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 " . $db->{my_complex}->[1]->{perl} . "!\n";
27 # (c) 2002-2005 Joseph Huckaby. All Rights Reserved.
28 # This program is free software; you can redistribute it and/or
29 # modify it under the same terms as Perl itself.
37 use vars qw/$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 $DATA_LENGTH_PACK = 'N';
58 my ($LONG_SIZE, $LONG_PACK, $DATA_LENGTH_SIZE, $DATA_LENGTH_PACK);
61 # Maximum number of buckets per list before another level of indexing is done.
62 # Increase this value for slightly greater speed, but larger database files.
63 # DO NOT decrease this value below 16, due to risk of recursive reindex overrun.
68 # Better not adjust anything below here, unless you're me :-)
72 # Setup digest function for keys
74 my ($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_NULL () { 'N' }
94 sub SIG_DATA () { 'D' }
95 sub SIG_INDEX () { 'I' }
96 sub SIG_BLIST () { 'B' }
100 # Setup constants for users to pass to new()
102 sub TYPE_HASH () { return SIG_HASH; }
103 sub TYPE_ARRAY () { return SIG_ARRAY; }
107 # Class constructor method for Perl OO interface.
108 # Calls tie() and returns blessed reference to tied hash or array,
109 # providing a hybrid OO/tie interface.
113 if (scalar(@_) > 1) { $args = {@_}; }
114 else { $args = { file => shift }; }
117 # Check if we want a tied hash or array.
120 if (defined($args->{type}) && $args->{type} eq TYPE_ARRAY) {
121 tie @$self, $class, %$args;
124 tie %$self, $class, %$args;
127 return bless $self, $class;
131 my @outer_params = qw( type base_offset );
134 # Setup $self and bless into this class.
141 base_offset => length(SIG_FILE),
146 foreach my $outer_parm ( @outer_params ) {
147 next unless exists $args->{$outer_parm};
148 $self->{$outer_parm} = delete $args->{$outer_parm}
151 $self->{root} = exists $args->{root}
153 : DBM::Deep::_::Root->new( $args );
155 if (!defined($self->fh)) { $self->_open(); }
162 tied( %{$_[0]} ) || $_[0]
167 # Tied hash constructor method, called by Perl's tie() function.
171 if (scalar(@_) > 1) { $args = {@_}; }
172 #XXX This use of ref() is bad and is a bug
173 elsif (ref($_[0])) { $args = $_[0]; }
174 else { $args = { file => shift }; }
176 $args->{type} = TYPE_HASH;
178 return $class->_init($args);
183 # Tied array constructor method, called by Perl's tie() function.
187 if (scalar(@_) > 1) { $args = {@_}; }
188 #XXX This use of ref() is bad and is a bug
189 elsif (ref($_[0])) { $args = $_[0]; }
190 else { $args = { file => shift }; }
192 $args->{type} = TYPE_ARRAY;
194 return $class->_init($args);
197 #XXX Unneeded now ...
201 my %translate_mode = (
211 # Open a FileHandle to the database, create if nonexistent.
212 # Make sure file signature matches DeepDB spec.
214 my $self = _get_self($_[0]);
216 if (defined($self->fh)) { $self->_close(); }
219 my $filename = $self->root->{file};
220 my $mode = $translate_mode{ $self->root->{mode} };
222 if (!(-e $filename) && $mode eq '+<') {
223 open( FH, '>', $filename );
228 open( $fh, $mode, $filename )
230 $self->root->{fh} = $fh;
231 }; if ($@ ) { $self->_throw_error( "Received error: $@\n" ); }
232 if (! defined($self->fh)) {
233 return $self->_throw_error("Cannot open file: " . $self->root->{file} . ": $!");
238 #XXX Can we remove this by using the right sysopen() flags?
239 binmode $fh; # for win32
241 if ($self->root->{autoflush}) {
242 my $old = select $fh;
249 my $bytes_read = read( $fh, $signature, length(SIG_FILE));
252 # File is empty -- write signature and master index
257 $self->root->{end} = length(SIG_FILE);
258 $self->_create_tag($self->base_offset, $self->type, chr(0) x $INDEX_SIZE);
260 my $plain_key = "[base]";
261 print($fh pack($DATA_LENGTH_PACK, length($plain_key)) . $plain_key );
262 $self->root->{end} += $DATA_LENGTH_SIZE + length($plain_key);
264 # Flush the filehandle
265 my $old_fh = select $fh;
275 # Check signature was valid
277 unless ($signature eq SIG_FILE) {
279 return $self->_throw_error("Signature not found -- file is not a Deep DB");
282 $self->root->{end} = (stat($fh))[7];
285 # Get our type from master index signature
287 my $tag = $self->_load_tag($self->base_offset);
289 #XXX We probably also want to store the hash algorithm name and not assume anything
292 return $self->_throw_error("Corrupted file, no master index record");
294 if ($self->{type} ne $tag->{signature}) {
295 return $self->_throw_error("File type mismatch");
303 # Close database FileHandle
305 my $self = _get_self($_[0]);
306 close $self->root->{fh};
311 # Given offset, signature and content, create tag and write to disk
313 my ($self, $offset, $sig, $content) = @_;
314 my $size = length($content);
318 seek($fh, $offset, 0);
319 print($fh $sig . pack($DATA_LENGTH_PACK, $size) . $content );
321 if ($offset == $self->root->{end}) {
322 $self->root->{end} += SIG_SIZE + $DATA_LENGTH_SIZE + $size;
328 offset => $offset + SIG_SIZE + $DATA_LENGTH_SIZE,
335 # Given offset, load single tag and return signature, size and data
342 seek($fh, $offset, 0);
343 if (eof $fh) { return undef; }
346 read( $fh, $sig, SIG_SIZE);
349 read( $fh, $size, $DATA_LENGTH_SIZE);
350 $size = unpack($DATA_LENGTH_PACK, $size);
353 read( $fh, $buffer, $size);
358 offset => $offset + SIG_SIZE + $DATA_LENGTH_SIZE,
365 # Given index tag, lookup single entry in index and return .
368 my ($tag, $index) = @_;
370 my $location = unpack($LONG_PACK, substr($tag->{content}, $index * $LONG_SIZE, $LONG_SIZE) );
371 if (!$location) { return; }
373 return $self->_load_tag( $location );
378 # Adds one key/value pair to bucket list, given offset, MD5 digest of key,
379 # plain (undigested) key and value.
382 my ($tag, $md5, $plain_key, $value) = @_;
383 my $keys = $tag->{content};
387 my $is_dbm_deep = eval { $value->isa( 'DBM::Deep' ) };
388 my $internal_ref = $is_dbm_deep && ($value->root eq $self->root);
393 # Iterate through buckets, seeing if this is a new entry or a replace.
395 for (my $i=0; $i<$MAX_BUCKETS; $i++) {
396 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
397 my $subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
400 # Found empty bucket (end of list). Populate and exit loop.
404 $location = $internal_ref
405 ? $value->base_offset
406 : $self->root->{end};
408 seek($fh, $tag->{offset} + ($i * $BUCKET_SIZE), 0);
409 print($fh $md5 . pack($LONG_PACK, $location) );
412 elsif ($md5 eq $key) {
414 # Found existing bucket with same key. Replace with new value.
419 $location = $value->base_offset;
420 seek($fh, $tag->{offset} + ($i * $BUCKET_SIZE), 0);
421 print($fh $md5 . pack($LONG_PACK, $location) );
424 seek($fh, $subloc + SIG_SIZE, 0);
426 read( $fh, $size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
429 # If value is a hash, array, or raw value with equal or less size, we can
430 # reuse the same content area of the database. Otherwise, we have to create
431 # a new content area at the EOF.
434 my $r = Scalar::Util::reftype( $value ) || '';
435 if ( $r eq 'HASH' || $r eq 'ARRAY' ) { $actual_length = $INDEX_SIZE; }
436 else { $actual_length = length($value); }
438 if ($actual_length <= $size) {
442 $location = $self->root->{end};
443 seek($fh, $tag->{offset} + ($i * $BUCKET_SIZE) + $HASH_SIZE, 0);
444 print($fh pack($LONG_PACK, $location) );
452 # If this is an internal reference, return now.
453 # No need to write value or plain key
460 # If bucket didn't fit into list, split into a new index level
463 seek($fh, $tag->{ref_loc}, 0);
464 print($fh pack($LONG_PACK, $self->root->{end}) );
466 my $index_tag = $self->_create_tag($self->root->{end}, SIG_INDEX, chr(0) x $INDEX_SIZE);
469 $keys .= $md5 . pack($LONG_PACK, 0);
471 for (my $i=0; $i<=$MAX_BUCKETS; $i++) {
472 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
474 my $old_subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
475 my $num = ord(substr($key, $tag->{ch} + 1, 1));
477 if ($offsets[$num]) {
478 my $offset = $offsets[$num] + SIG_SIZE + $DATA_LENGTH_SIZE;
479 seek($fh, $offset, 0);
481 read( $fh, $subkeys, $BUCKET_LIST_SIZE);
483 for (my $k=0; $k<$MAX_BUCKETS; $k++) {
484 my $subloc = unpack($LONG_PACK, substr($subkeys, ($k * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
486 seek($fh, $offset + ($k * $BUCKET_SIZE), 0);
487 print($fh $key . pack($LONG_PACK, $old_subloc || $self->root->{end}) );
493 $offsets[$num] = $self->root->{end};
494 seek($fh, $index_tag->{offset} + ($num * $LONG_SIZE), 0);
495 print($fh pack($LONG_PACK, $self->root->{end}) );
497 my $blist_tag = $self->_create_tag($self->root->{end}, SIG_BLIST, chr(0) x $BUCKET_LIST_SIZE);
499 seek($fh, $blist_tag->{offset}, 0);
500 print($fh $key . pack($LONG_PACK, $old_subloc || $self->root->{end}) );
505 $location ||= $self->root->{end};
506 } # re-index bucket list
509 # Seek to content area and store signature, value and plaintext key
513 seek($fh, $location, 0);
516 # Write signature based on content type, set content length and write actual value.
