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),
152 filter_store_key => undef,
153 filter_store_value => undef,
154 filter_fetch_key => undef,
155 filter_fetch_value => undef,
164 foreach my $outer_parm ( @outer_params ) {
165 next unless exists $args->{$outer_parm};
166 $self->{$outer_parm} = $args->{$outer_parm}
169 if ( exists $args->{root} ) {
170 $self->{root} = $args->{root};
173 # This is cleanup based on the fact that the $args
174 # coming in is for both the root and non-root items
175 delete $self->root->{$_} for @outer_params;
177 $self->root->{links}++;
179 if (!defined($self->fh)) { $self->open(); }
185 sub _get_self { tied( %{$_[0]} ) || $_[0] }
189 # Tied hash constructor method, called by Perl's tie() function.
193 if (scalar(@_) > 1) { $args = {@_}; }
194 #XXX This use of ref() is bad and is a bug
195 elsif (ref($_[0])) { $args = $_[0]; }
196 else { $args = { file => shift }; }
198 return $class->init($args);
203 # Tied array constructor method, called by Perl's tie() function.
207 if (scalar(@_) > 1) { $args = {@_}; }
208 #XXX This use of ref() is bad and is a bug
209 elsif (ref($_[0])) { $args = $_[0]; }
210 else { $args = { file => shift }; }
212 return $class->init($args);
217 # Class deconstructor. Close file handle if there are no more refs.
219 my $self = _get_self($_[0]);
222 $self->root->{links}--;
224 if (!$self->root->{links}) {
231 # Open a FileHandle to the database, create if nonexistent.
232 # Make sure file signature matches DeepDB spec.
234 my $self = _get_self($_[0]);
236 if (defined($self->fh)) { $self->close(); }
238 if (!(-e $self->root->{file}) && $self->root->{mode} eq 'r+') {
239 my $temp = FileHandle->new( $self->root->{file}, 'w' );
243 #XXX Convert to set_fh()
244 $self->root->{fh} = FileHandle->new( $self->root->{file}, $self->root->{mode} );
245 if (! defined($self->fh)) {
246 return $self->throw_error("Cannot open file: " . $self->root->{file} . ": $!");
249 binmode $self->fh; # for win32
250 if ($self->root->{autoflush}) {
251 $self->fh->autoflush();
255 seek($self->fh, 0, 0);
256 my $bytes_read = $self->fh->read($signature, length(SIG_FILE));
259 # File is empty -- write signature and master index
262 seek($self->fh, 0, 0);
263 $self->fh->print(SIG_FILE);
264 $self->root->{end} = length(SIG_FILE);
265 $self->create_tag($self->base_offset, $self->type, chr(0) x $INDEX_SIZE);
267 my $plain_key = "[base]";
268 $self->fh->print( pack($DATA_LENGTH_PACK, length($plain_key)) . $plain_key );
269 $self->root->{end} += $DATA_LENGTH_SIZE + length($plain_key);
276 # Check signature was valid
278 unless ($signature eq SIG_FILE) {
280 return $self->throw_error("Signature not found -- file is not a Deep DB");
283 $self->root->{end} = (stat($self->fh))[7];
286 # Get our type from master index signature
288 my $tag = $self->load_tag($self->base_offset);
289 #XXX This is a problem - need to verify type, not override it!
290 #XXX We probably also want to store the hash algorithm name, not assume anything
291 #XXX Convert to set_type() when one is written
292 $self->{type} = $tag->{signature};
299 # Close database FileHandle
301 my $self = _get_self($_[0]);
302 undef $self->root->{fh};
307 # Given offset, signature and content, create tag and write to disk
309 my ($self, $offset, $sig, $content) = @_;
310 my $size = length($content);
312 seek($self->fh, $offset, 0);
313 $self->fh->print( $sig . pack($DATA_LENGTH_PACK, $size) . $content );
315 if ($offset == $self->root->{end}) {
316 $self->root->{end} += SIG_SIZE + $DATA_LENGTH_SIZE + $size;
322 offset => $offset + SIG_SIZE + $DATA_LENGTH_SIZE,
329 # Given offset, load single tag and return signature, size and data
334 seek($self->fh, $offset, 0);
335 if ($self->fh->eof()) { return; }
338 $self->fh->read($sig, SIG_SIZE);
341 $self->fh->read($size, $DATA_LENGTH_SIZE);
342 $size = unpack($DATA_LENGTH_PACK, $size);
345 $self->fh->read($buffer, $size);
350 offset => $offset + SIG_SIZE + $DATA_LENGTH_SIZE,
357 # Given index tag, lookup single entry in index and return .
360 my ($tag, $index) = @_;
362 my $location = unpack($LONG_PACK, substr($tag->{content}, $index * $LONG_SIZE, $LONG_SIZE) );
363 if (!$location) { return; }
365 return $self->load_tag( $location );
370 # Adds one key/value pair to bucket list, given offset, MD5 digest of key,
371 # plain (undigested) key and value.
374 my ($tag, $md5, $plain_key, $value) = @_;
375 my $keys = $tag->{content};
379 my $is_dbm_deep = eval { $value->isa( 'DBM::Deep' ) };
380 my $internal_ref = $is_dbm_deep && ($value->root eq $self->root);
383 # Iterate through buckets, seeing if this is a new entry or a replace.
385 for (my $i=0; $i<$MAX_BUCKETS; $i++) {
386 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
387 my $subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
390 # Found empty bucket (end of list). Populate and exit loop.
394 if ($internal_ref) { $location = $value->base_offset; }
395 else { $location = $self->root->{end}; }
397 seek($self->fh, $tag->{offset} + ($i * $BUCKET_SIZE), 0);
398 $self->fh->print( $md5 . pack($LONG_PACK, $location) );
401 elsif ($md5 eq $key) {
403 # Found existing bucket with same key. Replace with new value.
408 $location = $value->base_offset;
409 seek($self->fh, $tag->{offset} + ($i * $BUCKET_SIZE), 0);
410 $self->fh->print( $md5 . pack($LONG_PACK, $location) );
413 seek($self->fh, $subloc + SIG_SIZE, 0);
415 $self->fh->read($size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
418 # If value is a hash, array, or raw value with equal or less size, we can
419 # reuse the same content area of the database. Otherwise, we have to create
420 # a new content area at the EOF.
423 my $r = Scalar::Util::reftype( $value ) || '';
424 if ( $r eq 'HASH' || $r eq 'ARRAY' ) { $actual_length = $INDEX_SIZE; }
425 else { $actual_length = length($value); }
427 if ($actual_length <= $size) {
431 $location = $self->root->{end};
432 seek($self->fh, $tag->{offset} + ($i * $BUCKET_SIZE) + $HASH_SIZE, 0);
433 $self->fh->print( pack($LONG_PACK, $location) );
441 # If this is an internal reference, return now.
442 # No need to write value or plain key
445 if ($internal_ref) { return $result; }
448 # If bucket didn't fit into list, split into a new index level
451 seek($self->fh, $tag->{ref_loc}, 0);
452 $self->fh->print( pack($LONG_PACK, $self->root->{end}) );
454 my $index_tag = $self->create_tag($self->root->{end}, SIG_INDEX, chr(0) x $INDEX_SIZE);
457 #XXX We've already guaranteed that this cannot be true at YYY
458 # if ($internal_ref) {
459 # $keys .= $md5 . pack($LONG_PACK, $value->base_offset);
460 # $location = $value->base_offset;
462 # else { $keys .= $md5 . pack($LONG_PACK, 0); }
463 $keys .= $md5 . pack($LONG_PACK, 0);
465 for (my $i=0; $i<=$MAX_BUCKETS; $i++) {
466 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
468 my $old_subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
469 my $num = ord(substr($key, $tag->{ch} + 1, 1));
471 if ($offsets[$num]) {
472 my $offset = $offsets[$num] + SIG_SIZE + $DATA_LENGTH_SIZE;
473 seek($self->fh, $offset, 0);
475 $self->fh->read($subkeys, $BUCKET_LIST_SIZE);
477 for (my $k=0; $k<$MAX_BUCKETS; $k++) {
478 my $subloc = unpack($LONG_PACK, substr($subkeys, ($k * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
480 seek($self->fh, $offset + ($k * $BUCKET_SIZE), 0);
481 $self->fh->print( $key . pack($LONG_PACK, $old_subloc || $self->root->{end}) );
487 $offsets[$num] = $self->root->{end};
488 seek($self->fh, $index_tag->{offset} + ($num * $LONG_SIZE), 0);
489 $self->fh->print( pack($LONG_PACK, $self->root->{end}) );
491 my $blist_tag = $self->create_tag($self->root->{end}, SIG_BLIST, chr(0) x $BUCKET_LIST_SIZE);
493 seek($self->fh, $blist_tag->{offset}, 0);
494 $self->fh->print( $key . pack($LONG_PACK, $old_subloc || $self->root->{end}) );
499 $location ||= $self->root->{end};
500 } # re-index bucket list
503 # Seek to content area and store signature, value and plaintext key
507 seek($self->fh, $location, 0);
510 # Write signature based on content type, set content length and write actual value.
512 my $r = Scalar::Util::reftype($value) || '';
514 $self->fh->print( TYPE_HASH );
515 $self->fh->print( pack($DATA_LENGTH_PACK, $INDEX_SIZE) . chr(0) x $INDEX_SIZE );
516 $content_length = $INDEX_SIZE;
518 elsif ($r eq 'ARRAY') {
519 $self->fh->print( TYPE_ARRAY );
520 $self->fh->print( pack($DATA_LENGTH_PACK, $INDEX_SIZE) . chr(0) x $INDEX_SIZE );
521 $content_length = $INDEX_SIZE;
523 elsif (!defined($value)) {
524 $self->fh->print( SIG_NULL );
525 $self->fh->print( pack($DATA_LENGTH_PACK, 0) );
529 $self->fh->print( SIG_DATA );
530 $self->fh->print( pack($DATA_LENGTH_PACK, length($value)) . $value );
531 $content_length = length($value);
535 # Plain key is stored AFTER value, as keys are typically fetched less often.
