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 { $args = { file => shift }; }
116 #print "Calling new()\n";
119 # Check if we want a tied hash or array.
122 if (defined($args->{type}) && $args->{type} eq TYPE_ARRAY) {
123 my $foo = tie @$self, $class, %$args;
124 #print "Tied '$foo' to array\n";
128 my $foo = tie %$self, $class, %$args;
129 #print "Tied '$foo' to hash\n";
134 #print "Created '$self'\n";
139 my @outer_params = qw( type base_offset );
142 # Setup $self and bless into this class.
149 base_offset => length(SIG_FILE),
160 filter_store_key => undef,
161 filter_store_value => undef,
162 filter_fetch_key => undef,
163 filter_fetch_value => undef,
172 foreach my $outer_parm ( @outer_params ) {
173 next unless exists $args->{$outer_parm};
174 $self->{$outer_parm} = $args->{$outer_parm}
177 if ( exists $args->{root} ) {
178 $self->{root} = $args->{root};
181 # This is cleanup based on the fact that the $args
182 # coming in is for both the root and non-root items
183 delete $self->root->{$_} for @outer_params;
185 $self->root->{links}++;
187 if (!defined($self->fh)) { $self->_open(); }
194 tied( %{$_[0]} ) || $_[0]
199 # Tied hash constructor method, called by Perl's tie() function.
203 if (scalar(@_) > 1) { $args = {@_}; }
204 #XXX This use of ref() is bad and is a bug
205 elsif (ref($_[0])) { $args = $_[0]; }
206 else { $args = { file => shift }; }
208 $args->{type} = TYPE_HASH;
210 return $class->_init($args);
215 # Tied array constructor method, called by Perl's tie() function.
219 if (scalar(@_) > 1) { $args = {@_}; }
220 #XXX This use of ref() is bad and is a bug
221 elsif (ref($_[0])) { $args = $_[0]; }
222 else { $args = { file => shift }; }
224 $args->{type} = TYPE_ARRAY;
226 return $class->_init($args);
231 # Class deconstructor. Close file handle if there are no more refs.
233 my $self = _get_self($_[0]);
236 $self->root->{links}--;
237 #print "DESTROY( $self ): ", $self->root, ':', $self->root->{links}, "\n";
239 if (!$self->root->{links}) {
244 my %translate_mode = (
254 # Open a FileHandle to the database, create if nonexistent.
255 # Make sure file signature matches DeepDB spec.
257 my $self = _get_self($_[0]);
259 if (defined($self->fh)) { $self->_close(); }
262 my $filename = $self->root->{file};
263 my $mode = $translate_mode{ $self->root->{mode} };
264 #print "Opening '$filename' as '$mode'\n";
266 #if (!(-e $filename) && $self->root->{mode} eq 'r+') {
267 if (!(-e $filename) && $mode eq '+<') {
268 #FileHandle->new( $filename, 'w' );
269 open( FH, '>', $filename );
273 #XXX Convert to set_fh()
274 $self->root->{fh} = FileHandle->new( $self->root->{file}, $self->root->{mode} );
276 # open( $fh, $mode, $filename )
278 # $self->root->{fh} = $fh;
279 }; if ($@ ) { $self->_throw_error( "Received error: $@\n" ); }
280 if (! defined($self->fh)) {
281 return $self->_throw_error("Cannot open file: " . $self->root->{file} . ": $!");
286 #XXX Can we remove this by using the right sysopen() flags?
287 binmode $fh; # for win32
289 if ($self->root->{autoflush}) {
290 # $self->fh->autoflush();
291 my $old = select $fh;
298 my $bytes_read = read( $fh, $signature, length(SIG_FILE));
301 # File is empty -- write signature and master index
306 $self->root->{end} = length(SIG_FILE);
307 $self->_create_tag($self->base_offset, $self->type, chr(0) x $INDEX_SIZE);
309 my $plain_key = "[base]";
310 print($fh pack($DATA_LENGTH_PACK, length($plain_key)) . $plain_key );
311 $self->root->{end} += $DATA_LENGTH_SIZE + length($plain_key);
314 my $old_fh = select $fh;
324 # Check signature was valid
326 unless ($signature eq SIG_FILE) {
328 return $self->_throw_error("Signature not found -- file is not a Deep DB");
331 $self->root->{end} = (stat($fh))[7];
334 # Get our type from master index signature
336 my $tag = $self->_load_tag($self->base_offset);
337 #XXX We probably also want to store the hash algorithm name and not assume anything
339 return $self->_throw_error("Corrupted file, no master index record");
341 if ($self->{type} ne $tag->{signature}) {
342 return $self->_throw_error("File type mismatch");
350 # Close database FileHandle
353 my $self = _get_self($_[0]);
354 undef $self->root->{fh};
359 # Given offset, signature and content, create tag and write to disk
361 my ($self, $offset, $sig, $content) = @_;
362 my $size = length($content);
366 seek($fh, $offset, 0);
367 print($fh $sig . pack($DATA_LENGTH_PACK, $size) . $content );
369 if ($offset == $self->root->{end}) {
370 $self->root->{end} += SIG_SIZE + $DATA_LENGTH_SIZE + $size;
376 offset => $offset + SIG_SIZE + $DATA_LENGTH_SIZE,
383 # Given offset, load single tag and return signature, size and data
390 seek($fh, $offset, 0);
391 if (eof $fh) { return undef; }
394 read( $fh, $sig, SIG_SIZE);
397 read( $fh, $size, $DATA_LENGTH_SIZE);
398 $size = unpack($DATA_LENGTH_PACK, $size);
401 read( $fh, $buffer, $size);
406 offset => $offset + SIG_SIZE + $DATA_LENGTH_SIZE,
413 # Given index tag, lookup single entry in index and return .
416 my ($tag, $index) = @_;
418 my $location = unpack($LONG_PACK, substr($tag->{content}, $index * $LONG_SIZE, $LONG_SIZE) );
419 if (!$location) { return; }
421 return $self->_load_tag( $location );
426 # Adds one key/value pair to bucket list, given offset, MD5 digest of key,
427 # plain (undigested) key and value.
430 my ($tag, $md5, $plain_key, $value) = @_;
431 my $keys = $tag->{content};
435 my $is_dbm_deep = eval { $value->isa( 'DBM::Deep' ) };
436 my $internal_ref = $is_dbm_deep && ($value->root eq $self->root);
442 # Iterate through buckets, seeing if this is a new entry or a replace.
444 for (my $i=0; $i<$MAX_BUCKETS; $i++) {
445 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
446 my $subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
449 # Found empty bucket (end of list). Populate and exit loop.
453 $location = $internal_ref
454 ? $value->base_offset
455 : $self->root->{end};
457 seek($fh, $tag->{offset} + ($i * $BUCKET_SIZE), 0);
458 print($fh $md5 . pack($LONG_PACK, $location) );
461 elsif ($md5 eq $key) {
463 # Found existing bucket with same key. Replace with new value.
468 $location = $value->base_offset;
469 seek($fh, $tag->{offset} + ($i * $BUCKET_SIZE), 0);
470 print($fh $md5 . pack($LONG_PACK, $location) );
473 seek($fh, $subloc + SIG_SIZE, 0);
475 read( $fh, $size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
478 # If value is a hash, array, or raw value with equal or less size, we can
479 # reuse the same content area of the database. Otherwise, we have to create
480 # a new content area at the EOF.
483 my $r = Scalar::Util::reftype( $value ) || '';
484 if ( $r eq 'HASH' || $r eq 'ARRAY' ) { $actual_length = $INDEX_SIZE; }
485 else { $actual_length = length($value); }
487 if ($actual_length <= $size) {
491 $location = $self->root->{end};
492 seek($fh, $tag->{offset} + ($i * $BUCKET_SIZE) + $HASH_SIZE, 0);
493 print($fh pack($LONG_PACK, $location) );
501 # If this is an internal reference, return now.