518 my $r = Scalar::Util::reftype($value) || '';
520 print($fh TYPE_HASH );
521 print($fh pack($DATA_LENGTH_PACK, $INDEX_SIZE) . chr(0) x $INDEX_SIZE );
522 $content_length = $INDEX_SIZE;
524 elsif ($r eq 'ARRAY') {
525 print($fh TYPE_ARRAY );
526 print($fh pack($DATA_LENGTH_PACK, $INDEX_SIZE) . chr(0) x $INDEX_SIZE );
527 $content_length = $INDEX_SIZE;
529 elsif (!defined($value)) {
530 print($fh SIG_NULL );
531 print($fh pack($DATA_LENGTH_PACK, 0) );
535 print($fh SIG_DATA );
536 print($fh pack($DATA_LENGTH_PACK, length($value)) . $value );
537 $content_length = length($value);
541 # Plain key is stored AFTER value, as keys are typically fetched less often.
543 print($fh pack($DATA_LENGTH_PACK, length($plain_key)) . $plain_key );
546 # If value is blessed, preserve class name
548 if ( $self->root->{autobless} ) {
549 my $value_class = Scalar::Util::blessed($value);
550 if ( defined $value_class && $value_class ne 'DBM::Deep' ) {
552 # Blessed ref -- will restore later
555 print($fh pack($DATA_LENGTH_PACK, length($value_class)) . $value_class );
556 $content_length += 1;
557 $content_length += $DATA_LENGTH_SIZE + length($value_class);
561 $content_length += 1;
566 # If this is a new content area, advance EOF counter
568 if ($location == $self->root->{end}) {
569 $self->root->{end} += SIG_SIZE;
570 $self->root->{end} += $DATA_LENGTH_SIZE + $content_length;
571 $self->root->{end} += $DATA_LENGTH_SIZE + length($plain_key);
575 # If content is a hash or array, create new child DeepDB object and
576 # pass each key or element to it.
579 my $branch = DBM::Deep->new(
581 base_offset => $location,
584 foreach my $key (keys %{$value}) {
585 #$branch->{$key} = $value->{$key};
586 $branch->STORE( $key, $value->{$key} );
589 elsif ($r eq 'ARRAY') {
590 my $branch = DBM::Deep->new(
592 base_offset => $location,
596 foreach my $element (@{$value}) {
597 #$branch->[$index] = $element;
598 $branch->STORE( $index, $element );
606 return $self->_throw_error("Fatal error: indexing failed -- possibly due to corruption in file");
609 sub _get_bucket_value {
611 # Fetch single value given tag and MD5 digested key.
614 my ($tag, $md5) = @_;
615 my $keys = $tag->{content};
620 # Iterate through buckets, looking for a key match
623 for (my $i=0; $i<$MAX_BUCKETS; $i++) {
624 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
625 my $subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
629 # Hit end of list, no match
634 if ( $md5 ne $key ) {
639 # Found match -- seek to offset and read signature
642 seek($fh, $subloc, 0);
643 read( $fh, $signature, SIG_SIZE);
646 # If value is a hash or array, return new DeepDB object with correct offset
648 if (($signature eq TYPE_HASH) || ($signature eq TYPE_ARRAY)) {
649 my $obj = DBM::Deep->new(
651 base_offset => $subloc,
655 if ($self->root->{autobless}) {
657 # Skip over value and plain key to see if object needs
660 seek($fh, $DATA_LENGTH_SIZE + $INDEX_SIZE, 1);
663 read( $fh, $size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
664 if ($size) { seek($fh, $size, 1); }
667 read( $fh, $bless_bit, 1);
668 if (ord($bless_bit)) {
670 # Yes, object needs to be re-blessed
673 read( $fh, $size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
674 if ($size) { read( $fh, $class_name, $size); }
675 if ($class_name) { $obj = bless( $obj, $class_name ); }
683 # Otherwise return actual value
685 elsif ($signature eq SIG_DATA) {
688 read( $fh, $size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
689 if ($size) { read( $fh, $value, $size); }
694 # Key exists, but content is null
704 # Delete single key/value pair given tag and MD5 digested key.
707 my ($tag, $md5) = @_;
708 my $keys = $tag->{content};
713 # Iterate through buckets, looking for a key match
716 for (my $i=0; $i<$MAX_BUCKETS; $i++) {
717 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
718 my $subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
722 # Hit end of list, no match
727 if ( $md5 ne $key ) {
732 # Matched key -- delete bucket and return
734 seek($fh, $tag->{offset} + ($i * $BUCKET_SIZE), 0);
735 print($fh substr($keys, ($i+1) * $BUCKET_SIZE ) );
736 print($fh chr(0) x $BUCKET_SIZE );
746 # Check existence of single key given tag and MD5 digested key.
749 my ($tag, $md5) = @_;
750 my $keys = $tag->{content};
753 # Iterate through buckets, looking for a key match
756 for (my $i=0; $i<$MAX_BUCKETS; $i++) {
757 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
758 my $subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
762 # Hit end of list, no match
767 if ( $md5 ne $key ) {
772 # Matched key -- return true
780 sub _find_bucket_list {
782 # Locate offset for bucket list, given digested key
788 # Locate offset for bucket list using digest index system
791 my $tag = $self->_load_tag($self->base_offset);
792 if (!$tag) { return; }
794 while ($tag->{signature} ne SIG_BLIST) {
795 $tag = $self->_index_lookup($tag, ord(substr($md5, $ch, 1)));
796 if (!$tag) { return; }
803 sub _traverse_index {
805 # Scan index and recursively step into deeper levels, looking for next key.
807 my ($self, $offset, $ch, $force_return_next) = @_;
808 $force_return_next = undef unless $force_return_next;
810 my $tag = $self->_load_tag( $offset );
814 if ($tag->{signature} ne SIG_BLIST) {
815 my $content = $tag->{content};
817 if ($self->{return_next}) { $start = 0; }
818 else { $start = ord(substr($self->{prev_md5}, $ch, 1)); }
820 for (my $index = $start; $index < 256; $index++) {
821 my $subloc = unpack($LONG_PACK, substr($content, $index * $LONG_SIZE, $LONG_SIZE) );
823 my $result = $self->_traverse_index( $subloc, $ch + 1, $force_return_next );
824 if (defined($result)) { return $result; }
828 $self->{return_next} = 1;
831 elsif ($tag->{signature} eq SIG_BLIST) {
832 my $keys = $tag->{content};
833 if ($force_return_next) { $self->{return_next} = 1; }
836 # Iterate through buckets, looking for a key match
838 for (my $i=0; $i<$MAX_BUCKETS; $i++) {
839 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
840 my $subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
844 # End of bucket list -- return to outer loop
846 $self->{return_next} = 1;
849 elsif ($key eq $self->{prev_md5}) {
851 # Located previous key -- return next one found
853 $self->{return_next} = 1;
856 elsif ($self->{return_next}) {
858 # Seek to bucket location and skip over signature
860 seek($fh, $subloc + SIG_SIZE, 0);
863 # Skip over value to get to plain key
866 read( $fh, $size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
867 if ($size) { seek($fh, $size, 1); }
870 # Read in plain key and return as scalar
873 read( $fh, $size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
874 if ($size) { read( $fh, $plain_key, $size); }
880 $self->{return_next} = 1;
881 } # tag is a bucket list
888 # Locate next key, given digested previous one
890 my $self = _get_self($_[0]);
892 $self->{prev_md5} = $_[1] ? $_[1] : undef;
893 $self->{return_next} = 0;
896 # If the previous key was not specifed, start at the top and
897 # return the first one found.
899 if (!$self->{prev_md5}) {
900 $self->{prev_md5} = chr(0) x $HASH_SIZE;
901 $self->{return_next} = 1;
904 return $self->_traverse_index( $self->base_offset, 0 );
909 # If db locking is set, flock() the db file. If called multiple
910 # times before unlock(), then the same number of unlocks() must
911 # be called before the lock is released.
913 my $self = _get_self($_[0]);
915 $type = LOCK_EX unless defined $type;
917 if ($self->root->{locking}) {
918 if (!$self->root->{locked}) { flock($self->fh, $type); }
919 $self->root->{locked}++;
929 # If db locking is set, unlock the db file. See note in lock()
930 # regarding calling lock() multiple times.
932 my $self = _get_self($_[0]);
934 if ($self->root->{locking} && $self->root->{locked} > 0) {
935 $self->root->{locked}--;
936 if (!$self->root->{locked}) { flock($self->fh, LOCK_UN); }
944 #XXX These uses of ref() need verified
947 # Copy single level of keys or elements to new DB handle.
948 # Recurse for nested structures
950 my $self = _get_self($_[0]);
953 if ($self->type eq TYPE_HASH) {
954 my $key = $self->first_key();
956 my $value = $self->get($key);
957 #XXX This doesn't work with autobless
958 if (!ref($value)) { $db_temp->{$key} = $value; }
960 my $type = $value->type;
961 if ($type eq TYPE_HASH) { $db_temp->{$key} = {}; }
962 else { $db_temp->{$key} = []; }
963 $value->_copy_node( $db_temp->{$key} );
965 $key = $self->next_key($key);
969 my $length = $self->length();
970 for (my $index = 0; $index < $length; $index++) {
971 my $value = $self->get($index);
972 if (!ref($value)) { $db_temp->[$index] = $value; }
973 #XXX NO tests for this code
975 my $type = $value->type;
976 if ($type eq TYPE_HASH) { $db_temp->[$index] = {}; }
977 else { $db_temp->[$index] = []; }
978 $value->_copy_node( $db_temp->[$index] );
986 # Recursively export into standard Perl hashes and arrays.
