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 " . $db->{my_complex}->[1]->{perl} . "!\n";
27 # (c) 2002-2005 Joseph Huckaby. All Rights Reserved.
28 # This program is free software; you can redistribute it and/or
29 # modify it under the same terms as Perl itself.
38 use vars qw/$VERSION/;
43 # Set to 4 and 'N' for 32-bit offset tags (default). Theoretical limit of 4 GB per file.
44 # (Perl must be compiled with largefile support for files > 2 GB)
46 # Set to 8 and 'Q' for 64-bit offsets. Theoretical limit of 16 XB per file.
47 # (Perl must be compiled with largefile and 64-bit long support)
53 # Set to 4 and 'N' for 32-bit data length prefixes. Limit of 4 GB for each key/value.
54 # Upgrading this is possible (see above) but probably not necessary. If you need
55 # more than 4 GB for a single key or value, this module is really not for you :-)
57 #my $DATA_LENGTH_SIZE = 4;
58 #my $DATA_LENGTH_PACK = 'N';
59 my ($LONG_SIZE, $LONG_PACK, $DATA_LENGTH_SIZE, $DATA_LENGTH_PACK);
62 # Maximum number of buckets per list before another level of indexing is done.
63 # Increase this value for slightly greater speed, but larger database files.
64 # DO NOT decrease this value below 16, due to risk of recursive reindex overrun.
69 # Better not adjust anything below here, unless you're me :-)
73 # Setup digest function for keys
75 my ($DIGEST_FUNC, $HASH_SIZE);
76 #my $DIGEST_FUNC = \&Digest::MD5::md5;
79 # Precalculate index and bucket sizes based on values above.
82 my ($INDEX_SIZE, $BUCKET_SIZE, $BUCKET_LIST_SIZE);
89 # Setup file and tag signatures. These should never change.
91 sub SIG_FILE () { 'DPDB' }
92 sub SIG_HASH () { 'H' }
93 sub SIG_ARRAY () { 'A' }
94 sub SIG_NULL () { 'N' }
95 sub SIG_DATA () { 'D' }
96 sub SIG_INDEX () { 'I' }
97 sub SIG_BLIST () { 'B' }
101 # Setup constants for users to pass to new()
103 sub TYPE_HASH () { return SIG_HASH; }
104 sub TYPE_ARRAY () { return SIG_ARRAY; }
108 # Class constructor method for Perl OO interface.
109 # Calls tie() and returns blessed reference to tied hash or array,
110 # providing a hybrid OO/tie interface.
114 if (scalar(@_) > 1) { $args = {@_}; }
115 else { $args = { file => shift }; }
118 # Check if we want a tied hash or array.
121 if (defined($args->{type}) && $args->{type} eq TYPE_ARRAY) {
122 tie @$self, $class, %$args;
125 tie %$self, $class, %$args;
128 return bless $self, $class;
132 my @outer_params = qw( type base_offset );
135 # Setup $self and bless into this class.
142 base_offset => length(SIG_FILE),
153 filter_store_key => undef,
154 filter_store_value => undef,
155 filter_fetch_key => undef,
156 filter_fetch_value => undef,
165 foreach my $outer_parm ( @outer_params ) {
166 next unless exists $args->{$outer_parm};
167 $self->{$outer_parm} = $args->{$outer_parm}
170 if ( exists $args->{root} ) {
171 $self->{root} = $args->{root};
174 # This is cleanup based on the fact that the $args
175 # coming in is for both the root and non-root items
176 delete $self->root->{$_} for @outer_params;
178 $self->root->{links}++;
180 if (!defined($self->fh)) { $self->_open(); }
186 sub _get_self { tied( %{$_[0]} ) || $_[0] }
190 # Tied hash constructor method, called by Perl's tie() function.
194 if (scalar(@_) > 1) { $args = {@_}; }
195 #XXX This use of ref() is bad and is a bug
196 elsif (ref($_[0])) { $args = $_[0]; }
197 else { $args = { file => shift }; }
199 $args->{type} = TYPE_HASH;
201 return $class->_init($args);
206 # Tied array constructor method, called by Perl's tie() function.
210 if (scalar(@_) > 1) { $args = {@_}; }
211 #XXX This use of ref() is bad and is a bug
212 elsif (ref($_[0])) { $args = $_[0]; }
213 else { $args = { file => shift }; }
215 $args->{type} = TYPE_ARRAY;
217 return $class->_init($args);
222 # Class deconstructor. Close file handle if there are no more refs.
224 my $self = _get_self($_[0]);
227 $self->root->{links}--;
229 if (!$self->root->{links}) {
236 # Open a FileHandle to the database, create if nonexistent.
237 # Make sure file signature matches DeepDB spec.
239 my $self = _get_self($_[0]);
241 if (defined($self->fh)) { $self->_close(); }
243 if (!(-e $self->root->{file}) && $self->root->{mode} eq 'r+') {
244 my $temp = FileHandle->new( $self->root->{file}, 'w' );
248 #XXX Convert to set_fh()
249 $self->root->{fh} = FileHandle->new( $self->root->{file}, $self->root->{mode} );
250 if (! defined($self->fh)) {
251 return $self->_throw_error("Cannot open file: " . $self->root->{file} . ": $!");
254 binmode $self->fh; # for win32
255 if ($self->root->{autoflush}) {
256 $self->fh->autoflush();
260 seek($self->fh, 0, 0);
261 my $bytes_read = $self->fh->read($signature, length(SIG_FILE));
264 # File is empty -- write signature and master index
267 seek($self->fh, 0, 0);
268 $self->fh->print(SIG_FILE);
269 $self->root->{end} = length(SIG_FILE);
270 $self->_create_tag($self->base_offset, $self->type, chr(0) x $INDEX_SIZE);
272 my $plain_key = "[base]";
273 $self->fh->print( pack($DATA_LENGTH_PACK, length($plain_key)) . $plain_key );
274 $self->root->{end} += $DATA_LENGTH_SIZE + length($plain_key);
281 # Check signature was valid
283 unless ($signature eq SIG_FILE) {
285 return $self->_throw_error("Signature not found -- file is not a Deep DB");
288 $self->root->{end} = (stat($self->fh))[7];
291 # Get our type from master index signature
293 my $tag = $self->_load_tag($self->base_offset);
294 #XXX We probably also want to store the hash algorithm name, not assume anything
295 #XXX Convert to set_type() when one is written
297 return $self->_throw_error("Corrupted file, no master index record");
299 if ($self->{type} ne $tag->{signature}) {
300 return $self->_throw_error("File type mismatch");
308 # Close database FileHandle
310 my $self = _get_self($_[0]);
311 undef $self->root->{fh};
316 # Given offset, signature and content, create tag and write to disk
318 my ($self, $offset, $sig, $content) = @_;
319 my $size = length($content);
321 seek($self->fh, $offset, 0);
322 $self->fh->print( $sig . pack($DATA_LENGTH_PACK, $size) . $content );
324 if ($offset == $self->root->{end}) {
325 $self->root->{end} += SIG_SIZE + $DATA_LENGTH_SIZE + $size;
331 offset => $offset + SIG_SIZE + $DATA_LENGTH_SIZE,
338 # Given offset, load single tag and return signature, size and data
343 seek($self->fh, $offset, 0);
344 if ($self->fh->eof()) { return undef; }
347 $self->fh->read($sig, SIG_SIZE);
350 $self->fh->read($size, $DATA_LENGTH_SIZE);
351 $size = unpack($DATA_LENGTH_PACK, $size);
354 $self->fh->read($buffer, $size);
359 offset => $offset + SIG_SIZE + $DATA_LENGTH_SIZE,
366 # Given index tag, lookup single entry in index and return .
369 my ($tag, $index) = @_;
371 my $location = unpack($LONG_PACK, substr($tag->{content}, $index * $LONG_SIZE, $LONG_SIZE) );
372 if (!$location) { return; }
374 return $self->_load_tag( $location );
379 # Adds one key/value pair to bucket list, given offset, MD5 digest of key,
380 # plain (undigested) key and value.
383 my ($tag, $md5, $plain_key, $value) = @_;
384 my $keys = $tag->{content};
388 my $is_dbm_deep = eval { $value->isa( 'DBM::Deep' ) };
389 my $internal_ref = $is_dbm_deep && ($value->root eq $self->root);
392 # Iterate through buckets, seeing if this is a new entry or a replace.
394 for (my $i=0; $i<$MAX_BUCKETS; $i++) {
395 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
396 my $subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
399 # Found empty bucket (end of list). Populate and exit loop.
403 if ($internal_ref) { $location = $value->base_offset; }
404 else { $location = $self->root->{end}; }
406 seek($self->fh, $tag->{offset} + ($i * $BUCKET_SIZE), 0);
407 $self->fh->print( $md5 . pack($LONG_PACK, $location) );
410 elsif ($md5 eq $key) {
412 # Found existing bucket with same key. Replace with new value.
417 $location = $value->base_offset;
418 seek($self->fh, $tag->{offset} + ($i * $BUCKET_SIZE), 0);
419 $self->fh->print( $md5 . pack($LONG_PACK, $location) );
422 seek($self->fh, $subloc + SIG_SIZE, 0);
424 $self->fh->read($size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
427 # If value is a hash, array, or raw value with equal or less size, we can
428 # reuse the same content area of the database. Otherwise, we have to create
429 # a new content area at the EOF.
432 my $r = Scalar::Util::reftype( $value ) || '';
433 if ( $r eq 'HASH' || $r eq 'ARRAY' ) { $actual_length = $INDEX_SIZE; }
434 else { $actual_length = length($value); }
436 if ($actual_length <= $size) {
440 $location = $self->root->{end};
441 seek($self->fh, $tag->{offset} + ($i * $BUCKET_SIZE) + $HASH_SIZE, 0);
442 $self->fh->print( pack($LONG_PACK, $location) );
450 # If this is an internal reference, return now.
