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 $fh;
262 my @stats = stat($fh);
263 $self->_root->{inode} = $stats[1];
264 $self->_root->{end} = $stats[7];
270 # Check signature was valid
272 unless ($signature eq SIG_FILE) {
274 return $self->_throw_error("Signature not found -- file is not a Deep DB");
277 my @stats = stat($fh);
278 $self->_root->{inode} = $stats[1];
279 $self->_root->{end} = $stats[7];
282 # Get our type from master index signature
284 my $tag = $self->_load_tag($self->_base_offset);
286 #XXX We probably also want to store the hash algorithm name and not assume anything
287 #XXX The cool thing would be to allow a different hashing algorithm at every level
290 return $self->_throw_error("Corrupted file, no master index record");
292 if ($self->{type} ne $tag->{signature}) {
293 return $self->_throw_error("File type mismatch");
303 my $self = $_[0]->_get_self;
304 close $self->_root->{fh} if $self->_root->{fh};
305 $self->_root->{fh} = undef;
310 # Given offset, signature and content, create tag and write to disk
312 my ($self, $offset, $sig, $content) = @_;
313 my $size = length($content);
317 seek($fh, $offset + $self->_root->{file_offset}, SEEK_SET);
318 print($fh $sig . pack($DATA_LENGTH_PACK, $size) . $content );
320 if ($offset == $self->_root->{end}) {
321 $self->_root->{end} += SIG_SIZE + $DATA_LENGTH_SIZE + $size;
327 offset => $offset + SIG_SIZE + $DATA_LENGTH_SIZE,
334 # Given offset, load single tag and return signature, size and data
341 seek($fh, $offset + $self->_root->{file_offset}, SEEK_SET);
342 if (eof $fh) { return undef; }
345 read( $fh, $sig, SIG_SIZE);
348 read( $fh, $size, $DATA_LENGTH_SIZE);
349 $size = unpack($DATA_LENGTH_PACK, $size);
352 read( $fh, $buffer, $size);
357 offset => $offset + SIG_SIZE + $DATA_LENGTH_SIZE,
364 # Given index tag, lookup single entry in index and return .
367 my ($tag, $index) = @_;
369 my $location = unpack($LONG_PACK, substr($tag->{content}, $index * $LONG_SIZE, $LONG_SIZE) );
370 if (!$location) { return; }
372 return $self->_load_tag( $location );
377 # Adds one key/value pair to bucket list, given offset, MD5 digest of key,
378 # plain (undigested) key and value.
381 my ($tag, $md5, $plain_key, $value) = @_;
382 my $keys = $tag->{content};
386 # added ref() check first to avoid eval and runtime exception for every
387 # scalar value being stored. performance tweak.
388 my $is_dbm_deep = eval { local $SIG{'__DIE__'}; $value->isa( 'DBM::Deep' ) };
390 my $internal_ref = $is_dbm_deep && ($value->_root eq $self->_root);
395 # Iterate through buckets, seeing if this is a new entry or a replace.
397 for (my $i=0; $i<$MAX_BUCKETS; $i++) {
398 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
399 my $subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
402 # Found empty bucket (end of list). Populate and exit loop.
406 $location = $internal_ref
407 ? $value->_base_offset
408 : $self->_root->{end};
410 seek($fh, $tag->{offset} + ($i * $BUCKET_SIZE) + $self->_root->{file_offset}, SEEK_SET);
411 print($fh $md5 . pack($LONG_PACK, $location) );
414 elsif ($md5 eq $key) {
416 # Found existing bucket with same key. Replace with new value.
421 $location = $value->_base_offset;
422 seek($fh, $tag->{offset} + ($i * $BUCKET_SIZE) + $self->_root->{file_offset}, SEEK_SET);
423 print($fh $md5 . pack($LONG_PACK, $location) );
426 seek($fh, $subloc + SIG_SIZE + $self->_root->{file_offset}, SEEK_SET);
428 read( $fh, $size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
431 # If value is a hash, array, or raw value with equal or less size, we can
432 # reuse the same content area of the database. Otherwise, we have to create
433 # a new content area at the EOF.
436 my $r = Scalar::Util::reftype( $value ) || '';
437 if ( $r eq 'HASH' || $r eq 'ARRAY' ) {
438 $actual_length = $INDEX_SIZE;
440 # if autobless is enabled, must also take into consideration
441 # the class name, as it is stored along with key/value.
442 if ( $self->_root->{autobless} ) {
443 my $value_class = Scalar::Util::blessed($value);
444 if ( defined $value_class && $value_class ne 'DBM::Deep' ) {
445 $actual_length += length($value_class);
449 else { $actual_length = length($value); }
451 if ($actual_length <= $size) {
455 $location = $self->_root->{end};
456 seek($fh, $tag->{offset} + ($i * $BUCKET_SIZE) + $HASH_SIZE + $self->_root->{file_offset}, SEEK_SET);
457 print($fh pack($LONG_PACK, $location) );
465 # If this is an internal reference, return now.
466 # No need to write value or plain key
473 # If bucket didn't fit into list, split into a new index level
476 seek($fh, $tag->{ref_loc} + $self->_root->{file_offset}, SEEK_SET);
477 print($fh pack($LONG_PACK, $self->_root->{end}) );
479 my $index_tag = $self->_create_tag($self->_root->{end}, SIG_INDEX, chr(0) x $INDEX_SIZE);
482 $keys .= $md5 . pack($LONG_PACK, 0);
484 for (my $i=0; $i<=$MAX_BUCKETS; $i++) {
485 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
487 my $old_subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
488 my $num = ord(substr($key, $tag->{ch} + 1, 1));
490 if ($offsets[$num]) {
491 my $offset = $offsets[$num] + SIG_SIZE + $DATA_LENGTH_SIZE;
492 seek($fh, $offset + $self->_root->{file_offset}, SEEK_SET);
494 read( $fh, $subkeys, $BUCKET_LIST_SIZE);
496 for (my $k=0; $k<$MAX_BUCKETS; $k++) {
497 my $subloc = unpack($LONG_PACK, substr($subkeys, ($k * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
499 seek($fh, $offset + ($k * $BUCKET_SIZE) + $self->_root->{file_offset}, SEEK_SET);
500 print($fh $key . pack($LONG_PACK, $old_subloc || $self->_root->{end}) );
506 $offsets[$num] = $self->_root->{end};
507 seek($fh, $index_tag->{offset} + ($num * $LONG_SIZE) + $self->_root->{file_offset}, SEEK_SET);
508 print($fh pack($LONG_PACK, $self->_root->{end}) );
510 my $blist_tag = $self->_create_tag($self->_root->{end}, SIG_BLIST, chr(0) x $BUCKET_LIST_SIZE);
512 seek($fh, $blist_tag->{offset} + $self->_root->{file_offset}, SEEK_SET);
513 print($fh $key . pack($LONG_PACK, $old_subloc || $self->_root->{end}) );
518 $location ||= $self->_root->{end};
519 } # re-index bucket list
522 # Seek to content area and store signature, value and plaintext key
526 seek($fh, $location + $self->_root->{file_offset}, SEEK_SET);
529 # Write signature based on content type, set content length and write actual value.
531 my $r = Scalar::Util::reftype($value) || '';
533 print($fh TYPE_HASH );
534 print($fh pack($DATA_LENGTH_PACK, $INDEX_SIZE) . chr(0) x $INDEX_SIZE );
535 $content_length = $INDEX_SIZE;
537 elsif ($r eq 'ARRAY') {
538 print($fh TYPE_ARRAY );
539 print($fh pack($DATA_LENGTH_PACK, $INDEX_SIZE) . chr(0) x $INDEX_SIZE );
540 $content_length = $INDEX_SIZE;
542 elsif (!defined($value)) {
543 print($fh SIG_NULL );
544 print($fh pack($DATA_LENGTH_PACK, 0) );
548 print($fh SIG_DATA );
549 print($fh pack($DATA_LENGTH_PACK, length($value)) . $value );
550 $content_length = length($value);
554 # Plain key is stored AFTER value, as keys are typically fetched less often.
556 print($fh pack($DATA_LENGTH_PACK, length($plain_key)) . $plain_key );
559 # If value is blessed, preserve class name
561 if ( $self->_root->{autobless} ) {
562 my $value_class = Scalar::Util::blessed($value);
563 if ( defined $value_class && $value_class ne 'DBM::Deep' ) {
565 # Blessed ref -- will restore later
568 print($fh pack($DATA_LENGTH_PACK, length($value_class)) . $value_class );
569 $content_length += 1;
570 $content_length += $DATA_LENGTH_SIZE + length($value_class);
574 $content_length += 1;
579 # If this is a new content area, advance EOF counter
581 if ($location == $self->_root->{end}) {
582 $self->_root->{end} += SIG_SIZE;
583 $self->_root->{end} += $DATA_LENGTH_SIZE + $content_length;
584 $self->_root->{end} += $DATA_LENGTH_SIZE + length($plain_key);
588 # If content is a hash or array, create new child DBM::Deep object and
589 # pass each key or element to it.