988 my $self = _get_self($_[0]);
991 if ($self->type eq TYPE_HASH) { $temp = {}; }
992 elsif ($self->type eq TYPE_ARRAY) { $temp = []; }
995 $self->_copy_node( $temp );
1003 # Recursively import Perl hash/array structure
1005 #XXX This use of ref() seems to be ok
1006 if (!ref($_[0])) { return; } # Perl calls import() on use -- ignore
1008 my $self = _get_self($_[0]);
1011 #XXX This use of ref() seems to be ok
1012 if (!ref($struct)) {
1014 # struct is not a reference, so just import based on our type
1018 if ($self->type eq TYPE_HASH) { $struct = {@_}; }
1019 elsif ($self->type eq TYPE_ARRAY) { $struct = [@_]; }
1022 my $r = Scalar::Util::reftype($struct) || '';
1023 if ($r eq "HASH" && $self->type eq TYPE_HASH) {
1024 foreach my $key (keys %$struct) { $self->put($key, $struct->{$key}); }
1026 elsif ($r eq "ARRAY" && $self->type eq TYPE_ARRAY) {
1027 $self->push( @$struct );
1030 return $self->_throw_error("Cannot import: type mismatch");
1038 # Rebuild entire database into new file, then move
1039 # it back on top of original.
1041 my $self = _get_self($_[0]);
1043 #XXX Need to create a new test for this
1044 # if ($self->root->{links} > 1) {
1045 # return $self->_throw_error("Cannot optimize: reference count is greater than 1");
1048 my $db_temp = DBM::Deep->new(
1049 file => $self->root->{file} . '.tmp',
1053 return $self->_throw_error("Cannot optimize: failed to open temp file: $!");
1057 $self->_copy_node( $db_temp );
1061 # Attempt to copy user, group and permissions over to new file
1063 my @stats = stat($self->fh);
1064 my $perms = $stats[2] & 07777;
1065 my $uid = $stats[4];
1066 my $gid = $stats[5];
1067 chown( $uid, $gid, $self->root->{file} . '.tmp' );
1068 chmod( $perms, $self->root->{file} . '.tmp' );
1070 # q.v. perlport for more information on this variable
1071 if ( $^O eq 'MSWin32' ) {
1073 # Potential race condition when optmizing on Win32 with locking.
1074 # The Windows filesystem requires that the filehandle be closed
1075 # before it is overwritten with rename(). This could be redone
1082 if (!rename $self->root->{file} . '.tmp', $self->root->{file}) {
1083 unlink $self->root->{file} . '.tmp';
1085 return $self->_throw_error("Optimize failed: Cannot copy temp file over original: $!");
1097 # Make copy of object and return
1099 my $self = _get_self($_[0]);
1101 return DBM::Deep->new(
1102 type => $self->type,
1103 base_offset => $self->base_offset,
1109 my %is_legal_filter = map {
1112 store_key store_value
1113 fetch_key fetch_value
1118 # Setup filter function for storing or fetching the key or value
1120 my $self = _get_self($_[0]);
1121 my $type = lc $_[1];
1122 my $func = $_[2] ? $_[2] : undef;
1124 if ( $is_legal_filter{$type} ) {
1125 $self->root->{"filter_$type"} = $func;
1139 # Get access to the root structure
1141 my $self = _get_self($_[0]);
1142 return $self->{root};
1147 # Get access to the raw FileHandle
1149 #XXX It will be useful, though, when we split out HASH and ARRAY
1150 my $self = _get_self($_[0]);
1151 return $self->root->{fh};
1156 # Get type of current node (TYPE_HASH or TYPE_ARRAY)
1158 my $self = _get_self($_[0]);
1159 return $self->{type};
1164 # Get base_offset of current node (TYPE_HASH or TYPE_ARRAY)
1166 my $self = _get_self($_[0]);
1167 return $self->{base_offset};
1172 # Get last error string, or undef if no error
1175 ? ( _get_self($_[0])->{root}->{error} or undef )
1185 # Store error string in self
1187 my $self = _get_self($_[0]);
1188 my $error_text = $_[1];
1190 $self->root->{error} = $error_text;
1192 unless ($self->root->{debug}) {
1193 die "DBM::Deep: $error_text\n";
1196 warn "DBM::Deep: $error_text\n";
1204 my $self = _get_self($_[0]);
1206 undef $self->root->{error};
1211 # Precalculate index, bucket and bucket list sizes
1214 #XXX I don't like this ...
1215 set_pack() unless defined $LONG_SIZE;
1217 $INDEX_SIZE = 256 * $LONG_SIZE;
1218 $BUCKET_SIZE = $HASH_SIZE + $LONG_SIZE;
1219 $BUCKET_LIST_SIZE = $MAX_BUCKETS * $BUCKET_SIZE;
1224 # Set pack/unpack modes (see file header for more)
1226 my ($long_s, $long_p, $data_s, $data_p) = @_;
1228 $LONG_SIZE = $long_s ? $long_s : 4;
1229 $LONG_PACK = $long_p ? $long_p : 'N';
1231 $DATA_LENGTH_SIZE = $data_s ? $data_s : 4;
1232 $DATA_LENGTH_PACK = $data_p ? $data_p : 'N';
1239 # Set key digest function (default is MD5)
1241 my ($digest_func, $hash_size) = @_;
1243 $DIGEST_FUNC = $digest_func ? $digest_func : \&Digest::MD5::md5;
1244 $HASH_SIZE = $hash_size ? $hash_size : 16;
1250 # tie() methods (hashes and arrays)
1255 # Store single hash key/value or array element in database.
1257 my $self = _get_self($_[0]);
1258 my $key = ($self->root->{filter_store_key} && $self->type eq TYPE_HASH) ? $self->root->{filter_store_key}->($_[1]) : $_[1];
1259 #XXX What is ref() checking here?
1260 #YYY User may be storing a hash, in which case we do not want it run
1261 #YYY through the filtering system
1262 my $value = ($self->root->{filter_store_value} && !ref($_[2])) ? $self->root->{filter_store_value}->($_[2]) : $_[2];
1264 my $unpacked_key = $key;
1265 if (($self->type eq TYPE_ARRAY) && ($key =~ /^\d+$/)) { $key = pack($LONG_PACK, $key); }
1266 my $md5 = $DIGEST_FUNC->($key);
1269 # Make sure file is open
1271 if (!defined($self->fh) && !$self->_open()) {
1279 # Request exclusive lock for writing
1281 $self->lock( LOCK_EX );
1284 # If locking is enabled, set 'end' parameter again, in case another
1285 # DB instance appended to our file while we were unlocked.
1287 if ($self->root->{locking} || $self->root->{volatile}) {
1288 $self->root->{end} = (stat($fh))[7];
1292 # Locate offset for bucket list using digest index system
1294 my $tag = $self->_load_tag($self->base_offset);
1296 $tag = $self->_create_tag($self->base_offset, SIG_INDEX, chr(0) x $INDEX_SIZE);
1300 while ($tag->{signature} ne SIG_BLIST) {
1301 my $num = ord(substr($md5, $ch, 1));
1302 my $new_tag = $self->_index_lookup($tag, $num);
1304 my $ref_loc = $tag->{offset} + ($num * $LONG_SIZE);
1305 seek($fh, $ref_loc, 0);
1306 print($fh pack($LONG_PACK, $self->root->{end}) );
1308 $tag = $self->_create_tag($self->root->{end}, SIG_BLIST, chr(0) x $BUCKET_LIST_SIZE);
1309 $tag->{ref_loc} = $ref_loc;
1314 my $ref_loc = $tag->{offset} + ($num * $LONG_SIZE);
1316 $tag->{ref_loc} = $ref_loc;
1323 # Add key/value to bucket list
1325 my $result = $self->_add_bucket( $tag, $md5, $key, $value );
1328 # If this object is an array, and bucket was not a replace, and key is numerical,
1329 # and index is equal or greater than current length, advance length variable.
1331 if (($result == 2) && ($self->type eq TYPE_ARRAY) && ($unpacked_key =~ /^\d+$/) && ($unpacked_key >= $self->FETCHSIZE())) {
1332 $self->STORESIZE( $unpacked_key + 1 );
1342 # Fetch single value or element given plain key or array index
1344 my $self = _get_self($_[0]);
1347 if ( $self->type eq TYPE_HASH ) {
1348 if ( my $filter = $self->root->{filter_store_key} ) {
1349 $key = $filter->( $key );
1352 elsif ( $self->type eq TYPE_ARRAY ) {
1353 if ( $key =~ /^\d+$/ ) {
1354 $key = pack($LONG_PACK, $key);
1358 my $md5 = $DIGEST_FUNC->($key);
1361 # Make sure file is open
1363 if (!defined($self->fh)) { $self->_open(); }
1366 # Request shared lock for reading
1368 $self->lock( LOCK_SH );
1370 my $tag = $self->_find_bucket_list( $md5 );
1377 # Get value from bucket list
1379 my $result = $self->_get_bucket_value( $tag, $md5 );
1383 #XXX What is ref() checking here?
1384 return ($result && !ref($result) && $self->root->{filter_fetch_value}) ? $self->root->{filter_fetch_value}->($result) : $result;
1389 # Delete single key/value pair or element given plain key or array index
1391 my $self = _get_self($_[0]);
1392 my $key = ($self->root->{filter_store_key} && $self->type eq TYPE_HASH) ? $self->root->{filter_store_key}->($_[1]) : $_[1];
1394 my $unpacked_key = $key;
1395 if (($self->type eq TYPE_ARRAY) && ($key =~ /^\d+$/)) { $key = pack($LONG_PACK, $key); }
1396 my $md5 = $DIGEST_FUNC->($key);
1399 # Make sure file is open
1401 if (!defined($self->fh)) { $self->_open(); }
1404 # Request exclusive lock for writing
1406 $self->lock( LOCK_EX );
1408 my $tag = $self->_find_bucket_list( $md5 );
1417 my $result = $self->_delete_bucket( $tag, $md5 );
1420 # If this object is an array and the key deleted was on the end of the stack,
1421 # decrement the length variable.