502 # No need to write value or plain key
509 # If bucket didn't fit into list, split into a new index level
512 seek($fh, $tag->{ref_loc}, 0);
513 print($fh pack($LONG_PACK, $self->root->{end}) );
515 my $index_tag = $self->_create_tag($self->root->{end}, SIG_INDEX, chr(0) x $INDEX_SIZE);
518 $keys .= $md5 . pack($LONG_PACK, 0);
520 for (my $i=0; $i<=$MAX_BUCKETS; $i++) {
521 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
523 my $old_subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
524 my $num = ord(substr($key, $tag->{ch} + 1, 1));
526 if ($offsets[$num]) {
527 my $offset = $offsets[$num] + SIG_SIZE + $DATA_LENGTH_SIZE;
528 seek($fh, $offset, 0);
530 read( $fh, $subkeys, $BUCKET_LIST_SIZE);
532 for (my $k=0; $k<$MAX_BUCKETS; $k++) {
533 my $subloc = unpack($LONG_PACK, substr($subkeys, ($k * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
535 seek($fh, $offset + ($k * $BUCKET_SIZE), 0);
536 print($fh $key . pack($LONG_PACK, $old_subloc || $self->root->{end}) );
542 $offsets[$num] = $self->root->{end};
543 seek($fh, $index_tag->{offset} + ($num * $LONG_SIZE), 0);
544 print($fh pack($LONG_PACK, $self->root->{end}) );
546 my $blist_tag = $self->_create_tag($self->root->{end}, SIG_BLIST, chr(0) x $BUCKET_LIST_SIZE);
548 seek($fh, $blist_tag->{offset}, 0);
549 print($fh $key . pack($LONG_PACK, $old_subloc || $self->root->{end}) );
554 $location ||= $self->root->{end};
555 } # re-index bucket list
558 # Seek to content area and store signature, value and plaintext key
562 seek($fh, $location, 0);
565 # Write signature based on content type, set content length and write actual value.
567 my $r = Scalar::Util::reftype($value) || '';
569 print($fh TYPE_HASH );
570 print($fh pack($DATA_LENGTH_PACK, $INDEX_SIZE) . chr(0) x $INDEX_SIZE );
571 $content_length = $INDEX_SIZE;
573 elsif ($r eq 'ARRAY') {
574 print($fh TYPE_ARRAY );
575 print($fh pack($DATA_LENGTH_PACK, $INDEX_SIZE) . chr(0) x $INDEX_SIZE );
576 $content_length = $INDEX_SIZE;
578 elsif (!defined($value)) {
579 print($fh SIG_NULL );
580 print($fh pack($DATA_LENGTH_PACK, 0) );
584 print($fh SIG_DATA );
585 print($fh pack($DATA_LENGTH_PACK, length($value)) . $value );
586 $content_length = length($value);
590 # Plain key is stored AFTER value, as keys are typically fetched less often.
592 print($fh pack($DATA_LENGTH_PACK, length($plain_key)) . $plain_key );
595 # If value is blessed, preserve class name
597 if ( $self->root->{autobless} ) {
598 my $value_class = Scalar::Util::blessed($value);
599 if ( defined $value_class && $value_class ne 'DBM::Deep' ) {
601 # Blessed ref -- will restore later
604 print($fh pack($DATA_LENGTH_PACK, length($value_class)) . $value_class );
605 $content_length += 1;
606 $content_length += $DATA_LENGTH_SIZE + length($value_class);
610 $content_length += 1;
615 # If this is a new content area, advance EOF counter
617 if ($location == $self->root->{end}) {
618 $self->root->{end} += SIG_SIZE;
619 $self->root->{end} += $DATA_LENGTH_SIZE + $content_length;
620 $self->root->{end} += $DATA_LENGTH_SIZE + length($plain_key);
624 # If content is a hash or array, create new child DeepDB object and
625 # pass each key or element to it.
629 my $branch = DBM::Deep->new(
631 base_offset => $location,
634 foreach my $key (keys %{$value}) {
635 #$branch->{$key} = $value->{$key};
636 $branch->STORE( $key, $value->{$key} );
639 elsif ($r eq 'ARRAY') {
640 #print "$self -> ", $self->root, $/;
641 my $branch = DBM::Deep->new(
643 base_offset => $location,
646 #print "After new - $branch -> ", $branch->root, "\n";
648 foreach my $element (@{$value}) {
649 #$branch->[$index] = $element;
650 $branch->STORE( $index, $element );
653 #print "After elements\n";
660 return $self->_throw_error("Fatal error: indexing failed -- possibly due to corruption in file");
663 sub _get_bucket_value {
665 # Fetch single value given tag and MD5 digested key.
668 my ($tag, $md5) = @_;
669 my $keys = $tag->{content};
674 # Iterate through buckets, looking for a key match
677 for (my $i=0; $i<$MAX_BUCKETS; $i++) {
678 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
679 my $subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
683 # Hit end of list, no match
688 if ( $md5 ne $key ) {
693 # Found match -- seek to offset and read signature
696 seek($fh, $subloc, 0);
697 read( $fh, $signature, SIG_SIZE);
700 # If value is a hash or array, return new DeepDB object with correct offset
702 if (($signature eq TYPE_HASH) || ($signature eq TYPE_ARRAY)) {
703 my $obj = DBM::Deep->new(
705 base_offset => $subloc,
709 if ($self->root->{autobless}) {
711 # Skip over value and plain key to see if object needs
714 seek($fh, $DATA_LENGTH_SIZE + $INDEX_SIZE, 1);
717 read( $fh, $size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
718 if ($size) { seek($fh, $size, 1); }
721 read( $fh, $bless_bit, 1);
722 if (ord($bless_bit)) {
724 # Yes, object needs to be re-blessed
727 read( $fh, $size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
728 if ($size) { read( $fh, $class_name, $size); }
729 if ($class_name) { $obj = bless( $obj, $class_name ); }
737 # Otherwise return actual value
739 elsif ($signature eq SIG_DATA) {
742 read( $fh, $size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
743 if ($size) { read( $fh, $value, $size); }
748 # Key exists, but content is null
758 # Delete single key/value pair given tag and MD5 digested key.
761 my ($tag, $md5) = @_;
762 my $keys = $tag->{content};
767 # Iterate through buckets, looking for a key match
770 for (my $i=0; $i<$MAX_BUCKETS; $i++) {
771 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
772 my $subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
776 # Hit end of list, no match
781 if ( $md5 ne $key ) {
786 # Matched key -- delete bucket and return
788 seek($fh, $tag->{offset} + ($i * $BUCKET_SIZE), 0);
789 print($fh substr($keys, ($i+1) * $BUCKET_SIZE ) );
790 print($fh chr(0) x $BUCKET_SIZE );
800 # Check existence of single key given tag and MD5 digested key.
803 my ($tag, $md5) = @_;
804 my $keys = $tag->{content};
807 # Iterate through buckets, looking for a key match
810 for (my $i=0; $i<$MAX_BUCKETS; $i++) {
811 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
812 my $subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
816 # Hit end of list, no match
821 if ( $md5 ne $key ) {
826 # Matched key -- return true
834 sub _find_bucket_list {
836 # Locate offset for bucket list, given digested key
842 # Locate offset for bucket list using digest index system
845 my $tag = $self->_load_tag($self->base_offset);
846 if (!$tag) { return; }
848 while ($tag->{signature} ne SIG_BLIST) {
849 $tag = $self->_index_lookup($tag, ord(substr($md5, $ch, 1)));
850 if (!$tag) { return; }
857 sub _traverse_index {
859 # Scan index and recursively step into deeper levels, looking for next key.
861 my ($self, $offset, $ch, $force_return_next) = @_;
862 $force_return_next = undef unless $force_return_next;
864 my $tag = $self->_load_tag( $offset );
868 if ($tag->{signature} ne SIG_BLIST) {
869 my $content = $tag->{content};
871 if ($self->{return_next}) { $start = 0; }
872 else { $start = ord(substr($self->{prev_md5}, $ch, 1)); }
874 for (my $index = $start; $index < 256; $index++) {
875 my $subloc = unpack($LONG_PACK, substr($content, $index * $LONG_SIZE, $LONG_SIZE) );
877 my $result = $self->_traverse_index( $subloc, $ch + 1, $force_return_next );
878 if (defined($result)) { return $result; }
882 $self->{return_next} = 1;
885 elsif ($tag->{signature} eq SIG_BLIST) {
886 my $keys = $tag->{content};
887 if ($force_return_next) { $self->{return_next} = 1; }
890 # Iterate through buckets, looking for a key match
892 for (my $i=0; $i<$MAX_BUCKETS; $i++) {
893 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
894 my $subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
898 # End of bucket list -- return to outer loop
900 $self->{return_next} = 1;
903 elsif ($key eq $self->{prev_md5}) {
905 # Located previous key -- return next one found
907 $self->{return_next} = 1;
910 elsif ($self->{return_next}) {
912 # Seek to bucket location and skip over signature
914 seek($fh, $subloc + SIG_SIZE, 0);
917 # Skip over value to get to plain key
920 read( $fh, $size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
921 if ($size) { seek($fh, $size, 1); }
924 # Read in plain key and return as scalar
927 read( $fh, $size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
928 if ($size) { read( $fh, $plain_key, $size); }
934 $self->{return_next} = 1;
935 } # tag is a bucket list
942 # Locate next key, given digested previous one
944 my $self = _get_self($_[0]);
946 $self->{prev_md5} = $_[1] ? $_[1] : undef;
947 $self->{return_next} = 0;
950 # If the previous key was not specifed, start at the top and
951 # return the first one found.