592 my $branch = DBM::Deep->new(
594 base_offset => $location,
595 root => $self->_root,
597 foreach my $key (keys %{$value}) {
598 $branch->STORE( $key, $value->{$key} );
601 elsif ($r eq 'ARRAY') {
602 my $branch = DBM::Deep->new(
604 base_offset => $location,
605 root => $self->_root,
608 foreach my $element (@{$value}) {
609 $branch->STORE( $index, $element );
617 return $self->_throw_error("Fatal error: indexing failed -- possibly due to corruption in file");
620 sub _get_bucket_value {
622 # Fetch single value given tag and MD5 digested key.
625 my ($tag, $md5) = @_;
626 my $keys = $tag->{content};
631 # Iterate through buckets, looking for a key match
634 for (my $i=0; $i<$MAX_BUCKETS; $i++) {
635 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
636 my $subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
640 # Hit end of list, no match
645 if ( $md5 ne $key ) {
650 # Found match -- seek to offset and read signature
653 seek($fh, $subloc + $self->_root->{file_offset}, SEEK_SET);
654 read( $fh, $signature, SIG_SIZE);
657 # If value is a hash or array, return new DBM::Deep object with correct offset
659 if (($signature eq TYPE_HASH) || ($signature eq TYPE_ARRAY)) {
660 my $obj = DBM::Deep->new(
662 base_offset => $subloc,
666 if ($self->_root->{autobless}) {
668 # Skip over value and plain key to see if object needs
671 seek($fh, $DATA_LENGTH_SIZE + $INDEX_SIZE, SEEK_CUR);
674 read( $fh, $size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
675 if ($size) { seek($fh, $size, SEEK_CUR); }
678 read( $fh, $bless_bit, 1);
679 if (ord($bless_bit)) {
681 # Yes, object needs to be re-blessed
684 read( $fh, $size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
685 if ($size) { read( $fh, $class_name, $size); }
686 if ($class_name) { $obj = bless( $obj, $class_name ); }
694 # Otherwise return actual value
696 elsif ($signature eq SIG_DATA) {
699 read( $fh, $size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
700 if ($size) { read( $fh, $value, $size); }
705 # Key exists, but content is null
715 # Delete single key/value pair given tag and MD5 digested key.
718 my ($tag, $md5) = @_;
719 my $keys = $tag->{content};
724 # Iterate through buckets, looking for a key match
727 for (my $i=0; $i<$MAX_BUCKETS; $i++) {
728 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
729 my $subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
733 # Hit end of list, no match
738 if ( $md5 ne $key ) {
743 # Matched key -- delete bucket and return
745 seek($fh, $tag->{offset} + ($i * $BUCKET_SIZE) + $self->_root->{file_offset}, SEEK_SET);
746 print($fh substr($keys, ($i+1) * $BUCKET_SIZE ) );
747 print($fh chr(0) x $BUCKET_SIZE );
757 # Check existence of single key given tag and MD5 digested key.
760 my ($tag, $md5) = @_;
761 my $keys = $tag->{content};
764 # Iterate through buckets, looking for a key match
767 for (my $i=0; $i<$MAX_BUCKETS; $i++) {
768 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
769 my $subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
773 # Hit end of list, no match
778 if ( $md5 ne $key ) {
783 # Matched key -- return true
791 sub _find_bucket_list {
793 # Locate offset for bucket list, given digested key
799 # Locate offset for bucket list using digest index system
802 my $tag = $self->_load_tag($self->_base_offset);
803 if (!$tag) { return; }
805 while ($tag->{signature} ne SIG_BLIST) {
806 $tag = $self->_index_lookup($tag, ord(substr($md5, $ch, 1)));
807 if (!$tag) { return; }
814 sub _traverse_index {
816 # Scan index and recursively step into deeper levels, looking for next key.
818 my ($self, $offset, $ch, $force_return_next) = @_;
819 $force_return_next = undef unless $force_return_next;
821 my $tag = $self->_load_tag( $offset );
825 if ($tag->{signature} ne SIG_BLIST) {
826 my $content = $tag->{content};
828 if ($self->{return_next}) { $start = 0; }
829 else { $start = ord(substr($self->{prev_md5}, $ch, 1)); }
831 for (my $index = $start; $index < 256; $index++) {
832 my $subloc = unpack($LONG_PACK, substr($content, $index * $LONG_SIZE, $LONG_SIZE) );
834 my $result = $self->_traverse_index( $subloc, $ch + 1, $force_return_next );
835 if (defined($result)) { return $result; }
839 $self->{return_next} = 1;
842 elsif ($tag->{signature} eq SIG_BLIST) {
843 my $keys = $tag->{content};
844 if ($force_return_next) { $self->{return_next} = 1; }
847 # Iterate through buckets, looking for a key match
849 for (my $i=0; $i<$MAX_BUCKETS; $i++) {
850 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
851 my $subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
855 # End of bucket list -- return to outer loop
857 $self->{return_next} = 1;
860 elsif ($key eq $self->{prev_md5}) {
862 # Located previous key -- return next one found
864 $self->{return_next} = 1;
867 elsif ($self->{return_next}) {
869 # Seek to bucket location and skip over signature
871 seek($fh, $subloc + SIG_SIZE + $self->_root->{file_offset}, SEEK_SET);
874 # Skip over value to get to plain key
877 read( $fh, $size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
878 if ($size) { seek($fh, $size, SEEK_CUR); }
881 # Read in plain key and return as scalar
884 read( $fh, $size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
885 if ($size) { read( $fh, $plain_key, $size); }
891 $self->{return_next} = 1;
892 } # tag is a bucket list
899 # Locate next key, given digested previous one
901 my $self = $_[0]->_get_self;
903 $self->{prev_md5} = $_[1] ? $_[1] : undef;
904 $self->{return_next} = 0;
907 # If the previous key was not specifed, start at the top and
908 # return the first one found.
910 if (!$self->{prev_md5}) {
911 $self->{prev_md5} = chr(0) x $HASH_SIZE;
912 $self->{return_next} = 1;
915 return $self->_traverse_index( $self->_base_offset, 0 );
920 # If db locking is set, flock() the db file. If called multiple
921 # times before unlock(), then the same number of unlocks() must
922 # be called before the lock is released.
924 my $self = $_[0]->_get_self;
926 $type = LOCK_EX unless defined $type;
928 if (!defined($self->_fh)) { return; }
930 if ($self->_root->{locking}) {
931 if (!$self->_root->{locked}) {
932 flock($self->_fh, $type);
934 # refresh end counter in case file has changed size
935 my @stats = stat($self->_root->{file});
936 $self->_root->{end} = $stats[7];
938 # double-check file inode, in case another process
939 # has optimize()d our file while we were waiting.
940 if ($stats[1] != $self->_root->{inode}) {
941 $self->_open(); # re-open
942 flock($self->_fh, $type); # re-lock
943 $self->_root->{end} = (stat($self->_fh))[7]; # re-end
946 $self->_root->{locked}++;
956 # If db locking is set, unlock the db file. See note in lock()
957 # regarding calling lock() multiple times.
959 my $self = $_[0]->_get_self;
961 if (!defined($self->_fh)) { return; }
963 if ($self->_root->{locking} && $self->_root->{locked} > 0) {
964 $self->_root->{locked}--;
965 if (!$self->_root->{locked}) { flock($self->_fh, LOCK_UN); }
974 my $self = shift->_get_self;
975 my ($spot, $value) = @_;
980 elsif ( eval { local $SIG{__DIE__}; $value->isa( 'DBM::Deep' ) } ) {
981 my $type = $value->_type;
982 ${$spot} = $type eq TYPE_HASH ? {} : [];
983 $value->_copy_node( ${$spot} );
986 my $r = Scalar::Util::reftype( $value );
987 my $c = Scalar::Util::blessed( $value );
988 if ( $r eq 'ARRAY' ) {
989 ${$spot} = [ @{$value} ];
992 ${$spot} = { %{$value} };
994 ${$spot} = bless ${$spot}, $c
1003 # Copy single level of keys or elements to new DB handle.
1004 # Recurse for nested structures
1006 my $self = shift->_get_self;
1009 if ($self->_type eq TYPE_HASH) {
1010 my $key = $self->first_key();
1012 my $value = $self->get($key);
1013 $self->_copy_value( \$db_temp->{$key}, $value );
1014 $key = $self->next_key($key);
1018 my $length = $self->length();
1019 for (my $index = 0; $index < $length; $index++) {
1020 my $value = $self->get($index);
1021 $self->_copy_value( \$db_temp->[$index], $value );
1030 # Recursively export into standard Perl hashes and arrays.