1423 if ($result && ($self->type eq TYPE_ARRAY) && ($unpacked_key == $self->FETCHSIZE() - 1)) {
1424 $self->STORESIZE( $unpacked_key );
1434 # Check if a single key or element exists given plain key or array index
1436 my $self = _get_self($_[0]);
1437 my $key = ($self->root->{filter_store_key} && $self->type eq TYPE_HASH) ? $self->root->{filter_store_key}->($_[1]) : $_[1];
1439 if (($self->type eq TYPE_ARRAY) && ($key =~ /^\d+$/)) { $key = pack($LONG_PACK, $key); }
1440 my $md5 = $DIGEST_FUNC->($key);
1443 # Make sure file is open
1445 if (!defined($self->fh)) { $self->_open(); }
1448 # Request shared lock for reading
1450 $self->lock( LOCK_SH );
1452 my $tag = $self->_find_bucket_list( $md5 );
1455 # For some reason, the built-in exists() function returns '' for false
1463 # Check if bucket exists and return 1 or ''
1465 my $result = $self->_bucket_exists( $tag, $md5 ) || '';
1474 # Clear all keys from hash, or all elements from array.
1476 my $self = _get_self($_[0]);
1479 # Make sure file is open
1481 if (!defined($self->fh)) { $self->_open(); }
1484 # Request exclusive lock for writing
1486 $self->lock( LOCK_EX );
1490 seek($fh, $self->base_offset, 0);
1496 $self->_create_tag($self->base_offset, $self->type, chr(0) x $INDEX_SIZE);
1505 # Locate and return first key (in no particular order)
1507 my $self = _get_self($_[0]);
1508 if ($self->type ne TYPE_HASH) {
1509 return $self->_throw_error("FIRSTKEY method only supported for hashes");
1513 # Make sure file is open
1515 if (!defined($self->fh)) { $self->_open(); }
1518 # Request shared lock for reading
1520 $self->lock( LOCK_SH );
1522 my $result = $self->_get_next_key();
1526 return ($result && $self->root->{filter_fetch_key}) ? $self->root->{filter_fetch_key}->($result) : $result;
1531 # Return next key (in no particular order), given previous one
1533 my $self = _get_self($_[0]);
1534 if ($self->type ne TYPE_HASH) {
1535 return $self->_throw_error("NEXTKEY method only supported for hashes");
1537 my $prev_key = ($self->root->{filter_store_key} && $self->type eq TYPE_HASH) ? $self->root->{filter_store_key}->($_[1]) : $_[1];
1538 my $prev_md5 = $DIGEST_FUNC->($prev_key);
1541 # Make sure file is open
1543 if (!defined($self->fh)) { $self->_open(); }
1546 # Request shared lock for reading
1548 $self->lock( LOCK_SH );
1550 my $result = $self->_get_next_key( $prev_md5 );
1554 return ($result && $self->root->{filter_fetch_key}) ? $self->root->{filter_fetch_key}->($result) : $result;
1558 # The following methods are for arrays only
1563 # Return the length of the array
1565 my $self = _get_self($_[0]);
1566 if ($self->type ne TYPE_ARRAY) {
1567 return $self->_throw_error("FETCHSIZE method only supported for arrays");
1570 my $SAVE_FILTER = $self->root->{filter_fetch_value};
1571 $self->root->{filter_fetch_value} = undef;
1573 my $packed_size = $self->FETCH('length');
1575 $self->root->{filter_fetch_value} = $SAVE_FILTER;
1577 if ($packed_size) { return int(unpack($LONG_PACK, $packed_size)); }
1583 # Set the length of the array
1585 my $self = _get_self($_[0]);
1586 if ($self->type ne TYPE_ARRAY) {
1587 return $self->_throw_error("STORESIZE method only supported for arrays");
1589 my $new_length = $_[1];
1591 my $SAVE_FILTER = $self->root->{filter_store_value};
1592 $self->root->{filter_store_value} = undef;
1594 my $result = $self->STORE('length', pack($LONG_PACK, $new_length));
1596 $self->root->{filter_store_value} = $SAVE_FILTER;
1603 # Remove and return the last element on the array
1605 my $self = _get_self($_[0]);
1606 if ($self->type ne TYPE_ARRAY) {
1607 return $self->_throw_error("POP method only supported for arrays");
1609 my $length = $self->FETCHSIZE();
1612 my $content = $self->FETCH( $length - 1 );
1613 $self->DELETE( $length - 1 );
1623 # Add new element(s) to the end of the array
1625 my $self = _get_self(shift);
1626 if ($self->type ne TYPE_ARRAY) {
1627 return $self->_throw_error("PUSH method only supported for arrays");
1629 my $length = $self->FETCHSIZE();
1631 while (my $content = shift @_) {
1632 $self->STORE( $length, $content );
1639 # Remove and return first element on the array.
1640 # Shift over remaining elements to take up space.
1642 my $self = _get_self($_[0]);
1643 if ($self->type ne TYPE_ARRAY) {
1644 return $self->_throw_error("SHIFT method only supported for arrays");
1646 my $length = $self->FETCHSIZE();
1649 my $content = $self->FETCH( 0 );
1652 # Shift elements over and remove last one.
1654 for (my $i = 0; $i < $length - 1; $i++) {
1655 $self->STORE( $i, $self->FETCH($i + 1) );
1657 $self->DELETE( $length - 1 );
1668 # Insert new element(s) at beginning of array.
1669 # Shift over other elements to make space.
1671 my $self = _get_self($_[0]);shift @_;
1672 if ($self->type ne TYPE_ARRAY) {
1673 return $self->_throw_error("UNSHIFT method only supported for arrays");
1675 my @new_elements = @_;
1676 my $length = $self->FETCHSIZE();
1677 my $new_size = scalar @new_elements;
1680 for (my $i = $length - 1; $i >= 0; $i--) {
1681 $self->STORE( $i + $new_size, $self->FETCH($i) );
1685 for (my $i = 0; $i < $new_size; $i++) {
1686 $self->STORE( $i, $new_elements[$i] );
1692 # Splices section of array with optional new section.
1693 # Returns deleted section, or last element deleted in scalar context.
1695 my $self = _get_self($_[0]);shift @_;
1696 if ($self->type ne TYPE_ARRAY) {
1697 return $self->_throw_error("SPLICE method only supported for arrays");
1699 my $length = $self->FETCHSIZE();
1702 # Calculate offset and length of splice
1704 my $offset = shift || 0;
1705 if ($offset < 0) { $offset += $length; }
1708 if (scalar @_) { $splice_length = shift; }
1709 else { $splice_length = $length - $offset; }
1710 if ($splice_length < 0) { $splice_length += ($length - $offset); }
1713 # Setup array with new elements, and copy out old elements for return
1715 my @new_elements = @_;
1716 my $new_size = scalar @new_elements;
1718 my @old_elements = ();
1719 for (my $i = $offset; $i < $offset + $splice_length; $i++) {
1720 push @old_elements, $self->FETCH( $i );
1724 # Adjust array length, and shift elements to accomodate new section.
1726 if ( $new_size != $splice_length ) {
1727 if ($new_size > $splice_length) {
1728 for (my $i = $length - 1; $i >= $offset + $splice_length; $i--) {
1729 $self->STORE( $i + ($new_size - $splice_length), $self->FETCH($i) );
1733 for (my $i = $offset + $splice_length; $i < $length; $i++) {
1734 $self->STORE( $i + ($new_size - $splice_length), $self->FETCH($i) );
1736 for (my $i = 0; $i < $splice_length - $new_size; $i++) {
1737 $self->DELETE( $length - 1 );
1744 # Insert new elements into array
1746 for (my $i = $offset; $i < $offset + $new_size; $i++) {
1747 $self->STORE( $i, shift @new_elements );
1751 # Return deleted section, or last element in scalar context.
1753 return wantarray ? @old_elements : $old_elements[-1];
1756 #XXX We don't need to define it.
1757 #XXX It will be useful, though, when we split out HASH and ARRAY
1760 # Perl will call EXTEND() when the array is likely to grow.
1761 # We don't care, but include it for compatibility.
1766 # Public method aliases
1768 *put = *store = *STORE;
1769 *get = *fetch = *FETCH;
1773 *first_key = *FIRSTKEY;
1774 *next_key = *NEXTKEY;
1775 *length = *FETCHSIZE;
1779 *unshift = *UNSHIFT;
1782 package DBM::Deep::_::Root;
1797 filter_store_key => undef,
1798 filter_store_value => undef,
1799 filter_fetch_key => undef,
1800 filter_fetch_value => undef,
1811 return unless $self;
1813 close $self->{fh} if $self->{fh};
1824 DBM::Deep - A pure perl multi-level hash/array DBM
1829 my $db = DBM::Deep->new( "foo.db" );
1831 $db->{key} = 'value'; # tie() style
1834 $db->put('key', 'value'); # OO style
1835 print $db->get('key');
1837 # true multi-level support
1838 $db->{my_complex} = [
1839 'hello', { perl => 'rules' },
1844 A unique flat-file database module, written in pure perl. True
1845 multi-level hash/array support (unlike MLDBM, which is faked), hybrid
1846 OO / tie() interface, cross-platform FTPable files, and quite fast. Can
1847 handle millions of keys and unlimited hash levels without significant
1848 slow-down. Written from the ground-up in pure perl -- this is NOT a
1849 wrapper around a C-based DBM. Out-of-the-box compatibility with Unix,
1850 Mac OS X and Windows.
1854 Hopefully you are using CPAN's excellent Perl module, which will download
1855 and install the module for you. If not, get the tarball, and run these
1867 Construction can be done OO-style (which is the recommended way), or using
1868 Perl's tie() function. Both are examined here.
1870 =head2 OO CONSTRUCTION
1872 The recommended way to construct a DBM::Deep object is to use the new()
1873 method, which gets you a blessed, tied hash or array reference.
1875 my $db = DBM::Deep->new( "foo.db" );
1877 This opens a new database handle, mapped to the file "foo.db". If this
1878 file does not exist, it will automatically be created. DB files are
1879 opened in "r+" (read/write) mode, and the type of object returned is a
1880 hash, unless otherwise specified (see L<OPTIONS> below).
1884 You can pass a number of options to the constructor to specify things like
1885 locking, autoflush, etc. This is done by passing an inline hash:
1887 my $db = DBM::Deep->new(
1893 Notice that the filename is now specified I<inside> the hash with
1894 the "file" parameter, as opposed to being the sole argument to the
1895 constructor. This is required if any options are specified.