537 $self->fh->print( pack($DATA_LENGTH_PACK, length($plain_key)) . $plain_key );
540 # If value is blessed, preserve class name
542 my $value_class = Scalar::Util::blessed($value);
543 #XXX NO tests for this
544 if ($self->root->{autobless} && defined $value_class) {
545 if ($value_class ne 'DBM::Deep') {
547 # Blessed ref -- will restore later
549 $self->fh->print( chr(1) );
550 $self->fh->print( pack($DATA_LENGTH_PACK, length($value_class)) . $value_class );
551 $content_length += 1;
552 $content_length += $DATA_LENGTH_SIZE + length($value_class);
556 # Simple unblessed ref -- no restore needed
558 $self->fh->print( chr(0) );
559 $content_length += 1;
564 # If this is a new content area, advance EOF counter
566 if ($location == $self->root->{end}) {
567 $self->root->{end} += SIG_SIZE;
568 $self->root->{end} += $DATA_LENGTH_SIZE + $content_length;
569 $self->root->{end} += $DATA_LENGTH_SIZE + length($plain_key);
573 # If content is a hash or array, create new child DeepDB object and
574 # pass each key or element to it.
577 my $branch = DBM::Deep->new(
579 base_offset => $location,
582 foreach my $key (keys %{$value}) {
583 $branch->{$key} = $value->{$key};
586 elsif ($r eq 'ARRAY') {
587 my $branch = DBM::Deep->new(
589 base_offset => $location,
593 foreach my $element (@{$value}) {
594 $branch->[$index] = $element;
602 return $self->throw_error("Fatal error: indexing failed -- possibly due to corruption in file");
605 sub get_bucket_value {
607 # Fetch single value given tag and MD5 digested key.
610 my ($tag, $md5) = @_;
611 my $keys = $tag->{content};
614 # Iterate through buckets, looking for a key match
617 for (my $i=0; $i<$MAX_BUCKETS; $i++) {
618 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
619 my $subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
623 # Hit end of list, no match
628 if ( $md5 ne $key ) {
633 # Found match -- seek to offset and read signature
636 seek($self->fh, $subloc, 0);
637 $self->fh->read($signature, SIG_SIZE);
640 # If value is a hash or array, return new DeepDB object with correct offset
642 if (($signature eq TYPE_HASH) || ($signature eq TYPE_ARRAY)) {
643 my $obj = DBM::Deep->new(
645 base_offset => $subloc,
649 #XXX NO tests for this
650 if ($self->root->{autobless}) {
652 # Skip over value and plain key to see if object needs
655 seek($self->fh, $DATA_LENGTH_SIZE + $INDEX_SIZE, 1);
658 $self->fh->read($size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
659 if ($size) { seek($self->fh, $size, 1); }
662 $self->fh->read($bless_bit, 1);
663 if (ord($bless_bit)) {
665 # Yes, object needs to be re-blessed
668 $self->fh->read($size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
669 if ($size) { $self->fh->read($class_name, $size); }
670 if ($class_name) { $obj = bless( $obj, $class_name ); }
678 # Otherwise return actual value
680 elsif ($signature eq SIG_DATA) {
683 $self->fh->read($size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
684 if ($size) { $self->fh->read($value, $size); }
689 # Key exists, but content is null
699 # Delete single key/value pair given tag and MD5 digested key.
702 my ($tag, $md5) = @_;
703 my $keys = $tag->{content};
706 # Iterate through buckets, looking for a key match
709 for (my $i=0; $i<$MAX_BUCKETS; $i++) {
710 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
711 my $subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
715 # Hit end of list, no match
720 if ( $md5 ne $key ) {
725 # Matched key -- delete bucket and return
727 seek($self->fh, $tag->{offset} + ($i * $BUCKET_SIZE), 0);
728 $self->fh->print( substr($keys, ($i+1) * $BUCKET_SIZE ) );
729 $self->fh->print( chr(0) x $BUCKET_SIZE );
739 # Check existence of single key given tag and MD5 digested key.
742 my ($tag, $md5) = @_;
743 my $keys = $tag->{content};
746 # Iterate through buckets, looking for a key match
749 for (my $i=0; $i<$MAX_BUCKETS; $i++) {
750 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
751 my $subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
755 # Hit end of list, no match
760 if ( $md5 ne $key ) {
765 # Matched key -- return true
773 sub find_bucket_list {
775 # Locate offset for bucket list, given digested key
781 # Locate offset for bucket list using digest index system
784 my $tag = $self->load_tag($self->base_offset);
785 if (!$tag) { return; }
787 while ($tag->{signature} ne SIG_BLIST) {
788 $tag = $self->index_lookup($tag, ord(substr($md5, $ch, 1)));
789 if (!$tag) { return; }
798 # Scan index and recursively step into deeper levels, looking for next key.
800 my ($self, $offset, $ch, $force_return_next) = @_;
801 $force_return_next = undef unless $force_return_next;
803 my $tag = $self->load_tag( $offset );
805 if ($tag->{signature} ne SIG_BLIST) {
806 my $content = $tag->{content};
808 if ($self->{return_next}) { $start = 0; }
809 else { $start = ord(substr($self->{prev_md5}, $ch, 1)); }
811 for (my $index = $start; $index < 256; $index++) {
812 my $subloc = unpack($LONG_PACK, substr($content, $index * $LONG_SIZE, $LONG_SIZE) );
814 my $result = $self->traverse_index( $subloc, $ch + 1, $force_return_next );
815 if (defined($result)) { return $result; }
819 $self->{return_next} = 1;
822 elsif ($tag->{signature} eq SIG_BLIST) {
823 my $keys = $tag->{content};
824 if ($force_return_next) { $self->{return_next} = 1; }
827 # Iterate through buckets, looking for a key match
829 for (my $i=0; $i<$MAX_BUCKETS; $i++) {
830 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
831 my $subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
835 # End of bucket list -- return to outer loop
837 $self->{return_next} = 1;
840 elsif ($key eq $self->{prev_md5}) {
842 # Located previous key -- return next one found
844 $self->{return_next} = 1;
847 elsif ($self->{return_next}) {
849 # Seek to bucket location and skip over signature
851 seek($self->fh, $subloc + SIG_SIZE, 0);
854 # Skip over value to get to plain key
857 $self->fh->read($size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
858 if ($size) { seek($self->fh, $size, 1); }
861 # Read in plain key and return as scalar
864 $self->fh->read($size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
865 if ($size) { $self->fh->read($plain_key, $size); }
871 $self->{return_next} = 1;
872 } # tag is a bucket list
879 # Locate next key, given digested previous one
881 my $self = _get_self($_[0]);
883 $self->{prev_md5} = $_[1] ? $_[1] : undef;
884 $self->{return_next} = 0;
887 # If the previous key was not specifed, start at the top and
888 # return the first one found.
890 if (!$self->{prev_md5}) {
891 $self->{prev_md5} = chr(0) x $HASH_SIZE;
892 $self->{return_next} = 1;
895 return $self->traverse_index( $self->base_offset, 0 );
900 # If db locking is set, flock() the db file. If called multiple
901 # times before unlock(), then the same number of unlocks() must
902 # be called before the lock is released.
904 my $self = _get_self($_[0]);
906 $type = LOCK_EX unless defined $type;
908 if ($self->root->{locking}) {
909 if (!$self->root->{locked}) { flock($self->fh, $type); }
910 $self->root->{locked}++;
916 # If db locking is set, unlock the db file. See note in lock()
917 # regarding calling lock() multiple times.
919 my $self = _get_self($_[0]);
922 if ($self->root->{locking} && $self->root->{locked} > 0) {
923 $self->root->{locked}--;
924 if (!$self->root->{locked}) { flock($self->fh, LOCK_UN); }
928 #XXX These uses of ref() need verified
931 # Copy single level of keys or elements to new DB handle.
932 # Recurse for nested structures
934 my $self = _get_self($_[0]);
937 if ($self->{type} eq TYPE_HASH) {
938 my $key = $self->first_key();
940 my $value = $self->get($key);
941 if (!ref($value)) { $db_temp->{$key} = $value; }
943 my $type = $value->type;
944 if ($type eq TYPE_HASH) { $db_temp->{$key} = {}; }
945 else { $db_temp->{$key} = []; }
946 $value->copy_node( $db_temp->{$key} );
948 $key = $self->next_key($key);
952 my $length = $self->length();
953 for (my $index = 0; $index < $length; $index++) {
954 my $value = $self->get($index);
955 if (!ref($value)) { $db_temp->[$index] = $value; }
957 my $type = $value->type;
958 if ($type eq TYPE_HASH) { $db_temp->[$index] = {}; }
959 else { $db_temp->[$index] = []; }
960 $value->copy_node( $db_temp->[$index] );
968 # Recursively export into standard Perl hashes and arrays.
970 my $self = _get_self($_[0]);
973 if ($self->type eq TYPE_HASH) { $temp = {}; }
974 elsif ($self->type eq TYPE_ARRAY) { $temp = []; }
977 $self->copy_node( $temp );
985 # Recursively import Perl hash/array structure
987 #XXX This use of ref() seems to be ok
988 if (!ref($_[0])) { return; } # Perl calls import() on use -- ignore
990 my $self = _get_self($_[0]);
993 #XXX This use of ref() seems to be ok
996 # struct is not a reference, so just import based on our type
1000 if ($self->type eq TYPE_HASH) { $struct = {@_}; }
1001 elsif ($self->type eq TYPE_ARRAY) { $struct = [@_]; }
1004 my $r = Scalar::Util::reftype($struct) || '';
1005 if ($r eq "HASH" && $self->type eq TYPE_HASH) {
1006 foreach my $key (keys %$struct) { $self->put($key, $struct->{$key}); }
1008 elsif ($r eq "ARRAY" && $self->type eq TYPE_ARRAY) {
1009 $self->push( @$struct );
1012 return $self->throw_error("Cannot import: type mismatch");
1020 # Rebuild entire database into new file, then move
1021 # it back on top of original.