1032 my $self = $_[0]->_get_self;
1035 if ($self->_type eq TYPE_HASH) { $temp = {}; }
1036 elsif ($self->_type eq TYPE_ARRAY) { $temp = []; }
1039 $self->_copy_node( $temp );
1047 # Recursively import Perl hash/array structure
1049 #XXX This use of ref() seems to be ok
1050 if (!ref($_[0])) { return; } # Perl calls import() on use -- ignore
1052 my $self = $_[0]->_get_self;
1055 #XXX This use of ref() seems to be ok
1056 if (!ref($struct)) {
1058 # struct is not a reference, so just import based on our type
1062 if ($self->_type eq TYPE_HASH) { $struct = {@_}; }
1063 elsif ($self->_type eq TYPE_ARRAY) { $struct = [@_]; }
1066 my $r = Scalar::Util::reftype($struct) || '';
1067 if ($r eq "HASH" && $self->_type eq TYPE_HASH) {
1068 foreach my $key (keys %$struct) { $self->put($key, $struct->{$key}); }
1070 elsif ($r eq "ARRAY" && $self->_type eq TYPE_ARRAY) {
1071 $self->push( @$struct );
1074 return $self->_throw_error("Cannot import: type mismatch");
1082 # Rebuild entire database into new file, then move
1083 # it back on top of original.
1085 my $self = $_[0]->_get_self;
1087 #XXX Need to create a new test for this
1088 # if ($self->_root->{links} > 1) {
1089 # return $self->_throw_error("Cannot optimize: reference count is greater than 1");
1092 my $db_temp = DBM::Deep->new(
1093 file => $self->_root->{file} . '.tmp',
1094 type => $self->_type
1097 return $self->_throw_error("Cannot optimize: failed to open temp file: $!");
1101 $self->_copy_node( $db_temp );
1105 # Attempt to copy user, group and permissions over to new file
1107 my @stats = stat($self->_fh);
1108 my $perms = $stats[2] & 07777;
1109 my $uid = $stats[4];
1110 my $gid = $stats[5];
1111 chown( $uid, $gid, $self->_root->{file} . '.tmp' );
1112 chmod( $perms, $self->_root->{file} . '.tmp' );
1114 # q.v. perlport for more information on this variable
1115 if ( $^O eq 'MSWin32' || $^O eq 'cygwin' ) {
1117 # Potential race condition when optmizing on Win32 with locking.
1118 # The Windows filesystem requires that the filehandle be closed
1119 # before it is overwritten with rename(). This could be redone
1126 if (!rename $self->_root->{file} . '.tmp', $self->_root->{file}) {
1127 unlink $self->_root->{file} . '.tmp';
1129 return $self->_throw_error("Optimize failed: Cannot copy temp file over original: $!");
1141 # Make copy of object and return
1143 my $self = $_[0]->_get_self;
1145 return DBM::Deep->new(
1146 type => $self->_type,
1147 base_offset => $self->_base_offset,
1148 root => $self->_root
1153 my %is_legal_filter = map {
1156 store_key store_value
1157 fetch_key fetch_value
1162 # Setup filter function for storing or fetching the key or value
1164 my $self = $_[0]->_get_self;
1165 my $type = lc $_[1];
1166 my $func = $_[2] ? $_[2] : undef;
1168 if ( $is_legal_filter{$type} ) {
1169 $self->_root->{"filter_$type"} = $func;
1183 # Get access to the root structure
1185 my $self = $_[0]->_get_self;
1186 return $self->{root};
1191 # Get access to the raw fh
1193 #XXX It will be useful, though, when we split out HASH and ARRAY
1194 my $self = $_[0]->_get_self;
1195 return $self->_root->{fh};
1200 # Get type of current node (TYPE_HASH or TYPE_ARRAY)
1202 my $self = $_[0]->_get_self;
1203 return $self->{type};
1208 # Get base_offset of current node (TYPE_HASH or TYPE_ARRAY)
1210 my $self = $_[0]->_get_self;
1211 return $self->{base_offset};
1216 # Get last error string, or undef if no error
1219 #? ( _get_self($_[0])->{root}->{error} or undef )
1220 ? ( $_[0]->_get_self->{root}->{error} or undef )
1230 # Store error string in self
1232 my $error_text = $_[1];
1234 if ( Scalar::Util::blessed $_[0] ) {
1235 my $self = $_[0]->_get_self;
1236 $self->_root->{error} = $error_text;
1238 unless ($self->_root->{debug}) {
1239 die "DBM::Deep: $error_text\n";
1242 warn "DBM::Deep: $error_text\n";
1246 die "DBM::Deep: $error_text\n";
1254 my $self = $_[0]->_get_self;
1256 undef $self->_root->{error};
1259 sub _precalc_sizes {
1261 # Precalculate index, bucket and bucket list sizes
1264 #XXX I don't like this ...
1265 set_pack() unless defined $LONG_SIZE;
1267 $INDEX_SIZE = 256 * $LONG_SIZE;
1268 $BUCKET_SIZE = $HASH_SIZE + $LONG_SIZE;
1269 $BUCKET_LIST_SIZE = $MAX_BUCKETS * $BUCKET_SIZE;
1274 # Set pack/unpack modes (see file header for more)
1276 my ($long_s, $long_p, $data_s, $data_p) = @_;
1278 $LONG_SIZE = $long_s ? $long_s : 4;
1279 $LONG_PACK = $long_p ? $long_p : 'N';
1281 $DATA_LENGTH_SIZE = $data_s ? $data_s : 4;
1282 $DATA_LENGTH_PACK = $data_p ? $data_p : 'N';
1289 # Set key digest function (default is MD5)
1291 my ($digest_func, $hash_size) = @_;
1293 $DIGEST_FUNC = $digest_func ? $digest_func : \&Digest::MD5::md5;
1294 $HASH_SIZE = $hash_size ? $hash_size : 16;
1300 # tie() methods (hashes and arrays)
1305 # Store single hash key/value or array element in database.
1307 my $self = $_[0]->_get_self;
1310 # User may be storing a hash, in which case we do not want it run
1311 # through the filtering system
1312 my $value = ($self->_root->{filter_store_value} && !ref($_[2]))
1313 ? $self->_root->{filter_store_value}->($_[2])
1316 my $md5 = $DIGEST_FUNC->($key);
1319 # Make sure file is open
1321 if (!defined($self->_fh) && !$self->_open()) {
1327 # Request exclusive lock for writing
1329 $self->lock( LOCK_EX );
1331 my $fh = $self->_fh;
1334 # Locate offset for bucket list using digest index system
1336 my $tag = $self->_load_tag($self->_base_offset);
1338 $tag = $self->_create_tag($self->_base_offset, SIG_INDEX, chr(0) x $INDEX_SIZE);
1342 while ($tag->{signature} ne SIG_BLIST) {
1343 my $num = ord(substr($md5, $ch, 1));
1344 my $new_tag = $self->_index_lookup($tag, $num);
1346 my $ref_loc = $tag->{offset} + ($num * $LONG_SIZE);
1347 seek($fh, $ref_loc + $self->_root->{file_offset}, SEEK_SET);
1348 print($fh pack($LONG_PACK, $self->_root->{end}) );
1350 $tag = $self->_create_tag($self->_root->{end}, SIG_BLIST, chr(0) x $BUCKET_LIST_SIZE);
1351 $tag->{ref_loc} = $ref_loc;
1356 my $ref_loc = $tag->{offset} + ($num * $LONG_SIZE);
1358 $tag->{ref_loc} = $ref_loc;
1365 # Add key/value to bucket list
1367 my $result = $self->_add_bucket( $tag, $md5, $key, $value );
1376 # Fetch single value or element given plain key or array index
1378 my $self = shift->_get_self;
1382 # Make sure file is open
1384 if (!defined($self->_fh)) { $self->_open(); }
1386 my $md5 = $DIGEST_FUNC->($key);
1389 # Request shared lock for reading
1391 $self->lock( LOCK_SH );
1393 my $tag = $self->_find_bucket_list( $md5 );
1400 # Get value from bucket list
1402 my $result = $self->_get_bucket_value( $tag, $md5 );
1406 #XXX What is ref() checking here?
1407 #YYY Filters only apply on scalar values, so the ref check is making
1408 #YYY sure the fetched bucket is a scalar, not a child hash or array.
1409 return ($result && !ref($result) && $self->_root->{filter_fetch_value})
1410 ? $self->_root->{filter_fetch_value}->($result)
1416 # Delete single key/value pair or element given plain key or array index
1418 my $self = $_[0]->_get_self;
1421 my $md5 = $DIGEST_FUNC->($key);
1424 # Make sure file is open
1426 if (!defined($self->_fh)) { $self->_open(); }
1429 # Request exclusive lock for writing
1431 $self->lock( LOCK_EX );
1433 my $tag = $self->_find_bucket_list( $md5 );
1442 my $value = $self->_get_bucket_value( $tag, $md5 );
1443 if ($value && !ref($value) && $self->_root->{filter_fetch_value}) {
1444 $value = $self->_root->{filter_fetch_value}->($value);
1447 my $result = $self->_delete_bucket( $tag, $md5 );
1450 # If this object is an array and the key deleted was on the end of the stack,
1451 # decrement the length variable.