1896 See L<OPTIONS> below for the complete list.
1900 You can also start with an array instead of a hash. For this, you must
1901 specify the C<type> parameter:
1903 my $db = DBM::Deep->new(
1905 type => DBM::Deep->TYPE_ARRAY
1908 B<Note:> Specifing the C<type> parameter only takes effect when beginning
1909 a new DB file. If you create a DBM::Deep object with an existing file, the
1910 C<type> will be loaded from the file header, and ignored if it is passed
1913 =head2 TIE CONSTRUCTION
1915 Alternatively, you can create a DBM::Deep handle by using Perl's built-in
1916 tie() function. This is not ideal, because you get only a basic, tied hash
1917 (or array) which is not blessed, so you can't call any functions on it.
1920 tie %hash, "DBM::Deep", "foo.db";
1923 tie @array, "DBM::Deep", "bar.db";
1925 As with the OO constructor, you can replace the DB filename parameter with
1926 a hash containing one or more options (see L<OPTIONS> just below for the
1929 tie %hash, "DBM::Deep", {
1937 There are a number of options that can be passed in when constructing your
1938 DBM::Deep objects. These apply to both the OO- and tie- based approaches.
1944 Filename of the DB file to link the handle to. You can pass a full absolute
1945 filesystem path, partial path, or a plain filename if the file is in the
1946 current working directory. This is a required parameter.
1950 File open mode (read-only, read-write, etc.) string passed to Perl's FileHandle
1951 module. This is an optional parameter, and defaults to "r+" (read/write).
1952 B<Note:> If the default (r+) mode is selected, the file will also be auto-
1953 created if it doesn't exist.
1957 This parameter specifies what type of object to create, a hash or array. Use
1958 one of these two constants: C<DBM::Deep-E<gt>TYPE_HASH> or C<DBM::Deep-E<gt>TYPE_ARRAY>.
1959 This only takes effect when beginning a new file. This is an optional
1960 parameter, and defaults to C<DBM::Deep-E<gt>TYPE_HASH>.
1964 Specifies whether locking is to be enabled. DBM::Deep uses Perl's Fnctl flock()
1965 function to lock the database in exclusive mode for writes, and shared mode for
1966 reads. Pass any true value to enable. This affects the base DB handle I<and
1967 any child hashes or arrays> that use the same DB file. This is an optional
1968 parameter, and defaults to 0 (disabled). See L<LOCKING> below for more.
1972 Specifies whether autoflush is to be enabled on the underlying FileHandle.
1973 This obviously slows down write operations, but is required if you may have
1974 multiple processes accessing the same DB file (also consider enable I<locking>
1975 or at least I<volatile>). Pass any true value to enable. This is an optional
1976 parameter, and defaults to 0 (disabled).
1980 If I<volatile> mode is enabled, DBM::Deep will stat() the DB file before each
1981 STORE() operation. This is required if an outside force may change the size of
1982 the file between transactions. Locking also implicitly enables volatile. This
1983 is useful if you want to use a different locking system or write your own. Pass
1984 any true value to enable. This is an optional parameter, and defaults to 0
1989 If I<autobless> mode is enabled, DBM::Deep will preserve blessed hashes, and
1990 restore them when fetched. This is an B<experimental> feature, and does have
1991 side-effects. Basically, when hashes are re-blessed into their original
1992 classes, they are no longer blessed into the DBM::Deep class! So you won't be
1993 able to call any DBM::Deep methods on them. You have been warned.
1994 This is an optional parameter, and defaults to 0 (disabled).
1998 See L<FILTERS> below.
2002 Setting I<debug> mode will make all errors non-fatal, dump them out to
2003 STDERR, and continue on. This is for debugging purposes only, and probably
2004 not what you want. This is an optional parameter, and defaults to 0 (disabled).
2008 Instead of passing a file path, you can instead pass a handle to an pre-opened
2009 filehandle. Note: Beware of using the magick *DATA handle, as this actually
2010 contains your entire Perl script, as well as the data following the __DATA__
2011 marker. This will not work, because DBM::Deep uses absolute seek()s into the
2012 file. Instead, consider reading *DATA into an IO::Scalar handle, then passing
2017 =head1 TIE INTERFACE
2019 With DBM::Deep you can access your databases using Perl's standard hash/array
2020 syntax. Because all DBM::Deep objects are I<tied> to hashes or arrays, you can treat
2021 them as such. DBM::Deep will intercept all reads/writes and direct them to the right
2022 place -- the DB file. This has nothing to do with the L<TIE CONSTRUCTION>
2023 section above. This simply tells you how to use DBM::Deep using regular hashes
2024 and arrays, rather than calling functions like C<get()> and C<put()> (although those
2025 work too). It is entirely up to you how to want to access your databases.
2029 You can treat any DBM::Deep object like a normal Perl hash reference. Add keys,
2030 or even nested hashes (or arrays) using standard Perl syntax:
2032 my $db = DBM::Deep->new( "foo.db" );
2034 $db->{mykey} = "myvalue";
2036 $db->{myhash}->{subkey} = "subvalue";
2038 print $db->{myhash}->{subkey} . "\n";
2040 You can even step through hash keys using the normal Perl C<keys()> function:
2042 foreach my $key (keys %$db) {
2043 print "$key: " . $db->{$key} . "\n";
2046 Remember that Perl's C<keys()> function extracts I<every> key from the hash and
2047 pushes them onto an array, all before the loop even begins. If you have an
2048 extra large hash, this may exhaust Perl's memory. Instead, consider using
2049 Perl's C<each()> function, which pulls keys/values one at a time, using very
2052 while (my ($key, $value) = each %$db) {
2053 print "$key: $value\n";
2056 Please note that when using C<each()>, you should always pass a direct
2057 hash reference, not a lookup. Meaning, you should B<never> do this:
2060 while (my ($key, $value) = each %{$db->{foo}}) { # BAD
2062 This causes an infinite loop, because for each iteration, Perl is calling
2063 FETCH() on the $db handle, resulting in a "new" hash for foo every time, so
2064 it effectively keeps returning the first key over and over again. Instead,
2065 assign a temporary variable to C<$db->{foo}>, then pass that to each().
2069 As with hashes, you can treat any DBM::Deep object like a normal Perl array
2070 reference. This includes inserting, removing and manipulating elements,
2071 and the C<push()>, C<pop()>, C<shift()>, C<unshift()> and C<splice()> functions.
2072 The object must have first been created using type C<DBM::Deep-E<gt>TYPE_ARRAY>,
2073 or simply be a nested array reference inside a hash. Example:
2075 my $db = DBM::Deep->new(
2076 file => "foo-array.db",
2077 type => DBM::Deep->TYPE_ARRAY
2081 push @$db, "bar", "baz";
2082 unshift @$db, "bah";
2084 my $last_elem = pop @$db; # baz
2085 my $first_elem = shift @$db; # bah
2086 my $second_elem = $db->[1]; # bar
2088 my $num_elements = scalar @$db;
2092 In addition to the I<tie()> interface, you can also use a standard OO interface
2093 to manipulate all aspects of DBM::Deep databases. Each type of object (hash or
2094 array) has its own methods, but both types share the following common methods:
2095 C<put()>, C<get()>, C<exists()>, C<delete()> and C<clear()>.
2101 Stores a new hash key/value pair, or sets an array element value. Takes two
2102 arguments, the hash key or array index, and the new value. The value can be
2103 a scalar, hash ref or array ref. Returns true on success, false on failure.
2105 $db->put("foo", "bar"); # for hashes
2106 $db->put(1, "bar"); # for arrays
2110 Fetches the value of a hash key or array element. Takes one argument: the hash
2111 key or array index. Returns a scalar, hash ref or array ref, depending on the
2114 my $value = $db->get("foo"); # for hashes
2115 my $value = $db->get(1); # for arrays
2119 Checks if a hash key or array index exists. Takes one argument: the hash key
2120 or array index. Returns true if it exists, false if not.
2122 if ($db->exists("foo")) { print "yay!\n"; } # for hashes
2123 if ($db->exists(1)) { print "yay!\n"; } # for arrays
2127 Deletes one hash key/value pair or array element. Takes one argument: the hash
2128 key or array index. Returns true on success, false if not found. For arrays,
2129 the remaining elements located after the deleted element are NOT moved over.
2130 The deleted element is essentially just undefined, which is exactly how Perl's
2131 internal arrays work. Please note that the space occupied by the deleted
2132 key/value or element is B<not> reused again -- see L<UNUSED SPACE RECOVERY>
2133 below for details and workarounds.
2135 $db->delete("foo"); # for hashes
2136 $db->delete(1); # for arrays
2140 Deletes B<all> hash keys or array elements. Takes no arguments. No return
2141 value. Please note that the space occupied by the deleted keys/values or
2142 elements is B<not> reused again -- see L<UNUSED SPACE RECOVERY> below for
2143 details and workarounds.
2145 $db->clear(); # hashes or arrays
2151 For hashes, DBM::Deep supports all the common methods described above, and the
2152 following additional methods: C<first_key()> and C<next_key()>.
2158 Returns the "first" key in the hash. As with built-in Perl hashes, keys are
2159 fetched in an undefined order (which appears random). Takes no arguments,
2160 returns the key as a scalar value.
2162 my $key = $db->first_key();
2166 Returns the "next" key in the hash, given the previous one as the sole argument.
2167 Returns undef if there are no more keys to be fetched.
2169 $key = $db->next_key($key);
2173 Here are some examples of using hashes:
2175 my $db = DBM::Deep->new( "foo.db" );
2177 $db->put("foo", "bar");
2178 print "foo: " . $db->get("foo") . "\n";
2180 $db->put("baz", {}); # new child hash ref
2181 $db->get("baz")->put("buz", "biz");
2182 print "buz: " . $db->get("baz")->get("buz") . "\n";
2184 my $key = $db->first_key();
2186 print "$key: " . $db->get($key) . "\n";
2187 $key = $db->next_key($key);
2190 if ($db->exists("foo")) { $db->delete("foo"); }
2194 For arrays, DBM::Deep supports all the common methods described above, and the
2195 following additional methods: C<length()>, C<push()>, C<pop()>, C<shift()>,
2196 C<unshift()> and C<splice()>.
2202 Returns the number of elements in the array. Takes no arguments.
2204 my $len = $db->length();
2208 Adds one or more elements onto the end of the array. Accepts scalars, hash
2209 refs or array refs. No return value.