451 # No need to write value or plain key
458 # If bucket didn't fit into list, split into a new index level
461 seek($self->fh, $tag->{ref_loc}, 0);
462 $self->fh->print( pack($LONG_PACK, $self->root->{end}) );
464 my $index_tag = $self->_create_tag($self->root->{end}, SIG_INDEX, chr(0) x $INDEX_SIZE);
467 $keys .= $md5 . pack($LONG_PACK, 0);
469 for (my $i=0; $i<=$MAX_BUCKETS; $i++) {
470 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
472 my $old_subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
473 my $num = ord(substr($key, $tag->{ch} + 1, 1));
475 if ($offsets[$num]) {
476 my $offset = $offsets[$num] + SIG_SIZE + $DATA_LENGTH_SIZE;
477 seek($self->fh, $offset, 0);
479 $self->fh->read($subkeys, $BUCKET_LIST_SIZE);
481 for (my $k=0; $k<$MAX_BUCKETS; $k++) {
482 my $subloc = unpack($LONG_PACK, substr($subkeys, ($k * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
484 seek($self->fh, $offset + ($k * $BUCKET_SIZE), 0);
485 $self->fh->print( $key . pack($LONG_PACK, $old_subloc || $self->root->{end}) );
491 $offsets[$num] = $self->root->{end};
492 seek($self->fh, $index_tag->{offset} + ($num * $LONG_SIZE), 0);
493 $self->fh->print( pack($LONG_PACK, $self->root->{end}) );
495 my $blist_tag = $self->_create_tag($self->root->{end}, SIG_BLIST, chr(0) x $BUCKET_LIST_SIZE);
497 seek($self->fh, $blist_tag->{offset}, 0);
498 $self->fh->print( $key . pack($LONG_PACK, $old_subloc || $self->root->{end}) );
503 $location ||= $self->root->{end};
504 } # re-index bucket list
507 # Seek to content area and store signature, value and plaintext key
511 seek($self->fh, $location, 0);
514 # Write signature based on content type, set content length and write actual value.
516 my $r = Scalar::Util::reftype($value) || '';
518 $self->fh->print( TYPE_HASH );
519 $self->fh->print( pack($DATA_LENGTH_PACK, $INDEX_SIZE) . chr(0) x $INDEX_SIZE );
520 $content_length = $INDEX_SIZE;
522 elsif ($r eq 'ARRAY') {
523 $self->fh->print( TYPE_ARRAY );
524 $self->fh->print( pack($DATA_LENGTH_PACK, $INDEX_SIZE) . chr(0) x $INDEX_SIZE );
525 $content_length = $INDEX_SIZE;
527 elsif (!defined($value)) {
528 $self->fh->print( SIG_NULL );
529 $self->fh->print( pack($DATA_LENGTH_PACK, 0) );
533 $self->fh->print( SIG_DATA );
534 $self->fh->print( pack($DATA_LENGTH_PACK, length($value)) . $value );
535 $content_length = length($value);
539 # Plain key is stored AFTER value, as keys are typically fetched less often.
541 $self->fh->print( pack($DATA_LENGTH_PACK, length($plain_key)) . $plain_key );
544 # If value is blessed, preserve class name
546 my $value_class = Scalar::Util::blessed($value);
547 if ($self->root->{autobless} && defined $value_class && $value_class ne 'DBM::Deep' ) {
549 # Blessed ref -- will restore later
551 $self->fh->print( chr(1) );
552 $self->fh->print( pack($DATA_LENGTH_PACK, length($value_class)) . $value_class );
553 $content_length += 1;
554 $content_length += $DATA_LENGTH_SIZE + length($value_class);
558 # If this is a new content area, advance EOF counter
560 if ($location == $self->root->{end}) {
561 $self->root->{end} += SIG_SIZE;
562 $self->root->{end} += $DATA_LENGTH_SIZE + $content_length;
563 $self->root->{end} += $DATA_LENGTH_SIZE + length($plain_key);
567 # If content is a hash or array, create new child DeepDB object and
568 # pass each key or element to it.
571 my $branch = DBM::Deep->new(
573 base_offset => $location,
576 foreach my $key (keys %{$value}) {
577 $branch->{$key} = $value->{$key};
580 elsif ($r eq 'ARRAY') {
581 my $branch = DBM::Deep->new(
583 base_offset => $location,
587 foreach my $element (@{$value}) {
588 $branch->[$index] = $element;
596 return $self->_throw_error("Fatal error: indexing failed -- possibly due to corruption in file");
599 sub _get_bucket_value {
601 # Fetch single value given tag and MD5 digested key.
604 my ($tag, $md5) = @_;
605 my $keys = $tag->{content};
608 # Iterate through buckets, looking for a key match
611 for (my $i=0; $i<$MAX_BUCKETS; $i++) {
612 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
613 my $subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
617 # Hit end of list, no match
622 if ( $md5 ne $key ) {
627 # Found match -- seek to offset and read signature
630 seek($self->fh, $subloc, 0);
631 $self->fh->read($signature, SIG_SIZE);
634 # If value is a hash or array, return new DeepDB object with correct offset
636 if (($signature eq TYPE_HASH) || ($signature eq TYPE_ARRAY)) {
637 my $obj = DBM::Deep->new(
639 base_offset => $subloc,
643 if ($self->root->{autobless}) {
645 # Skip over value and plain key to see if object needs
648 seek($self->fh, $DATA_LENGTH_SIZE + $INDEX_SIZE, 1);
651 $self->fh->read($size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
652 if ($size) { seek($self->fh, $size, 1); }
655 $self->fh->read($bless_bit, 1);
656 if (ord($bless_bit)) {
658 # Yes, object needs to be re-blessed
661 $self->fh->read($size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
662 if ($size) { $self->fh->read($class_name, $size); }
663 if ($class_name) { $obj = bless( $obj, $class_name ); }
671 # Otherwise return actual value
673 elsif ($signature eq SIG_DATA) {
676 $self->fh->read($size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
677 if ($size) { $self->fh->read($value, $size); }
682 # Key exists, but content is null
692 # Delete single key/value pair given tag and MD5 digested key.
695 my ($tag, $md5) = @_;
696 my $keys = $tag->{content};
699 # Iterate through buckets, looking for a key match
702 for (my $i=0; $i<$MAX_BUCKETS; $i++) {
703 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
704 my $subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
708 # Hit end of list, no match
713 if ( $md5 ne $key ) {
718 # Matched key -- delete bucket and return
720 seek($self->fh, $tag->{offset} + ($i * $BUCKET_SIZE), 0);
721 $self->fh->print( substr($keys, ($i+1) * $BUCKET_SIZE ) );
722 $self->fh->print( chr(0) x $BUCKET_SIZE );
732 # Check existence of single key given tag and MD5 digested key.
735 my ($tag, $md5) = @_;
736 my $keys = $tag->{content};
739 # Iterate through buckets, looking for a key match
742 for (my $i=0; $i<$MAX_BUCKETS; $i++) {
743 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
744 my $subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
748 # Hit end of list, no match
753 if ( $md5 ne $key ) {
758 # Matched key -- return true
766 sub _find_bucket_list {
768 # Locate offset for bucket list, given digested key
774 # Locate offset for bucket list using digest index system
777 my $tag = $self->_load_tag($self->base_offset);
778 if (!$tag) { return; }
780 while ($tag->{signature} ne SIG_BLIST) {
781 $tag = $self->_index_lookup($tag, ord(substr($md5, $ch, 1)));
782 if (!$tag) { return; }
789 sub _traverse_index {
791 # Scan index and recursively step into deeper levels, looking for next key.
793 my ($self, $offset, $ch, $force_return_next) = @_;
794 $force_return_next = undef unless $force_return_next;
796 my $tag = $self->_load_tag( $offset );
798 if ($tag->{signature} ne SIG_BLIST) {
799 my $content = $tag->{content};
801 if ($self->{return_next}) { $start = 0; }
802 else { $start = ord(substr($self->{prev_md5}, $ch, 1)); }
804 for (my $index = $start; $index < 256; $index++) {
805 my $subloc = unpack($LONG_PACK, substr($content, $index * $LONG_SIZE, $LONG_SIZE) );
807 my $result = $self->_traverse_index( $subloc, $ch + 1, $force_return_next );
808 if (defined($result)) { return $result; }
812 $self->{return_next} = 1;
815 elsif ($tag->{signature} eq SIG_BLIST) {
816 my $keys = $tag->{content};
817 if ($force_return_next) { $self->{return_next} = 1; }
820 # Iterate through buckets, looking for a key match
822 for (my $i=0; $i<$MAX_BUCKETS; $i++) {
823 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
824 my $subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
828 # End of bucket list -- return to outer loop
830 $self->{return_next} = 1;
833 elsif ($key eq $self->{prev_md5}) {
835 # Located previous key -- return next one found
837 $self->{return_next} = 1;
840 elsif ($self->{return_next}) {
842 # Seek to bucket location and skip over signature
844 seek($self->fh, $subloc + SIG_SIZE, 0);
847 # Skip over value to get to plain key
850 $self->fh->read($size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
851 if ($size) { seek($self->fh, $size, 1); }
854 # Read in plain key and return as scalar
857 $self->fh->read($size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
858 if ($size) { $self->fh->read($plain_key, $size); }
864 $self->{return_next} = 1;
865 } # tag is a bucket list
872 # Locate next key, given digested previous one
874 my $self = _get_self($_[0]);
876 $self->{prev_md5} = $_[1] ? $_[1] : undef;
877 $self->{return_next} = 0;
880 # If the previous key was not specifed, start at the top and
881 # return the first one found.
883 if (!$self->{prev_md5}) {
884 $self->{prev_md5} = chr(0) x $HASH_SIZE;
885 $self->{return_next} = 1;
888 return $self->_traverse_index( $self->base_offset, 0 );
893 # If db locking is set, flock() the db file. If called multiple
894 # times before unlock(), then the same number of unlocks() must
895 # be called before the lock is released.
897 my $self = _get_self($_[0]);
899 $type = LOCK_EX unless defined $type;
901 if ($self->root->{locking}) {
902 if (!$self->root->{locked}) { flock($self->fh, $type); }
903 $self->root->{locked}++;
909 # If db locking is set, unlock the db file. See note in lock()
910 # regarding calling lock() multiple times.
912 my $self = _get_self($_[0]);
914 if ($self->root->{locking} && $self->root->{locked} > 0) {
915 $self->root->{locked}--;
916 if (!$self->root->{locked}) { flock($self->fh, LOCK_UN); }
920 #XXX These uses of ref() need verified
923 # Copy single level of keys or elements to new DB handle.