1023 my $self = _get_self($_[0]);
1024 if ($self->root->{links} > 1) {
1025 return $self->throw_error("Cannot optimize: reference count is greater than 1");
1028 my $db_temp = DBM::Deep->new(
1029 file => $self->root->{file} . '.tmp',
1033 return $self->throw_error("Cannot optimize: failed to open temp file: $!");
1037 $self->copy_node( $db_temp );
1041 # Attempt to copy user, group and permissions over to new file
1043 my @stats = stat($self->fh);
1044 my $perms = $stats[2] & 07777;
1045 my $uid = $stats[4];
1046 my $gid = $stats[5];
1047 chown( $uid, $gid, $self->root->{file} . '.tmp' );
1048 chmod( $perms, $self->root->{file} . '.tmp' );
1050 # q.v. perlport for more information on this variable
1051 if ( $^O eq 'MSWin32' ) {
1053 # Potential race condition when optmizing on Win32 with locking.
1054 # The Windows filesystem requires that the filehandle be closed
1055 # before it is overwritten with rename(). This could be redone
1062 if (!rename $self->root->{file} . '.tmp', $self->root->{file}) {
1063 unlink $self->root->{file} . '.tmp';
1065 return $self->throw_error("Optimize failed: Cannot copy temp file over original: $!");
1077 # Make copy of object and return
1079 my $self = _get_self($_[0]);
1081 return DBM::Deep->new(
1082 type => $self->type,
1083 base_offset => $self->base_offset,
1089 my %is_legal_filter = map {
1092 store_key store_value
1093 fetch_key fetch_value
1098 # Setup filter function for storing or fetching the key or value
1100 my $self = _get_self($_[0]);
1101 my $type = lc $_[1];
1102 my $func = $_[2] ? $_[2] : undef;
1104 if ( $is_legal_filter{$type} ) {
1105 $self->root->{"filter_$type"} = $func;
1119 # Get access to the root structure
1121 my $self = _get_self($_[0]);
1122 return $self->{root};
1127 # Get access to the raw FileHandle
1129 my $self = _get_self($_[0]);
1130 return $self->root->{fh};
1135 # Get type of current node (TYPE_HASH or TYPE_ARRAY)
1137 my $self = _get_self($_[0]);
1138 return $self->{type};
1143 # Get base_offset of current node (TYPE_HASH or TYPE_ARRAY)
1145 my $self = _get_self($_[0]);
1146 return $self->{base_offset};
1151 # Get last error string, or undef if no error
1154 ? ( _get_self($_[0])->{root}->{error} or undef )
1164 # Store error string in self
1166 my $self = _get_self($_[0]);
1167 my $error_text = $_[1];
1169 $self->root->{error} = $error_text;
1171 unless ($self->root->{debug}) {
1172 die "DBM::Deep: $error_text\n";
1175 warn "DBM::Deep: $error_text\n";
1183 my $self = _get_self($_[0]);
1185 undef $self->root->{error};
1190 # Precalculate index, bucket and bucket list sizes
1193 #XXX I don't like this ...
1194 set_pack() unless defined $LONG_SIZE;
1196 $INDEX_SIZE = 256 * $LONG_SIZE;
1197 $BUCKET_SIZE = $HASH_SIZE + $LONG_SIZE;
1198 $BUCKET_LIST_SIZE = $MAX_BUCKETS * $BUCKET_SIZE;
1203 # Set pack/unpack modes (see file header for more)
1205 my ($long_s, $long_p, $data_s, $data_p) = @_;
1207 $LONG_SIZE = $long_s ? $long_s : 4;
1208 $LONG_PACK = $long_p ? $long_p : 'N';
1210 $DATA_LENGTH_SIZE = $data_s ? $data_s : 4;
1211 $DATA_LENGTH_PACK = $data_p ? $data_p : 'N';
1218 # Set key digest function (default is MD5)
1220 my ($digest_func, $hash_size) = @_;
1222 $DIGEST_FUNC = $digest_func ? $digest_func : \&Digest::MD5::md5;
1223 $HASH_SIZE = $hash_size ? $hash_size : 16;
1229 # tie() methods (hashes and arrays)
1234 # Store single hash key/value or array element in database.
1236 my $self = _get_self($_[0]);
1237 my $key = ($self->root->{filter_store_key} && $self->type eq TYPE_HASH) ? $self->root->{filter_store_key}->($_[1]) : $_[1];
1238 #XXX What is ref() checking here?
1239 my $value = ($self->root->{filter_store_value} && !ref($_[2])) ? $self->root->{filter_store_value}->($_[2]) : $_[2];
1241 my $unpacked_key = $key;
1242 if (($self->type eq TYPE_ARRAY) && ($key =~ /^\d+$/)) { $key = pack($LONG_PACK, $key); }
1243 my $md5 = $DIGEST_FUNC->($key);
1246 # Make sure file is open
1248 if (!defined($self->fh) && !$self->open()) {
1253 # Request exclusive lock for writing
1255 $self->lock( LOCK_EX );
1258 # If locking is enabled, set 'end' parameter again, in case another
1259 # DB instance appended to our file while we were unlocked.
1261 if ($self->root->{locking} || $self->root->{volatile}) {
1262 $self->root->{end} = (stat($self->fh))[7];
1266 # Locate offset for bucket list using digest index system
1268 my $tag = $self->load_tag($self->base_offset);
1270 $tag = $self->create_tag($self->base_offset, SIG_INDEX, chr(0) x $INDEX_SIZE);
1274 while ($tag->{signature} ne SIG_BLIST) {
1275 my $num = ord(substr($md5, $ch, 1));
1276 my $new_tag = $self->index_lookup($tag, $num);
1278 my $ref_loc = $tag->{offset} + ($num * $LONG_SIZE);
1279 seek($self->fh, $ref_loc, 0);
1280 $self->fh->print( pack($LONG_PACK, $self->root->{end}) );
1282 $tag = $self->create_tag($self->root->{end}, SIG_BLIST, chr(0) x $BUCKET_LIST_SIZE);
1283 $tag->{ref_loc} = $ref_loc;
1288 my $ref_loc = $tag->{offset} + ($num * $LONG_SIZE);
1290 $tag->{ref_loc} = $ref_loc;
1297 # Add key/value to bucket list
1299 my $result = $self->add_bucket( $tag, $md5, $key, $value );
1302 # If this object is an array, and bucket was not a replace, and key is numerical,
1303 # and index is equal or greater than current length, advance length variable.
1305 if (($result == 2) && ($self->type eq TYPE_ARRAY) && ($unpacked_key =~ /^\d+$/) && ($unpacked_key >= $self->FETCHSIZE())) {
1306 $self->STORESIZE( $unpacked_key + 1 );
1316 # Fetch single value or element given plain key or array index
1318 my $self = _get_self($_[0]);
1321 if ( $self->type eq TYPE_HASH ) {
1322 if ( my $filter = $self->root->{filter_store_key} ) {
1323 $key = $filter->( $key );
1326 elsif ( $self->type eq TYPE_ARRAY ) {
1327 if ( $key =~ /^\d+$/ ) {
1328 $key = pack($LONG_PACK, $key);
1332 my $md5 = $DIGEST_FUNC->($key);
1335 # Make sure file is open
1337 if (!defined($self->fh)) { $self->open(); }
1340 # Request shared lock for reading
1342 $self->lock( LOCK_SH );
1344 my $tag = $self->find_bucket_list( $md5 );
1351 # Get value from bucket list
1353 my $result = $self->get_bucket_value( $tag, $md5 );
1357 #XXX What is ref() checking here?
1358 return ($result && !ref($result) && $self->root->{filter_fetch_value}) ? $self->root->{filter_fetch_value}->($result) : $result;
1363 # Delete single key/value pair or element given plain key or array index
1365 my $self = _get_self($_[0]);
1366 my $key = ($self->root->{filter_store_key} && $self->type eq TYPE_HASH) ? $self->root->{filter_store_key}->($_[1]) : $_[1];
1368 my $unpacked_key = $key;
1369 if (($self->type eq TYPE_ARRAY) && ($key =~ /^\d+$/)) { $key = pack($LONG_PACK, $key); }
1370 my $md5 = $DIGEST_FUNC->($key);
1373 # Make sure file is open
1375 if (!defined($self->fh)) { $self->open(); }
1378 # Request exclusive lock for writing
1380 $self->lock( LOCK_EX );
1382 my $tag = $self->find_bucket_list( $md5 );
1391 my $result = $self->delete_bucket( $tag, $md5 );
1394 # If this object is an array and the key deleted was on the end of the stack,
1395 # decrement the length variable.
1397 if ($result && ($self->type eq TYPE_ARRAY) && ($unpacked_key == $self->FETCHSIZE() - 1)) {
1398 $self->STORESIZE( $unpacked_key );
1408 # Check if a single key or element exists given plain key or array index
1410 my $self = _get_self($_[0]);
1411 my $key = ($self->root->{filter_store_key} && $self->type eq TYPE_HASH) ? $self->root->{filter_store_key}->($_[1]) : $_[1];
1413 if (($self->type eq TYPE_ARRAY) && ($key =~ /^\d+$/)) { $key = pack($LONG_PACK, $key); }
1414 my $md5 = $DIGEST_FUNC->($key);
1417 # Make sure file is open
1419 if (!defined($self->fh)) { $self->open(); }
1422 # Request shared lock for reading
1424 $self->lock( LOCK_SH );
1426 my $tag = $self->find_bucket_list( $md5 );
1429 # For some reason, the built-in exists() function returns '' for false
1437 # Check if bucket exists and return 1 or ''
1439 my $result = $self->bucket_exists( $tag, $md5 ) || '';
1448 # Clear all keys from hash, or all elements from array.