2211 $db->push("foo", "bar", {});
2215 Fetches the last element in the array, and deletes it. Takes no arguments.
2216 Returns undef if array is empty. Returns the element value.
2218 my $elem = $db->pop();
2222 Fetches the first element in the array, deletes it, then shifts all the
2223 remaining elements over to take up the space. Returns the element value. This
2224 method is not recommended with large arrays -- see L<LARGE ARRAYS> below for
2227 my $elem = $db->shift();
2231 Inserts one or more elements onto the beginning of the array, shifting all
2232 existing elements over to make room. Accepts scalars, hash refs or array refs.
2233 No return value. This method is not recommended with large arrays -- see
2234 <LARGE ARRAYS> below for details.
2236 $db->unshift("foo", "bar", {});
2240 Performs exactly like Perl's built-in function of the same name. See L<perldoc
2241 -f splice> for usage -- it is too complicated to document here. This method is
2242 not recommended with large arrays -- see L<LARGE ARRAYS> below for details.
2246 Here are some examples of using arrays:
2248 my $db = DBM::Deep->new(
2250 type => DBM::Deep->TYPE_ARRAY
2253 $db->push("bar", "baz");
2254 $db->unshift("foo");
2257 my $len = $db->length();
2258 print "length: $len\n"; # 4
2260 for (my $k=0; $k<$len; $k++) {
2261 print "$k: " . $db->get($k) . "\n";
2264 $db->splice(1, 2, "biz", "baf");
2266 while (my $elem = shift @$db) {
2267 print "shifted: $elem\n";
2272 Enable automatic file locking by passing a true value to the C<locking>
2273 parameter when constructing your DBM::Deep object (see L<SETUP> above).
2275 my $db = DBM::Deep->new(
2280 This causes DBM::Deep to C<flock()> the underlying FileHandle object with exclusive
2281 mode for writes, and shared mode for reads. This is required if you have
2282 multiple processes accessing the same database file, to avoid file corruption.
2283 Please note that C<flock()> does NOT work for files over NFS. See L<DB OVER
2284 NFS> below for more.
2286 =head2 EXPLICIT LOCKING
2288 You can explicitly lock a database, so it remains locked for multiple
2289 transactions. This is done by calling the C<lock()> method, and passing an
2290 optional lock mode argument (defaults to exclusive mode). This is particularly
2291 useful for things like counters, where the current value needs to be fetched,
2292 then incremented, then stored again.
2295 my $counter = $db->get("counter");
2297 $db->put("counter", $counter);
2306 You can pass C<lock()> an optional argument, which specifies which mode to use
2307 (exclusive or shared). Use one of these two constants: C<DBM::Deep-E<gt>LOCK_EX>
2308 or C<DBM::Deep-E<gt>LOCK_SH>. These are passed directly to C<flock()>, and are the
2309 same as the constants defined in Perl's C<Fcntl> module.
2311 $db->lock( DBM::Deep->LOCK_SH );
2315 If you want to implement your own file locking scheme, be sure to create your
2316 DBM::Deep objects setting the C<volatile> option to true. This hints to DBM::Deep
2317 that the DB file may change between transactions. See L<LOW-LEVEL ACCESS>
2320 =head1 IMPORTING/EXPORTING
2322 You can import existing complex structures by calling the C<import()> method,
2323 and export an entire database into an in-memory structure using the C<export()>
2324 method. Both are examined here.
2328 Say you have an existing hash with nested hashes/arrays inside it. Instead of
2329 walking the structure and adding keys/elements to the database as you go,
2330 simply pass a reference to the C<import()> method. This recursively adds
2331 everything to an existing DBM::Deep object for you. Here is an example:
2336 array1 => [ "elem0", "elem1", "elem2" ],
2338 subkey1 => "subvalue1",
2339 subkey2 => "subvalue2"
2343 my $db = DBM::Deep->new( "foo.db" );
2344 $db->import( $struct );
2346 print $db->{key1} . "\n"; # prints "value1"
2348 This recursively imports the entire C<$struct> object into C<$db>, including
2349 all nested hashes and arrays. If the DBM::Deep object contains exsiting data,
2350 keys are merged with the existing ones, replacing if they already exist.
2351 The C<import()> method can be called on any database level (not just the base
2352 level), and works with both hash and array DB types.
2356 B<Note:> Make sure your existing structure has no circular references in it.
2357 These will cause an infinite loop when importing.
2361 Calling the C<export()> method on an existing DBM::Deep object will return
2362 a reference to a new in-memory copy of the database. The export is done
2363 recursively, so all nested hashes/arrays are all exported to standard Perl
2364 objects. Here is an example:
2366 my $db = DBM::Deep->new( "foo.db" );
2368 $db->{key1} = "value1";
2369 $db->{key2} = "value2";
2371 $db->{hash1}->{subkey1} = "subvalue1";
2372 $db->{hash1}->{subkey2} = "subvalue2";
2374 my $struct = $db->export();
2376 print $struct->{key1} . "\n"; # prints "value1"
2378 This makes a complete copy of the database in memory, and returns a reference
2379 to it. The C<export()> method can be called on any database level (not just
2380 the base level), and works with both hash and array DB types. Be careful of
2381 large databases -- you can store a lot more data in a DBM::Deep object than an
2382 in-memory Perl structure.
2386 B<Note:> Make sure your database has no circular references in it.
2387 These will cause an infinite loop when exporting.
2391 DBM::Deep has a number of hooks where you can specify your own Perl function
2392 to perform filtering on incoming or outgoing data. This is a perfect
2393 way to extend the engine, and implement things like real-time compression or
2394 encryption. Filtering applies to the base DB level, and all child hashes /
2395 arrays. Filter hooks can be specified when your DBM::Deep object is first
2396 constructed, or by calling the C<set_filter()> method at any time. There are
2397 four available filter hooks, described below:
2401 =item * filter_store_key
2403 This filter is called whenever a hash key is stored. It
2404 is passed the incoming key, and expected to return a transformed key.
2406 =item * filter_store_value
2408 This filter is called whenever a hash key or array element is stored. It
2409 is passed the incoming value, and expected to return a transformed value.
2411 =item * filter_fetch_key
2413 This filter is called whenever a hash key is fetched (i.e. via
2414 C<first_key()> or C<next_key()>). It is passed the transformed key,
2415 and expected to return the plain key.
2417 =item * filter_fetch_value
2419 This filter is called whenever a hash key or array element is fetched.
2420 It is passed the transformed value, and expected to return the plain value.
2424 Here are the two ways to setup a filter hook:
2426 my $db = DBM::Deep->new(
2428 filter_store_value => \&my_filter_store,
2429 filter_fetch_value => \&my_filter_fetch
2434 $db->set_filter( "filter_store_value", \&my_filter_store );
2435 $db->set_filter( "filter_fetch_value", \&my_filter_fetch );
2437 Your filter function will be called only when dealing with SCALAR keys or
2438 values. When nested hashes and arrays are being stored/fetched, filtering
2439 is bypassed. Filters are called as static functions, passed a single SCALAR
2440 argument, and expected to return a single SCALAR value. If you want to
2441 remove a filter, set the function reference to C<undef>:
2443 $db->set_filter( "filter_store_value", undef );
2445 =head2 REAL-TIME ENCRYPTION EXAMPLE
2447 Here is a working example that uses the I<Crypt::Blowfish> module to
2448 do real-time encryption / decryption of keys & values with DBM::Deep Filters.
2449 Please visit L<http://search.cpan.org/search?module=Crypt::Blowfish> for more
2450 on I<Crypt::Blowfish>. You'll also need the I<Crypt::CBC> module.
2453 use Crypt::Blowfish;
2456 my $cipher = Crypt::CBC->new({
2457 'key' => 'my secret key',
2458 'cipher' => 'Blowfish',
2460 'regenerate_key' => 0,
2461 'padding' => 'space',
2465 my $db = DBM::Deep->new(
2466 file => "foo-encrypt.db",
2467 filter_store_key => \&my_encrypt,
2468 filter_store_value => \&my_encrypt,
2469 filter_fetch_key => \&my_decrypt,
2470 filter_fetch_value => \&my_decrypt,
2473 $db->{key1} = "value1";
2474 $db->{key2} = "value2";
2475 print "key1: " . $db->{key1} . "\n";
2476 print "key2: " . $db->{key2} . "\n";
2482 return $cipher->encrypt( $_[0] );
2485 return $cipher->decrypt( $_[0] );
2488 =head2 REAL-TIME COMPRESSION EXAMPLE
2490 Here is a working example that uses the I<Compress::Zlib> module to do real-time
2491 compression / decompression of keys & values with DBM::Deep Filters.
2492 Please visit L<http://search.cpan.org/search?module=Compress::Zlib> for
2493 more on I<Compress::Zlib>.
2498 my $db = DBM::Deep->new(
2499 file => "foo-compress.db",
2500 filter_store_key => \&my_compress,
2501 filter_store_value => \&my_compress,
2502 filter_fetch_key => \&my_decompress,
2503 filter_fetch_value => \&my_decompress,
2506 $db->{key1} = "value1";
2507 $db->{key2} = "value2";
2508 print "key1: " . $db->{key1} . "\n";
2509 print "key2: " . $db->{key2} . "\n";
2515 return Compress::Zlib::memGzip( $_[0] ) ;
2518 return Compress::Zlib::memGunzip( $_[0] ) ;
2521 B<Note:> Filtering of keys only applies to hashes. Array "keys" are
2522 actually numerical index numbers, and are not filtered.
2524 =head1 ERROR HANDLING
2526 Most DBM::Deep methods return a true value for success, and call die() on
2527 failure. You can wrap calls in an eval block to catch the die. Also, the
2528 actual error message is stored in an internal scalar, which can be fetched by
2529 calling the C<error()> method.