924 # Recurse for nested structures
926 my $self = _get_self($_[0]);
929 if ($self->type eq TYPE_HASH) {
930 my $key = $self->first_key();
932 my $value = $self->get($key);
933 if (!ref($value)) { $db_temp->{$key} = $value; }
935 my $type = $value->type;
936 if ($type eq TYPE_HASH) { $db_temp->{$key} = {}; }
937 else { $db_temp->{$key} = []; }
938 $value->_copy_node( $db_temp->{$key} );
940 $key = $self->next_key($key);
944 my $length = $self->length();
945 for (my $index = 0; $index < $length; $index++) {
946 my $value = $self->get($index);
947 if (!ref($value)) { $db_temp->[$index] = $value; }
948 #XXX NO tests for this code
950 my $type = $value->type;
951 if ($type eq TYPE_HASH) { $db_temp->[$index] = {}; }
952 else { $db_temp->[$index] = []; }
953 $value->_copy_node( $db_temp->[$index] );
961 # Recursively export into standard Perl hashes and arrays.
963 my $self = _get_self($_[0]);
966 if ($self->type eq TYPE_HASH) { $temp = {}; }
967 elsif ($self->type eq TYPE_ARRAY) { $temp = []; }
970 $self->_copy_node( $temp );
978 # Recursively import Perl hash/array structure
980 #XXX This use of ref() seems to be ok
981 if (!ref($_[0])) { return; } # Perl calls import() on use -- ignore
983 my $self = _get_self($_[0]);
986 #XXX This use of ref() seems to be ok
989 # struct is not a reference, so just import based on our type
993 if ($self->type eq TYPE_HASH) { $struct = {@_}; }
994 elsif ($self->type eq TYPE_ARRAY) { $struct = [@_]; }
997 my $r = Scalar::Util::reftype($struct) || '';
998 if ($r eq "HASH" && $self->type eq TYPE_HASH) {
999 foreach my $key (keys %$struct) { $self->put($key, $struct->{$key}); }
1001 elsif ($r eq "ARRAY" && $self->type eq TYPE_ARRAY) {
1002 $self->push( @$struct );
1005 return $self->_throw_error("Cannot import: type mismatch");
1013 # Rebuild entire database into new file, then move
1014 # it back on top of original.
1016 my $self = _get_self($_[0]);
1017 if ($self->root->{links} > 1) {
1018 return $self->_throw_error("Cannot optimize: reference count is greater than 1");
1021 my $db_temp = DBM::Deep->new(
1022 file => $self->root->{file} . '.tmp',
1026 return $self->_throw_error("Cannot optimize: failed to open temp file: $!");
1030 $self->_copy_node( $db_temp );
1034 # Attempt to copy user, group and permissions over to new file
1036 my @stats = stat($self->fh);
1037 my $perms = $stats[2] & 07777;
1038 my $uid = $stats[4];
1039 my $gid = $stats[5];
1040 chown( $uid, $gid, $self->root->{file} . '.tmp' );
1041 chmod( $perms, $self->root->{file} . '.tmp' );
1043 # q.v. perlport for more information on this variable
1044 if ( $^O eq 'MSWin32' ) {
1046 # Potential race condition when optmizing on Win32 with locking.
1047 # The Windows filesystem requires that the filehandle be closed
1048 # before it is overwritten with rename(). This could be redone
1055 if (!rename $self->root->{file} . '.tmp', $self->root->{file}) {
1056 unlink $self->root->{file} . '.tmp';
1058 return $self->_throw_error("Optimize failed: Cannot copy temp file over original: $!");
1070 # Make copy of object and return
1072 my $self = _get_self($_[0]);
1074 return DBM::Deep->new(
1075 type => $self->type,
1076 base_offset => $self->base_offset,
1082 my %is_legal_filter = map {
1085 store_key store_value
1086 fetch_key fetch_value
1091 # Setup filter function for storing or fetching the key or value
1093 my $self = _get_self($_[0]);
1094 my $type = lc $_[1];
1095 my $func = $_[2] ? $_[2] : undef;
1097 if ( $is_legal_filter{$type} ) {
1098 $self->root->{"filter_$type"} = $func;
1112 # Get access to the root structure
1114 my $self = _get_self($_[0]);
1115 return $self->{root};
1120 # Get access to the raw FileHandle
1122 #XXX It will be useful, though, when we split out HASH and ARRAY
1123 my $self = _get_self($_[0]);
1124 return $self->root->{fh};
1129 # Get type of current node (TYPE_HASH or TYPE_ARRAY)
1131 my $self = _get_self($_[0]);
1132 return $self->{type};
1137 # Get base_offset of current node (TYPE_HASH or TYPE_ARRAY)
1139 my $self = _get_self($_[0]);
1140 return $self->{base_offset};
1145 # Get last error string, or undef if no error
1148 ? ( _get_self($_[0])->{root}->{error} or undef )
1158 # Store error string in self
1160 my $self = _get_self($_[0]);
1161 my $error_text = $_[1];
1163 $self->root->{error} = $error_text;
1165 unless ($self->root->{debug}) {
1166 die "DBM::Deep: $error_text\n";
1169 warn "DBM::Deep: $error_text\n";
1177 my $self = _get_self($_[0]);
1179 undef $self->root->{error};
1184 # Precalculate index, bucket and bucket list sizes
1187 #XXX I don't like this ...
1188 set_pack() unless defined $LONG_SIZE;
1190 $INDEX_SIZE = 256 * $LONG_SIZE;
1191 $BUCKET_SIZE = $HASH_SIZE + $LONG_SIZE;
1192 $BUCKET_LIST_SIZE = $MAX_BUCKETS * $BUCKET_SIZE;
1197 # Set pack/unpack modes (see file header for more)
1199 my ($long_s, $long_p, $data_s, $data_p) = @_;
1201 $LONG_SIZE = $long_s ? $long_s : 4;
1202 $LONG_PACK = $long_p ? $long_p : 'N';
1204 $DATA_LENGTH_SIZE = $data_s ? $data_s : 4;
1205 $DATA_LENGTH_PACK = $data_p ? $data_p : 'N';
1212 # Set key digest function (default is MD5)
1214 my ($digest_func, $hash_size) = @_;
1216 $DIGEST_FUNC = $digest_func ? $digest_func : \&Digest::MD5::md5;
1217 $HASH_SIZE = $hash_size ? $hash_size : 16;
1223 # tie() methods (hashes and arrays)
1228 # Store single hash key/value or array element in database.
1230 my $self = _get_self($_[0]);
1231 my $key = ($self->root->{filter_store_key} && $self->type eq TYPE_HASH) ? $self->root->{filter_store_key}->($_[1]) : $_[1];
1232 #XXX What is ref() checking here?
1233 #YYY User may be storing a hash, in which case we do not want it run
1234 #YYY through the filtering system
1235 my $value = ($self->root->{filter_store_value} && !ref($_[2])) ? $self->root->{filter_store_value}->($_[2]) : $_[2];
1237 my $unpacked_key = $key;
1238 if (($self->type eq TYPE_ARRAY) && ($key =~ /^\d+$/)) { $key = pack($LONG_PACK, $key); }
1239 my $md5 = $DIGEST_FUNC->($key);
1242 # Make sure file is open
1244 if (!defined($self->fh) && !$self->_open()) {
1249 # Request exclusive lock for writing
1251 $self->lock( LOCK_EX );
1254 # If locking is enabled, set 'end' parameter again, in case another
1255 # DB instance appended to our file while we were unlocked.
1257 if ($self->root->{locking} || $self->root->{volatile}) {
1258 $self->root->{end} = (stat($self->fh))[7];
1262 # Locate offset for bucket list using digest index system
1264 my $tag = $self->_load_tag($self->base_offset);
1266 $tag = $self->_create_tag($self->base_offset, SIG_INDEX, chr(0) x $INDEX_SIZE);
1270 while ($tag->{signature} ne SIG_BLIST) {
1271 my $num = ord(substr($md5, $ch, 1));
1272 my $new_tag = $self->_index_lookup($tag, $num);
1274 my $ref_loc = $tag->{offset} + ($num * $LONG_SIZE);
1275 seek($self->fh, $ref_loc, 0);
1276 $self->fh->print( pack($LONG_PACK, $self->root->{end}) );
1278 $tag = $self->_create_tag($self->root->{end}, SIG_BLIST, chr(0) x $BUCKET_LIST_SIZE);
1279 $tag->{ref_loc} = $ref_loc;
1284 my $ref_loc = $tag->{offset} + ($num * $LONG_SIZE);
1286 $tag->{ref_loc} = $ref_loc;
1293 # Add key/value to bucket list
1295 my $result = $self->_add_bucket( $tag, $md5, $key, $value );
1298 # If this object is an array, and bucket was not a replace, and key is numerical,
1299 # and index is equal or greater than current length, advance length variable.
1301 if (($result == 2) && ($self->type eq TYPE_ARRAY) && ($unpacked_key =~ /^\d+$/) && ($unpacked_key >= $self->FETCHSIZE())) {
1302 $self->STORESIZE( $unpacked_key + 1 );
1312 # Fetch single value or element given plain key or array index
1314 my $self = _get_self($_[0]);
1317 if ( $self->type eq TYPE_HASH ) {
1318 if ( my $filter = $self->root->{filter_store_key} ) {
1319 $key = $filter->( $key );
1322 elsif ( $self->type eq TYPE_ARRAY ) {
1323 if ( $key =~ /^\d+$/ ) {
1324 $key = pack($LONG_PACK, $key);
1328 my $md5 = $DIGEST_FUNC->($key);
1331 # Make sure file is open
1333 if (!defined($self->fh)) { $self->_open(); }
1336 # Request shared lock for reading
1338 $self->lock( LOCK_SH );
1340 my $tag = $self->_find_bucket_list( $md5 );
1347 # Get value from bucket list
1349 my $result = $self->_get_bucket_value( $tag, $md5 );
1353 #XXX What is ref() checking here?