1461 # Check if a single key or element exists given plain key or array index
1463 my $self = $_[0]->_get_self;
1466 my $md5 = $DIGEST_FUNC->($key);
1469 # Make sure file is open
1471 if (!defined($self->_fh)) { $self->_open(); }
1474 # Request shared lock for reading
1476 $self->lock( LOCK_SH );
1478 my $tag = $self->_find_bucket_list( $md5 );
1481 # For some reason, the built-in exists() function returns '' for false
1489 # Check if bucket exists and return 1 or ''
1491 my $result = $self->_bucket_exists( $tag, $md5 ) || '';
1500 # Clear all keys from hash, or all elements from array.
1502 my $self = $_[0]->_get_self;
1505 # Make sure file is open
1507 if (!defined($self->_fh)) { $self->_open(); }
1510 # Request exclusive lock for writing
1512 $self->lock( LOCK_EX );
1514 my $fh = $self->_fh;
1516 seek($fh, $self->_base_offset + $self->_root->{file_offset}, SEEK_SET);
1522 $self->_create_tag($self->_base_offset, $self->_type, chr(0) x $INDEX_SIZE);
1530 # Public method aliases
1532 sub put { (shift)->STORE( @_ ) }
1533 sub store { (shift)->STORE( @_ ) }
1534 sub get { (shift)->FETCH( @_ ) }
1535 sub fetch { (shift)->FETCH( @_ ) }
1536 sub delete { (shift)->DELETE( @_ ) }
1537 sub exists { (shift)->EXISTS( @_ ) }
1538 sub clear { (shift)->CLEAR( @_ ) }
1540 package DBM::Deep::_::Root;
1554 filter_store_key => undef,
1555 filter_store_value => undef,
1556 filter_fetch_key => undef,
1557 filter_fetch_value => undef,
1563 if ( $self->{fh} && !$self->{file_offset} ) {
1564 $self->{file_offset} = tell( $self->{fh} );
1572 return unless $self;
1574 close $self->{fh} if $self->{fh};
1585 DBM::Deep - A pure perl multi-level hash/array DBM
1590 my $db = DBM::Deep->new( "foo.db" );
1592 $db->{key} = 'value'; # tie() style
1595 $db->put('key' => 'value'); # OO style
1596 print $db->get('key');
1598 # true multi-level support
1599 $db->{my_complex} = [
1600 'hello', { perl => 'rules' },
1606 A unique flat-file database module, written in pure perl. True
1607 multi-level hash/array support (unlike MLDBM, which is faked), hybrid
1608 OO / tie() interface, cross-platform FTPable files, and quite fast. Can
1609 handle millions of keys and unlimited hash levels without significant
1610 slow-down. Written from the ground-up in pure perl -- this is NOT a
1611 wrapper around a C-based DBM. Out-of-the-box compatibility with Unix,
1612 Mac OS X and Windows.
1616 Hopefully you are using Perl's excellent CPAN module, which will download
1617 and install the module for you. If not, get the tarball, and run these
1629 Construction can be done OO-style (which is the recommended way), or using
1630 Perl's tie() function. Both are examined here.
1632 =head2 OO CONSTRUCTION
1634 The recommended way to construct a DBM::Deep object is to use the new()
1635 method, which gets you a blessed, tied hash or array reference.
1637 my $db = DBM::Deep->new( "foo.db" );
1639 This opens a new database handle, mapped to the file "foo.db". If this
1640 file does not exist, it will automatically be created. DB files are
1641 opened in "r+" (read/write) mode, and the type of object returned is a
1642 hash, unless otherwise specified (see L<OPTIONS> below).
1644 You can pass a number of options to the constructor to specify things like
1645 locking, autoflush, etc. This is done by passing an inline hash:
1647 my $db = DBM::Deep->new(
1653 Notice that the filename is now specified I<inside> the hash with
1654 the "file" parameter, as opposed to being the sole argument to the
1655 constructor. This is required if any options are specified.
1656 See L<OPTIONS> below for the complete list.
1660 You can also start with an array instead of a hash. For this, you must
1661 specify the C<type> parameter:
1663 my $db = DBM::Deep->new(
1665 type => DBM::Deep->TYPE_ARRAY
1668 B<Note:> Specifing the C<type> parameter only takes effect when beginning
1669 a new DB file. If you create a DBM::Deep object with an existing file, the
1670 C<type> will be loaded from the file header, and an error will be thrown if
1671 the wrong type is passed in.
1673 =head2 TIE CONSTRUCTION
1675 Alternately, you can create a DBM::Deep handle by using Perl's built-in
1676 tie() function. The object returned from tie() can be used to call methods,
1677 such as lock() and unlock(), but cannot be used to assign to the DBM::Deep
1678 file (as expected with most tie'd objects).
1681 my $db = tie %hash, "DBM::Deep", "foo.db";
1684 my $db = tie @array, "DBM::Deep", "bar.db";
1686 As with the OO constructor, you can replace the DB filename parameter with
1687 a hash containing one or more options (see L<OPTIONS> just below for the
1690 tie %hash, "DBM::Deep", {
1698 There are a number of options that can be passed in when constructing your
1699 DBM::Deep objects. These apply to both the OO- and tie- based approaches.
1705 Filename of the DB file to link the handle to. You can pass a full absolute
1706 filesystem path, partial path, or a plain filename if the file is in the
1707 current working directory. This is a required parameter (though q.v. fh).
1711 If you want, you can pass in the fh instead of the file. This is most useful for doing
1714 my $db = DBM::Deep->new( { fh => \*DATA } );
1716 You are responsible for making sure that the fh has been opened appropriately for your
1717 needs. If you open it read-only and attempt to write, an exception will be thrown. If you
1718 open it write-only or append-only, an exception will be thrown immediately as DBM::Deep
1719 needs to read from the fh.
1723 This is the offset within the file that the DBM::Deep db starts. Most of the time, you will
1724 not need to set this. However, it's there if you want it.
1726 If you pass in fh and do not set this, it will be set appropriately.
1730 This parameter specifies what type of object to create, a hash or array. Use
1731 one of these two constants: C<DBM::Deep-E<gt>TYPE_HASH> or C<DBM::Deep-E<gt>TYPE_ARRAY>.
1732 This only takes effect when beginning a new file. This is an optional
1733 parameter, and defaults to C<DBM::Deep-E<gt>TYPE_HASH>.
1737 Specifies whether locking is to be enabled. DBM::Deep uses Perl's Fnctl flock()
1738 function to lock the database in exclusive mode for writes, and shared mode for
1739 reads. Pass any true value to enable. This affects the base DB handle I<and
1740 any child hashes or arrays> that use the same DB file. This is an optional
1741 parameter, and defaults to 0 (disabled). See L<LOCKING> below for more.
1745 Specifies whether autoflush is to be enabled on the underlying filehandle.
1746 This obviously slows down write operations, but is required if you may have
1747 multiple processes accessing the same DB file (also consider enable I<locking>).
1748 Pass any true value to enable. This is an optional parameter, and defaults to 0
1753 If I<autobless> mode is enabled, DBM::Deep will preserve blessed hashes, and
1754 restore them when fetched. This is an B<experimental> feature, and does have
1755 side-effects. Basically, when hashes are re-blessed into their original
1756 classes, they are no longer blessed into the DBM::Deep class! So you won't be
1757 able to call any DBM::Deep methods on them. You have been warned.
1758 This is an optional parameter, and defaults to 0 (disabled).
1762 See L<FILTERS> below.
1766 Setting I<debug> mode will make all errors non-fatal, dump them out to
1767 STDERR, and continue on. This is for debugging purposes only, and probably
1768 not what you want. This is an optional parameter, and defaults to 0 (disabled).
1772 Instead of passing a file path, you can instead pass a handle to an pre-opened
1773 filehandle. Note: Beware of using the magick *DATA handle, as this actually
1774 contains your entire Perl script, as well as the data following the __DATA__
1775 marker. This will not work, because DBM::Deep uses absolute seek()s into the
1776 file. Instead, consider reading *DATA into an IO::Scalar handle, then passing
1777 in that. Also please note optimize() will NOT work when passing in only a
1778 handle. Pass in a real filename in order to use optimize().
1782 =head1 TIE INTERFACE
1784 With DBM::Deep you can access your databases using Perl's standard hash/array
1785 syntax. Because all DBM::Deep objects are I<tied> to hashes or arrays, you can
1786 treat them as such. DBM::Deep will intercept all reads/writes and direct them
1787 to the right place -- the DB file. This has nothing to do with the
1788 L<TIE CONSTRUCTION> section above. This simply tells you how to use DBM::Deep
1789 using regular hashes and arrays, rather than calling functions like C<get()>
1790 and C<put()> (although those work too). It is entirely up to you how to want
1791 to access your databases.