1450 my $self = _get_self($_[0]);
1453 # Make sure file is open
1455 if (!defined($self->fh)) { $self->open(); }
1458 # Request exclusive lock for writing
1460 $self->lock( LOCK_EX );
1462 seek($self->fh, $self->base_offset, 0);
1463 if ($self->fh->eof()) {
1468 $self->create_tag($self->base_offset, $self->type, chr(0) x $INDEX_SIZE);
1477 # Locate and return first key (in no particular order)
1479 my $self = _get_self($_[0]);
1480 if ($self->type ne TYPE_HASH) {
1481 return $self->throw_error("FIRSTKEY method only supported for hashes");
1485 # Make sure file is open
1487 if (!defined($self->fh)) { $self->open(); }
1490 # Request shared lock for reading
1492 $self->lock( LOCK_SH );
1494 my $result = $self->get_next_key();
1498 return ($result && $self->root->{filter_fetch_key}) ? $self->root->{filter_fetch_key}->($result) : $result;
1503 # Return next key (in no particular order), given previous one
1505 my $self = _get_self($_[0]);
1506 if ($self->type ne TYPE_HASH) {
1507 return $self->throw_error("NEXTKEY method only supported for hashes");
1509 my $prev_key = ($self->root->{filter_store_key} && $self->type eq TYPE_HASH) ? $self->root->{filter_store_key}->($_[1]) : $_[1];
1510 my $prev_md5 = $DIGEST_FUNC->($prev_key);
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( $prev_md5 );
1526 return ($result && $self->root->{filter_fetch_key}) ? $self->root->{filter_fetch_key}->($result) : $result;
1530 # The following methods are for arrays only
1535 # Return the length of the array
1537 my $self = _get_self($_[0]);
1538 if ($self->type ne TYPE_ARRAY) {
1539 return $self->throw_error("FETCHSIZE method only supported for arrays");
1542 my $SAVE_FILTER = $self->root->{filter_fetch_value};
1543 $self->root->{filter_fetch_value} = undef;
1545 my $packed_size = $self->FETCH('length');
1547 $self->root->{filter_fetch_value} = $SAVE_FILTER;
1549 if ($packed_size) { return int(unpack($LONG_PACK, $packed_size)); }
1555 # Set the length of the array
1557 my $self = _get_self($_[0]);
1558 if ($self->type ne TYPE_ARRAY) {
1559 return $self->throw_error("STORESIZE method only supported for arrays");
1561 my $new_length = $_[1];
1563 my $SAVE_FILTER = $self->root->{filter_store_value};
1564 $self->root->{filter_store_value} = undef;
1566 my $result = $self->STORE('length', pack($LONG_PACK, $new_length));
1568 $self->root->{filter_store_value} = $SAVE_FILTER;
1575 # Remove and return the last element on the array
1577 my $self = _get_self($_[0]);
1578 if ($self->type ne TYPE_ARRAY) {
1579 return $self->throw_error("POP method only supported for arrays");
1581 my $length = $self->FETCHSIZE();
1584 my $content = $self->FETCH( $length - 1 );
1585 $self->DELETE( $length - 1 );
1595 # Add new element(s) to the end of the array
1597 my $self = _get_self(shift);
1598 if ($self->type ne TYPE_ARRAY) {
1599 return $self->throw_error("PUSH method only supported for arrays");
1601 my $length = $self->FETCHSIZE();
1603 while (my $content = shift @_) {
1604 $self->STORE( $length, $content );
1611 # Remove and return first element on the array.
1612 # Shift over remaining elements to take up space.
1614 my $self = _get_self($_[0]);
1615 if ($self->type ne TYPE_ARRAY) {
1616 return $self->throw_error("SHIFT method only supported for arrays");
1618 my $length = $self->FETCHSIZE();
1621 my $content = $self->FETCH( 0 );
1624 # Shift elements over and remove last one.
1626 for (my $i = 0; $i < $length - 1; $i++) {
1627 $self->STORE( $i, $self->FETCH($i + 1) );
1629 $self->DELETE( $length - 1 );
1640 # Insert new element(s) at beginning of array.
1641 # Shift over other elements to make space.
1643 my $self = _get_self($_[0]);shift @_;
1644 if ($self->type ne TYPE_ARRAY) {
1645 return $self->throw_error("UNSHIFT method only supported for arrays");
1647 my @new_elements = @_;
1648 my $length = $self->FETCHSIZE();
1649 my $new_size = scalar @new_elements;
1652 for (my $i = $length - 1; $i >= 0; $i--) {
1653 $self->STORE( $i + $new_size, $self->FETCH($i) );
1657 for (my $i = 0; $i < $new_size; $i++) {
1658 $self->STORE( $i, $new_elements[$i] );
1664 # Splices section of array with optional new section.
1665 # Returns deleted section, or last element deleted in scalar context.
1667 my $self = _get_self($_[0]);shift @_;
1668 if ($self->type ne TYPE_ARRAY) {
1669 return $self->throw_error("SPLICE method only supported for arrays");
1671 my $length = $self->FETCHSIZE();
1674 # Calculate offset and length of splice
1676 my $offset = shift || 0;
1677 if ($offset < 0) { $offset += $length; }
1680 if (scalar @_) { $splice_length = shift; }
1681 else { $splice_length = $length - $offset; }
1682 if ($splice_length < 0) { $splice_length += ($length - $offset); }
1685 # Setup array with new elements, and copy out old elements for return
1687 my @new_elements = @_;
1688 my $new_size = scalar @new_elements;
1690 my @old_elements = ();
1691 for (my $i = $offset; $i < $offset + $splice_length; $i++) {
1692 push @old_elements, $self->FETCH( $i );
1696 # Adjust array length, and shift elements to accomodate new section.
1698 if ( $new_size != $splice_length ) {
1699 if ($new_size > $splice_length) {
1700 for (my $i = $length - 1; $i >= $offset + $splice_length; $i--) {
1701 $self->STORE( $i + ($new_size - $splice_length), $self->FETCH($i) );
1705 for (my $i = $offset + $splice_length; $i < $length; $i++) {
1706 $self->STORE( $i + ($new_size - $splice_length), $self->FETCH($i) );
1708 for (my $i = 0; $i < $splice_length - $new_size; $i++) {
1709 $self->DELETE( $length - 1 );
1716 # Insert new elements into array
1718 for (my $i = $offset; $i < $offset + $new_size; $i++) {
1719 $self->STORE( $i, shift @new_elements );
1723 # Return deleted section, or last element in scalar context.
1725 return wantarray ? @old_elements : $old_elements[-1];
1728 #XXX We don't need to define it.
1731 # Perl will call EXTEND() when the array is likely to grow.
1732 # We don't care, but include it for compatibility.
1737 # Public method aliases
1739 *put = *store = *STORE;
1740 *get = *fetch = *FETCH;
1744 *first_key = *FIRSTKEY;
1745 *next_key = *NEXTKEY;
1746 *length = *FETCHSIZE;
1750 *unshift = *UNSHIFT;
1759 DBM::Deep - A pure perl multi-level hash/array DBM
1764 my $db = DBM::Deep->new( "foo.db" );
1766 $db->{key} = 'value'; # tie() style
1769 $db->put('key', 'value'); # OO style
1770 print $db->get('key');
1772 # true multi-level support
1773 $db->{my_complex} = [
1774 'hello', { perl => 'rules' },
1779 A unique flat-file database module, written in pure perl. True
1780 multi-level hash/array support (unlike MLDBM, which is faked), hybrid
1781 OO / tie() interface, cross-platform FTPable files, and quite fast. Can
1782 handle millions of keys and unlimited hash levels without significant
1783 slow-down. Written from the ground-up in pure perl -- this is NOT a
1784 wrapper around a C-based DBM. Out-of-the-box compatibility with Unix,
1785 Mac OS X and Windows.
1789 Hopefully you are using CPAN's excellent Perl module, which will download
1790 and install the module for you. If not, get the tarball, and run these
1802 Construction can be done OO-style (which is the recommended way), or using
1803 Perl's tie() function. Both are examined here.
1805 =head2 OO CONSTRUCTION
1807 The recommended way to construct a DBM::Deep object is to use the new()
1808 method, which gets you a blessed, tied hash or array reference.
1810 my $db = DBM::Deep->new( "foo.db" );
1812 This opens a new database handle, mapped to the file "foo.db". If this
1813 file does not exist, it will automatically be created. DB files are
1814 opened in "r+" (read/write) mode, and the type of object returned is a
1815 hash, unless otherwise specified (see L<OPTIONS> below).
1819 You can pass a number of options to the constructor to specify things like
1820 locking, autoflush, etc. This is done by passing an inline hash:
1822 my $db = DBM::Deep->new(
1828 Notice that the filename is now specified I<inside> the hash with
1829 the "file" parameter, as opposed to being the sole argument to the
1830 constructor. This is required if any options are specified.
1831 See L<OPTIONS> below for the complete list.
1835 You can also start with an array instead of a hash. For this, you must
1836 specify the C<type> parameter:
1838 my $db = DBM::Deep->new(
1840 type => DBM::Deep->TYPE_ARRAY
1843 B<Note:> Specifing the C<type> parameter only takes effect when beginning
1844 a new DB file. If you create a DBM::Deep object with an existing file, the
1845 C<type> will be loaded from the file header, and ignored if it is passed
1848 =head2 TIE CONSTRUCTION
1850 Alternatively, you can create a DBM::Deep handle by using Perl's built-in
1851 tie() function. This is not ideal, because you get only a basic, tied hash
1852 (or array) which is not blessed, so you can't call any functions on it.