953 if (!$self->{prev_md5}) {
954 $self->{prev_md5} = chr(0) x $HASH_SIZE;
955 $self->{return_next} = 1;
958 return $self->_traverse_index( $self->base_offset, 0 );
963 # If db locking is set, flock() the db file. If called multiple
964 # times before unlock(), then the same number of unlocks() must
965 # be called before the lock is released.
967 my $self = _get_self($_[0]);
969 $type = LOCK_EX unless defined $type;
971 if ($self->root->{locking}) {
972 if (!$self->root->{locked}) { flock($self->fh, $type); }
973 $self->root->{locked}++;
979 # If db locking is set, unlock the db file. See note in lock()
980 # regarding calling lock() multiple times.
982 my $self = _get_self($_[0]);
984 if ($self->root->{locking} && $self->root->{locked} > 0) {
985 $self->root->{locked}--;
986 if (!$self->root->{locked}) { flock($self->fh, LOCK_UN); }
990 #XXX These uses of ref() need verified
993 # Copy single level of keys or elements to new DB handle.
994 # Recurse for nested structures
996 my $self = _get_self($_[0]);
999 if ($self->type eq TYPE_HASH) {
1000 my $key = $self->first_key();
1002 my $value = $self->get($key);
1003 #XXX This doesn't work with autobless
1004 if (!ref($value)) { $db_temp->{$key} = $value; }
1006 my $type = $value->type;
1007 if ($type eq TYPE_HASH) { $db_temp->{$key} = {}; }
1008 else { $db_temp->{$key} = []; }
1009 $value->_copy_node( $db_temp->{$key} );
1011 $key = $self->next_key($key);
1015 my $length = $self->length();
1016 for (my $index = 0; $index < $length; $index++) {
1017 my $value = $self->get($index);
1018 if (!ref($value)) { $db_temp->[$index] = $value; }
1019 #XXX NO tests for this code
1021 my $type = $value->type;
1022 if ($type eq TYPE_HASH) { $db_temp->[$index] = {}; }
1023 else { $db_temp->[$index] = []; }
1024 $value->_copy_node( $db_temp->[$index] );
1032 # Recursively export into standard Perl hashes and arrays.
1034 my $self = _get_self($_[0]);
1037 if ($self->type eq TYPE_HASH) { $temp = {}; }
1038 elsif ($self->type eq TYPE_ARRAY) { $temp = []; }
1041 $self->_copy_node( $temp );
1049 # Recursively import Perl hash/array structure
1051 #XXX This use of ref() seems to be ok
1052 if (!ref($_[0])) { return; } # Perl calls import() on use -- ignore
1054 my $self = _get_self($_[0]);
1057 #XXX This use of ref() seems to be ok
1058 if (!ref($struct)) {
1060 # struct is not a reference, so just import based on our type
1064 if ($self->type eq TYPE_HASH) { $struct = {@_}; }
1065 elsif ($self->type eq TYPE_ARRAY) { $struct = [@_]; }
1068 my $r = Scalar::Util::reftype($struct) || '';
1069 if ($r eq "HASH" && $self->type eq TYPE_HASH) {
1070 foreach my $key (keys %$struct) { $self->put($key, $struct->{$key}); }
1072 elsif ($r eq "ARRAY" && $self->type eq TYPE_ARRAY) {
1073 $self->push( @$struct );
1076 return $self->_throw_error("Cannot import: type mismatch");
1084 # Rebuild entire database into new file, then move
1085 # it back on top of original.
1087 my $self = _get_self($_[0]);
1088 if ($self->root->{links} > 1) {
1089 return $self->_throw_error("Cannot optimize: reference count is greater than 1");
1092 my $db_temp = DBM::Deep->new(
1093 file => $self->root->{file} . '.tmp',
1097 return $self->_throw_error("Cannot optimize: failed to open temp file: $!");
1101 $self->_copy_node( $db_temp );
1105 # Attempt to copy user, group and permissions over to new file
1107 my @stats = stat($self->fh);
1108 my $perms = $stats[2] & 07777;
1109 my $uid = $stats[4];
1110 my $gid = $stats[5];
1111 chown( $uid, $gid, $self->root->{file} . '.tmp' );
1112 chmod( $perms, $self->root->{file} . '.tmp' );
1114 # q.v. perlport for more information on this variable
1115 if ( $^O eq 'MSWin32' ) {
1117 # Potential race condition when optmizing on Win32 with locking.
1118 # The Windows filesystem requires that the filehandle be closed
1119 # before it is overwritten with rename(). This could be redone
1126 if (!rename $self->root->{file} . '.tmp', $self->root->{file}) {
1127 unlink $self->root->{file} . '.tmp';
1129 return $self->_throw_error("Optimize failed: Cannot copy temp file over original: $!");
1141 # Make copy of object and return
1143 my $self = _get_self($_[0]);
1145 return DBM::Deep->new(
1146 type => $self->type,
1147 base_offset => $self->base_offset,
1153 my %is_legal_filter = map {
1156 store_key store_value
1157 fetch_key fetch_value
1162 # Setup filter function for storing or fetching the key or value
1164 my $self = _get_self($_[0]);
1165 my $type = lc $_[1];
1166 my $func = $_[2] ? $_[2] : undef;
1168 if ( $is_legal_filter{$type} ) {
1169 $self->root->{"filter_$type"} = $func;
1183 # Get access to the root structure
1185 my $self = _get_self($_[0]);
1186 return $self->{root};
1191 # Get access to the raw FileHandle
1193 #XXX It will be useful, though, when we split out HASH and ARRAY
1194 my $self = _get_self($_[0]);
1195 return $self->root->{fh};
1200 # Get type of current node (TYPE_HASH or TYPE_ARRAY)
1202 my $self = _get_self($_[0]);
1203 return $self->{type};
1208 # Get base_offset of current node (TYPE_HASH or TYPE_ARRAY)
1210 my $self = _get_self($_[0]);
1211 return $self->{base_offset};
1216 # Get last error string, or undef if no error
1219 ? ( _get_self($_[0])->{root}->{error} or undef )
1229 # Store error string in self
1231 my $self = _get_self($_[0]);
1232 my $error_text = $_[1];
1234 $self->root->{error} = $error_text;
1236 unless ($self->root->{debug}) {
1237 die "DBM::Deep: $error_text\n";
1240 warn "DBM::Deep: $error_text\n";
1248 my $self = _get_self($_[0]);
1250 undef $self->root->{error};
1255 # Precalculate index, bucket and bucket list sizes
1258 #XXX I don't like this ...
1259 set_pack() unless defined $LONG_SIZE;
1261 $INDEX_SIZE = 256 * $LONG_SIZE;
1262 $BUCKET_SIZE = $HASH_SIZE + $LONG_SIZE;
1263 $BUCKET_LIST_SIZE = $MAX_BUCKETS * $BUCKET_SIZE;
1268 # Set pack/unpack modes (see file header for more)
1270 my ($long_s, $long_p, $data_s, $data_p) = @_;
1272 $LONG_SIZE = $long_s ? $long_s : 4;
1273 $LONG_PACK = $long_p ? $long_p : 'N';
1275 $DATA_LENGTH_SIZE = $data_s ? $data_s : 4;
1276 $DATA_LENGTH_PACK = $data_p ? $data_p : 'N';
1283 # Set key digest function (default is MD5)
1285 my ($digest_func, $hash_size) = @_;
1287 $DIGEST_FUNC = $digest_func ? $digest_func : \&Digest::MD5::md5;
1288 $HASH_SIZE = $hash_size ? $hash_size : 16;
1294 # tie() methods (hashes and arrays)
1299 # Store single hash key/value or array element in database.
1301 my $self = _get_self($_[0]);
1302 #print "STORE: $self ... $_[0]\n";
1303 my $key = ($self->root->{filter_store_key} && $self->type eq TYPE_HASH) ? $self->root->{filter_store_key}->($_[1]) : $_[1];
1304 #XXX What is ref() checking here?