1795 You can treat any DBM::Deep object like a normal Perl hash reference. Add keys,
1796 or even nested hashes (or arrays) using standard Perl syntax:
1798 my $db = DBM::Deep->new( "foo.db" );
1800 $db->{mykey} = "myvalue";
1802 $db->{myhash}->{subkey} = "subvalue";
1804 print $db->{myhash}->{subkey} . "\n";
1806 You can even step through hash keys using the normal Perl C<keys()> function:
1808 foreach my $key (keys %$db) {
1809 print "$key: " . $db->{$key} . "\n";
1812 Remember that Perl's C<keys()> function extracts I<every> key from the hash and
1813 pushes them onto an array, all before the loop even begins. If you have an
1814 extra large hash, this may exhaust Perl's memory. Instead, consider using
1815 Perl's C<each()> function, which pulls keys/values one at a time, using very
1818 while (my ($key, $value) = each %$db) {
1819 print "$key: $value\n";
1822 Please note that when using C<each()>, you should always pass a direct
1823 hash reference, not a lookup. Meaning, you should B<never> do this:
1826 while (my ($key, $value) = each %{$db->{foo}}) { # BAD
1828 This causes an infinite loop, because for each iteration, Perl is calling
1829 FETCH() on the $db handle, resulting in a "new" hash for foo every time, so
1830 it effectively keeps returning the first key over and over again. Instead,
1831 assign a temporary variable to C<$db->{foo}>, then pass that to each().
1835 As with hashes, you can treat any DBM::Deep object like a normal Perl array
1836 reference. This includes inserting, removing and manipulating elements,
1837 and the C<push()>, C<pop()>, C<shift()>, C<unshift()> and C<splice()> functions.
1838 The object must have first been created using type C<DBM::Deep-E<gt>TYPE_ARRAY>,
1839 or simply be a nested array reference inside a hash. Example:
1841 my $db = DBM::Deep->new(
1842 file => "foo-array.db",
1843 type => DBM::Deep->TYPE_ARRAY
1847 push @$db, "bar", "baz";
1848 unshift @$db, "bah";
1850 my $last_elem = pop @$db; # baz
1851 my $first_elem = shift @$db; # bah
1852 my $second_elem = $db->[1]; # bar
1854 my $num_elements = scalar @$db;
1858 In addition to the I<tie()> interface, you can also use a standard OO interface
1859 to manipulate all aspects of DBM::Deep databases. Each type of object (hash or
1860 array) has its own methods, but both types share the following common methods:
1861 C<put()>, C<get()>, C<exists()>, C<delete()> and C<clear()>.
1865 =item * new() / clone()
1867 These are the constructor and copy-functions.
1869 =item * put() / store()
1871 Stores a new hash key/value pair, or sets an array element value. Takes two
1872 arguments, the hash key or array index, and the new value. The value can be
1873 a scalar, hash ref or array ref. Returns true on success, false on failure.
1875 $db->put("foo", "bar"); # for hashes
1876 $db->put(1, "bar"); # for arrays
1878 =item * get() / fetch()
1880 Fetches the value of a hash key or array element. Takes one argument: the hash
1881 key or array index. Returns a scalar, hash ref or array ref, depending on the
1884 my $value = $db->get("foo"); # for hashes
1885 my $value = $db->get(1); # for arrays
1889 Checks if a hash key or array index exists. Takes one argument: the hash key
1890 or array index. Returns true if it exists, false if not.
1892 if ($db->exists("foo")) { print "yay!\n"; } # for hashes
1893 if ($db->exists(1)) { print "yay!\n"; } # for arrays
1897 Deletes one hash key/value pair or array element. Takes one argument: the hash
1898 key or array index. Returns true on success, false if not found. For arrays,
1899 the remaining elements located after the deleted element are NOT moved over.
1900 The deleted element is essentially just undefined, which is exactly how Perl's
1901 internal arrays work. Please note that the space occupied by the deleted
1902 key/value or element is B<not> reused again -- see L<UNUSED SPACE RECOVERY>
1903 below for details and workarounds.
1905 $db->delete("foo"); # for hashes
1906 $db->delete(1); # for arrays
1910 Deletes B<all> hash keys or array elements. Takes no arguments. No return
1911 value. Please note that the space occupied by the deleted keys/values or
1912 elements is B<not> reused again -- see L<UNUSED SPACE RECOVERY> below for
1913 details and workarounds.
1915 $db->clear(); # hashes or arrays
1917 =item * lock() / unlock()
1923 Recover lost disk space.
1925 =item * import() / export()
1927 Data going in and out.
1929 =item * set_digest() / set_pack() / set_filter()
1931 q.v. adjusting the interal parameters.
1933 =item * error() / clear_error()
1935 Error handling methods (may be deprecated).
1941 For hashes, DBM::Deep supports all the common methods described above, and the
1942 following additional methods: C<first_key()> and C<next_key()>.
1948 Returns the "first" key in the hash. As with built-in Perl hashes, keys are
1949 fetched in an undefined order (which appears random). Takes no arguments,
1950 returns the key as a scalar value.
1952 my $key = $db->first_key();
1956 Returns the "next" key in the hash, given the previous one as the sole argument.
1957 Returns undef if there are no more keys to be fetched.
1959 $key = $db->next_key($key);
1963 Here are some examples of using hashes:
1965 my $db = DBM::Deep->new( "foo.db" );
1967 $db->put("foo", "bar");
1968 print "foo: " . $db->get("foo") . "\n";
1970 $db->put("baz", {}); # new child hash ref
1971 $db->get("baz")->put("buz", "biz");
1972 print "buz: " . $db->get("baz")->get("buz") . "\n";
1974 my $key = $db->first_key();
1976 print "$key: " . $db->get($key) . "\n";
1977 $key = $db->next_key($key);
1980 if ($db->exists("foo")) { $db->delete("foo"); }
1984 For arrays, DBM::Deep supports all the common methods described above, and the
1985 following additional methods: C<length()>, C<push()>, C<pop()>, C<shift()>,
1986 C<unshift()> and C<splice()>.
1992 Returns the number of elements in the array. Takes no arguments.
1994 my $len = $db->length();
1998 Adds one or more elements onto the end of the array. Accepts scalars, hash
1999 refs or array refs. No return value.
2001 $db->push("foo", "bar", {});
2005 Fetches the last element in the array, and deletes it. Takes no arguments.
2006 Returns undef if array is empty. Returns the element value.
2008 my $elem = $db->pop();
2012 Fetches the first element in the array, deletes it, then shifts all the
2013 remaining elements over to take up the space. Returns the element value. This
2014 method is not recommended with large arrays -- see L<LARGE ARRAYS> below for
2017 my $elem = $db->shift();
2021 Inserts one or more elements onto the beginning of the array, shifting all
2022 existing elements over to make room. Accepts scalars, hash refs or array refs.
2023 No return value. This method is not recommended with large arrays -- see
2024 <LARGE ARRAYS> below for details.
2026 $db->unshift("foo", "bar", {});
2030 Performs exactly like Perl's built-in function of the same name. See L<perldoc
2031 -f splice> for usage -- it is too complicated to document here. This method is
2032 not recommended with large arrays -- see L<LARGE ARRAYS> below for details.
2036 Here are some examples of using arrays:
2038 my $db = DBM::Deep->new(
2040 type => DBM::Deep->TYPE_ARRAY
2043 $db->push("bar", "baz");
2044 $db->unshift("foo");
2047 my $len = $db->length();
2048 print "length: $len\n"; # 4
2050 for (my $k=0; $k<$len; $k++) {
2051 print "$k: " . $db->get($k) . "\n";
2054 $db->splice(1, 2, "biz", "baf");
2056 while (my $elem = shift @$db) {
2057 print "shifted: $elem\n";
2062 Enable automatic file locking by passing a true value to the C<locking>
2063 parameter when constructing your DBM::Deep object (see L<SETUP> above).
2065 my $db = DBM::Deep->new(
2070 This causes DBM::Deep to C<flock()> the underlying filehandle with exclusive
2071 mode for writes, and shared mode for reads. This is required if you have
2072 multiple processes accessing the same database file, to avoid file corruption.
2073 Please note that C<flock()> does NOT work for files over NFS. See L<DB OVER
2074 NFS> below for more.
2076 =head2 EXPLICIT LOCKING
2078 You can explicitly lock a database, so it remains locked for multiple
2079 transactions. This is done by calling the C<lock()> method, and passing an
2080 optional lock mode argument (defaults to exclusive mode). This is particularly
2081 useful for things like counters, where the current value needs to be fetched,
2082 then incremented, then stored again.
2085 my $counter = $db->get("counter");
2087 $db->put("counter", $counter);
2096 You can pass C<lock()> an optional argument, which specifies which mode to use
2097 (exclusive or shared). Use one of these two constants: C<DBM::Deep-E<gt>LOCK_EX>
2098 or C<DBM::Deep-E<gt>LOCK_SH>. These are passed directly to C<flock()>, and are the
2099 same as the constants defined in Perl's C<Fcntl> module.
2101 $db->lock( DBM::Deep->LOCK_SH );
2105 =head1 IMPORTING/EXPORTING
2107 You can import existing complex structures by calling the C<import()> method,
2108 and export an entire database into an in-memory structure using the C<export()>
2109 method. Both are examined here.