1855 tie %hash, "DBM::Deep", "foo.db";
1858 tie @array, "DBM::Deep", "bar.db";
1860 As with the OO constructor, you can replace the DB filename parameter with
1861 a hash containing one or more options (see L<OPTIONS> just below for the
1864 tie %hash, "DBM::Deep", {
1872 There are a number of options that can be passed in when constructing your
1873 DBM::Deep objects. These apply to both the OO- and tie- based approaches.
1879 Filename of the DB file to link the handle to. You can pass a full absolute
1880 filesystem path, partial path, or a plain filename if the file is in the
1881 current working directory. This is a required parameter.
1885 File open mode (read-only, read-write, etc.) string passed to Perl's FileHandle
1886 module. This is an optional parameter, and defaults to "r+" (read/write).
1887 B<Note:> If the default (r+) mode is selected, the file will also be auto-
1888 created if it doesn't exist.
1892 This parameter specifies what type of object to create, a hash or array. Use
1893 one of these two constants: C<DBM::Deep-E<gt>TYPE_HASH> or C<DBM::Deep-E<gt>TYPE_ARRAY>.
1894 This only takes effect when beginning a new file. This is an optional
1895 parameter, and defaults to C<DBM::Deep-E<gt>TYPE_HASH>.
1899 Specifies whether locking is to be enabled. DBM::Deep uses Perl's Fnctl flock()
1900 function to lock the database in exclusive mode for writes, and shared mode for
1901 reads. Pass any true value to enable. This affects the base DB handle I<and
1902 any child hashes or arrays> that use the same DB file. This is an optional
1903 parameter, and defaults to 0 (disabled). See L<LOCKING> below for more.
1907 Specifies whether autoflush is to be enabled on the underlying FileHandle.
1908 This obviously slows down write operations, but is required if you may have
1909 multiple processes accessing the same DB file (also consider enable I<locking>
1910 or at least I<volatile>). Pass any true value to enable. This is an optional
1911 parameter, and defaults to 0 (disabled).
1915 If I<volatile> mode is enabled, DBM::Deep will stat() the DB file before each
1916 STORE() operation. This is required if an outside force may change the size of
1917 the file between transactions. Locking also implicitly enables volatile. This
1918 is useful if you want to use a different locking system or write your own. Pass
1919 any true value to enable. This is an optional parameter, and defaults to 0
1924 If I<autobless> mode is enabled, DBM::Deep will preserve blessed hashes, and
1925 restore them when fetched. This is an B<experimental> feature, and does have
1926 side-effects. Basically, when hashes are re-blessed into their original
1927 classes, they are no longer blessed into the DBM::Deep class! So you won't be
1928 able to call any DBM::Deep methods on them. You have been warned.
1929 This is an optional parameter, and defaults to 0 (disabled).
1933 See L<FILTERS> below.
1937 Setting I<debug> mode will make all errors non-fatal, dump them out to
1938 STDERR, and continue on. This is for debugging purposes only, and probably
1939 not what you want. This is an optional parameter, and defaults to 0 (disabled).
1943 Instead of passing a file path, you can instead pass a handle to an pre-opened
1944 filehandle. Note: Beware of using the magick *DATA handle, as this actually
1945 contains your entire Perl script, as well as the data following the __DATA__
1946 marker. This will not work, because DBM::Deep uses absolute seek()s into the
1947 file. Instead, consider reading *DATA into an IO::Scalar handle, then passing
1952 =head1 TIE INTERFACE
1954 With DBM::Deep you can access your databases using Perl's standard hash/array
1955 syntax. Because all Deep objects are I<tied> to hashes or arrays, you can treat
1956 them as such. Deep will intercept all reads/writes and direct them to the right
1957 place -- the DB file. This has nothing to do with the L<TIE CONSTRUCTION>
1958 section above. This simply tells you how to use DBM::Deep using regular hashes
1959 and arrays, rather than calling functions like C<get()> and C<put()> (although those
1960 work too). It is entirely up to you how to want to access your databases.
1964 You can treat any DBM::Deep object like a normal Perl hash reference. Add keys,
1965 or even nested hashes (or arrays) using standard Perl syntax:
1967 my $db = DBM::Deep->new( "foo.db" );
1969 $db->{mykey} = "myvalue";
1971 $db->{myhash}->{subkey} = "subvalue";
1973 print $db->{myhash}->{subkey} . "\n";
1975 You can even step through hash keys using the normal Perl C<keys()> function:
1977 foreach my $key (keys %$db) {
1978 print "$key: " . $db->{$key} . "\n";
1981 Remember that Perl's C<keys()> function extracts I<every> key from the hash and
1982 pushes them onto an array, all before the loop even begins. If you have an
1983 extra large hash, this may exhaust Perl's memory. Instead, consider using
1984 Perl's C<each()> function, which pulls keys/values one at a time, using very
1987 while (my ($key, $value) = each %$db) {
1988 print "$key: $value\n";
1991 Please note that when using C<each()>, you should always pass a direct
1992 hash reference, not a lookup. Meaning, you should B<never> do this:
1995 while (my ($key, $value) = each %{$db->{foo}}) { # BAD
1997 This causes an infinite loop, because for each iteration, Perl is calling
1998 FETCH() on the $db handle, resulting in a "new" hash for foo every time, so
1999 it effectively keeps returning the first key over and over again. Instead,
2000 assign a temporary variable to C<$db->{foo}>, then pass that to each().
2004 As with hashes, you can treat any DBM::Deep object like a normal Perl array
2005 reference. This includes inserting, removing and manipulating elements,
2006 and the C<push()>, C<pop()>, C<shift()>, C<unshift()> and C<splice()> functions.
2007 The object must have first been created using type C<DBM::Deep-E<gt>TYPE_ARRAY>,
2008 or simply be a nested array reference inside a hash. Example:
2010 my $db = DBM::Deep->new(
2011 file => "foo-array.db",
2012 type => DBM::Deep->TYPE_ARRAY
2016 push @$db, "bar", "baz";
2017 unshift @$db, "bah";
2019 my $last_elem = pop @$db; # baz
2020 my $first_elem = shift @$db; # bah
2021 my $second_elem = $db->[1]; # bar
2023 my $num_elements = scalar @$db;
2027 In addition to the I<tie()> interface, you can also use a standard OO interface
2028 to manipulate all aspects of DBM::Deep databases. Each type of object (hash or
2029 array) has its own methods, but both types share the following common methods:
2030 C<put()>, C<get()>, C<exists()>, C<delete()> and C<clear()>.
2036 Stores a new hash key/value pair, or sets an array element value. Takes two
2037 arguments, the hash key or array index, and the new value. The value can be
2038 a scalar, hash ref or array ref. Returns true on success, false on failure.
2040 $db->put("foo", "bar"); # for hashes
2041 $db->put(1, "bar"); # for arrays
2045 Fetches the value of a hash key or array element. Takes one argument: the hash
2046 key or array index. Returns a scalar, hash ref or array ref, depending on the
2049 my $value = $db->get("foo"); # for hashes
2050 my $value = $db->get(1); # for arrays
2054 Checks if a hash key or array index exists. Takes one argument: the hash key
2055 or array index. Returns true if it exists, false if not.
2057 if ($db->exists("foo")) { print "yay!\n"; } # for hashes
2058 if ($db->exists(1)) { print "yay!\n"; } # for arrays
2062 Deletes one hash key/value pair or array element. Takes one argument: the hash
2063 key or array index. Returns true on success, false if not found. For arrays,
2064 the remaining elements located after the deleted element are NOT moved over.
2065 The deleted element is essentially just undefined, which is exactly how Perl's
2066 internal arrays work. Please note that the space occupied by the deleted
2067 key/value or element is B<not> reused again -- see L<UNUSED SPACE RECOVERY>
2068 below for details and workarounds.
2070 $db->delete("foo"); # for hashes
2071 $db->delete(1); # for arrays
2075 Deletes B<all> hash keys or array elements. Takes no arguments. No return
2076 value. Please note that the space occupied by the deleted keys/values or
2077 elements is B<not> reused again -- see L<UNUSED SPACE RECOVERY> below for
2078 details and workarounds.
2080 $db->clear(); # hashes or arrays
2086 For hashes, DBM::Deep supports all the common methods described above, and the
2087 following additional methods: C<first_key()> and C<next_key()>.
2093 Returns the "first" key in the hash. As with built-in Perl hashes, keys are
2094 fetched in an undefined order (which appears random). Takes no arguments,
2095 returns the key as a scalar value.
2097 my $key = $db->first_key();
2101 Returns the "next" key in the hash, given the previous one as the sole argument.
2102 Returns undef if there are no more keys to be fetched.
2104 $key = $db->next_key($key);
2108 Here are some examples of using hashes:
2110 my $db = DBM::Deep->new( "foo.db" );
2112 $db->put("foo", "bar");
2113 print "foo: " . $db->get("foo") . "\n";
2115 $db->put("baz", {}); # new child hash ref
2116 $db->get("baz")->put("buz", "biz");
2117 print "buz: " . $db->get("baz")->get("buz") . "\n";
2119 my $key = $db->first_key();
2121 print "$key: " . $db->get($key) . "\n";
2122 $key = $db->next_key($key);
2125 if ($db->exists("foo")) { $db->delete("foo"); }
2129 For arrays, DBM::Deep supports all the common methods described above, and the
2130 following additional methods: C<length()>, C<push()>, C<pop()>, C<shift()>,
2131 C<unshift()> and C<splice()>.
2137 Returns the number of elements in the array. Takes no arguments.
2139 my $len = $db->length();
2143 Adds one or more elements onto the end of the array. Accepts scalars, hash
2144 refs or array refs. No return value.