1305 #YYY User may be storing a hash, in which case we do not want it run
1306 #YYY through the filtering system
1307 my $value = ($self->root->{filter_store_value} && !ref($_[2])) ? $self->root->{filter_store_value}->($_[2]) : $_[2];
1309 my $unpacked_key = $key;
1310 if (($self->type eq TYPE_ARRAY) && ($key =~ /^\d+$/)) { $key = pack($LONG_PACK, $key); }
1311 my $md5 = $DIGEST_FUNC->($key);
1315 # Make sure file is open
1317 if (!defined($self->fh) && !$self->_open()) {
1325 # Request exclusive lock for writing
1327 $self->lock( LOCK_EX );
1330 # If locking is enabled, set 'end' parameter again, in case another
1331 # DB instance appended to our file while we were unlocked.
1333 if ($self->root->{locking} || $self->root->{volatile}) {
1334 $self->root->{end} = (stat($fh))[7];
1338 # Locate offset for bucket list using digest index system
1340 my $tag = $self->_load_tag($self->base_offset);
1342 $tag = $self->_create_tag($self->base_offset, SIG_INDEX, chr(0) x $INDEX_SIZE);
1346 while ($tag->{signature} ne SIG_BLIST) {
1347 my $num = ord(substr($md5, $ch, 1));
1348 my $new_tag = $self->_index_lookup($tag, $num);
1350 my $ref_loc = $tag->{offset} + ($num * $LONG_SIZE);
1351 seek($fh, $ref_loc, 0);
1352 print($fh pack($LONG_PACK, $self->root->{end}) );
1354 $tag = $self->_create_tag($self->root->{end}, SIG_BLIST, chr(0) x $BUCKET_LIST_SIZE);
1355 $tag->{ref_loc} = $ref_loc;
1360 my $ref_loc = $tag->{offset} + ($num * $LONG_SIZE);
1362 $tag->{ref_loc} = $ref_loc;
1369 # Add key/value to bucket list
1371 my $result = $self->_add_bucket( $tag, $md5, $key, $value );
1375 # If this object is an array, and bucket was not a replace, and key is numerical,
1376 # and index is equal or greater than current length, advance length variable.
1378 if (($result == 2) && ($self->type eq TYPE_ARRAY) && ($unpacked_key =~ /^\d+$/) && ($unpacked_key >= $self->FETCHSIZE())) {
1379 $self->STORESIZE( $unpacked_key + 1 );
1389 # Fetch single value or element given plain key or array index
1391 my $self = _get_self($_[0]);
1392 #print "FETCH: $self ... $_[0]\n";
1395 if ( $self->type eq TYPE_HASH ) {
1396 if ( my $filter = $self->root->{filter_store_key} ) {
1397 $key = $filter->( $key );
1400 elsif ( $self->type eq TYPE_ARRAY ) {
1401 if ( $key =~ /^\d+$/ ) {
1402 $key = pack($LONG_PACK, $key);
1406 my $md5 = $DIGEST_FUNC->($key);
1409 # Make sure file is open
1411 if (!defined($self->fh)) {
1412 #print "Calling _open from FETCH for '$key'\n";
1416 # Request shared lock for reading
1418 $self->lock( LOCK_SH );
1420 my $tag = $self->_find_bucket_list( $md5 );
1427 # Get value from bucket list
1429 my $result = $self->_get_bucket_value( $tag, $md5 );
1433 #XXX What is ref() checking here?
1434 return ($result && !ref($result) && $self->root->{filter_fetch_value}) ? $self->root->{filter_fetch_value}->($result) : $result;
1439 # Delete single key/value pair or element given plain key or array index
1441 my $self = _get_self($_[0]);
1442 my $key = ($self->root->{filter_store_key} && $self->type eq TYPE_HASH) ? $self->root->{filter_store_key}->($_[1]) : $_[1];
1444 my $unpacked_key = $key;
1445 if (($self->type eq TYPE_ARRAY) && ($key =~ /^\d+$/)) { $key = pack($LONG_PACK, $key); }
1446 my $md5 = $DIGEST_FUNC->($key);
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 my $tag = $self->_find_bucket_list( $md5 );
1467 my $result = $self->_delete_bucket( $tag, $md5 );
1470 # If this object is an array and the key deleted was on the end of the stack,
1471 # decrement the length variable.
1473 if ($result && ($self->type eq TYPE_ARRAY) && ($unpacked_key == $self->FETCHSIZE() - 1)) {
1474 $self->STORESIZE( $unpacked_key );
1484 # Check if a single key or element exists given plain key or array index
1486 my $self = _get_self($_[0]);
1487 my $key = ($self->root->{filter_store_key} && $self->type eq TYPE_HASH) ? $self->root->{filter_store_key}->($_[1]) : $_[1];
1489 if (($self->type eq TYPE_ARRAY) && ($key =~ /^\d+$/)) { $key = pack($LONG_PACK, $key); }
1490 my $md5 = $DIGEST_FUNC->($key);
1493 # Make sure file is open
1495 if (!defined($self->fh)) { $self->_open(); }
1498 # Request shared lock for reading
1500 $self->lock( LOCK_SH );
1502 my $tag = $self->_find_bucket_list( $md5 );
1505 # For some reason, the built-in exists() function returns '' for false
1513 # Check if bucket exists and return 1 or ''
1515 my $result = $self->_bucket_exists( $tag, $md5 ) || '';
1524 # Clear all keys from hash, or all elements from array.
1526 my $self = _get_self($_[0]);
1529 # Make sure file is open
1531 if (!defined($self->fh)) { $self->_open(); }
1534 # Request exclusive lock for writing
1536 $self->lock( LOCK_EX );
1540 seek($fh, $self->base_offset, 0);
1546 $self->_create_tag($self->base_offset, $self->type, chr(0) x $INDEX_SIZE);
1555 # Locate and return first key (in no particular order)
1557 my $self = _get_self($_[0]);
1558 if ($self->type ne TYPE_HASH) {
1559 return $self->_throw_error("FIRSTKEY method only supported for hashes");
1563 # Make sure file is open
1565 if (!defined($self->fh)) { $self->_open(); }
1568 # Request shared lock for reading
1570 $self->lock( LOCK_SH );
1572 my $result = $self->_get_next_key();
1576 return ($result && $self->root->{filter_fetch_key}) ? $self->root->{filter_fetch_key}->($result) : $result;
1581 # Return next key (in no particular order), given previous one
1583 my $self = _get_self($_[0]);
1584 if ($self->type ne TYPE_HASH) {
1585 return $self->_throw_error("NEXTKEY method only supported for hashes");
1587 my $prev_key = ($self->root->{filter_store_key} && $self->type eq TYPE_HASH) ? $self->root->{filter_store_key}->($_[1]) : $_[1];
1588 my $prev_md5 = $DIGEST_FUNC->($prev_key);
1591 # Make sure file is open
1593 if (!defined($self->fh)) { $self->_open(); }
1596 # Request shared lock for reading
1598 $self->lock( LOCK_SH );
1600 my $result = $self->_get_next_key( $prev_md5 );
1604 return ($result && $self->root->{filter_fetch_key}) ? $self->root->{filter_fetch_key}->($result) : $result;
1608 # The following methods are for arrays only
1613 # Return the length of the array
1615 my $self = _get_self($_[0]);
1616 if ($self->type ne TYPE_ARRAY) {
1617 return $self->_throw_error("FETCHSIZE method only supported for arrays");
1620 my $SAVE_FILTER = $self->root->{filter_fetch_value};
1621 $self->root->{filter_fetch_value} = undef;
1623 #print "Fetching size ...\n";
1624 my $packed_size = $self->FETCH('length');
1625 #print "size is '$packed_size'\n";
1627 $self->root->{filter_fetch_value} = $SAVE_FILTER;
1629 if ($packed_size) { return int(unpack($LONG_PACK, $packed_size)); }
1635 # Set the length of the array
1637 my $self = _get_self($_[0]);
1638 if ($self->type ne TYPE_ARRAY) {
1639 return $self->_throw_error("STORESIZE method only supported for arrays");
1641 my $new_length = $_[1];
1643 my $SAVE_FILTER = $self->root->{filter_store_value};
1644 $self->root->{filter_store_value} = undef;
1646 my $result = $self->STORE('length', pack($LONG_PACK, $new_length));
1648 $self->root->{filter_store_value} = $SAVE_FILTER;
1655 # Remove and return the last element on the array
1657 my $self = _get_self($_[0]);
1658 if ($self->type ne TYPE_ARRAY) {
1659 return $self->_throw_error("POP method only supported for arrays");
1661 my $length = $self->FETCHSIZE();
1664 my $content = $self->FETCH( $length - 1 );
1665 $self->DELETE( $length - 1 );
1675 # Add new element(s) to the end of the array
1677 my $self = _get_self(shift);
1678 if ($self->type ne TYPE_ARRAY) {
1679 return $self->_throw_error("PUSH method only supported for arrays");
1681 my $length = $self->FETCHSIZE();
1683 while (my $content = shift @_) {
1684 $self->STORE( $length, $content );
1691 # Remove and return first element on the array.