2113 Say you have an existing hash with nested hashes/arrays inside it. Instead of
2114 walking the structure and adding keys/elements to the database as you go,
2115 simply pass a reference to the C<import()> method. This recursively adds
2116 everything to an existing DBM::Deep object for you. Here is an example:
2121 array1 => [ "elem0", "elem1", "elem2" ],
2123 subkey1 => "subvalue1",
2124 subkey2 => "subvalue2"
2128 my $db = DBM::Deep->new( "foo.db" );
2129 $db->import( $struct );
2131 print $db->{key1} . "\n"; # prints "value1"
2133 This recursively imports the entire C<$struct> object into C<$db>, including
2134 all nested hashes and arrays. If the DBM::Deep object contains exsiting data,
2135 keys are merged with the existing ones, replacing if they already exist.
2136 The C<import()> method can be called on any database level (not just the base
2137 level), and works with both hash and array DB types.
2139 B<Note:> Make sure your existing structure has no circular references in it.
2140 These will cause an infinite loop when importing.
2144 Calling the C<export()> method on an existing DBM::Deep object will return
2145 a reference to a new in-memory copy of the database. The export is done
2146 recursively, so all nested hashes/arrays are all exported to standard Perl
2147 objects. Here is an example:
2149 my $db = DBM::Deep->new( "foo.db" );
2151 $db->{key1} = "value1";
2152 $db->{key2} = "value2";
2154 $db->{hash1}->{subkey1} = "subvalue1";
2155 $db->{hash1}->{subkey2} = "subvalue2";
2157 my $struct = $db->export();
2159 print $struct->{key1} . "\n"; # prints "value1"
2161 This makes a complete copy of the database in memory, and returns a reference
2162 to it. The C<export()> method can be called on any database level (not just
2163 the base level), and works with both hash and array DB types. Be careful of
2164 large databases -- you can store a lot more data in a DBM::Deep object than an
2165 in-memory Perl structure.
2167 B<Note:> Make sure your database has no circular references in it.
2168 These will cause an infinite loop when exporting.
2172 DBM::Deep has a number of hooks where you can specify your own Perl function
2173 to perform filtering on incoming or outgoing data. This is a perfect
2174 way to extend the engine, and implement things like real-time compression or
2175 encryption. Filtering applies to the base DB level, and all child hashes /
2176 arrays. Filter hooks can be specified when your DBM::Deep object is first
2177 constructed, or by calling the C<set_filter()> method at any time. There are
2178 four available filter hooks, described below:
2182 =item * filter_store_key
2184 This filter is called whenever a hash key is stored. It
2185 is passed the incoming key, and expected to return a transformed key.
2187 =item * filter_store_value
2189 This filter is called whenever a hash key or array element is stored. It
2190 is passed the incoming value, and expected to return a transformed value.
2192 =item * filter_fetch_key
2194 This filter is called whenever a hash key is fetched (i.e. via
2195 C<first_key()> or C<next_key()>). It is passed the transformed key,
2196 and expected to return the plain key.
2198 =item * filter_fetch_value
2200 This filter is called whenever a hash key or array element is fetched.
2201 It is passed the transformed value, and expected to return the plain value.
2205 Here are the two ways to setup a filter hook:
2207 my $db = DBM::Deep->new(
2209 filter_store_value => \&my_filter_store,
2210 filter_fetch_value => \&my_filter_fetch
2215 $db->set_filter( "filter_store_value", \&my_filter_store );
2216 $db->set_filter( "filter_fetch_value", \&my_filter_fetch );
2218 Your filter function will be called only when dealing with SCALAR keys or
2219 values. When nested hashes and arrays are being stored/fetched, filtering
2220 is bypassed. Filters are called as static functions, passed a single SCALAR
2221 argument, and expected to return a single SCALAR value. If you want to
2222 remove a filter, set the function reference to C<undef>:
2224 $db->set_filter( "filter_store_value", undef );
2226 =head2 REAL-TIME ENCRYPTION EXAMPLE
2228 Here is a working example that uses the I<Crypt::Blowfish> module to
2229 do real-time encryption / decryption of keys & values with DBM::Deep Filters.
2230 Please visit L<http://search.cpan.org/search?module=Crypt::Blowfish> for more
2231 on I<Crypt::Blowfish>. You'll also need the I<Crypt::CBC> module.
2234 use Crypt::Blowfish;
2237 my $cipher = Crypt::CBC->new({
2238 'key' => 'my secret key',
2239 'cipher' => 'Blowfish',
2241 'regenerate_key' => 0,
2242 'padding' => 'space',
2246 my $db = DBM::Deep->new(
2247 file => "foo-encrypt.db",
2248 filter_store_key => \&my_encrypt,
2249 filter_store_value => \&my_encrypt,
2250 filter_fetch_key => \&my_decrypt,
2251 filter_fetch_value => \&my_decrypt,
2254 $db->{key1} = "value1";
2255 $db->{key2} = "value2";
2256 print "key1: " . $db->{key1} . "\n";
2257 print "key2: " . $db->{key2} . "\n";
2263 return $cipher->encrypt( $_[0] );
2266 return $cipher->decrypt( $_[0] );
2269 =head2 REAL-TIME COMPRESSION EXAMPLE
2271 Here is a working example that uses the I<Compress::Zlib> module to do real-time
2272 compression / decompression of keys & values with DBM::Deep Filters.
2273 Please visit L<http://search.cpan.org/search?module=Compress::Zlib> for
2274 more on I<Compress::Zlib>.
2279 my $db = DBM::Deep->new(
2280 file => "foo-compress.db",
2281 filter_store_key => \&my_compress,
2282 filter_store_value => \&my_compress,
2283 filter_fetch_key => \&my_decompress,
2284 filter_fetch_value => \&my_decompress,
2287 $db->{key1} = "value1";
2288 $db->{key2} = "value2";
2289 print "key1: " . $db->{key1} . "\n";
2290 print "key2: " . $db->{key2} . "\n";
2296 return Compress::Zlib::memGzip( $_[0] ) ;
2299 return Compress::Zlib::memGunzip( $_[0] ) ;
2302 B<Note:> Filtering of keys only applies to hashes. Array "keys" are
2303 actually numerical index numbers, and are not filtered.
2305 =head1 ERROR HANDLING
2307 Most DBM::Deep methods return a true value for success, and call die() on
2308 failure. You can wrap calls in an eval block to catch the die. Also, the
2309 actual error message is stored in an internal scalar, which can be fetched by
2310 calling the C<error()> method.
2312 my $db = DBM::Deep->new( "foo.db" ); # create hash
2313 eval { $db->push("foo"); }; # ILLEGAL -- push is array-only call
2315 print $@; # prints error message
2316 print $db->error(); # prints error message
2318 You can then call C<clear_error()> to clear the current error state.
2322 If you set the C<debug> option to true when creating your DBM::Deep object,
2323 all errors are considered NON-FATAL, and dumped to STDERR. This should only
2324 be used for debugging purposes and not production work. DBM::Deep expects errors
2325 to be thrown, not propagated back up the stack.
2327 =head1 LARGEFILE SUPPORT
2329 If you have a 64-bit system, and your Perl is compiled with both LARGEFILE
2330 and 64-bit support, you I<may> be able to create databases larger than 2 GB.
2331 DBM::Deep by default uses 32-bit file offset tags, but these can be changed
2332 by calling the static C<set_pack()> method before you do anything else.
2334 DBM::Deep::set_pack(8, 'Q');
2336 This tells DBM::Deep to pack all file offsets with 8-byte (64-bit) quad words
2337 instead of 32-bit longs. After setting these values your DB files have a
2338 theoretical maximum size of 16 XB (exabytes).
2340 B<Note:> Changing these values will B<NOT> work for existing database files.
2341 Only change this for new files, and make sure it stays set consistently
2342 throughout the file's life. If you do set these values, you can no longer
2343 access 32-bit DB files. You can, however, call C<set_pack(4, 'N')> to change
2344 back to 32-bit mode.
2346 B<Note:> I have not personally tested files > 2 GB -- all my systems have
2347 only a 32-bit Perl. However, I have received user reports that this does
2350 =head1 LOW-LEVEL ACCESS
2352 If you require low-level access to the underlying filehandle that DBM::Deep uses,
2353 you can call the C<_fh()> method, which returns the handle:
2355 my $fh = $db->_fh();
2357 This method can be called on the root level of the datbase, or any child
2358 hashes or arrays. All levels share a I<root> structure, which contains things
2359 like the filehandle, a reference counter, and all the options specified
2360 when you created the object. You can get access to this root structure by
2361 calling the C<root()> method.
2363 my $root = $db->_root();
2365 This is useful for changing options after the object has already been created,
2366 such as enabling/disabling locking, or debug modes. You can also
2367 store your own temporary user data in this structure (be wary of name
2368 collision), which is then accessible from any child hash or array.
2370 =head1 CUSTOM DIGEST ALGORITHM
2372 DBM::Deep by default uses the I<Message Digest 5> (MD5) algorithm for hashing
2373 keys. However you can override this, and use another algorithm (such as SHA-256)
2374 or even write your own. But please note that DBM::Deep currently expects zero
2375 collisions, so your algorithm has to be I<perfect>, so to speak.
