in the file is
2656 exhausted. Then, when another key/element comes in, the list is converted to a
2657 new index record. However, this index will look at the next char in the MD5
2658 hash, and arrange new Bucket List pointers accordingly. This process is called
2659 I<Re-Indexing>. Basically, a new index tag is created at the file EOF, and all
2660 17 (16 + new one) keys/elements are removed from the old Bucket List and
2661 inserted into the new index. Several new Bucket Lists are created in the
2662 process, as a new MD5 char from the key is being examined (it is unlikely that
2663 the keys will all share the same next char of their MD5s).
2665 Because of the way the I<MD5> algorithm works, it is impossible to tell exactly
2666 when the Bucket Lists will turn into indexes, but the first round tends to
2667 happen right around 4,000 keys. You will see a I<slight> decrease in
2668 performance here, but it picks back up pretty quick (see L<SPEED> above). Then
2669 it takes B<a lot> more keys to exhaust the next level of Bucket Lists. It's
2670 right around 900,000 keys. This process can continue nearly indefinitely --
2671 right up until the point the I<MD5> signatures start colliding with each other,
2672 and this is B<EXTREMELY> rare -- like winning the lottery 5 times in a row AND
2673 getting struck by lightning while you are walking to cash in your tickets.
2674 Theoretically, since I<MD5> hashes are 128-bit values, you I<could> have up to
2675 340,282,366,921,000,000,000,000,000,000,000,000,000 keys/elements (I believe
2676 this is 340 unodecillion, but don't quote me).
2680 When a new key/element is stored, the key (or index number) is first run through
2681 I<Digest::MD5> to get a 128-bit signature (example, in hex:
2682 b05783b0773d894396d475ced9d2f4f6). Then, the I<Master Index> record is checked
2683 for the first char of the signature (in this case I<b0>). If it does not exist,
2684 a new I<Bucket List> is created for our key (and the next 15 future keys that
2685 happen to also have I<b> as their first MD5 char). The entire MD5 is written
2686 to the I<Bucket List> along with the offset of the new I<Bucket> record (EOF at
2687 this point, unless we are replacing an existing I<Bucket>), where the actual
2688 data will be stored.
2692 Fetching an existing key/element involves getting a I<Digest::MD5> of the key
2693 (or index number), then walking along the indexes. If there are enough
2694 keys/elements in this DB level, there might be nested indexes, each linked to
2695 a particular char of the MD5. Finally, a I<Bucket List> is pointed to, which
2696 contains up to 16 full MD5 hashes. Each is checked for equality to the key in
2697 question. If we found a match, the I<Bucket> tag is loaded, where the value and
2698 plain key are stored.
2700 Fetching the plain key occurs when calling the I<first_key()> and I<next_key()>
2701 methods. In this process the indexes are walked systematically, and each key
2702 fetched in increasing MD5 order (which is why it appears random). Once the
2703 I<Bucket> is found, the value is skipped and the plain key returned instead.
2704 B<Note:> Do not count on keys being fetched as if the MD5 hashes were
2705 alphabetically sorted. This only happens on an index-level -- as soon as the
2706 I<Bucket Lists> are hit, the keys will come out in the order they went in --
2707 so it's pretty much undefined how the keys will come out -- just like Perl's
2710 =head1 CODE COVERAGE
2712 We use B<Devel::Cover> to test the code coverage of our tests, below is the
2713 B<Devel::Cover> report on this module's test suite.
2715 ---------------------------- ------ ------ ------ ------ ------ ------ ------
2716 File stmt bran cond sub pod time total
2717 ---------------------------- ------ ------ ------ ------ ------ ------ ------
2718 blib/lib/DBM/Deep.pm 95.0 83.2 68.7 98.2 100.0 57.8 90.7
2719 blib/lib/DBM/Deep/Array.pm 98.9 88.9 87.5 100.0 n/a 27.4 96.4
2720 blib/lib/DBM/Deep/Hash.pm 95.3 80.0 100.0 100.0 n/a 14.8 92.4
2721 Total 95.8 83.9 72.8 98.8 100.0 100.0 91.8
2722 ---------------------------- ------ ------ ------ ------ ------ ------ ------
2724 =head1 MORE INFORMATION
2726 Check out the DBM::Deep Google Group at L<http://groups.google.com/group/DBM-Deep>
2727 or send email to L<DBM-Deep@googlegroups.com>.
2731 Joseph Huckaby, L<jhuckaby@cpan.org>
2733 Rob Kinyon, L<rkinyon@cpan.org>
2735 Special thanks to Adam Sah and Rich Gaushell! You know why :-)
2739 perltie(1), Tie::Hash(3), Digest::MD5(3), Fcntl(3), flock(2), lockf(3), nfs(5),
2740 Digest::SHA256(3), Crypt::Blowfish(3), Compress::Zlib(3)
2744 Copyright (c) 2002-2006 Joseph Huckaby. All Rights Reserved.
2745 This is free software, you may use it and distribute it under the
2746 same terms as Perl itself.