1692 # Shift over remaining elements to take up space.
1694 my $self = _get_self($_[0]);
1695 if ($self->type ne TYPE_ARRAY) {
1696 return $self->_throw_error("SHIFT method only supported for arrays");
1698 my $length = $self->FETCHSIZE();
1701 my $content = $self->FETCH( 0 );
1704 # Shift elements over and remove last one.
1706 for (my $i = 0; $i < $length - 1; $i++) {
1707 $self->STORE( $i, $self->FETCH($i + 1) );
1709 $self->DELETE( $length - 1 );
1720 # Insert new element(s) at beginning of array.
1721 # Shift over other elements to make space.
1723 my $self = _get_self($_[0]);shift @_;
1724 if ($self->type ne TYPE_ARRAY) {
1725 return $self->_throw_error("UNSHIFT method only supported for arrays");
1727 my @new_elements = @_;
1728 my $length = $self->FETCHSIZE();
1729 my $new_size = scalar @new_elements;
1732 for (my $i = $length - 1; $i >= 0; $i--) {
1733 $self->STORE( $i + $new_size, $self->FETCH($i) );
1737 for (my $i = 0; $i < $new_size; $i++) {
1738 $self->STORE( $i, $new_elements[$i] );
1744 # Splices section of array with optional new section.
1745 # Returns deleted section, or last element deleted in scalar context.
1747 my $self = _get_self($_[0]);shift @_;
1748 if ($self->type ne TYPE_ARRAY) {
1749 return $self->_throw_error("SPLICE method only supported for arrays");
1751 my $length = $self->FETCHSIZE();
1754 # Calculate offset and length of splice
1756 my $offset = shift || 0;
1757 if ($offset < 0) { $offset += $length; }
1760 if (scalar @_) { $splice_length = shift; }
1761 else { $splice_length = $length - $offset; }
1762 if ($splice_length < 0) { $splice_length += ($length - $offset); }
1765 # Setup array with new elements, and copy out old elements for return
1767 my @new_elements = @_;
1768 my $new_size = scalar @new_elements;
1770 my @old_elements = ();
1771 for (my $i = $offset; $i < $offset + $splice_length; $i++) {
1772 push @old_elements, $self->FETCH( $i );
1776 # Adjust array length, and shift elements to accomodate new section.
1778 if ( $new_size != $splice_length ) {
1779 if ($new_size > $splice_length) {
1780 for (my $i = $length - 1; $i >= $offset + $splice_length; $i--) {
1781 $self->STORE( $i + ($new_size - $splice_length), $self->FETCH($i) );
1785 for (my $i = $offset + $splice_length; $i < $length; $i++) {
1786 $self->STORE( $i + ($new_size - $splice_length), $self->FETCH($i) );
1788 for (my $i = 0; $i < $splice_length - $new_size; $i++) {
1789 $self->DELETE( $length - 1 );
1796 # Insert new elements into array
1798 for (my $i = $offset; $i < $offset + $new_size; $i++) {
1799 $self->STORE( $i, shift @new_elements );
1803 # Return deleted section, or last element in scalar context.
1805 return wantarray ? @old_elements : $old_elements[-1];
1808 #XXX We don't need to define it.
1809 #XXX It will be useful, though, when we split out HASH and ARRAY
1812 # Perl will call EXTEND() when the array is likely to grow.
1813 # We don't care, but include it for compatibility.
1818 # Public method aliases
1820 *put = *store = *STORE;
1821 *get = *fetch = *FETCH;
1825 *first_key = *FIRSTKEY;
1826 *next_key = *NEXTKEY;
1827 *length = *FETCHSIZE;
1831 *unshift = *UNSHIFT;
1840 DBM::Deep - A pure perl multi-level hash/array DBM
1845 my $db = DBM::Deep->new( "foo.db" );
1847 $db->{key} = 'value'; # tie() style
1850 $db->put('key', 'value'); # OO style
1851 print $db->get('key');
1853 # true multi-level support
1854 $db->{my_complex} = [
1855 'hello', { perl => 'rules' },
1860 A unique flat-file database module, written in pure perl. True
1861 multi-level hash/array support (unlike MLDBM, which is faked), hybrid
1862 OO / tie() interface, cross-platform FTPable files, and quite fast. Can
1863 handle millions of keys and unlimited hash levels without significant
1864 slow-down. Written from the ground-up in pure perl -- this is NOT a
1865 wrapper around a C-based DBM. Out-of-the-box compatibility with Unix,
1866 Mac OS X and Windows.
1870 Hopefully you are using CPAN's excellent Perl module, which will download
1871 and install the module for you. If not, get the tarball, and run these
1883 Construction can be done OO-style (which is the recommended way), or using
1884 Perl's tie() function. Both are examined here.
1886 =head2 OO CONSTRUCTION
1888 The recommended way to construct a DBM::Deep object is to use the new()
1889 method, which gets you a blessed, tied hash or array reference.
1891 my $db = DBM::Deep->new( "foo.db" );
1893 This opens a new database handle, mapped to the file "foo.db". If this
1894 file does not exist, it will automatically be created. DB files are
1895 opened in "r+" (read/write) mode, and the type of object returned is a
1896 hash, unless otherwise specified (see L<OPTIONS> below).
1900 You can pass a number of options to the constructor to specify things like
1901 locking, autoflush, etc. This is done by passing an inline hash:
1903 my $db = DBM::Deep->new(
1909 Notice that the filename is now specified I<inside> the hash with
1910 the "file" parameter, as opposed to being the sole argument to the
1911 constructor. This is required if any options are specified.
1912 See L<OPTIONS> below for the complete list.
1916 You can also start with an array instead of a hash. For this, you must
1917 specify the C<type> parameter:
1919 my $db = DBM::Deep->new(
1921 type => DBM::Deep->TYPE_ARRAY
1924 B<Note:> Specifing the C<type> parameter only takes effect when beginning
1925 a new DB file. If you create a DBM::Deep object with an existing file, the
1926 C<type> will be loaded from the file header, and ignored if it is passed
1929 =head2 TIE CONSTRUCTION
1931 Alternatively, you can create a DBM::Deep handle by using Perl's built-in
1932 tie() function. This is not ideal, because you get only a basic, tied hash
1933 (or array) which is not blessed, so you can't call any functions on it.
1936 tie %hash, "DBM::Deep", "foo.db";
1939 tie @array, "DBM::Deep", "bar.db";
1941 As with the OO constructor, you can replace the DB filename parameter with
1942 a hash containing one or more options (see L<OPTIONS> just below for the
1945 tie %hash, "DBM::Deep", {
1953 There are a number of options that can be passed in when constructing your
1954 DBM::Deep objects. These apply to both the OO- and tie- based approaches.
1960 Filename of the DB file to link the handle to. You can pass a full absolute
1961 filesystem path, partial path, or a plain filename if the file is in the
1962 current working directory. This is a required parameter.
1966 File open mode (read-only, read-write, etc.) string passed to Perl's FileHandle
1967 module. This is an optional parameter, and defaults to "r+" (read/write).
1968 B<Note:> If the default (r+) mode is selected, the file will also be auto-
1969 created if it doesn't exist.
1973 This parameter specifies what type of object to create, a hash or array. Use
1974 one of these two constants: C<DBM::Deep-E<gt>TYPE_HASH> or C<DBM::Deep-E<gt>TYPE_ARRAY>.
1975 This only takes effect when beginning a new file. This is an optional
1976 parameter, and defaults to C<DBM::Deep-E<gt>TYPE_HASH>.
1980 Specifies whether locking is to be enabled. DBM::Deep uses Perl's Fnctl flock()
1981 function to lock the database in exclusive mode for writes, and shared mode for
1982 reads. Pass any true value to enable. This affects the base DB handle I<and
1983 any child hashes or arrays> that use the same DB file. This is an optional
1984 parameter, and defaults to 0 (disabled). See L<LOCKING> below for more.