2376 Collision detection may be introduced in a later version.
2380 You can specify a custom digest algorithm by calling the static C<set_digest()>
2381 function, passing a reference to a subroutine, and the length of the algorithm's
2382 hashes (in bytes). This is a global static function, which affects ALL DBM::Deep
2383 objects. Here is a working example that uses a 256-bit hash from the
2384 I<Digest::SHA256> module. Please see
2385 L<http://search.cpan.org/search?module=Digest::SHA256> for more.
2390 my $context = Digest::SHA256::new(256);
2392 DBM::Deep::set_digest( \&my_digest, 32 );
2394 my $db = DBM::Deep->new( "foo-sha.db" );
2396 $db->{key1} = "value1";
2397 $db->{key2} = "value2";
2398 print "key1: " . $db->{key1} . "\n";
2399 print "key2: " . $db->{key2} . "\n";
2405 return substr( $context->hash($_[0]), 0, 32 );
2408 B<Note:> Your returned digest strings must be B<EXACTLY> the number
2409 of bytes you specify in the C<set_digest()> function (in this case 32).
2411 =head1 CIRCULAR REFERENCES
2413 DBM::Deep has B<experimental> support for circular references. Meaning you
2414 can have a nested hash key or array element that points to a parent object.
2415 This relationship is stored in the DB file, and is preserved between sessions.
2418 my $db = DBM::Deep->new( "foo.db" );
2421 $db->{circle} = $db; # ref to self
2423 print $db->{foo} . "\n"; # prints "foo"
2424 print $db->{circle}->{foo} . "\n"; # prints "foo" again
2426 One catch is, passing the object to a function that recursively walks the
2427 object tree (such as I<Data::Dumper> or even the built-in C<optimize()> or
2428 C<export()> methods) will result in an infinite loop. The other catch is,
2429 if you fetch the I<key> of a circular reference (i.e. using the C<first_key()>
2430 or C<next_key()> methods), you will get the I<target object's key>, not the
2431 ref's key. This gets even more interesting with the above example, where
2432 the I<circle> key points to the base DB object, which technically doesn't
2433 have a key. So I made DBM::Deep return "[base]" as the key name in that
2436 =head1 CAVEATS / ISSUES / BUGS
2438 This section describes all the known issues with DBM::Deep. It you have found
2439 something that is not listed here, please send e-mail to L<jhuckaby@cpan.org>.
2441 =head2 UNUSED SPACE RECOVERY
2443 One major caveat with DBM::Deep is that space occupied by existing keys and
2444 values is not recovered when they are deleted. Meaning if you keep deleting
2445 and adding new keys, your file will continuously grow. I am working on this,
2446 but in the meantime you can call the built-in C<optimize()> method from time to
2447 time (perhaps in a crontab or something) to recover all your unused space.
2449 $db->optimize(); # returns true on success
2451 This rebuilds the ENTIRE database into a new file, then moves it on top of
2452 the original. The new file will have no unused space, thus it will take up as
2453 little disk space as possible. Please note that this operation can take
2454 a long time for large files, and you need enough disk space to temporarily hold
2455 2 copies of your DB file. The temporary file is created in the same directory
2456 as the original, named with a ".tmp" extension, and is deleted when the
2457 operation completes. Oh, and if locking is enabled, the DB is automatically
2458 locked for the entire duration of the copy.
2460 B<WARNING:> Only call optimize() on the top-level node of the database, and
2461 make sure there are no child references lying around. DBM::Deep keeps a reference
2462 counter, and if it is greater than 1, optimize() will abort and return undef.
2464 =head2 AUTOVIVIFICATION
2466 Unfortunately, autovivification doesn't work with tied hashes. This appears to
2467 be a bug in Perl's tie() system, as I<Jakob Schmidt> encountered the very same
2468 issue with his I<DWH_FIle> module (see L<http://search.cpan.org/search?module=DWH_File>),
2469 and it is also mentioned in the BUGS section for the I<MLDBM> module <see
2470 L<http://search.cpan.org/search?module=MLDBM>). Basically, on a new db file,
2473 $db->{foo}->{bar} = "hello";
2475 Since "foo" doesn't exist, you cannot add "bar" to it. You end up with "foo"
2476 being an empty hash. Try this instead, which works fine:
2478 $db->{foo} = { bar => "hello" };
2480 As of Perl 5.8.7, this bug still exists. I have walked very carefully through
2481 the execution path, and Perl indeed passes an empty hash to the STORE() method.
2482 Probably a bug in Perl.
2484 =head2 FILE CORRUPTION
2486 The current level of error handling in DBM::Deep is minimal. Files I<are> checked
2487 for a 32-bit signature when opened, but other corruption in files can cause
2488 segmentation faults. DBM::Deep may try to seek() past the end of a file, or get
2489 stuck in an infinite loop depending on the level of corruption. File write
2490 operations are not checked for failure (for speed), so if you happen to run
2491 out of disk space, DBM::Deep will probably fail in a bad way. These things will
2492 be addressed in a later version of DBM::Deep.
2496 Beware of using DB files over NFS. DBM::Deep uses flock(), which works well on local
2497 filesystems, but will NOT protect you from file corruption over NFS. I've heard
2498 about setting up your NFS server with a locking daemon, then using lockf() to
2499 lock your files, but your mileage may vary there as well. From what I
2500 understand, there is no real way to do it. However, if you need access to the
2501 underlying filehandle in DBM::Deep for using some other kind of locking scheme like
2502 lockf(), see the L<LOW-LEVEL ACCESS> section above.
2504 =head2 COPYING OBJECTS
2506 Beware of copying tied objects in Perl. Very strange things can happen.
2507 Instead, use DBM::Deep's C<clone()> method which safely copies the object and
2508 returns a new, blessed, tied hash or array to the same level in the DB.
2510 my $copy = $db->clone();
2512 B<Note>: Since clone() here is cloning the object, not the database location, any
2513 modifications to either $db or $copy will be visible in both.
2517 Beware of using C<shift()>, C<unshift()> or C<splice()> with large arrays.
2518 These functions cause every element in the array to move, which can be murder
2519 on DBM::Deep, as every element has to be fetched from disk, then stored again in
2520 a different location. This will be addressed in the forthcoming version 1.00.
2524 This section discusses DBM::Deep's speed and memory usage.
2528 Obviously, DBM::Deep isn't going to be as fast as some C-based DBMs, such as
2529 the almighty I<BerkeleyDB>. But it makes up for it in features like true
2530 multi-level hash/array support, and cross-platform FTPable files. Even so,
2531 DBM::Deep is still pretty fast, and the speed stays fairly consistent, even
2532 with huge databases. Here is some test data:
2534 Adding 1,000,000 keys to new DB file...
2536 At 100 keys, avg. speed is 2,703 keys/sec
2537 At 200 keys, avg. speed is 2,642 keys/sec
2538 At 300 keys, avg. speed is 2,598 keys/sec
2539 At 400 keys, avg. speed is 2,578 keys/sec
2540 At 500 keys, avg. speed is 2,722 keys/sec
2541 At 600 keys, avg. speed is 2,628 keys/sec
2542 At 700 keys, avg. speed is 2,700 keys/sec
2543 At 800 keys, avg. speed is 2,607 keys/sec
2544 At 900 keys, avg. speed is 2,190 keys/sec
2545 At 1,000 keys, avg. speed is 2,570 keys/sec
2546 At 2,000 keys, avg. speed is 2,417 keys/sec
2547 At 3,000 keys, avg. speed is 1,982 keys/sec
2548 At 4,000 keys, avg. speed is 1,568 keys/sec
2549 At 5,000 keys, avg. speed is 1,533 keys/sec
2550 At 6,000 keys, avg. speed is 1,787 keys/sec
2551 At 7,000 keys, avg. speed is 1,977 keys/sec
2552 At 8,000 keys, avg. speed is 2,028 keys/sec
2553 At 9,000 keys, avg. speed is 2,077 keys/sec
2554 At 10,000 keys, avg. speed is 2,031 keys/sec
2555 At 20,000 keys, avg. speed is 1,970 keys/sec
2556 At 30,000 keys, avg. speed is 2,050 keys/sec
2557 At 40,000 keys, avg. speed is 2,073 keys/sec
2558 At 50,000 keys, avg. speed is 1,973 keys/sec
2559 At 60,000 keys, avg. speed is 1,914 keys/sec
2560 At 70,000 keys, avg. speed is 2,091 keys/sec
2561 At 80,000 keys, avg. speed is 2,103 keys/sec
2562 At 90,000 keys, avg. speed is 1,886 keys/sec
2563 At 100,000 keys, avg. speed is 1,970 keys/sec
2564 At 200,000 keys, avg. speed is 2,053 keys/sec
2565 At 300,000 keys, avg. speed is 1,697 keys/sec
2566 At 400,000 keys, avg. speed is 1,838 keys/sec
2567 At 500,000 keys, avg. speed is 1,941 keys/sec
2568 At 600,000 keys, avg. speed is 1,930 keys/sec
2569 At 700,000 keys, avg. speed is 1,735 keys/sec
2570 At 800,000 keys, avg. speed is 1,795 keys/sec
2571 At 900,000 keys, avg. speed is 1,221 keys/sec
2572 At 1,000,000 keys, avg. speed is 1,077 keys/sec
2574 This test was performed on a PowerMac G4 1gHz running Mac OS X 10.3.2 & Perl
2575 5.8.1, with an 80GB Ultra ATA/100 HD spinning at 7200RPM. The hash keys and
2576 values were between 6 - 12 chars in length. The DB file ended up at 210MB.