1354 return ($result && !ref($result) && $self->root->{filter_fetch_value}) ? $self->root->{filter_fetch_value}->($result) : $result;
1359 # Delete single key/value pair or element given plain key or array index
1361 my $self = _get_self($_[0]);
1362 my $key = ($self->root->{filter_store_key} && $self->type eq TYPE_HASH) ? $self->root->{filter_store_key}->($_[1]) : $_[1];
1364 my $unpacked_key = $key;
1365 if (($self->type eq TYPE_ARRAY) && ($key =~ /^\d+$/)) { $key = pack($LONG_PACK, $key); }
1366 my $md5 = $DIGEST_FUNC->($key);
1369 # Make sure file is open
1371 if (!defined($self->fh)) { $self->_open(); }
1374 # Request exclusive lock for writing
1376 $self->lock( LOCK_EX );
1378 my $tag = $self->_find_bucket_list( $md5 );
1387 my $result = $self->_delete_bucket( $tag, $md5 );
1390 # If this object is an array and the key deleted was on the end of the stack,
1391 # decrement the length variable.
1393 if ($result && ($self->type eq TYPE_ARRAY) && ($unpacked_key == $self->FETCHSIZE() - 1)) {
1394 $self->STORESIZE( $unpacked_key );
1404 # Check if a single key or element exists given plain key or array index
1406 my $self = _get_self($_[0]);
1407 my $key = ($self->root->{filter_store_key} && $self->type eq TYPE_HASH) ? $self->root->{filter_store_key}->($_[1]) : $_[1];
1409 if (($self->type eq TYPE_ARRAY) && ($key =~ /^\d+$/)) { $key = pack($LONG_PACK, $key); }
1410 my $md5 = $DIGEST_FUNC->($key);
1413 # Make sure file is open
1415 if (!defined($self->fh)) { $self->_open(); }
1418 # Request shared lock for reading
1420 $self->lock( LOCK_SH );
1422 my $tag = $self->_find_bucket_list( $md5 );
1425 # For some reason, the built-in exists() function returns '' for false
1433 # Check if bucket exists and return 1 or ''
1435 my $result = $self->_bucket_exists( $tag, $md5 ) || '';
1444 # Clear all keys from hash, or all elements from array.
1446 my $self = _get_self($_[0]);
1449 # Make sure file is open
1451 if (!defined($self->fh)) { $self->_open(); }
1454 # Request exclusive lock for writing
1456 $self->lock( LOCK_EX );
1458 seek($self->fh, $self->base_offset, 0);
1459 if ($self->fh->eof()) {
1464 $self->_create_tag($self->base_offset, $self->type, chr(0) x $INDEX_SIZE);
1473 # Locate and return first key (in no particular order)
1475 my $self = _get_self($_[0]);
1476 if ($self->type ne TYPE_HASH) {
1477 return $self->_throw_error("FIRSTKEY method only supported for hashes");
1481 # Make sure file is open
1483 if (!defined($self->fh)) { $self->_open(); }
1486 # Request shared lock for reading
1488 $self->lock( LOCK_SH );
1490 my $result = $self->_get_next_key();
1494 return ($result && $self->root->{filter_fetch_key}) ? $self->root->{filter_fetch_key}->($result) : $result;
1499 # Return next key (in no particular order), given previous one
1501 my $self = _get_self($_[0]);
1502 if ($self->type ne TYPE_HASH) {
1503 return $self->_throw_error("NEXTKEY method only supported for hashes");
1505 my $prev_key = ($self->root->{filter_store_key} && $self->type eq TYPE_HASH) ? $self->root->{filter_store_key}->($_[1]) : $_[1];
1506 my $prev_md5 = $DIGEST_FUNC->($prev_key);
1509 # Make sure file is open
1511 if (!defined($self->fh)) { $self->_open(); }
1514 # Request shared lock for reading
1516 $self->lock( LOCK_SH );
1518 my $result = $self->_get_next_key( $prev_md5 );
1522 return ($result && $self->root->{filter_fetch_key}) ? $self->root->{filter_fetch_key}->($result) : $result;
1526 # The following methods are for arrays only
1531 # Return the length of the array
1533 my $self = _get_self($_[0]);
1534 if ($self->type ne TYPE_ARRAY) {
1535 return $self->_throw_error("FETCHSIZE method only supported for arrays");
1538 my $SAVE_FILTER = $self->root->{filter_fetch_value};
1539 $self->root->{filter_fetch_value} = undef;
1541 my $packed_size = $self->FETCH('length');
1543 $self->root->{filter_fetch_value} = $SAVE_FILTER;
1545 if ($packed_size) { return int(unpack($LONG_PACK, $packed_size)); }
1551 # Set the length of the array
1553 my $self = _get_self($_[0]);
1554 if ($self->type ne TYPE_ARRAY) {
1555 return $self->_throw_error("STORESIZE method only supported for arrays");
1557 my $new_length = $_[1];
1559 my $SAVE_FILTER = $self->root->{filter_store_value};
1560 $self->root->{filter_store_value} = undef;
1562 my $result = $self->STORE('length', pack($LONG_PACK, $new_length));
1564 $self->root->{filter_store_value} = $SAVE_FILTER;
1571 # Remove and return the last element on the array
1573 my $self = _get_self($_[0]);
1574 if ($self->type ne TYPE_ARRAY) {
1575 return $self->_throw_error("POP method only supported for arrays");
1577 my $length = $self->FETCHSIZE();
1580 my $content = $self->FETCH( $length - 1 );
1581 $self->DELETE( $length - 1 );
1591 # Add new element(s) to the end of the array
1593 my $self = _get_self(shift);
1594 if ($self->type ne TYPE_ARRAY) {
1595 return $self->_throw_error("PUSH method only supported for arrays");
1597 my $length = $self->FETCHSIZE();
1599 while (my $content = shift @_) {
1600 $self->STORE( $length, $content );
1607 # Remove and return first element on the array.
1608 # Shift over remaining elements to take up space.
1610 my $self = _get_self($_[0]);
1611 if ($self->type ne TYPE_ARRAY) {
1612 return $self->_throw_error("SHIFT method only supported for arrays");
1614 my $length = $self->FETCHSIZE();
1617 my $content = $self->FETCH( 0 );
1620 # Shift elements over and remove last one.
1622 for (my $i = 0; $i < $length - 1; $i++) {
1623 $self->STORE( $i, $self->FETCH($i + 1) );
1625 $self->DELETE( $length - 1 );
1636 # Insert new element(s) at beginning of array.
1637 # Shift over other elements to make space.
1639 my $self = _get_self($_[0]);shift @_;
1640 if ($self->type ne TYPE_ARRAY) {
1641 return $self->_throw_error("UNSHIFT method only supported for arrays");
1643 my @new_elements = @_;
1644 my $length = $self->FETCHSIZE();
1645 my $new_size = scalar @new_elements;
1648 for (my $i = $length - 1; $i >= 0; $i--) {
1649 $self->STORE( $i + $new_size, $self->FETCH($i) );
1653 for (my $i = 0; $i < $new_size; $i++) {
1654 $self->STORE( $i, $new_elements[$i] );
1660 # Splices section of array with optional new section.
1661 # Returns deleted section, or last element deleted in scalar context.
1663 my $self = _get_self($_[0]);shift @_;
1664 if ($self->type ne TYPE_ARRAY) {
1665 return $self->_throw_error("SPLICE method only supported for arrays");
1667 my $length = $self->FETCHSIZE();
1670 # Calculate offset and length of splice
1672 my $offset = shift || 0;
1673 if ($offset < 0) { $offset += $length; }
1676 if (scalar @_) { $splice_length = shift; }
1677 else { $splice_length = $length - $offset; }
1678 if ($splice_length < 0) { $splice_length += ($length - $offset); }
1681 # Setup array with new elements, and copy out old elements for return
1683 my @new_elements = @_;
1684 my $new_size = scalar @new_elements;
1686 my @old_elements = ();
1687 for (my $i = $offset; $i < $offset + $splice_length; $i++) {
1688 push @old_elements, $self->FETCH( $i );
1692 # Adjust array length, and shift elements to accomodate new section.
1694 if ( $new_size != $splice_length ) {
1695 if ($new_size > $splice_length) {
1696 for (my $i = $length - 1; $i >= $offset + $splice_length; $i--) {
1697 $self->STORE( $i + ($new_size - $splice_length), $self->FETCH($i) );
1701 for (my $i = $offset + $splice_length; $i < $length; $i++) {
1702 $self->STORE( $i + ($new_size - $splice_length), $self->FETCH($i) );
1704 for (my $i = 0; $i < $splice_length - $new_size; $i++) {
1705 $self->DELETE( $length - 1 );
1712 # Insert new elements into array
1714 for (my $i = $offset; $i < $offset + $new_size; $i++) {
1715 $self->STORE( $i, shift @new_elements );
1719 # Return deleted section, or last element in scalar context.
1721 return wantarray ? @old_elements : $old_elements[-1];
1724 #XXX We don't need to define it.
1725 #XXX It will be useful, though, when we split out HASH and ARRAY
1728 # Perl will call EXTEND() when the array is likely to grow.
1729 # We don't care, but include it for compatibility.
1734 # Public method aliases
1736 *put = *store = *STORE;
1737 *get = *fetch = *FETCH;
1741 *first_key = *FIRSTKEY;
1742 *next_key = *NEXTKEY;
1743 *length = *FETCHSIZE;
1747 *unshift = *UNSHIFT;
1756 DBM::Deep - A pure perl multi-level hash/array DBM
1761 my $db = DBM::Deep->new( "foo.db" );
1763 $db->{key} = 'value'; # tie() style
1766 $db->put('key', 'value'); # OO style
1767 print $db->get('key');
1769 # true multi-level support
1770 $db->{my_complex} = [
1771 'hello', { perl => 'rules' },
1776 A unique flat-file database module, written in pure perl. True
1777 multi-level hash/array support (unlike MLDBM, which is faked), hybrid
1778 OO / tie() interface, cross-platform FTPable files, and quite fast. Can
1779 handle millions of keys and unlimited hash levels without significant
1780 slow-down. Written from the ground-up in pure perl -- this is NOT a
1781 wrapper around a C-based DBM. Out-of-the-box compatibility with Unix,
1782 Mac OS X and Windows.
1786 Hopefully you are using CPAN's excellent Perl module, which will download
1787 and install the module for you. If not, get the tarball, and run these
1799 Construction can be done OO-style (which is the recommended way), or using
1800 Perl's tie() function. Both are examined here.
1802 =head2 OO CONSTRUCTION
1804 The recommended way to construct a DBM::Deep object is to use the new()
1805 method, which gets you a blessed, tied hash or array reference.