1988 Specifies whether autoflush is to be enabled on the underlying FileHandle.
1989 This obviously slows down write operations, but is required if you may have
1990 multiple processes accessing the same DB file (also consider enable I<locking>
1991 or at least I<volatile>). Pass any true value to enable. This is an optional
1992 parameter, and defaults to 0 (disabled).
1996 If I<volatile> mode is enabled, DBM::Deep will stat() the DB file before each
1997 STORE() operation. This is required if an outside force may change the size of
1998 the file between transactions. Locking also implicitly enables volatile. This
1999 is useful if you want to use a different locking system or write your own. Pass
2000 any true value to enable. This is an optional parameter, and defaults to 0
2005 If I<autobless> mode is enabled, DBM::Deep will preserve blessed hashes, and
2006 restore them when fetched. This is an B<experimental> feature, and does have
2007 side-effects. Basically, when hashes are re-blessed into their original
2008 classes, they are no longer blessed into the DBM::Deep class! So you won't be
2009 able to call any DBM::Deep methods on them. You have been warned.
2010 This is an optional parameter, and defaults to 0 (disabled).
2014 See L<FILTERS> below.
2018 Setting I<debug> mode will make all errors non-fatal, dump them out to
2019 STDERR, and continue on. This is for debugging purposes only, and probably
2020 not what you want. This is an optional parameter, and defaults to 0 (disabled).
2024 Instead of passing a file path, you can instead pass a handle to an pre-opened
2025 filehandle. Note: Beware of using the magick *DATA handle, as this actually
2026 contains your entire Perl script, as well as the data following the __DATA__
2027 marker. This will not work, because DBM::Deep uses absolute seek()s into the
2028 file. Instead, consider reading *DATA into an IO::Scalar handle, then passing
2033 =head1 TIE INTERFACE
2035 With DBM::Deep you can access your databases using Perl's standard hash/array
2036 syntax. Because all DBM::Deep objects are I<tied> to hashes or arrays, you can treat
2037 them as such. DBM::Deep will intercept all reads/writes and direct them to the right
2038 place -- the DB file. This has nothing to do with the L<TIE CONSTRUCTION>
2039 section above. This simply tells you how to use DBM::Deep using regular hashes
2040 and arrays, rather than calling functions like C<get()> and C<put()> (although those
2041 work too). It is entirely up to you how to want to access your databases.
2045 You can treat any DBM::Deep object like a normal Perl hash reference. Add keys,
2046 or even nested hashes (or arrays) using standard Perl syntax:
2048 my $db = DBM::Deep->new( "foo.db" );
2050 $db->{mykey} = "myvalue";
2052 $db->{myhash}->{subkey} = "subvalue";
2054 print $db->{myhash}->{subkey} . "\n";
2056 You can even step through hash keys using the normal Perl C<keys()> function:
2058 foreach my $key (keys %$db) {
2059 print "$key: " . $db->{$key} . "\n";
2062 Remember that Perl's C<keys()> function extracts I<every> key from the hash and
2063 pushes them onto an array, all before the loop even begins. If you have an
2064 extra large hash, this may exhaust Perl's memory. Instead, consider using
2065 Perl's C<each()> function, which pulls keys/values one at a time, using very
2068 while (my ($key, $value) = each %$db) {
2069 print "$key: $value\n";
2072 Please note that when using C<each()>, you should always pass a direct
2073 hash reference, not a lookup. Meaning, you should B<never> do this:
2076 while (my ($key, $value) = each %{$db->{foo}}) { # BAD
2078 This causes an infinite loop, because for each iteration, Perl is calling
2079 FETCH() on the $db handle, resulting in a "new" hash for foo every time, so
2080 it effectively keeps returning the first key over and over again. Instead,
2081 assign a temporary variable to C<$db->{foo}>, then pass that to each().
2085 As with hashes, you can treat any DBM::Deep object like a normal Perl array
2086 reference. This includes inserting, removing and manipulating elements,
2087 and the C<push()>, C<pop()>, C<shift()>, C<unshift()> and C<splice()> functions.
2088 The object must have first been created using type C<DBM::Deep-E<gt>TYPE_ARRAY>,
2089 or simply be a nested array reference inside a hash. Example:
2091 my $db = DBM::Deep->new(
2092 file => "foo-array.db",
2093 type => DBM::Deep->TYPE_ARRAY
2097 push @$db, "bar", "baz";
2098 unshift @$db, "bah";
2100 my $last_elem = pop @$db; # baz
2101 my $first_elem = shift @$db; # bah
2102 my $second_elem = $db->[1]; # bar
2104 my $num_elements = scalar @$db;
2108 In addition to the I<tie()> interface, you can also use a standard OO interface
2109 to manipulate all aspects of DBM::Deep databases. Each type of object (hash or
2110 array) has its own methods, but both types share the following common methods:
2111 C<put()>, C<get()>, C<exists()>, C<delete()> and C<clear()>.
2117 Stores a new hash key/value pair, or sets an array element value. Takes two
2118 arguments, the hash key or array index, and the new value. The value can be
2119 a scalar, hash ref or array ref. Returns true on success, false on failure.
2121 $db->put("foo", "bar"); # for hashes
2122 $db->put(1, "bar"); # for arrays
2126 Fetches the value of a hash key or array element. Takes one argument: the hash
2127 key or array index. Returns a scalar, hash ref or array ref, depending on the
2130 my $value = $db->get("foo"); # for hashes
2131 my $value = $db->get(1); # for arrays
2135 Checks if a hash key or array index exists. Takes one argument: the hash key
2136 or array index. Returns true if it exists, false if not.
2138 if ($db->exists("foo")) { print "yay!\n"; } # for hashes
2139 if ($db->exists(1)) { print "yay!\n"; } # for arrays
2143 Deletes one hash key/value pair or array element. Takes one argument: the hash
2144 key or array index. Returns true on success, false if not found. For arrays,
2145 the remaining elements located after the deleted element are NOT moved over.
2146 The deleted element is essentially just undefined, which is exactly how Perl's
2147 internal arrays work. Please note that the space occupied by the deleted
2148 key/value or element is B<not> reused again -- see L<UNUSED SPACE RECOVERY>
2149 below for details and workarounds.
2151 $db->delete("foo"); # for hashes
2152 $db->delete(1); # for arrays
2156 Deletes B<all> hash keys or array elements. Takes no arguments. No return
2157 value. Please note that the space occupied by the deleted keys/values or
2158 elements is B<not> reused again -- see L<UNUSED SPACE RECOVERY> below for
2159 details and workarounds.
2161 $db->clear(); # hashes or arrays
2167 For hashes, DBM::Deep supports all the common methods described above, and the
2168 following additional methods: C<first_key()> and C<next_key()>.
2174 Returns the "first" key in the hash. As with built-in Perl hashes, keys are
2175 fetched in an undefined order (which appears random). Takes no arguments,
2176 returns the key as a scalar value.
2178 my $key = $db->first_key();
2182 Returns the "next" key in the hash, given the previous one as the sole argument.
2183 Returns undef if there are no more keys to be fetched.
2185 $key = $db->next_key($key);
2189 Here are some examples of using hashes:
2191 my $db = DBM::Deep->new( "foo.db" );
2193 $db->put("foo", "bar");
2194 print "foo: " . $db->get("foo") . "\n";
2196 $db->put("baz", {}); # new child hash ref
2197 $db->get("baz")->put("buz", "biz");
2198 print "buz: " . $db->get("baz")->get("buz") . "\n";
2200 my $key = $db->first_key();
2202 print "$key: " . $db->get($key) . "\n";
2203 $key = $db->next_key($key);
2206 if ($db->exists("foo")) { $db->delete("foo"); }
2210 For arrays, DBM::Deep supports all the common methods described above, and the
2211 following additional methods: C<length()>, C<push()>, C<pop()>, C<shift()>,
2212 C<unshift()> and C<splice()>.
2218 Returns the number of elements in the array. Takes no arguments.
2220 my $len = $db->length();
2224 Adds one or more elements onto the end of the array. Accepts scalars, hash
2225 refs or array refs. No return value.