2577 Run time was 12 min 3 sec.
2581 One of the great things about DBM::Deep is that it uses very little memory.
2582 Even with huge databases (1,000,000+ keys) you will not see much increased
2583 memory on your process. DBM::Deep relies solely on the filesystem for storing
2584 and fetching data. Here is output from I</usr/bin/top> before even opening a
2587 PID USER PRI NI SIZE RSS SHARE STAT %CPU %MEM TIME COMMAND
2588 22831 root 11 0 2716 2716 1296 R 0.0 0.2 0:07 perl
2590 Basically the process is taking 2,716K of memory. And here is the same
2591 process after storing and fetching 1,000,000 keys:
2593 PID USER PRI NI SIZE RSS SHARE STAT %CPU %MEM TIME COMMAND
2594 22831 root 14 0 2772 2772 1328 R 0.0 0.2 13:32 perl
2596 Notice the memory usage increased by only 56K. Test was performed on a 700mHz
2597 x86 box running Linux RedHat 7.2 & Perl 5.6.1.
2599 =head1 DB FILE FORMAT
2601 In case you were interested in the underlying DB file format, it is documented
2602 here in this section. You don't need to know this to use the module, it's just
2603 included for reference.
2607 DBM::Deep files always start with a 32-bit signature to identify the file type.
2608 This is at offset 0. The signature is "DPDB" in network byte order. This is
2609 checked for when the file is opened and an error will be thrown if it's not found.
2613 The DBM::Deep file is in a I<tagged format>, meaning each section of the file
2614 has a standard header containing the type of data, the length of data, and then
2615 the data itself. The type is a single character (1 byte), the length is a
2616 32-bit unsigned long in network byte order, and the data is, well, the data.
2617 Here is how it unfolds:
2621 Immediately after the 32-bit file signature is the I<Master Index> record.
2622 This is a standard tag header followed by 1024 bytes (in 32-bit mode) or 2048
2623 bytes (in 64-bit mode) of data. The type is I<H> for hash or I<A> for array,
2624 depending on how the DBM::Deep object was constructed.
2626 The index works by looking at a I<MD5 Hash> of the hash key (or array index
2627 number). The first 8-bit char of the MD5 signature is the offset into the
2628 index, multipled by 4 in 32-bit mode, or 8 in 64-bit mode. The value of the
2629 index element is a file offset of the next tag for the key/element in question,
2630 which is usually a I<Bucket List> tag (see below).
2632 The next tag I<could> be another index, depending on how many keys/elements
2633 exist. See L<RE-INDEXING> below for details.
2637 A I<Bucket List> is a collection of 16 MD5 hashes for keys/elements, plus
2638 file offsets to where the actual data is stored. It starts with a standard
2639 tag header, with type I<B>, and a data size of 320 bytes in 32-bit mode, or
2640 384 bytes in 64-bit mode. Each MD5 hash is stored in full (16 bytes), plus
2641 the 32-bit or 64-bit file offset for the I<Bucket> containing the actual data.
2642 When the list fills up, a I<Re-Index> operation is performed (See
2643 L<RE-INDEXING> below).
2647 A I<Bucket> is a tag containing a key/value pair (in hash mode), or a
2648 index/value pair (in array mode). It starts with a standard tag header with
2649 type I<D> for scalar data (string, binary, etc.), or it could be a nested
2650 hash (type I<H>) or array (type I<A>). The value comes just after the tag
2651 header. The size reported in the tag header is only for the value, but then,
2652 just after the value is another size (32-bit unsigned long) and then the plain
2653 key itself. Since the value is likely to be fetched more often than the plain
2654 key, I figured it would be I<slightly> faster to store the value first.
2656 If the type is I<H> (hash) or I<A> (array), the value is another I<Master Index>
2657 record for the nested structure, where the process begins all over again.
2661 After a I<Bucket List> grows to 16 records, its allocated space in the file is
2662 exhausted. Then, when another key/element comes in, the list is converted to a
2663 new index record. However, this index will look at the next char in the MD5
2664 hash, and arrange new Bucket List pointers accordingly. This process is called
2665 I<Re-Indexing>. Basically, a new index tag is created at the file EOF, and all
2666 17 (16 + new one) keys/elements are removed from the old Bucket List and
2667 inserted into the new index. Several new Bucket Lists are created in the
2668 process, as a new MD5 char from the key is being examined (it is unlikely that
2669 the keys will all share the same next char of their MD5s).
2671 Because of the way the I<MD5> algorithm works, it is impossible to tell exactly
2672 when the Bucket Lists will turn into indexes, but the first round tends to
2673 happen right around 4,000 keys. You will see a I<slight> decrease in
2674 performance here, but it picks back up pretty quick (see L<SPEED> above). Then
2675 it takes B<a lot> more keys to exhaust the next level of Bucket Lists. It's
2676 right around 900,000 keys. This process can continue nearly indefinitely --
2677 right up until the point the I<MD5> signatures start colliding with each other,
2678 and this is B<EXTREMELY> rare -- like winning the lottery 5 times in a row AND
2679 getting struck by lightning while you are walking to cash in your tickets.
2680 Theoretically, since I<MD5> hashes are 128-bit values, you I<could> have up to
2681 340,282,366,921,000,000,000,000,000,000,000,000,000 keys/elements (I believe
2682 this is 340 unodecillion, but don't quote me).
2686 When a new key/element is stored, the key (or index number) is first run through
2687 I<Digest::MD5> to get a 128-bit signature (example, in hex:
2688 b05783b0773d894396d475ced9d2f4f6). Then, the I<Master Index> record is checked
2689 for the first char of the signature (in this case I<b0>). If it does not exist,
2690 a new I<Bucket List> is created for our key (and the next 15 future keys that
2691 happen to also have I<b> as their first MD5 char). The entire MD5 is written
2692 to the I<Bucket List> along with the offset of the new I<Bucket> record (EOF at
2693 this point, unless we are replacing an existing I<Bucket>), where the actual
2694 data will be stored.
2698 Fetching an existing key/element involves getting a I<Digest::MD5> of the key
2699 (or index number), then walking along the indexes. If there are enough
2700 keys/elements in this DB level, there might be nested indexes, each linked to
2701 a particular char of the MD5. Finally, a I<Bucket List> is pointed to, which
2702 contains up to 16 full MD5 hashes. Each is checked for equality to the key in
2703 question. If we found a match, the I<Bucket> tag is loaded, where the value and
2704 plain key are stored.
2706 Fetching the plain key occurs when calling the I<first_key()> and I<next_key()>
2707 methods. In this process the indexes are walked systematically, and each key
2708 fetched in increasing MD5 order (which is why it appears random). Once the
2709 I<Bucket> is found, the value is skipped and the plain key returned instead.
2710 B<Note:> Do not count on keys being fetched as if the MD5 hashes were
2711 alphabetically sorted. This only happens on an index-level -- as soon as the
2712 I<Bucket Lists> are hit, the keys will come out in the order they went in --
2713 so it's pretty much undefined how the keys will come out -- just like Perl's
2716 =head1 CODE COVERAGE
2718 We use B<Devel::Cover> to test the code coverage of our tests, below is the
2719 B<Devel::Cover> report on this module's test suite.
2721 ---------------------------- ------ ------ ------ ------ ------ ------ ------
2722 File stmt bran cond sub pod time total
2723 ---------------------------- ------ ------ ------ ------ ------ ------ ------
2724 blib/lib/DBM/Deep.pm 95.0 83.2 68.7 98.2 100.0 57.8 90.7
2725 blib/lib/DBM/Deep/Array.pm 98.9 88.9 87.5 100.0 n/a 27.4 96.4
2726 blib/lib/DBM/Deep/Hash.pm 95.3 80.0 100.0 100.0 n/a 14.8 92.4
2727 Total 95.8 83.9 72.8 98.8 100.0 100.0 91.8
2728 ---------------------------- ------ ------ ------ ------ ------ ------ ------
2730 =head1 MORE INFORMATION
2732 Check out the DBM::Deep Google Group at L<http://groups.google.com/group/DBM-Deep>
2733 or send email to L<DBM-Deep@googlegroups.com>.
2737 Joseph Huckaby, L<jhuckaby@cpan.org>
2739 Rob Kinyon, L<rkinyon@cpan.org>
2741 Special thanks to Adam Sah and Rich Gaushell! You know why :-)
2745 perltie(1), Tie::Hash(3), Digest::MD5(3), Fcntl(3), flock(2), lockf(3), nfs(5),
2746 Digest::SHA256(3), Crypt::Blowfish(3), Compress::Zlib(3)
2750 Copyright (c) 2002-2006 Joseph Huckaby. All Rights Reserved.
2751 This is free software, you may use it and distribute it under the
2752 same terms as Perl itself.