2146 $db->push("foo", "bar", {});
2150 Fetches the last element in the array, and deletes it. Takes no arguments.
2151 Returns undef if array is empty. Returns the element value.
2153 my $elem = $db->pop();
2157 Fetches the first element in the array, deletes it, then shifts all the
2158 remaining elements over to take up the space. Returns the element value. This
2159 method is not recommended with large arrays -- see L<LARGE ARRAYS> below for
2162 my $elem = $db->shift();
2166 Inserts one or more elements onto the beginning of the array, shifting all
2167 existing elements over to make room. Accepts scalars, hash refs or array refs.
2168 No return value. This method is not recommended with large arrays -- see
2169 <LARGE ARRAYS> below for details.
2171 $db->unshift("foo", "bar", {});
2175 Performs exactly like Perl's built-in function of the same name. See L<perldoc
2176 -f splice> for usage -- it is too complicated to document here. This method is
2177 not recommended with large arrays -- see L<LARGE ARRAYS> below for details.
2181 Here are some examples of using arrays:
2183 my $db = DBM::Deep->new(
2185 type => DBM::Deep->TYPE_ARRAY
2188 $db->push("bar", "baz");
2189 $db->unshift("foo");
2192 my $len = $db->length();
2193 print "length: $len\n"; # 4
2195 for (my $k=0; $k<$len; $k++) {
2196 print "$k: " . $db->get($k) . "\n";
2199 $db->splice(1, 2, "biz", "baf");
2201 while (my $elem = shift @$db) {
2202 print "shifted: $elem\n";
2207 Enable automatic file locking by passing a true value to the C<locking>
2208 parameter when constructing your DBM::Deep object (see L<SETUP> above).
2210 my $db = DBM::Deep->new(
2215 This causes Deep to C<flock()> the underlying FileHandle object with exclusive
2216 mode for writes, and shared mode for reads. This is required if you have
2217 multiple processes accessing the same database file, to avoid file corruption.
2218 Please note that C<flock()> does NOT work for files over NFS. See L<DB OVER
2219 NFS> below for more.
2221 =head2 EXPLICIT LOCKING
2223 You can explicitly lock a database, so it remains locked for multiple
2224 transactions. This is done by calling the C<lock()> method, and passing an
2225 optional lock mode argument (defaults to exclusive mode). This is particularly
2226 useful for things like counters, where the current value needs to be fetched,
2227 then incremented, then stored again.
2230 my $counter = $db->get("counter");
2232 $db->put("counter", $counter);
2241 You can pass C<lock()> an optional argument, which specifies which mode to use
2242 (exclusive or shared). Use one of these two constants: C<DBM::Deep-E<gt>LOCK_EX>
2243 or C<DBM::Deep-E<gt>LOCK_SH>. These are passed directly to C<flock()>, and are the
2244 same as the constants defined in Perl's C<Fcntl> module.
2246 $db->lock( DBM::Deep->LOCK_SH );
2250 If you want to implement your own file locking scheme, be sure to create your
2251 DBM::Deep objects setting the C<volatile> option to true. This hints to Deep
2252 that the DB file may change between transactions. See L<LOW-LEVEL ACCESS>
2255 =head1 IMPORTING/EXPORTING
2257 You can import existing complex structures by calling the C<import()> method,
2258 and export an entire database into an in-memory structure using the C<export()>
2259 method. Both are examined here.
2263 Say you have an existing hash with nested hashes/arrays inside it. Instead of
2264 walking the structure and adding keys/elements to the database as you go,
2265 simply pass a reference to the C<import()> method. This recursively adds
2266 everything to an existing DBM::Deep object for you. Here is an example:
2271 array1 => [ "elem0", "elem1", "elem2" ],
2273 subkey1 => "subvalue1",
2274 subkey2 => "subvalue2"
2278 my $db = DBM::Deep->new( "foo.db" );
2279 $db->import( $struct );
2281 print $db->{key1} . "\n"; # prints "value1"
2283 This recursively imports the entire C<$struct> object into C<$db>, including
2284 all nested hashes and arrays. If the DBM::Deep object contains exsiting data,
2285 keys are merged with the existing ones, replacing if they already exist.
2286 The C<import()> method can be called on any database level (not just the base
2287 level), and works with both hash and array DB types.
2291 B<Note:> Make sure your existing structure has no circular references in it.
2292 These will cause an infinite loop when importing.
2296 Calling the C<export()> method on an existing DBM::Deep object will return
2297 a reference to a new in-memory copy of the database. The export is done
2298 recursively, so all nested hashes/arrays are all exported to standard Perl
2299 objects. Here is an example:
2301 my $db = DBM::Deep->new( "foo.db" );
2303 $db->{key1} = "value1";
2304 $db->{key2} = "value2";
2306 $db->{hash1}->{subkey1} = "subvalue1";
2307 $db->{hash1}->{subkey2} = "subvalue2";
2309 my $struct = $db->export();
2311 print $struct->{key1} . "\n"; # prints "value1"
2313 This makes a complete copy of the database in memory, and returns a reference
2314 to it. The C<export()> method can be called on any database level (not just
2315 the base level), and works with both hash and array DB types. Be careful of
2316 large databases -- you can store a lot more data in a DBM::Deep object than an
2317 in-memory Perl structure.
2321 B<Note:> Make sure your database has no circular references in it.
2322 These will cause an infinite loop when exporting.
2326 DBM::Deep has a number of hooks where you can specify your own Perl function
2327 to perform filtering on incoming or outgoing data. This is a perfect
2328 way to extend the engine, and implement things like real-time compression or
2329 encryption. Filtering applies to the base DB level, and all child hashes /
2330 arrays. Filter hooks can be specified when your DBM::Deep object is first
2331 constructed, or by calling the C<set_filter()> method at any time. There are
2332 four available filter hooks, described below:
2336 =item * filter_store_key
2338 This filter is called whenever a hash key is stored. It
2339 is passed the incoming key, and expected to return a transformed key.
2341 =item * filter_store_value
2343 This filter is called whenever a hash key or array element is stored. It
2344 is passed the incoming value, and expected to return a transformed value.
2346 =item * filter_fetch_key
2348 This filter is called whenever a hash key is fetched (i.e. via
2349 C<first_key()> or C<next_key()>). It is passed the transformed key,
2350 and expected to return the plain key.
2352 =item * filter_fetch_value
2354 This filter is called whenever a hash key or array element is fetched.
2355 It is passed the transformed value, and expected to return the plain value.
2359 Here are the two ways to setup a filter hook:
2361 my $db = DBM::Deep->new(
2363 filter_store_value => \&my_filter_store,
2364 filter_fetch_value => \&my_filter_fetch
2369 $db->set_filter( "filter_store_value", \&my_filter_store );
2370 $db->set_filter( "filter_fetch_value", \&my_filter_fetch );
2372 Your filter function will be called only when dealing with SCALAR keys or
2373 values. When nested hashes and arrays are being stored/fetched, filtering
2374 is bypassed. Filters are called as static functions, passed a single SCALAR
2375 argument, and expected to return a single SCALAR value. If you want to
2376 remove a filter, set the function reference to C<undef>:
2378 $db->set_filter( "filter_store_value", undef );
2380 =head2 REAL-TIME ENCRYPTION EXAMPLE
2382 Here is a working example that uses the I<Crypt::Blowfish> module to
2383 do real-time encryption / decryption of keys & values with DBM::Deep Filters.
2384 Please visit L<http://search.cpan.org/search?module=Crypt::Blowfish> for more
2385 on I<Crypt::Blowfish>. You'll also need the I<Crypt::CBC> module.
2388 use Crypt::Blowfish;
2391 my $cipher = Crypt::CBC->new({
2392 'key' => 'my secret key',
2393 'cipher' => 'Blowfish',
2395 'regenerate_key' => 0,
2396 'padding' => 'space',
2400 my $db = DBM::Deep->new(
2401 file => "foo-encrypt.db",
2402 filter_store_key => \&my_encrypt,
2403 filter_store_value => \&my_encrypt,
2404 filter_fetch_key => \&my_decrypt,
2405 filter_fetch_value => \&my_decrypt,
2408 $db->{key1} = "value1";
2409 $db->{key2} = "value2";
2410 print "key1: " . $db->{key1} . "\n";
2411 print "key2: " . $db->{key2} . "\n";
2417 return $cipher->encrypt( $_[0] );
2420 return $cipher->decrypt( $_[0] );
2423 =head2 REAL-TIME COMPRESSION EXAMPLE
2425 Here is a working example that uses the I<Compress::Zlib> module to do real-time
2426 compression / decompression of keys & values with DBM::Deep Filters.
2427 Please visit L<http://search.cpan.org/search?module=Compress::Zlib> for
2428 more on I<Compress::Zlib>.
2433 my $db = DBM::Deep->new(
2434 file => "foo-compress.db",
2435 filter_store_key => \&my_compress,
2436 filter_store_value => \&my_compress,
2437 filter_fetch_key => \&my_decompress,
2438 filter_fetch_value => \&my_decompress,
2441 $db->{key1} = "value1";
2442 $db->{key2} = "value2";
2443 print "key1: " . $db->{key1} . "\n";
2444 print "key2: " . $db->{key2} . "\n";
2450 return Compress::Zlib::memGzip( $_[0] ) ;
2453 return Compress::Zlib::memGunzip( $_[0] ) ;
2456 B<Note:> Filtering of keys only applies to hashes. Array "keys" are
2457 actually numerical index numbers, and are not filtered.