2227 $db->push("foo", "bar", {});
2231 Fetches the last element in the array, and deletes it. Takes no arguments.
2232 Returns undef if array is empty. Returns the element value.
2234 my $elem = $db->pop();
2238 Fetches the first element in the array, deletes it, then shifts all the
2239 remaining elements over to take up the space. Returns the element value. This
2240 method is not recommended with large arrays -- see L<LARGE ARRAYS> below for
2243 my $elem = $db->shift();
2247 Inserts one or more elements onto the beginning of the array, shifting all
2248 existing elements over to make room. Accepts scalars, hash refs or array refs.
2249 No return value. This method is not recommended with large arrays -- see
2250 <LARGE ARRAYS> below for details.
2252 $db->unshift("foo", "bar", {});
2256 Performs exactly like Perl's built-in function of the same name. See L<perldoc
2257 -f splice> for usage -- it is too complicated to document here. This method is
2258 not recommended with large arrays -- see L<LARGE ARRAYS> below for details.
2262 Here are some examples of using arrays:
2264 my $db = DBM::Deep->new(
2266 type => DBM::Deep->TYPE_ARRAY
2269 $db->push("bar", "baz");
2270 $db->unshift("foo");
2273 my $len = $db->length();
2274 print "length: $len\n"; # 4
2276 for (my $k=0; $k<$len; $k++) {
2277 print "$k: " . $db->get($k) . "\n";
2280 $db->splice(1, 2, "biz", "baf");
2282 while (my $elem = shift @$db) {
2283 print "shifted: $elem\n";
2288 Enable automatic file locking by passing a true value to the C<locking>
2289 parameter when constructing your DBM::Deep object (see L<SETUP> above).
2291 my $db = DBM::Deep->new(
2296 This causes DBM::Deep to C<flock()> the underlying FileHandle object with exclusive
2297 mode for writes, and shared mode for reads. This is required if you have
2298 multiple processes accessing the same database file, to avoid file corruption.
2299 Please note that C<flock()> does NOT work for files over NFS. See L<DB OVER
2300 NFS> below for more.
2302 =head2 EXPLICIT LOCKING
2304 You can explicitly lock a database, so it remains locked for multiple
2305 transactions. This is done by calling the C<lock()> method, and passing an
2306 optional lock mode argument (defaults to exclusive mode). This is particularly
2307 useful for things like counters, where the current value needs to be fetched,
2308 then incremented, then stored again.
2311 my $counter = $db->get("counter");
2313 $db->put("counter", $counter);
2322 You can pass C<lock()> an optional argument, which specifies which mode to use
2323 (exclusive or shared). Use one of these two constants: C<DBM::Deep-E<gt>LOCK_EX>
2324 or C<DBM::Deep-E<gt>LOCK_SH>. These are passed directly to C<flock()>, and are the
2325 same as the constants defined in Perl's C<Fcntl> module.
2327 $db->lock( DBM::Deep->LOCK_SH );
2331 If you want to implement your own file locking scheme, be sure to create your
2332 DBM::Deep objects setting the C<volatile> option to true. This hints to DBM::Deep
2333 that the DB file may change between transactions. See L<LOW-LEVEL ACCESS>
2336 =head1 IMPORTING/EXPORTING
2338 You can import existing complex structures by calling the C<import()> method,
2339 and export an entire database into an in-memory structure using the C<export()>
2340 method. Both are examined here.
2344 Say you have an existing hash with nested hashes/arrays inside it. Instead of
2345 walking the structure and adding keys/elements to the database as you go,
2346 simply pass a reference to the C<import()> method. This recursively adds
2347 everything to an existing DBM::Deep object for you. Here is an example:
2352 array1 => [ "elem0", "elem1", "elem2" ],
2354 subkey1 => "subvalue1",
2355 subkey2 => "subvalue2"
2359 my $db = DBM::Deep->new( "foo.db" );
2360 $db->import( $struct );
2362 print $db->{key1} . "\n"; # prints "value1"
2364 This recursively imports the entire C<$struct> object into C<$db>, including
2365 all nested hashes and arrays. If the DBM::Deep object contains exsiting data,
2366 keys are merged with the existing ones, replacing if they already exist.
2367 The C<import()> method can be called on any database level (not just the base
2368 level), and works with both hash and array DB types.
2372 B<Note:> Make sure your existing structure has no circular references in it.
2373 These will cause an infinite loop when importing.
2377 Calling the C<export()> method on an existing DBM::Deep object will return
2378 a reference to a new in-memory copy of the database. The export is done
2379 recursively, so all nested hashes/arrays are all exported to standard Perl
2380 objects. Here is an example:
2382 my $db = DBM::Deep->new( "foo.db" );
2384 $db->{key1} = "value1";
2385 $db->{key2} = "value2";
2387 $db->{hash1}->{subkey1} = "subvalue1";
2388 $db->{hash1}->{subkey2} = "subvalue2";
2390 my $struct = $db->export();
2392 print $struct->{key1} . "\n"; # prints "value1"
2394 This makes a complete copy of the database in memory, and returns a reference
2395 to it. The C<export()> method can be called on any database level (not just
2396 the base level), and works with both hash and array DB types. Be careful of
2397 large databases -- you can store a lot more data in a DBM::Deep object than an
2398 in-memory Perl structure.
2402 B<Note:> Make sure your database has no circular references in it.
2403 These will cause an infinite loop when exporting.
2407 DBM::Deep has a number of hooks where you can specify your own Perl function
2408 to perform filtering on incoming or outgoing data. This is a perfect
2409 way to extend the engine, and implement things like real-time compression or
2410 encryption. Filtering applies to the base DB level, and all child hashes /
2411 arrays. Filter hooks can be specified when your DBM::Deep object is first
2412 constructed, or by calling the C<set_filter()> method at any time. There are
2413 four available filter hooks, described below:
2417 =item * filter_store_key
2419 This filter is called whenever a hash key is stored. It
2420 is passed the incoming key, and expected to return a transformed key.
2422 =item * filter_store_value
2424 This filter is called whenever a hash key or array element is stored. It
2425 is passed the incoming value, and expected to return a transformed value.
2427 =item * filter_fetch_key
2429 This filter is called whenever a hash key is fetched (i.e. via
2430 C<first_key()> or C<next_key()>). It is passed the transformed key,
2431 and expected to return the plain key.
2433 =item * filter_fetch_value
2435 This filter is called whenever a hash key or array element is fetched.
2436 It is passed the transformed value, and expected to return the plain value.
2440 Here are the two ways to setup a filter hook:
2442 my $db = DBM::Deep->new(
2444 filter_store_value => \&my_filter_store,
2445 filter_fetch_value => \&my_filter_fetch
2450 $db->set_filter( "filter_store_value", \&my_filter_store );
2451 $db->set_filter( "filter_fetch_value", \&my_filter_fetch );
2453 Your filter function will be called only when dealing with SCALAR keys or
2454 values. When nested hashes and arrays are being stored/fetched, filtering
2455 is bypassed. Filters are called as static functions, passed a single SCALAR
2456 argument, and expected to return a single SCALAR value. If you want to
2457 remove a filter, set the function reference to C<undef>:
2459 $db->set_filter( "filter_store_value", undef );
2461 =head2 REAL-TIME ENCRYPTION EXAMPLE
2463 Here is a working example that uses the I<Crypt::Blowfish> module to
2464 do real-time encryption / decryption of keys & values with DBM::Deep Filters.
2465 Please visit L<http://search.cpan.org/search?module=Crypt::Blowfish> for more
2466 on I<Crypt::Blowfish>. You'll also need the I<Crypt::CBC> module.
2469 use Crypt::Blowfish;
2472 my $cipher = Crypt::CBC->new({
2473 'key' => 'my secret key',
2474 'cipher' => 'Blowfish',
2476 'regenerate_key' => 0,
2477 'padding' => 'space',
2481 my $db = DBM::Deep->new(
2482 file => "foo-encrypt.db",
2483 filter_store_key => \&my_encrypt,
2484 filter_store_value => \&my_encrypt,
2485 filter_fetch_key => \&my_decrypt,
2486 filter_fetch_value => \&my_decrypt,
2489 $db->{key1} = "value1";
2490 $db->{key2} = "value2";
2491 print "key1: " . $db->{key1} . "\n";
2492 print "key2: " . $db->{key2} . "\n";
2498 return $cipher->encrypt( $_[0] );
2501 return $cipher->decrypt( $_[0] );
2504 =head2 REAL-TIME COMPRESSION EXAMPLE
2506 Here is a working example that uses the I<Compress::Zlib> module to do real-time
2507 compression / decompression of keys & values with DBM::Deep Filters.