1807 my $db = DBM::Deep->new( "foo.db" );
1809 This opens a new database handle, mapped to the file "foo.db". If this
1810 file does not exist, it will automatically be created. DB files are
1811 opened in "r+" (read/write) mode, and the type of object returned is a
1812 hash, unless otherwise specified (see L<OPTIONS> below).
1816 You can pass a number of options to the constructor to specify things like
1817 locking, autoflush, etc. This is done by passing an inline hash:
1819 my $db = DBM::Deep->new(
1825 Notice that the filename is now specified I<inside> the hash with
1826 the "file" parameter, as opposed to being the sole argument to the
1827 constructor. This is required if any options are specified.
1828 See L<OPTIONS> below for the complete list.
1832 You can also start with an array instead of a hash. For this, you must
1833 specify the C<type> parameter:
1835 my $db = DBM::Deep->new(
1837 type => DBM::Deep->TYPE_ARRAY
1840 B<Note:> Specifing the C<type> parameter only takes effect when beginning
1841 a new DB file. If you create a DBM::Deep object with an existing file, the
1842 C<type> will be loaded from the file header, and ignored if it is passed
1845 =head2 TIE CONSTRUCTION
1847 Alternatively, you can create a DBM::Deep handle by using Perl's built-in
1848 tie() function. This is not ideal, because you get only a basic, tied hash
1849 (or array) which is not blessed, so you can't call any functions on it.
1852 tie %hash, "DBM::Deep", "foo.db";
1855 tie @array, "DBM::Deep", "bar.db";
1857 As with the OO constructor, you can replace the DB filename parameter with
1858 a hash containing one or more options (see L<OPTIONS> just below for the
1861 tie %hash, "DBM::Deep", {
1869 There are a number of options that can be passed in when constructing your
1870 DBM::Deep objects. These apply to both the OO- and tie- based approaches.
1876 Filename of the DB file to link the handle to. You can pass a full absolute
1877 filesystem path, partial path, or a plain filename if the file is in the
1878 current working directory. This is a required parameter.
1882 File open mode (read-only, read-write, etc.) string passed to Perl's FileHandle
1883 module. This is an optional parameter, and defaults to "r+" (read/write).
1884 B<Note:> If the default (r+) mode is selected, the file will also be auto-
1885 created if it doesn't exist.
1889 This parameter specifies what type of object to create, a hash or array. Use
1890 one of these two constants: C<DBM::Deep-E<gt>TYPE_HASH> or C<DBM::Deep-E<gt>TYPE_ARRAY>.
1891 This only takes effect when beginning a new file. This is an optional
1892 parameter, and defaults to C<DBM::Deep-E<gt>TYPE_HASH>.
1896 Specifies whether locking is to be enabled. DBM::Deep uses Perl's Fnctl flock()
1897 function to lock the database in exclusive mode for writes, and shared mode for
1898 reads. Pass any true value to enable. This affects the base DB handle I<and
1899 any child hashes or arrays> that use the same DB file. This is an optional
1900 parameter, and defaults to 0 (disabled). See L<LOCKING> below for more.
1904 Specifies whether autoflush is to be enabled on the underlying FileHandle.
1905 This obviously slows down write operations, but is required if you may have
1906 multiple processes accessing the same DB file (also consider enable I<locking>
1907 or at least I<volatile>). Pass any true value to enable. This is an optional
1908 parameter, and defaults to 0 (disabled).
1912 If I<volatile> mode is enabled, DBM::Deep will stat() the DB file before each
1913 STORE() operation. This is required if an outside force may change the size of
1914 the file between transactions. Locking also implicitly enables volatile. This
1915 is useful if you want to use a different locking system or write your own. Pass
1916 any true value to enable. This is an optional parameter, and defaults to 0
1921 If I<autobless> mode is enabled, DBM::Deep will preserve blessed hashes, and
1922 restore them when fetched. This is an B<experimental> feature, and does have
1923 side-effects. Basically, when hashes are re-blessed into their original
1924 classes, they are no longer blessed into the DBM::Deep class! So you won't be
1925 able to call any DBM::Deep methods on them. You have been warned.
1926 This is an optional parameter, and defaults to 0 (disabled).
1930 See L<FILTERS> below.
1934 Setting I<debug> mode will make all errors non-fatal, dump them out to
1935 STDERR, and continue on. This is for debugging purposes only, and probably
1936 not what you want. This is an optional parameter, and defaults to 0 (disabled).
1940 Instead of passing a file path, you can instead pass a handle to an pre-opened
1941 filehandle. Note: Beware of using the magick *DATA handle, as this actually
1942 contains your entire Perl script, as well as the data following the __DATA__
1943 marker. This will not work, because DBM::Deep uses absolute seek()s into the
1944 file. Instead, consider reading *DATA into an IO::Scalar handle, then passing
1949 =head1 TIE INTERFACE
1951 With DBM::Deep you can access your databases using Perl's standard hash/array
1952 syntax. Because all DBM::Deep objects are I<tied> to hashes or arrays, you can treat
1953 them as such. DBM::Deep will intercept all reads/writes and direct them to the right
1954 place -- the DB file. This has nothing to do with the L<TIE CONSTRUCTION>
1955 section above. This simply tells you how to use DBM::Deep using regular hashes
1956 and arrays, rather than calling functions like C<get()> and C<put()> (although those
1957 work too). It is entirely up to you how to want to access your databases.
1961 You can treat any DBM::Deep object like a normal Perl hash reference. Add keys,
1962 or even nested hashes (or arrays) using standard Perl syntax:
1964 my $db = DBM::Deep->new( "foo.db" );
1966 $db->{mykey} = "myvalue";
1968 $db->{myhash}->{subkey} = "subvalue";
1970 print $db->{myhash}->{subkey} . "\n";
1972 You can even step through hash keys using the normal Perl C<keys()> function:
1974 foreach my $key (keys %$db) {
1975 print "$key: " . $db->{$key} . "\n";
1978 Remember that Perl's C<keys()> function extracts I<every> key from the hash and
1979 pushes them onto an array, all before the loop even begins. If you have an
1980 extra large hash, this may exhaust Perl's memory. Instead, consider using
1981 Perl's C<each()> function, which pulls keys/values one at a time, using very
1984 while (my ($key, $value) = each %$db) {
1985 print "$key: $value\n";
1988 Please note that when using C<each()>, you should always pass a direct
1989 hash reference, not a lookup. Meaning, you should B<never> do this:
1992 while (my ($key, $value) = each %{$db->{foo}}) { # BAD
1994 This causes an infinite loop, because for each iteration, Perl is calling
1995 FETCH() on the $db handle, resulting in a "new" hash for foo every time, so
1996 it effectively keeps returning the first key over and over again. Instead,
1997 assign a temporary variable to C<$db->{foo}>, then pass that to each().
2001 As with hashes, you can treat any DBM::Deep object like a normal Perl array
2002 reference. This includes inserting, removing and manipulating elements,
2003 and the C<push()>, C<pop()>, C<shift()>, C<unshift()> and C<splice()> functions.
2004 The object must have first been created using type C<DBM::Deep-E<gt>TYPE_ARRAY>,
2005 or simply be a nested array reference inside a hash. Example:
2007 my $db = DBM::Deep->new(
2008 file => "foo-array.db",
2009 type => DBM::Deep->TYPE_ARRAY
2013 push @$db, "bar", "baz";
2014 unshift @$db, "bah";
2016 my $last_elem = pop @$db; # baz
2017 my $first_elem = shift @$db; # bah
2018 my $second_elem = $db->[1]; # bar
2020 my $num_elements = scalar @$db;
2024 In addition to the I<tie()> interface, you can also use a standard OO interface
2025 to manipulate all aspects of DBM::Deep databases. Each type of object (hash or
2026 array) has its own methods, but both types share the following common methods:
2027 C<put()>, C<get()>, C<exists()>, C<delete()> and C<clear()>.
2033 Stores a new hash key/value pair, or sets an array element value. Takes two
2034 arguments, the hash key or array index, and the new value. The value can be
2035 a scalar, hash ref or array ref. Returns true on success, false on failure.
2037 $db->put("foo", "bar"); # for hashes
2038 $db->put(1, "bar"); # for arrays
2042 Fetches the value of a hash key or array element. Takes one argument: the hash
2043 key or array index. Returns a scalar, hash ref or array ref, depending on the
2046 my $value = $db->get("foo"); # for hashes
2047 my $value = $db->get(1); # for arrays
2051 Checks if a hash key or array index exists. Takes one argument: the hash key
2052 or array index. Returns true if it exists, false if not.
2054 if ($db->exists("foo")) { print "yay!\n"; } # for hashes
2055 if ($db->exists(1)) { print "yay!\n"; } # for arrays
2059 Deletes one hash key/value pair or array element. Takes one argument: the hash
2060 key or array index. Returns true on success, false if not found. For arrays,
2061 the remaining elements located after the deleted element are NOT moved over.
2062 The deleted element is essentially just undefined, which is exactly how Perl's
2063 internal arrays work. Please note that the space occupied by the deleted
2064 key/value or element is B<not> reused again -- see L<UNUSED SPACE RECOVERY>
2065 below for details and workarounds.
2067 $db->delete("foo"); # for hashes
2068 $db->delete(1); # for arrays
2072 Deletes B<all> hash keys or array elements. Takes no arguments. No return
2073 value. Please note that the space occupied by the deleted keys/values or
2074 elements is B<not> reused again -- see L<UNUSED SPACE RECOVERY> below for
2075 details and workarounds.
2077 $db->clear(); # hashes or arrays
2083 For hashes, DBM::Deep supports all the common methods described above, and the
2084 following additional methods: C<first_key()> and C<next_key()>.
2090 Returns the "first" key in the hash. As with built-in Perl hashes, keys are
2091 fetched in an undefined order (which appears random). Takes no arguments,
2092 returns the key as a scalar value.