2459 =head1 ERROR HANDLING
2461 Most DBM::Deep methods return a true value for success, and call die() on
2462 failure. You can wrap calls in an eval block to catch the die. Also, the
2463 actual error message is stored in an internal scalar, which can be fetched by
2464 calling the C<error()> method.
2466 my $db = DBM::Deep->new( "foo.db" ); # create hash
2467 eval { $db->push("foo"); }; # ILLEGAL -- push is array-only call
2469 print $db->error(); # prints error message
2471 You can then call C<clear_error()> to clear the current error state.
2475 If you set the C<debug> option to true when creating your DBM::Deep object,
2476 all errors are considered NON-FATAL, and dumped to STDERR. This is only
2477 for debugging purposes.
2479 =head1 LARGEFILE SUPPORT
2481 If you have a 64-bit system, and your Perl is compiled with both LARGEFILE
2482 and 64-bit support, you I<may> be able to create databases larger than 2 GB.
2483 DBM::Deep by default uses 32-bit file offset tags, but these can be changed
2484 by calling the static C<set_pack()> method before you do anything else.
2486 DBM::Deep::set_pack(8, 'Q');
2488 This tells DBM::Deep to pack all file offsets with 8-byte (64-bit) quad words
2489 instead of 32-bit longs. After setting these values your DB files have a
2490 theoretical maximum size of 16 XB (exabytes).
2494 B<Note:> Changing these values will B<NOT> work for existing database files.
2495 Only change this for new files, and make sure it stays set consistently
2496 throughout the file's life. If you do set these values, you can no longer
2497 access 32-bit DB files. You can, however, call C<set_pack(4, 'N')> to change
2498 back to 32-bit mode.
2502 B<Note:> I have not personally tested files > 2 GB -- all my systems have
2503 only a 32-bit Perl. However, I have received user reports that this does
2506 =head1 LOW-LEVEL ACCESS
2508 If you require low-level access to the underlying FileHandle that Deep uses,
2509 you can call the C<fh()> method, which returns the handle:
2513 This method can be called on the root level of the datbase, or any child
2514 hashes or arrays. All levels share a I<root> structure, which contains things
2515 like the FileHandle, a reference counter, and all your options you specified
2516 when you created the object. You can get access to this root structure by
2517 calling the C<root()> method.
2519 my $root = $db->root();
2521 This is useful for changing options after the object has already been created,
2522 such as enabling/disabling locking, volatile or debug modes. You can also
2523 store your own temporary user data in this structure (be wary of name
2524 collision), which is then accessible from any child hash or array.
2526 =head1 CUSTOM DIGEST ALGORITHM
2528 DBM::Deep by default uses the I<Message Digest 5> (MD5) algorithm for hashing
2529 keys. However you can override this, and use another algorithm (such as SHA-256)
2530 or even write your own. But please note that Deep currently expects zero
2531 collisions, so your algorithm has to be I<perfect>, so to speak.
2532 Collision detection may be introduced in a later version.
2536 You can specify a custom digest algorithm by calling the static C<set_digest()>
2537 function, passing a reference to a subroutine, and the length of the algorithm's
2538 hashes (in bytes). This is a global static function, which affects ALL Deep
2539 objects. Here is a working example that uses a 256-bit hash from the
2540 I<Digest::SHA256> module. Please see
2541 L<http://search.cpan.org/search?module=Digest::SHA256> for more.
2546 my $context = Digest::SHA256::new(256);
2548 DBM::Deep::set_digest( \&my_digest, 32 );
2550 my $db = DBM::Deep->new( "foo-sha.db" );
2552 $db->{key1} = "value1";
2553 $db->{key2} = "value2";
2554 print "key1: " . $db->{key1} . "\n";
2555 print "key2: " . $db->{key2} . "\n";
2561 return substr( $context->hash($_[0]), 0, 32 );
2564 B<Note:> Your returned digest strings must be B<EXACTLY> the number
2565 of bytes you specify in the C<set_digest()> function (in this case 32).
2567 =head1 CIRCULAR REFERENCES
2569 DBM::Deep has B<experimental> support for circular references. Meaning you
2570 can have a nested hash key or array element that points to a parent object.
2571 This relationship is stored in the DB file, and is preserved between sessions.
2574 my $db = DBM::Deep->new( "foo.db" );
2577 $db->{circle} = $db; # ref to self
2579 print $db->{foo} . "\n"; # prints "foo"
2580 print $db->{circle}->{foo} . "\n"; # prints "foo" again
2582 One catch is, passing the object to a function that recursively walks the
2583 object tree (such as I<Data::Dumper> or even the built-in C<optimize()> or
2584 C<export()> methods) will result in an infinite loop. The other catch is,
2585 if you fetch the I<key> of a circular reference (i.e. using the C<first_key()>
2586 or C<next_key()> methods), you will get the I<target object's key>, not the
2587 ref's key. This gets even more interesting with the above example, where
2588 the I<circle> key points to the base DB object, which technically doesn't
2589 have a key. So I made DBM::Deep return "[base]" as the key name in that
2592 =head1 CAVEATS / ISSUES / BUGS
2594 This section describes all the known issues with DBM::Deep. It you have found
2595 something that is not listed here, please send e-mail to L<jhuckaby@cpan.org>.
2597 =head2 UNUSED SPACE RECOVERY
2599 One major caveat with Deep is that space occupied by existing keys and
2600 values is not recovered when they are deleted. Meaning if you keep deleting
2601 and adding new keys, your file will continuously grow. I am working on this,
2602 but in the meantime you can call the built-in C<optimize()> method from time to
2603 time (perhaps in a crontab or something) to recover all your unused space.
2605 $db->optimize(); # returns true on success
2607 This rebuilds the ENTIRE database into a new file, then moves it on top of
2608 the original. The new file will have no unused space, thus it will take up as
2609 little disk space as possible. Please note that this operation can take
2610 a long time for large files, and you need enough disk space to temporarily hold
2611 2 copies of your DB file. The temporary file is created in the same directory
2612 as the original, named with a ".tmp" extension, and is deleted when the
2613 operation completes. Oh, and if locking is enabled, the DB is automatically
2614 locked for the entire duration of the copy.
2618 B<WARNING:> Only call optimize() on the top-level node of the database, and
2619 make sure there are no child references lying around. Deep keeps a reference
2620 counter, and if it is greater than 1, optimize() will abort and return undef.
2622 =head2 AUTOVIVIFICATION
2624 Unfortunately, autovivification doesn't work with tied hashes. This appears to
2625 be a bug in Perl's tie() system, as I<Jakob Schmidt> encountered the very same
2626 issue with his I<DWH_FIle> module (see L<http://search.cpan.org/search?module=DWH_File>),
2627 and it is also mentioned in the BUGS section for the I<MLDBM> module <see
2628 L<http://search.cpan.org/search?module=MLDBM>). Basically, on a new db file,
2631 $db->{foo}->{bar} = "hello";
2633 Since "foo" doesn't exist, you cannot add "bar" to it. You end up with "foo"
2634 being an empty hash. Try this instead, which works fine:
2636 $db->{foo} = { bar => "hello" };
2638 As of Perl 5.8.7, this bug still exists. I have walked very carefully through
2639 the execution path, and Perl indeed passes an empty hash to the STORE() method.
2640 Probably a bug in Perl.
2642 =head2 FILE CORRUPTION
2644 The current level of error handling in Deep is minimal. Files I<are> checked
2645 for a 32-bit signature on open(), but other corruption in files can cause
2646 segmentation faults. Deep may try to seek() past the end of a file, or get
2647 stuck in an infinite loop depending on the level of corruption. File write
2648 operations are not checked for failure (for speed), so if you happen to run
2649 out of disk space, Deep will probably fail in a bad way. These things will
2650 be addressed in a later version of DBM::Deep.
2654 Beware of using DB files over NFS. Deep uses flock(), which works well on local
2655 filesystems, but will NOT protect you from file corruption over NFS. I've heard
2656 about setting up your NFS server with a locking daemon, then using lockf() to
2657 lock your files, but your milage may vary there as well. From what I
2658 understand, there is no real way to do it. However, if you need access to the
2659 underlying FileHandle in Deep for using some other kind of locking scheme like
2660 lockf(), see the L<LOW-LEVEL ACCESS> section above.
2662 =head2 COPYING OBJECTS
2664 Beware of copying tied objects in Perl. Very strange things can happen.
2665 Instead, use Deep's C<clone()> method which safely copies the object and
2666 returns a new, blessed, tied hash or array to the same level in the DB.
2668 my $copy = $db->clone();
2672 Beware of using C<shift()>, C<unshift()> or C<splice()> with large arrays.
2673 These functions cause every element in the array to move, which can be murder
2674 on DBM::Deep, as every element has to be fetched from disk, then stored again in
2675 a different location. This may be addressed in a later version.