2531 my $db = DBM::Deep->new( "foo.db" ); # create hash
2532 eval { $db->push("foo"); }; # ILLEGAL -- push is array-only call
2534 print $db->error(); # prints error message
2536 You can then call C<clear_error()> to clear the current error state.
2540 If you set the C<debug> option to true when creating your DBM::Deep object,
2541 all errors are considered NON-FATAL, and dumped to STDERR. This is only
2542 for debugging purposes.
2544 =head1 LARGEFILE SUPPORT
2546 If you have a 64-bit system, and your Perl is compiled with both LARGEFILE
2547 and 64-bit support, you I<may> be able to create databases larger than 2 GB.
2548 DBM::Deep by default uses 32-bit file offset tags, but these can be changed
2549 by calling the static C<set_pack()> method before you do anything else.
2551 DBM::Deep::set_pack(8, 'Q');
2553 This tells DBM::Deep to pack all file offsets with 8-byte (64-bit) quad words
2554 instead of 32-bit longs. After setting these values your DB files have a
2555 theoretical maximum size of 16 XB (exabytes).
2559 B<Note:> Changing these values will B<NOT> work for existing database files.
2560 Only change this for new files, and make sure it stays set consistently
2561 throughout the file's life. If you do set these values, you can no longer
2562 access 32-bit DB files. You can, however, call C<set_pack(4, 'N')> to change
2563 back to 32-bit mode.
2567 B<Note:> I have not personally tested files > 2 GB -- all my systems have
2568 only a 32-bit Perl. However, I have received user reports that this does
2571 =head1 LOW-LEVEL ACCESS
2573 If you require low-level access to the underlying FileHandle that DBM::Deep uses,
2574 you can call the C<fh()> method, which returns the handle:
2578 This method can be called on the root level of the datbase, or any child
2579 hashes or arrays. All levels share a I<root> structure, which contains things
2580 like the FileHandle, a reference counter, and all your options you specified
2581 when you created the object. You can get access to this root structure by
2582 calling the C<root()> method.
2584 my $root = $db->root();
2586 This is useful for changing options after the object has already been created,
2587 such as enabling/disabling locking, volatile or debug modes. You can also
2588 store your own temporary user data in this structure (be wary of name
2589 collision), which is then accessible from any child hash or array.
2591 =head1 CUSTOM DIGEST ALGORITHM
2593 DBM::Deep by default uses the I<Message Digest 5> (MD5) algorithm for hashing
2594 keys. However you can override this, and use another algorithm (such as SHA-256)
2595 or even write your own. But please note that DBM::Deep currently expects zero
2596 collisions, so your algorithm has to be I<perfect>, so to speak.
2597 Collision detection may be introduced in a later version.
2601 You can specify a custom digest algorithm by calling the static C<set_digest()>
2602 function, passing a reference to a subroutine, and the length of the algorithm's
2603 hashes (in bytes). This is a global static function, which affects ALL DBM::Deep
2604 objects. Here is a working example that uses a 256-bit hash from the
2605 I<Digest::SHA256> module. Please see
2606 L<http://search.cpan.org/search?module=Digest::SHA256> for more.
2611 my $context = Digest::SHA256::new(256);
2613 DBM::Deep::set_digest( \&my_digest, 32 );
2615 my $db = DBM::Deep->new( "foo-sha.db" );
2617 $db->{key1} = "value1";
2618 $db->{key2} = "value2";
2619 print "key1: " . $db->{key1} . "\n";
2620 print "key2: " . $db->{key2} . "\n";
2626 return substr( $context->hash($_[0]), 0, 32 );
2629 B<Note:> Your returned digest strings must be B<EXACTLY> the number
2630 of bytes you specify in the C<set_digest()> function (in this case 32).
2632 =head1 CIRCULAR REFERENCES
2634 DBM::Deep has B<experimental> support for circular references. Meaning you
2635 can have a nested hash key or array element that points to a parent object.
2636 This relationship is stored in the DB file, and is preserved between sessions.
2639 my $db = DBM::Deep->new( "foo.db" );
2642 $db->{circle} = $db; # ref to self
2644 print $db->{foo} . "\n"; # prints "foo"
2645 print $db->{circle}->{foo} . "\n"; # prints "foo" again
2647 One catch is, passing the object to a function that recursively walks the
2648 object tree (such as I<Data::Dumper> or even the built-in C<optimize()> or
2649 C<export()> methods) will result in an infinite loop. The other catch is,
2650 if you fetch the I<key> of a circular reference (i.e. using the C<first_key()>
2651 or C<next_key()> methods), you will get the I<target object's key>, not the
2652 ref's key. This gets even more interesting with the above example, where
2653 the I<circle> key points to the base DB object, which technically doesn't
2654 have a key. So I made DBM::Deep return "[base]" as the key name in that
2657 =head1 CAVEATS / ISSUES / BUGS
2659 This section describes all the known issues with DBM::Deep. It you have found
2660 something that is not listed here, please send e-mail to L<jhuckaby@cpan.org>.
2662 =head2 UNUSED SPACE RECOVERY
2664 One major caveat with DBM::Deep is that space occupied by existing keys and
2665 values is not recovered when they are deleted. Meaning if you keep deleting
2666 and adding new keys, your file will continuously grow. I am working on this,
2667 but in the meantime you can call the built-in C<optimize()> method from time to
2668 time (perhaps in a crontab or something) to recover all your unused space.
2670 $db->optimize(); # returns true on success
2672 This rebuilds the ENTIRE database into a new file, then moves it on top of
2673 the original. The new file will have no unused space, thus it will take up as
2674 little disk space as possible. Please note that this operation can take
2675 a long time for large files, and you need enough disk space to temporarily hold
2676 2 copies of your DB file. The temporary file is created in the same directory
2677 as the original, named with a ".tmp" extension, and is deleted when the
2678 operation completes. Oh, and if locking is enabled, the DB is automatically
2679 locked for the entire duration of the copy.
2683 B<WARNING:> Only call optimize() on the top-level node of the database, and
2684 make sure there are no child references lying around. DBM::Deep keeps a reference
2685 counter, and if it is greater than 1, optimize() will abort and return undef.
2687 =head2 AUTOVIVIFICATION
2689 Unfortunately, autovivification doesn't work with tied hashes. This appears to
2690 be a bug in Perl's tie() system, as I<Jakob Schmidt> encountered the very same
2691 issue with his I<DWH_FIle> module (see L<http://search.cpan.org/search?module=DWH_File>),
2692 and it is also mentioned in the BUGS section for the I<MLDBM> module <see
2693 L<http://search.cpan.org/search?module=MLDBM>). Basically, on a new db file,
2696 $db->{foo}->{bar} = "hello";
2698 Since "foo" doesn't exist, you cannot add "bar" to it. You end up with "foo"
2699 being an empty hash. Try this instead, which works fine:
2701 $db->{foo} = { bar => "hello" };
2703 As of Perl 5.8.7, this bug still exists. I have walked very carefully through
2704 the execution path, and Perl indeed passes an empty hash to the STORE() method.
2705 Probably a bug in Perl.
2707 =head2 FILE CORRUPTION
2709 The current level of error handling in DBM::Deep is minimal. Files I<are> checked
2710 for a 32-bit signature when opened, but other corruption in files can cause
2711 segmentation faults. DBM::Deep may try to seek() past the end of a file, or get
2712 stuck in an infinite loop depending on the level of corruption. File write
2713 operations are not checked for failure (for speed), so if you happen to run
2714 out of disk space, DBM::Deep will probably fail in a bad way. These things will
2715 be addressed in a later version of DBM::Deep.
2719 Beware of using DB files over NFS. DBM::Deep uses flock(), which works well on local
2720 filesystems, but will NOT protect you from file corruption over NFS. I've heard
2721 about setting up your NFS server with a locking daemon, then using lockf() to
2722 lock your files, but your milage may vary there as well. From what I
2723 understand, there is no real way to do it. However, if you need access to the
2724 underlying FileHandle in DBM::Deep for using some other kind of locking scheme like
2725 lockf(), see the L<LOW-LEVEL ACCESS> section above.
2727 =head2 COPYING OBJECTS
2729 Beware of copying tied objects in Perl. Very strange things can happen.
2730 Instead, use DBM::Deep's C<clone()> method which safely copies the object and
2731 returns a new, blessed, tied hash or array to the same level in the DB.
2733 my $copy = $db->clone();
2737 Beware of using C<shift()>, C<unshift()> or C<splice()> with large arrays.
2738 These functions cause every element in the array to move, which can be murder
2739 on DBM::Deep, as every element has to be fetched from disk, then stored again in
2740 a different location. This may be addressed in a later version.