2094 my $key = $db->first_key();
2098 Returns the "next" key in the hash, given the previous one as the sole argument.
2099 Returns undef if there are no more keys to be fetched.
2101 $key = $db->next_key($key);
2105 Here are some examples of using hashes:
2107 my $db = DBM::Deep->new( "foo.db" );
2109 $db->put("foo", "bar");
2110 print "foo: " . $db->get("foo") . "\n";
2112 $db->put("baz", {}); # new child hash ref
2113 $db->get("baz")->put("buz", "biz");
2114 print "buz: " . $db->get("baz")->get("buz") . "\n";
2116 my $key = $db->first_key();
2118 print "$key: " . $db->get($key) . "\n";
2119 $key = $db->next_key($key);
2122 if ($db->exists("foo")) { $db->delete("foo"); }
2126 For arrays, DBM::Deep supports all the common methods described above, and the
2127 following additional methods: C<length()>, C<push()>, C<pop()>, C<shift()>,
2128 C<unshift()> and C<splice()>.
2134 Returns the number of elements in the array. Takes no arguments.
2136 my $len = $db->length();
2140 Adds one or more elements onto the end of the array. Accepts scalars, hash
2141 refs or array refs. No return value.
2143 $db->push("foo", "bar", {});
2147 Fetches the last element in the array, and deletes it. Takes no arguments.
2148 Returns undef if array is empty. Returns the element value.
2150 my $elem = $db->pop();
2154 Fetches the first element in the array, deletes it, then shifts all the
2155 remaining elements over to take up the space. Returns the element value. This
2156 method is not recommended with large arrays -- see L<LARGE ARRAYS> below for
2159 my $elem = $db->shift();
2163 Inserts one or more elements onto the beginning of the array, shifting all
2164 existing elements over to make room. Accepts scalars, hash refs or array refs.
2165 No return value. This method is not recommended with large arrays -- see
2166 <LARGE ARRAYS> below for details.
2168 $db->unshift("foo", "bar", {});
2172 Performs exactly like Perl's built-in function of the same name. See L<perldoc
2173 -f splice> for usage -- it is too complicated to document here. This method is
2174 not recommended with large arrays -- see L<LARGE ARRAYS> below for details.
2178 Here are some examples of using arrays:
2180 my $db = DBM::Deep->new(
2182 type => DBM::Deep->TYPE_ARRAY
2185 $db->push("bar", "baz");
2186 $db->unshift("foo");
2189 my $len = $db->length();
2190 print "length: $len\n"; # 4
2192 for (my $k=0; $k<$len; $k++) {
2193 print "$k: " . $db->get($k) . "\n";
2196 $db->splice(1, 2, "biz", "baf");
2198 while (my $elem = shift @$db) {
2199 print "shifted: $elem\n";
2204 Enable automatic file locking by passing a true value to the C<locking>
2205 parameter when constructing your DBM::Deep object (see L<SETUP> above).
2207 my $db = DBM::Deep->new(
2212 This causes DBM::Deep to C<flock()> the underlying FileHandle object with exclusive
2213 mode for writes, and shared mode for reads. This is required if you have
2214 multiple processes accessing the same database file, to avoid file corruption.
2215 Please note that C<flock()> does NOT work for files over NFS. See L<DB OVER
2216 NFS> below for more.
2218 =head2 EXPLICIT LOCKING
2220 You can explicitly lock a database, so it remains locked for multiple
2221 transactions. This is done by calling the C<lock()> method, and passing an
2222 optional lock mode argument (defaults to exclusive mode). This is particularly
2223 useful for things like counters, where the current value needs to be fetched,
2224 then incremented, then stored again.
2227 my $counter = $db->get("counter");
2229 $db->put("counter", $counter);
2238 You can pass C<lock()> an optional argument, which specifies which mode to use
2239 (exclusive or shared). Use one of these two constants: C<DBM::Deep-E<gt>LOCK_EX>
2240 or C<DBM::Deep-E<gt>LOCK_SH>. These are passed directly to C<flock()>, and are the
2241 same as the constants defined in Perl's C<Fcntl> module.
2243 $db->lock( DBM::Deep->LOCK_SH );
2247 If you want to implement your own file locking scheme, be sure to create your
2248 DBM::Deep objects setting the C<volatile> option to true. This hints to DBM::Deep
2249 that the DB file may change between transactions. See L<LOW-LEVEL ACCESS>
2252 =head1 IMPORTING/EXPORTING
2254 You can import existing complex structures by calling the C<import()> method,
2255 and export an entire database into an in-memory structure using the C<export()>
2256 method. Both are examined here.
2260 Say you have an existing hash with nested hashes/arrays inside it. Instead of
2261 walking the structure and adding keys/elements to the database as you go,
2262 simply pass a reference to the C<import()> method. This recursively adds
2263 everything to an existing DBM::Deep object for you. Here is an example:
2268 array1 => [ "elem0", "elem1", "elem2" ],
2270 subkey1 => "subvalue1",
2271 subkey2 => "subvalue2"
2275 my $db = DBM::Deep->new( "foo.db" );
2276 $db->import( $struct );
2278 print $db->{key1} . "\n"; # prints "value1"
2280 This recursively imports the entire C<$struct> object into C<$db>, including
2281 all nested hashes and arrays. If the DBM::Deep object contains exsiting data,
2282 keys are merged with the existing ones, replacing if they already exist.
2283 The C<import()> method can be called on any database level (not just the base
2284 level), and works with both hash and array DB types.
2288 B<Note:> Make sure your existing structure has no circular references in it.
2289 These will cause an infinite loop when importing.
2293 Calling the C<export()> method on an existing DBM::Deep object will return
2294 a reference to a new in-memory copy of the database. The export is done
2295 recursively, so all nested hashes/arrays are all exported to standard Perl
2296 objects. Here is an example:
2298 my $db = DBM::Deep->new( "foo.db" );
2300 $db->{key1} = "value1";
2301 $db->{key2} = "value2";
2303 $db->{hash1}->{subkey1} = "subvalue1";
2304 $db->{hash1}->{subkey2} = "subvalue2";
2306 my $struct = $db->export();
2308 print $struct->{key1} . "\n"; # prints "value1"
2310 This makes a complete copy of the database in memory, and returns a reference
2311 to it. The C<export()> method can be called on any database level (not just
2312 the base level), and works with both hash and array DB types. Be careful of
2313 large databases -- you can store a lot more data in a DBM::Deep object than an
2314 in-memory Perl structure.
2318 B<Note:> Make sure your database has no circular references in it.
2319 These will cause an infinite loop when exporting.
2323 DBM::Deep has a number of hooks where you can specify your own Perl function
2324 to perform filtering on incoming or outgoing data. This is a perfect
2325 way to extend the engine, and implement things like real-time compression or
2326 encryption. Filtering applies to the base DB level, and all child hashes /
2327 arrays. Filter hooks can be specified when your DBM::Deep object is first
2328 constructed, or by calling the C<set_filter()> method at any time. There are
2329 four available filter hooks, described below:
2333 =item * filter_store_key
2335 This filter is called whenever a hash key is stored. It
2336 is passed the incoming key, and expected to return a transformed key.
2338 =item * filter_store_value
2340 This filter is called whenever a hash key or array element is stored. It
2341 is passed the incoming value, and expected to return a transformed value.
2343 =item * filter_fetch_key
2345 This filter is called whenever a hash key is fetched (i.e. via
2346 C<first_key()> or C<next_key()>). It is passed the transformed key,
2347 and expected to return the plain key.
2349 =item * filter_fetch_value
2351 This filter is called whenever a hash key or array element is fetched.
2352 It is passed the transformed value, and expected to return the plain value.
2356 Here are the two ways to setup a filter hook:
2358 my $db = DBM::Deep->new(
2360 filter_store_value => \&my_filter_store,
2361 filter_fetch_value => \&my_filter_fetch
2366 $db->set_filter( "filter_store_value", \&my_filter_store );
2367 $db->set_filter( "filter_fetch_value", \&my_filter_fetch );
2369 Your filter function will be called only when dealing with SCALAR keys or
2370 values. When nested hashes and arrays are being stored/fetched, filtering
2371 is bypassed. Filters are called as static functions, passed a single SCALAR
2372 argument, and expected to return a single SCALAR value. If you want to
2373 remove a filter, set the function reference to C<undef>:
2375 $db->set_filter( "filter_store_value", undef );
2377 =head2 REAL-TIME ENCRYPTION EXAMPLE
2379 Here is a working example that uses the I<Crypt::Blowfish> module to
2380 do real-time encryption / decryption of keys & values with DBM::Deep Filters.
2381 Please visit L<http://search.cpan.org/search?module=Crypt::Blowfish> for more
2382 on I<Crypt::Blowfish>. You'll also need the I<Crypt::CBC> module.
2385 use Crypt::Blowfish;
2388 my $cipher = Crypt::CBC->new({
2389 'key' => 'my secret key',
2390 'cipher' => 'Blowfish',
2392 'regenerate_key' => 0,
2393 'padding' => 'space',
2397 my $db = DBM::Deep->new(
2398 file => "foo-encrypt.db",
2399 filter_store_key => \&my_encrypt,
2400 filter_store_value => \&my_encrypt,
2401 filter_fetch_key => \&my_decrypt,
2402 filter_fetch_value => \&my_decrypt,
2405 $db->{key1} = "value1";
2406 $db->{key2} = "value2";
2407 print "key1: " . $db->{key1} . "\n";
2408 print "key2: " . $db->{key2} . "\n";
2414 return $cipher->encrypt( $_[0] );
2417 return $cipher->decrypt( $_[0] );
2420 =head2 REAL-TIME COMPRESSION EXAMPLE
2422 Here is a working example that uses the I<Compress::Zlib> module to do real-time
2423 compression / decompression of keys & values with DBM::Deep Filters.
2424 Please visit L<http://search.cpan.org/search?module=Compress::Zlib> for
2425 more on I<Compress::Zlib>.
2430 my $db = DBM::Deep->new(
2431 file => "foo-compress.db",
2432 filter_store_key => \&my_compress,
2433 filter_store_value => \&my_compress,
2434 filter_fetch_key => \&my_decompress,
2435 filter_fetch_value => \&my_decompress,
2438 $db->{key1} = "value1";
2439 $db->{key2} = "value2";
2440 print "key1: " . $db->{key1} . "\n";
2441 print "key2: " . $db->{key2} . "\n";
2447 return Compress::Zlib::memGzip( $_[0] ) ;
2450 return Compress::Zlib::memGunzip( $_[0] ) ;
2453 B<Note:> Filtering of keys only applies to hashes. Array "keys" are
2454 actually numerical index numbers, and are not filtered.