2679 This section discusses DBM::Deep's speed and memory usage.
2683 Obviously, DBM::Deep isn't going to be as fast as some C-based DBMs, such as
2684 the almighty I<BerkeleyDB>. But it makes up for it in features like true
2685 multi-level hash/array support, and cross-platform FTPable files. Even so,
2686 DBM::Deep is still pretty fast, and the speed stays fairly consistent, even
2687 with huge databases. Here is some test data:
2689 Adding 1,000,000 keys to new DB file...
2691 At 100 keys, avg. speed is 2,703 keys/sec
2692 At 200 keys, avg. speed is 2,642 keys/sec
2693 At 300 keys, avg. speed is 2,598 keys/sec
2694 At 400 keys, avg. speed is 2,578 keys/sec
2695 At 500 keys, avg. speed is 2,722 keys/sec
2696 At 600 keys, avg. speed is 2,628 keys/sec
2697 At 700 keys, avg. speed is 2,700 keys/sec
2698 At 800 keys, avg. speed is 2,607 keys/sec
2699 At 900 keys, avg. speed is 2,190 keys/sec
2700 At 1,000 keys, avg. speed is 2,570 keys/sec
2701 At 2,000 keys, avg. speed is 2,417 keys/sec
2702 At 3,000 keys, avg. speed is 1,982 keys/sec
2703 At 4,000 keys, avg. speed is 1,568 keys/sec
2704 At 5,000 keys, avg. speed is 1,533 keys/sec
2705 At 6,000 keys, avg. speed is 1,787 keys/sec
2706 At 7,000 keys, avg. speed is 1,977 keys/sec
2707 At 8,000 keys, avg. speed is 2,028 keys/sec
2708 At 9,000 keys, avg. speed is 2,077 keys/sec
2709 At 10,000 keys, avg. speed is 2,031 keys/sec
2710 At 20,000 keys, avg. speed is 1,970 keys/sec
2711 At 30,000 keys, avg. speed is 2,050 keys/sec
2712 At 40,000 keys, avg. speed is 2,073 keys/sec
2713 At 50,000 keys, avg. speed is 1,973 keys/sec
2714 At 60,000 keys, avg. speed is 1,914 keys/sec
2715 At 70,000 keys, avg. speed is 2,091 keys/sec
2716 At 80,000 keys, avg. speed is 2,103 keys/sec
2717 At 90,000 keys, avg. speed is 1,886 keys/sec
2718 At 100,000 keys, avg. speed is 1,970 keys/sec
2719 At 200,000 keys, avg. speed is 2,053 keys/sec
2720 At 300,000 keys, avg. speed is 1,697 keys/sec
2721 At 400,000 keys, avg. speed is 1,838 keys/sec
2722 At 500,000 keys, avg. speed is 1,941 keys/sec
2723 At 600,000 keys, avg. speed is 1,930 keys/sec
2724 At 700,000 keys, avg. speed is 1,735 keys/sec
2725 At 800,000 keys, avg. speed is 1,795 keys/sec
2726 At 900,000 keys, avg. speed is 1,221 keys/sec
2727 At 1,000,000 keys, avg. speed is 1,077 keys/sec
2729 This test was performed on a PowerMac G4 1gHz running Mac OS X 10.3.2 & Perl
2730 5.8.1, with an 80GB Ultra ATA/100 HD spinning at 7200RPM. The hash keys and
2731 values were between 6 - 12 chars in length. The DB file ended up at 210MB.
2732 Run time was 12 min 3 sec.
2736 One of the great things about DBM::Deep is that it uses very little memory.
2737 Even with huge databases (1,000,000+ keys) you will not see much increased
2738 memory on your process. Deep relies solely on the filesystem for storing
2739 and fetching data. Here is output from I</usr/bin/top> before even opening a
2742 PID USER PRI NI SIZE RSS SHARE STAT %CPU %MEM TIME COMMAND
2743 22831 root 11 0 2716 2716 1296 R 0.0 0.2 0:07 perl
2745 Basically the process is taking 2,716K of memory. And here is the same
2746 process after storing and fetching 1,000,000 keys:
2748 PID USER PRI NI SIZE RSS SHARE STAT %CPU %MEM TIME COMMAND
2749 22831 root 14 0 2772 2772 1328 R 0.0 0.2 13:32 perl
2751 Notice the memory usage increased by only 56K. Test was performed on a 700mHz
2752 x86 box running Linux RedHat 7.2 & Perl 5.6.1.
2754 =head1 DB FILE FORMAT
2756 In case you were interested in the underlying DB file format, it is documented
2757 here in this section. You don't need to know this to use the module, it's just
2758 included for reference.
2762 DBM::Deep files always start with a 32-bit signature to identify the file type.
2763 This is at offset 0. The signature is "DPDB" in network byte order. This is
2764 checked upon each file open().
2768 The DBM::Deep file is in a I<tagged format>, meaning each section of the file
2769 has a standard header containing the type of data, the length of data, and then
2770 the data itself. The type is a single character (1 byte), the length is a
2771 32-bit unsigned long in network byte order, and the data is, well, the data.
2772 Here is how it unfolds:
2776 Immediately after the 32-bit file signature is the I<Master Index> record.
2777 This is a standard tag header followed by 1024 bytes (in 32-bit mode) or 2048
2778 bytes (in 64-bit mode) of data. The type is I<H> for hash or I<A> for array,
2779 depending on how the DBM::Deep object was constructed.
2783 The index works by looking at a I<MD5 Hash> of the hash key (or array index
2784 number). The first 8-bit char of the MD5 signature is the offset into the
2785 index, multipled by 4 in 32-bit mode, or 8 in 64-bit mode. The value of the
2786 index element is a file offset of the next tag for the key/element in question,
2787 which is usually a I<Bucket List> tag (see below).
2791 The next tag I<could> be another index, depending on how many keys/elements
2792 exist. See L<RE-INDEXING> below for details.
2796 A I<Bucket List> is a collection of 16 MD5 hashes for keys/elements, plus
2797 file offsets to where the actual data is stored. It starts with a standard
2798 tag header, with type I<B>, and a data size of 320 bytes in 32-bit mode, or
2799 384 bytes in 64-bit mode. Each MD5 hash is stored in full (16 bytes), plus
2800 the 32-bit or 64-bit file offset for the I<Bucket> containing the actual data.
2801 When the list fills up, a I<Re-Index> operation is performed (See
2802 L<RE-INDEXING> below).
2806 A I<Bucket> is a tag containing a key/value pair (in hash mode), or a
2807 index/value pair (in array mode). It starts with a standard tag header with
2808 type I<D> for scalar data (string, binary, etc.), or it could be a nested
2809 hash (type I<H>) or array (type I<A>). The value comes just after the tag
2810 header. The size reported in the tag header is only for the value, but then,
2811 just after the value is another size (32-bit unsigned long) and then the plain
2812 key itself. Since the value is likely to be fetched more often than the plain
2813 key, I figured it would be I<slightly> faster to store the value first.
2817 If the type is I<H> (hash) or I<A> (array), the value is another I<Master Index>
2818 record for the nested structure, where the process begins all over again.
2822 After a I<Bucket List> grows to 16 records, its allocated space in the file is
2823 exhausted. Then, when another key/element comes in, the list is converted to a
2824 new index record. However, this index will look at the next char in the MD5
2825 hash, and arrange new Bucket List pointers accordingly. This process is called
2826 I<Re-Indexing>. Basically, a new index tag is created at the file EOF, and all
2827 17 (16 + new one) keys/elements are removed from the old Bucket List and
2828 inserted into the new index. Several new Bucket Lists are created in the
2829 process, as a new MD5 char from the key is being examined (it is unlikely that
2830 the keys will all share the same next char of their MD5s).
2834 Because of the way the I<MD5> algorithm works, it is impossible to tell exactly
2835 when the Bucket Lists will turn into indexes, but the first round tends to
2836 happen right around 4,000 keys. You will see a I<slight> decrease in
2837 performance here, but it picks back up pretty quick (see L<SPEED> above). Then
2838 it takes B<a lot> more keys to exhaust the next level of Bucket Lists. It's
2839 right around 900,000 keys. This process can continue nearly indefinitely --
2840 right up until the point the I<MD5> signatures start colliding with each other,
2841 and this is B<EXTREMELY> rare -- like winning the lottery 5 times in a row AND
2842 getting struck by lightning while you are walking to cash in your tickets.
2843 Theoretically, since I<MD5> hashes are 128-bit values, you I<could> have up to
2844 340,282,366,921,000,000,000,000,000,000,000,000,000 keys/elements (I believe
2845 this is 340 unodecillion, but don't quote me).
2849 When a new key/element is stored, the key (or index number) is first ran through
2850 I<Digest::MD5> to get a 128-bit signature (example, in hex:
2851 b05783b0773d894396d475ced9d2f4f6). Then, the I<Master Index> record is checked
2852 for the first char of the signature (in this case I<b>). If it does not exist,
2853 a new I<Bucket List> is created for our key (and the next 15 future keys that
2854 happen to also have I<b> as their first MD5 char). The entire MD5 is written
2855 to the I<Bucket List> along with the offset of the new I<Bucket> record (EOF at
2856 this point, unless we are replacing an existing I<Bucket>), where the actual
2857 data will be stored.
2861 Fetching an existing key/element involves getting a I<Digest::MD5> of the key
2862 (or index number), then walking along the indexes. If there are enough
2863 keys/elements in this DB level, there might be nested indexes, each linked to
2864 a particular char of the MD5. Finally, a I<Bucket List> is pointed to, which
2865 contains up to 16 full MD5 hashes. Each is checked for equality to the key in
2866 question. If we found a match, the I<Bucket> tag is loaded, where the value and
2867 plain key are stored.
2871 Fetching the plain key occurs when calling the I<first_key()> and I<next_key()>
2872 methods. In this process the indexes are walked systematically, and each key
2873 fetched in increasing MD5 order (which is why it appears random). Once the
2874 I<Bucket> is found, the value is skipped the plain key returned instead.
2875 B<Note:> Do not count on keys being fetched as if the MD5 hashes were
2876 alphabetically sorted. This only happens on an index-level -- as soon as the
2877 I<Bucket Lists> are hit, the keys will come out in the order they went in --
2878 so it's pretty much undefined how the keys will come out -- just like Perl's
2883 Joseph Huckaby, L<jhuckaby@cpan.org>
2885 Special thanks to Adam Sah and Rich Gaushell! You know why :-)
2889 perltie(1), Tie::Hash(3), Digest::MD5(3), Fcntl(3), flock(2), lockf(3), nfs(5),
2890 Digest::SHA256(3), Crypt::Blowfish(3), Compress::Zlib(3)
2894 Copyright (c) 2002-2005 Joseph Huckaby. All Rights Reserved.
2895 This is free software, you may use it and distribute it under the
2896 same terms as Perl itself.