2744 This section discusses DBM::Deep's speed and memory usage.
2748 Obviously, DBM::Deep isn't going to be as fast as some C-based DBMs, such as
2749 the almighty I<BerkeleyDB>. But it makes up for it in features like true
2750 multi-level hash/array support, and cross-platform FTPable files. Even so,
2751 DBM::Deep is still pretty fast, and the speed stays fairly consistent, even
2752 with huge databases. Here is some test data:
2754 Adding 1,000,000 keys to new DB file...
2756 At 100 keys, avg. speed is 2,703 keys/sec
2757 At 200 keys, avg. speed is 2,642 keys/sec
2758 At 300 keys, avg. speed is 2,598 keys/sec
2759 At 400 keys, avg. speed is 2,578 keys/sec
2760 At 500 keys, avg. speed is 2,722 keys/sec
2761 At 600 keys, avg. speed is 2,628 keys/sec
2762 At 700 keys, avg. speed is 2,700 keys/sec
2763 At 800 keys, avg. speed is 2,607 keys/sec
2764 At 900 keys, avg. speed is 2,190 keys/sec
2765 At 1,000 keys, avg. speed is 2,570 keys/sec
2766 At 2,000 keys, avg. speed is 2,417 keys/sec
2767 At 3,000 keys, avg. speed is 1,982 keys/sec
2768 At 4,000 keys, avg. speed is 1,568 keys/sec
2769 At 5,000 keys, avg. speed is 1,533 keys/sec
2770 At 6,000 keys, avg. speed is 1,787 keys/sec
2771 At 7,000 keys, avg. speed is 1,977 keys/sec
2772 At 8,000 keys, avg. speed is 2,028 keys/sec
2773 At 9,000 keys, avg. speed is 2,077 keys/sec
2774 At 10,000 keys, avg. speed is 2,031 keys/sec
2775 At 20,000 keys, avg. speed is 1,970 keys/sec
2776 At 30,000 keys, avg. speed is 2,050 keys/sec
2777 At 40,000 keys, avg. speed is 2,073 keys/sec
2778 At 50,000 keys, avg. speed is 1,973 keys/sec
2779 At 60,000 keys, avg. speed is 1,914 keys/sec
2780 At 70,000 keys, avg. speed is 2,091 keys/sec
2781 At 80,000 keys, avg. speed is 2,103 keys/sec
2782 At 90,000 keys, avg. speed is 1,886 keys/sec
2783 At 100,000 keys, avg. speed is 1,970 keys/sec
2784 At 200,000 keys, avg. speed is 2,053 keys/sec
2785 At 300,000 keys, avg. speed is 1,697 keys/sec
2786 At 400,000 keys, avg. speed is 1,838 keys/sec
2787 At 500,000 keys, avg. speed is 1,941 keys/sec
2788 At 600,000 keys, avg. speed is 1,930 keys/sec
2789 At 700,000 keys, avg. speed is 1,735 keys/sec
2790 At 800,000 keys, avg. speed is 1,795 keys/sec
2791 At 900,000 keys, avg. speed is 1,221 keys/sec
2792 At 1,000,000 keys, avg. speed is 1,077 keys/sec
2794 This test was performed on a PowerMac G4 1gHz running Mac OS X 10.3.2 & Perl
2795 5.8.1, with an 80GB Ultra ATA/100 HD spinning at 7200RPM. The hash keys and
2796 values were between 6 - 12 chars in length. The DB file ended up at 210MB.
2797 Run time was 12 min 3 sec.
2801 One of the great things about DBM::Deep is that it uses very little memory.
2802 Even with huge databases (1,000,000+ keys) you will not see much increased
2803 memory on your process. DBM::Deep relies solely on the filesystem for storing
2804 and fetching data. Here is output from I</usr/bin/top> before even opening a
2807 PID USER PRI NI SIZE RSS SHARE STAT %CPU %MEM TIME COMMAND
2808 22831 root 11 0 2716 2716 1296 R 0.0 0.2 0:07 perl
2810 Basically the process is taking 2,716K of memory. And here is the same
2811 process after storing and fetching 1,000,000 keys:
2813 PID USER PRI NI SIZE RSS SHARE STAT %CPU %MEM TIME COMMAND
2814 22831 root 14 0 2772 2772 1328 R 0.0 0.2 13:32 perl
2816 Notice the memory usage increased by only 56K. Test was performed on a 700mHz
2817 x86 box running Linux RedHat 7.2 & Perl 5.6.1.
2819 =head1 DB FILE FORMAT
2821 In case you were interested in the underlying DB file format, it is documented
2822 here in this section. You don't need to know this to use the module, it's just
2823 included for reference.
2827 DBM::Deep files always start with a 32-bit signature to identify the file type.
2828 This is at offset 0. The signature is "DPDB" in network byte order. This is
2829 checked when the file is opened.
2833 The DBM::Deep file is in a I<tagged format>, meaning each section of the file
2834 has a standard header containing the type of data, the length of data, and then
2835 the data itself. The type is a single character (1 byte), the length is a
2836 32-bit unsigned long in network byte order, and the data is, well, the data.
2837 Here is how it unfolds:
2841 Immediately after the 32-bit file signature is the I<Master Index> record.
2842 This is a standard tag header followed by 1024 bytes (in 32-bit mode) or 2048
2843 bytes (in 64-bit mode) of data. The type is I<H> for hash or I<A> for array,
2844 depending on how the DBM::Deep object was constructed.
2848 The index works by looking at a I<MD5 Hash> of the hash key (or array index
2849 number). The first 8-bit char of the MD5 signature is the offset into the
2850 index, multipled by 4 in 32-bit mode, or 8 in 64-bit mode. The value of the
2851 index element is a file offset of the next tag for the key/element in question,
2852 which is usually a I<Bucket List> tag (see below).
2856 The next tag I<could> be another index, depending on how many keys/elements
2857 exist. See L<RE-INDEXING> below for details.
2861 A I<Bucket List> is a collection of 16 MD5 hashes for keys/elements, plus
2862 file offsets to where the actual data is stored. It starts with a standard
2863 tag header, with type I<B>, and a data size of 320 bytes in 32-bit mode, or
2864 384 bytes in 64-bit mode. Each MD5 hash is stored in full (16 bytes), plus
2865 the 32-bit or 64-bit file offset for the I<Bucket> containing the actual data.
2866 When the list fills up, a I<Re-Index> operation is performed (See
2867 L<RE-INDEXING> below).
2871 A I<Bucket> is a tag containing a key/value pair (in hash mode), or a
2872 index/value pair (in array mode). It starts with a standard tag header with
2873 type I<D> for scalar data (string, binary, etc.), or it could be a nested
2874 hash (type I<H>) or array (type I<A>). The value comes just after the tag
2875 header. The size reported in the tag header is only for the value, but then,
2876 just after the value is another size (32-bit unsigned long) and then the plain
2877 key itself. Since the value is likely to be fetched more often than the plain
2878 key, I figured it would be I<slightly> faster to store the value first.
2882 If the type is I<H> (hash) or I<A> (array), the value is another I<Master Index>
2883 record for the nested structure, where the process begins all over again.
2887 After a I<Bucket List> grows to 16 records, its allocated space in the file is
2888 exhausted. Then, when another key/element comes in, the list is converted to a
2889 new index record. However, this index will look at the next char in the MD5
2890 hash, and arrange new Bucket List pointers accordingly. This process is called
2891 I<Re-Indexing>. Basically, a new index tag is created at the file EOF, and all
2892 17 (16 + new one) keys/elements are removed from the old Bucket List and
2893 inserted into the new index. Several new Bucket Lists are created in the
2894 process, as a new MD5 char from the key is being examined (it is unlikely that
2895 the keys will all share the same next char of their MD5s).
2899 Because of the way the I<MD5> algorithm works, it is impossible to tell exactly
2900 when the Bucket Lists will turn into indexes, but the first round tends to
2901 happen right around 4,000 keys. You will see a I<slight> decrease in
2902 performance here, but it picks back up pretty quick (see L<SPEED> above). Then
2903 it takes B<a lot> more keys to exhaust the next level of Bucket Lists. It's
2904 right around 900,000 keys. This process can continue nearly indefinitely --
2905 right up until the point the I<MD5> signatures start colliding with each other,
2906 and this is B<EXTREMELY> rare -- like winning the lottery 5 times in a row AND
2907 getting struck by lightning while you are walking to cash in your tickets.
2908 Theoretically, since I<MD5> hashes are 128-bit values, you I<could> have up to
2909 340,282,366,921,000,000,000,000,000,000,000,000,000 keys/elements (I believe
2910 this is 340 unodecillion, but don't quote me).
2914 When a new key/element is stored, the key (or index number) is first ran through
2915 I<Digest::MD5> to get a 128-bit signature (example, in hex:
2916 b05783b0773d894396d475ced9d2f4f6). Then, the I<Master Index> record is checked
2917 for the first char of the signature (in this case I<b>). If it does not exist,
2918 a new I<Bucket List> is created for our key (and the next 15 future keys that
2919 happen to also have I<b> as their first MD5 char). The entire MD5 is written
2920 to the I<Bucket List> along with the offset of the new I<Bucket> record (EOF at
2921 this point, unless we are replacing an existing I<Bucket>), where the actual
2922 data will be stored.
2926 Fetching an existing key/element involves getting a I<Digest::MD5> of the key
2927 (or index number), then walking along the indexes. If there are enough
2928 keys/elements in this DB level, there might be nested indexes, each linked to
2929 a particular char of the MD5. Finally, a I<Bucket List> is pointed to, which
2930 contains up to 16 full MD5 hashes. Each is checked for equality to the key in
2931 question. If we found a match, the I<Bucket> tag is loaded, where the value and
2932 plain key are stored.
2936 Fetching the plain key occurs when calling the I<first_key()> and I<next_key()>
2937 methods. In this process the indexes are walked systematically, and each key
2938 fetched in increasing MD5 order (which is why it appears random). Once the
2939 I<Bucket> is found, the value is skipped the plain key returned instead.
2940 B<Note:> Do not count on keys being fetched as if the MD5 hashes were
2941 alphabetically sorted. This only happens on an index-level -- as soon as the
2942 I<Bucket Lists> are hit, the keys will come out in the order they went in --
2943 so it's pretty much undefined how the keys will come out -- just like Perl's
2946 =head1 CODE COVERAGE
2948 I use B<Devel::Cover> to test the code coverage of my tests, below is the B<Devel::Cover> report on this
2949 module's test suite.
2951 ---------------------------- ------ ------ ------ ------ ------ ------ ------
2952 File stmt bran cond sub pod time total
2953 ---------------------------- ------ ------ ------ ------ ------ ------ ------
2954 blib/lib/DBM/Deep.pm 94.9 84.5 77.8 100.0 11.1 100.0 89.7
2955 Total 94.9 84.5 77.8 100.0 11.1 100.0 89.7
2956 ---------------------------- ------ ------ ------ ------ ------ ------ ------
2960 Joseph Huckaby, L<jhuckaby@cpan.org>
2962 Special thanks to Adam Sah and Rich Gaushell! You know why :-)
2966 perltie(1), Tie::Hash(3), Digest::MD5(3), Fcntl(3), flock(2), lockf(3), nfs(5),
2967 Digest::SHA256(3), Crypt::Blowfish(3), Compress::Zlib(3)
2971 Copyright (c) 2002-2005 Joseph Huckaby. All Rights Reserved.
2972 This is free software, you may use it and distribute it under the
2973 same terms as Perl itself.