2456 =head1 ERROR HANDLING
2458 Most DBM::Deep methods return a true value for success, and call die() on
2459 failure. You can wrap calls in an eval block to catch the die. Also, the
2460 actual error message is stored in an internal scalar, which can be fetched by
2461 calling the C<error()> method.
2463 my $db = DBM::Deep->new( "foo.db" ); # create hash
2464 eval { $db->push("foo"); }; # ILLEGAL -- push is array-only call
2466 print $db->error(); # prints error message
2468 You can then call C<clear_error()> to clear the current error state.
2472 If you set the C<debug> option to true when creating your DBM::Deep object,
2473 all errors are considered NON-FATAL, and dumped to STDERR. This is only
2474 for debugging purposes.
2476 =head1 LARGEFILE SUPPORT
2478 If you have a 64-bit system, and your Perl is compiled with both LARGEFILE
2479 and 64-bit support, you I<may> be able to create databases larger than 2 GB.
2480 DBM::Deep by default uses 32-bit file offset tags, but these can be changed
2481 by calling the static C<set_pack()> method before you do anything else.
2483 DBM::Deep::set_pack(8, 'Q');
2485 This tells DBM::Deep to pack all file offsets with 8-byte (64-bit) quad words
2486 instead of 32-bit longs. After setting these values your DB files have a
2487 theoretical maximum size of 16 XB (exabytes).
2491 B<Note:> Changing these values will B<NOT> work for existing database files.
2492 Only change this for new files, and make sure it stays set consistently
2493 throughout the file's life. If you do set these values, you can no longer
2494 access 32-bit DB files. You can, however, call C<set_pack(4, 'N')> to change
2495 back to 32-bit mode.
2499 B<Note:> I have not personally tested files > 2 GB -- all my systems have
2500 only a 32-bit Perl. However, I have received user reports that this does
2503 =head1 LOW-LEVEL ACCESS
2505 If you require low-level access to the underlying FileHandle that DBM::Deep uses,
2506 you can call the C<fh()> method, which returns the handle:
2510 This method can be called on the root level of the datbase, or any child
2511 hashes or arrays. All levels share a I<root> structure, which contains things
2512 like the FileHandle, a reference counter, and all your options you specified
2513 when you created the object. You can get access to this root structure by
2514 calling the C<root()> method.
2516 my $root = $db->root();
2518 This is useful for changing options after the object has already been created,
2519 such as enabling/disabling locking, volatile or debug modes. You can also
2520 store your own temporary user data in this structure (be wary of name
2521 collision), which is then accessible from any child hash or array.
2523 =head1 CUSTOM DIGEST ALGORITHM
2525 DBM::Deep by default uses the I<Message Digest 5> (MD5) algorithm for hashing
2526 keys. However you can override this, and use another algorithm (such as SHA-256)
2527 or even write your own. But please note that DBM::Deep currently expects zero
2528 collisions, so your algorithm has to be I<perfect>, so to speak.
2529 Collision detection may be introduced in a later version.
2533 You can specify a custom digest algorithm by calling the static C<set_digest()>
2534 function, passing a reference to a subroutine, and the length of the algorithm's
2535 hashes (in bytes). This is a global static function, which affects ALL DBM::Deep
2536 objects. Here is a working example that uses a 256-bit hash from the
2537 I<Digest::SHA256> module. Please see
2538 L<http://search.cpan.org/search?module=Digest::SHA256> for more.
2543 my $context = Digest::SHA256::new(256);
2545 DBM::Deep::set_digest( \&my_digest, 32 );
2547 my $db = DBM::Deep->new( "foo-sha.db" );
2549 $db->{key1} = "value1";
2550 $db->{key2} = "value2";
2551 print "key1: " . $db->{key1} . "\n";
2552 print "key2: " . $db->{key2} . "\n";
2558 return substr( $context->hash($_[0]), 0, 32 );
2561 B<Note:> Your returned digest strings must be B<EXACTLY> the number
2562 of bytes you specify in the C<set_digest()> function (in this case 32).
2564 =head1 CIRCULAR REFERENCES
2566 DBM::Deep has B<experimental> support for circular references. Meaning you
2567 can have a nested hash key or array element that points to a parent object.
2568 This relationship is stored in the DB file, and is preserved between sessions.
2571 my $db = DBM::Deep->new( "foo.db" );
2574 $db->{circle} = $db; # ref to self
2576 print $db->{foo} . "\n"; # prints "foo"
2577 print $db->{circle}->{foo} . "\n"; # prints "foo" again
2579 One catch is, passing the object to a function that recursively walks the
2580 object tree (such as I<Data::Dumper> or even the built-in C<optimize()> or
2581 C<export()> methods) will result in an infinite loop. The other catch is,
2582 if you fetch the I<key> of a circular reference (i.e. using the C<first_key()>
2583 or C<next_key()> methods), you will get the I<target object's key>, not the
2584 ref's key. This gets even more interesting with the above example, where
2585 the I<circle> key points to the base DB object, which technically doesn't
2586 have a key. So I made DBM::Deep return "[base]" as the key name in that
2589 =head1 CAVEATS / ISSUES / BUGS
2591 This section describes all the known issues with DBM::Deep. It you have found
2592 something that is not listed here, please send e-mail to L<jhuckaby@cpan.org>.
2594 =head2 UNUSED SPACE RECOVERY
2596 One major caveat with DBM::Deep is that space occupied by existing keys and
2597 values is not recovered when they are deleted. Meaning if you keep deleting
2598 and adding new keys, your file will continuously grow. I am working on this,
2599 but in the meantime you can call the built-in C<optimize()> method from time to
2600 time (perhaps in a crontab or something) to recover all your unused space.
2602 $db->optimize(); # returns true on success
2604 This rebuilds the ENTIRE database into a new file, then moves it on top of
2605 the original. The new file will have no unused space, thus it will take up as
2606 little disk space as possible. Please note that this operation can take
2607 a long time for large files, and you need enough disk space to temporarily hold
2608 2 copies of your DB file. The temporary file is created in the same directory
2609 as the original, named with a ".tmp" extension, and is deleted when the
2610 operation completes. Oh, and if locking is enabled, the DB is automatically
2611 locked for the entire duration of the copy.
2615 B<WARNING:> Only call optimize() on the top-level node of the database, and
2616 make sure there are no child references lying around. DBM::Deep keeps a reference
2617 counter, and if it is greater than 1, optimize() will abort and return undef.
2619 =head2 AUTOVIVIFICATION
2621 Unfortunately, autovivification doesn't work with tied hashes. This appears to
2622 be a bug in Perl's tie() system, as I<Jakob Schmidt> encountered the very same
2623 issue with his I<DWH_FIle> module (see L<http://search.cpan.org/search?module=DWH_File>),
2624 and it is also mentioned in the BUGS section for the I<MLDBM> module <see
2625 L<http://search.cpan.org/search?module=MLDBM>). Basically, on a new db file,
2628 $db->{foo}->{bar} = "hello";
2630 Since "foo" doesn't exist, you cannot add "bar" to it. You end up with "foo"
2631 being an empty hash. Try this instead, which works fine:
2633 $db->{foo} = { bar => "hello" };
2635 As of Perl 5.8.7, this bug still exists. I have walked very carefully through
2636 the execution path, and Perl indeed passes an empty hash to the STORE() method.
2637 Probably a bug in Perl.
2639 =head2 FILE CORRUPTION
2641 The current level of error handling in DBM::Deep is minimal. Files I<are> checked
2642 for a 32-bit signature when opened, but other corruption in files can cause
2643 segmentation faults. DBM::Deep may try to seek() past the end of a file, or get
2644 stuck in an infinite loop depending on the level of corruption. File write
2645 operations are not checked for failure (for speed), so if you happen to run
2646 out of disk space, DBM::Deep will probably fail in a bad way. These things will
2647 be addressed in a later version of DBM::Deep.
2651 Beware of using DB files over NFS. DBM::Deep uses flock(), which works well on local
2652 filesystems, but will NOT protect you from file corruption over NFS. I've heard
2653 about setting up your NFS server with a locking daemon, then using lockf() to
2654 lock your files, but your milage may vary there as well. From what I
2655 understand, there is no real way to do it. However, if you need access to the
2656 underlying FileHandle in DBM::Deep for using some other kind of locking scheme like
2657 lockf(), see the L<LOW-LEVEL ACCESS> section above.
2659 =head2 COPYING OBJECTS
2661 Beware of copying tied objects in Perl. Very strange things can happen.
2662 Instead, use DBM::Deep's C<clone()> method which safely copies the object and
2663 returns a new, blessed, tied hash or array to the same level in the DB.
2665 my $copy = $db->clone();
2669 Beware of using C<shift()>, C<unshift()> or C<splice()> with large arrays.
2670 These functions cause every element in the array to move, which can be murder
2671 on DBM::Deep, as every element has to be fetched from disk, then stored again in
2672 a different location. This may be addressed in a later version.
2676 This section discusses DBM::Deep's speed and memory usage.
2680 Obviously, DBM::Deep isn't going to be as fast as some C-based DBMs, such as
2681 the almighty I<BerkeleyDB>. But it makes up for it in features like true
2682 multi-level hash/array support, and cross-platform FTPable files. Even so,
2683 DBM::Deep is still pretty fast, and the speed stays fairly consistent, even
2684 with huge databases. Here is some test data:
2686 Adding 1,000,000 keys to new DB file...
2688 At 100 keys, avg. speed is 2,703 keys/sec
2689 At 200 keys, avg. speed is 2,642 keys/sec
2690 At 300 keys, avg. speed is 2,598 keys/sec
2691 At 400 keys, avg. speed is 2,578 keys/sec
2692 At 500 keys, avg. speed is 2,722 keys/sec
2693 At 600 keys, avg. speed is 2,628 keys/sec
2694 At 700 keys, avg. speed is 2,700 keys/sec
2695 At 800 keys, avg. speed is 2,607 keys/sec
2696 At 900 keys, avg. speed is 2,190 keys/sec
2697 At 1,000 keys, avg. speed is 2,570 keys/sec
2698 At 2,000 keys, avg. speed is 2,417 keys/sec
2699 At 3,000 keys, avg. speed is 1,982 keys/sec
2700 At 4,000 keys, avg. speed is 1,568 keys/sec
2701 At 5,000 keys, avg. speed is 1,533 keys/sec
2702 At 6,000 keys, avg. speed is 1,787 keys/sec
2703 At 7,000 keys, avg. speed is 1,977 keys/sec
2704 At 8,000 keys, avg. speed is 2,028 keys/sec
2705 At 9,000 keys, avg. speed is 2,077 keys/sec
2706 At 10,000 keys, avg. speed is 2,031 keys/sec
2707 At 20,000 keys, avg. speed is 1,970 keys/sec
2708 At 30,000 keys, avg. speed is 2,050 keys/sec
2709 At 40,000 keys, avg. speed is 2,073 keys/sec
2710 At 50,000 keys, avg. speed is 1,973 keys/sec
2711 At 60,000 keys, avg. speed is 1,914 keys/sec
2712 At 70,000 keys, avg. speed is 2,091 keys/sec
2713 At 80,000 keys, avg. speed is 2,103 keys/sec
2714 At 90,000 keys, avg. speed is 1,886 keys/sec
2715 At 100,000 keys, avg. speed is 1,970 keys/sec
2716 At 200,000 keys, avg. speed is 2,053 keys/sec
2717 At 300,000 keys, avg. speed is 1,697 keys/sec
2718 At 400,000 keys, avg. speed is 1,838 keys/sec
2719 At 500,000 keys, avg. speed is 1,941 keys/sec
2720 At 600,000 keys, avg. speed is 1,930 keys/sec
2721 At 700,000 keys, avg. speed is 1,735 keys/sec
2722 At 800,000 keys, avg. speed is 1,795 keys/sec
2723 At 900,000 keys, avg. speed is 1,221 keys/sec
2724 At 1,000,000 keys, avg. speed is 1,077 keys/sec
2726 This test was performed on a PowerMac G4 1gHz running Mac OS X 10.3.2 & Perl
2727 5.8.1, with an 80GB Ultra ATA/100 HD spinning at 7200RPM. The hash keys and
2728 values were between 6 - 12 chars in length. The DB file ended up at 210MB.
2729 Run time was 12 min 3 sec.
2733 One of the great things about DBM::Deep is that it uses very little memory.
2734 Even with huge databases (1,000,000+ keys) you will not see much increased
2735 memory on your process. DBM::Deep relies solely on the filesystem for storing
2736 and fetching data. Here is output from I</usr/bin/top> before even opening a
2739 PID USER PRI NI SIZE RSS SHARE STAT %CPU %MEM TIME COMMAND
2740 22831 root 11 0 2716 2716 1296 R 0.0 0.2 0:07 perl
2742 Basically the process is taking 2,716K of memory. And here is the same
2743 process after storing and fetching 1,000,000 keys:
2745 PID USER PRI NI SIZE RSS SHARE STAT %CPU %MEM TIME COMMAND
2746 22831 root 14 0 2772 2772 1328 R 0.0 0.2 13:32 perl
2748 Notice the memory usage increased by only 56K. Test was performed on a 700mHz
2749 x86 box running Linux RedHat 7.2 & Perl 5.6.1.
2751 =head1 DB FILE FORMAT
2753 In case you were interested in the underlying DB file format, it is documented
2754 here in this section. You don't need to know this to use the module, it's just
2755 included for reference.
2759 DBM::Deep files always start with a 32-bit signature to identify the file type.
2760 This is at offset 0. The signature is "DPDB" in network byte order. This is
2761 checked when the file is opened.
2765 The DBM::Deep file is in a I<tagged format>, meaning each section of the file
2766 has a standard header containing the type of data, the length of data, and then
2767 the data itself. The type is a single character (1 byte), the length is a
2768 32-bit unsigned long in network byte order, and the data is, well, the data.
2769 Here is how it unfolds:
2773 Immediately after the 32-bit file signature is the I<Master Index> record.
2774 This is a standard tag header followed by 1024 bytes (in 32-bit mode) or 2048
2775 bytes (in 64-bit mode) of data. The type is I<H> for hash or I<A> for array,
2776 depending on how the DBM::Deep object was constructed.
2780 The index works by looking at a I<MD5 Hash> of the hash key (or array index
2781 number). The first 8-bit char of the MD5 signature is the offset into the
2782 index, multipled by 4 in 32-bit mode, or 8 in 64-bit mode. The value of the
2783 index element is a file offset of the next tag for the key/element in question,
2784 which is usually a I<Bucket List> tag (see below).
2788 The next tag I<could> be another index, depending on how many keys/elements
2789 exist. See L<RE-INDEXING> below for details.
2793 A I<Bucket List> is a collection of 16 MD5 hashes for keys/elements, plus
2794 file offsets to where the actual data is stored. It starts with a standard
2795 tag header, with type I<B>, and a data size of 320 bytes in 32-bit mode, or
2796 384 bytes in 64-bit mode. Each MD5 hash is stored in full (16 bytes), plus
2797 the 32-bit or 64-bit file offset for the I<Bucket> containing the actual data.
2798 When the list fills up, a I<Re-Index> operation is performed (See
2799 L<RE-INDEXING> below).
2803 A I<Bucket> is a tag containing a key/value pair (in hash mode), or a
2804 index/value pair (in array mode). It starts with a standard tag header with
2805 type I<D> for scalar data (string, binary, etc.), or it could be a nested
2806 hash (type I<H>) or array (type I<A>). The value comes just after the tag
2807 header. The size reported in the tag header is only for the value, but then,
2808 just after the value is another size (32-bit unsigned long) and then the plain
2809 key itself. Since the value is likely to be fetched more often than the plain
2810 key, I figured it would be I<slightly> faster to store the value first.
2814 If the type is I<H> (hash) or I<A> (array), the value is another I<Master Index>
2815 record for the nested structure, where the process begins all over again.
2819 After a I<Bucket List> grows to 16 records, its allocated space in the file is
2820 exhausted. Then, when another key/element comes in, the list is converted to a
2821 new index record. However, this index will look at the next char in the MD5
2822 hash, and arrange new Bucket List pointers accordingly. This process is called
2823 I<Re-Indexing>. Basically, a new index tag is created at the file EOF, and all
2824 17 (16 + new one) keys/elements are removed from the old Bucket List and
2825 inserted into the new index. Several new Bucket Lists are created in the
2826 process, as a new MD5 char from the key is being examined (it is unlikely that
2827 the keys will all share the same next char of their MD5s).
2831 Because of the way the I<MD5> algorithm works, it is impossible to tell exactly
2832 when the Bucket Lists will turn into indexes, but the first round tends to
2833 happen right around 4,000 keys. You will see a I<slight> decrease in
2834 performance here, but it picks back up pretty quick (see L<SPEED> above). Then
2835 it takes B<a lot> more keys to exhaust the next level of Bucket Lists. It's
2836 right around 900,000 keys. This process can continue nearly indefinitely --
2837 right up until the point the I<MD5> signatures start colliding with each other,
2838 and this is B<EXTREMELY> rare -- like winning the lottery 5 times in a row AND
2839 getting struck by lightning while you are walking to cash in your tickets.
2840 Theoretically, since I<MD5> hashes are 128-bit values, you I<could> have up to
2841 340,282,366,921,000,000,000,000,000,000,000,000,000 keys/elements (I believe
2842 this is 340 unodecillion, but don't quote me).
2846 When a new key/element is stored, the key (or index number) is first ran through
2847 I<Digest::MD5> to get a 128-bit signature (example, in hex:
2848 b05783b0773d894396d475ced9d2f4f6). Then, the I<Master Index> record is checked
2849 for the first char of the signature (in this case I<b>). If it does not exist,
2850 a new I<Bucket List> is created for our key (and the next 15 future keys that
2851 happen to also have I<b> as their first MD5 char). The entire MD5 is written
2852 to the I<Bucket List> along with the offset of the new I<Bucket> record (EOF at
2853 this point, unless we are replacing an existing I<Bucket>), where the actual
2854 data will be stored.
2858 Fetching an existing key/element involves getting a I<Digest::MD5> of the key
2859 (or index number), then walking along the indexes. If there are enough
2860 keys/elements in this DB level, there might be nested indexes, each linked to
2861 a particular char of the MD5. Finally, a I<Bucket List> is pointed to, which
2862 contains up to 16 full MD5 hashes. Each is checked for equality to the key in
2863 question. If we found a match, the I<Bucket> tag is loaded, where the value and
2864 plain key are stored.
2868 Fetching the plain key occurs when calling the I<first_key()> and I<next_key()>
2869 methods. In this process the indexes are walked systematically, and each key
2870 fetched in increasing MD5 order (which is why it appears random). Once the
2871 I<Bucket> is found, the value is skipped the plain key returned instead.
2872 B<Note:> Do not count on keys being fetched as if the MD5 hashes were
2873 alphabetically sorted. This only happens on an index-level -- as soon as the
2874 I<Bucket Lists> are hit, the keys will come out in the order they went in --
2875 so it's pretty much undefined how the keys will come out -- just like Perl's
2878 =head1 CODE COVERAGE
2880 I use B<Devel::Cover> to test the code coverage of my tests, below is the B<Devel::Cover> report on this
2881 module's test suite.
2883 ---------------------------- ------ ------ ------ ------ ------ ------ ------
2884 File stmt bran cond sub pod time total
2885 ---------------------------- ------ ------ ------ ------ ------ ------ ------
2886 blib/lib/DBM/Deep.pm 94.9 84.5 77.8 100.0 11.1 100.0 89.7
2887 Total 94.9 84.5 77.8 100.0 11.1 100.0 89.7
2888 ---------------------------- ------ ------ ------ ------ ------ ------ ------
2892 Joseph Huckaby, L<jhuckaby@cpan.org>
2894 Special thanks to Adam Sah and Rich Gaushell! You know why :-)
2898 perltie(1), Tie::Hash(3), Digest::MD5(3), Fcntl(3), flock(2), lockf(3), nfs(5),
2899 Digest::SHA256(3), Crypt::Blowfish(3), Compress::Zlib(3)
2903 Copyright (c) 2002-2005 Joseph Huckaby. All Rights Reserved.
2904 This is free software, you may use it and distribute it under the
2905 same terms as Perl itself.