2508 Please visit L<http://search.cpan.org/search?module=Compress::Zlib> for
2509 more on I<Compress::Zlib>.
2514 my $db = DBM::Deep->new(
2515 file => "foo-compress.db",
2516 filter_store_key => \&my_compress,
2517 filter_store_value => \&my_compress,
2518 filter_fetch_key => \&my_decompress,
2519 filter_fetch_value => \&my_decompress,
2522 $db->{key1} = "value1";
2523 $db->{key2} = "value2";
2524 print "key1: " . $db->{key1} . "\n";
2525 print "key2: " . $db->{key2} . "\n";
2531 return Compress::Zlib::memGzip( $_[0] ) ;
2534 return Compress::Zlib::memGunzip( $_[0] ) ;
2537 B<Note:> Filtering of keys only applies to hashes. Array "keys" are
2538 actually numerical index numbers, and are not filtered.
2540 =head1 ERROR HANDLING
2542 Most DBM::Deep methods return a true value for success, and call die() on
2543 failure. You can wrap calls in an eval block to catch the die. Also, the
2544 actual error message is stored in an internal scalar, which can be fetched by
2545 calling the C<error()> method.
2547 my $db = DBM::Deep->new( "foo.db" ); # create hash
2548 eval { $db->push("foo"); }; # ILLEGAL -- push is array-only call
2550 print $db->error(); # prints error message
2552 You can then call C<clear_error()> to clear the current error state.
2556 If you set the C<debug> option to true when creating your DBM::Deep object,
2557 all errors are considered NON-FATAL, and dumped to STDERR. This is only
2558 for debugging purposes.
2560 =head1 LARGEFILE SUPPORT
2562 If you have a 64-bit system, and your Perl is compiled with both LARGEFILE
2563 and 64-bit support, you I<may> be able to create databases larger than 2 GB.
2564 DBM::Deep by default uses 32-bit file offset tags, but these can be changed
2565 by calling the static C<set_pack()> method before you do anything else.
2567 DBM::Deep::set_pack(8, 'Q');
2569 This tells DBM::Deep to pack all file offsets with 8-byte (64-bit) quad words
2570 instead of 32-bit longs. After setting these values your DB files have a
2571 theoretical maximum size of 16 XB (exabytes).
2575 B<Note:> Changing these values will B<NOT> work for existing database files.
2576 Only change this for new files, and make sure it stays set consistently
2577 throughout the file's life. If you do set these values, you can no longer
2578 access 32-bit DB files. You can, however, call C<set_pack(4, 'N')> to change
2579 back to 32-bit mode.
2583 B<Note:> I have not personally tested files > 2 GB -- all my systems have
2584 only a 32-bit Perl. However, I have received user reports that this does
2587 =head1 LOW-LEVEL ACCESS
2589 If you require low-level access to the underlying FileHandle that DBM::Deep uses,
2590 you can call the C<fh()> method, which returns the handle:
2594 This method can be called on the root level of the datbase, or any child
2595 hashes or arrays. All levels share a I<root> structure, which contains things
2596 like the FileHandle, a reference counter, and all your options you specified
2597 when you created the object. You can get access to this root structure by
2598 calling the C<root()> method.
2600 my $root = $db->root();
2602 This is useful for changing options after the object has already been created,
2603 such as enabling/disabling locking, volatile or debug modes. You can also
2604 store your own temporary user data in this structure (be wary of name
2605 collision), which is then accessible from any child hash or array.
2607 =head1 CUSTOM DIGEST ALGORITHM
2609 DBM::Deep by default uses the I<Message Digest 5> (MD5) algorithm for hashing
2610 keys. However you can override this, and use another algorithm (such as SHA-256)
2611 or even write your own. But please note that DBM::Deep currently expects zero
2612 collisions, so your algorithm has to be I<perfect>, so to speak.
2613 Collision detection may be introduced in a later version.
2617 You can specify a custom digest algorithm by calling the static C<set_digest()>
2618 function, passing a reference to a subroutine, and the length of the algorithm's
2619 hashes (in bytes). This is a global static function, which affects ALL DBM::Deep
2620 objects. Here is a working example that uses a 256-bit hash from the
2621 I<Digest::SHA256> module. Please see
2622 L<http://search.cpan.org/search?module=Digest::SHA256> for more.
2627 my $context = Digest::SHA256::new(256);
2629 DBM::Deep::set_digest( \&my_digest, 32 );
2631 my $db = DBM::Deep->new( "foo-sha.db" );
2633 $db->{key1} = "value1";
2634 $db->{key2} = "value2";
2635 print "key1: " . $db->{key1} . "\n";
2636 print "key2: " . $db->{key2} . "\n";
2642 return substr( $context->hash($_[0]), 0, 32 );
2645 B<Note:> Your returned digest strings must be B<EXACTLY> the number
2646 of bytes you specify in the C<set_digest()> function (in this case 32).
2648 =head1 CIRCULAR REFERENCES
2650 DBM::Deep has B<experimental> support for circular references. Meaning you
2651 can have a nested hash key or array element that points to a parent object.
2652 This relationship is stored in the DB file, and is preserved between sessions.
2655 my $db = DBM::Deep->new( "foo.db" );
2658 $db->{circle} = $db; # ref to self
2660 print $db->{foo} . "\n"; # prints "foo"
2661 print $db->{circle}->{foo} . "\n"; # prints "foo" again
2663 One catch is, passing the object to a function that recursively walks the
2664 object tree (such as I<Data::Dumper> or even the built-in C<optimize()> or
2665 C<export()> methods) will result in an infinite loop. The other catch is,
2666 if you fetch the I<key> of a circular reference (i.e. using the C<first_key()>
2667 or C<next_key()> methods), you will get the I<target object's key>, not the
2668 ref's key. This gets even more interesting with the above example, where
2669 the I<circle> key points to the base DB object, which technically doesn't
2670 have a key. So I made DBM::Deep return "[base]" as the key name in that
2673 =head1 CAVEATS / ISSUES / BUGS
2675 This section describes all the known issues with DBM::Deep. It you have found
2676 something that is not listed here, please send e-mail to L<jhuckaby@cpan.org>.
2678 =head2 UNUSED SPACE RECOVERY
2680 One major caveat with DBM::Deep is that space occupied by existing keys and
2681 values is not recovered when they are deleted. Meaning if you keep deleting
2682 and adding new keys, your file will continuously grow. I am working on this,
2683 but in the meantime you can call the built-in C<optimize()> method from time to
2684 time (perhaps in a crontab or something) to recover all your unused space.
2686 $db->optimize(); # returns true on success
2688 This rebuilds the ENTIRE database into a new file, then moves it on top of
2689 the original. The new file will have no unused space, thus it will take up as
2690 little disk space as possible. Please note that this operation can take
2691 a long time for large files, and you need enough disk space to temporarily hold
2692 2 copies of your DB file. The temporary file is created in the same directory
2693 as the original, named with a ".tmp" extension, and is deleted when the
2694 operation completes. Oh, and if locking is enabled, the DB is automatically
2695 locked for the entire duration of the copy.
2699 B<WARNING:> Only call optimize() on the top-level node of the database, and
2700 make sure there are no child references lying around. DBM::Deep keeps a reference
2701 counter, and if it is greater than 1, optimize() will abort and return undef.
2703 =head2 AUTOVIVIFICATION
2705 Unfortunately, autovivification doesn't work with tied hashes. This appears to
2706 be a bug in Perl's tie() system, as I<Jakob Schmidt> encountered the very same
2707 issue with his I<DWH_FIle> module (see L<http://search.cpan.org/search?module=DWH_File>),
2708 and it is also mentioned in the BUGS section for the I<MLDBM> module <see
2709 L<http://search.cpan.org/search?module=MLDBM>). Basically, on a new db file,
2712 $db->{foo}->{bar} = "hello";
2714 Since "foo" doesn't exist, you cannot add "bar" to it. You end up with "foo"
2715 being an empty hash. Try this instead, which works fine:
2717 $db->{foo} = { bar => "hello" };
2719 As of Perl 5.8.7, this bug still exists. I have walked very carefully through
2720 the execution path, and Perl indeed passes an empty hash to the STORE() method.
2721 Probably a bug in Perl.
2723 =head2 FILE CORRUPTION
2725 The current level of error handling in DBM::Deep is minimal. Files I<are> checked
2726 for a 32-bit signature when opened, but other corruption in files can cause
2727 segmentation faults. DBM::Deep may try to seek() past the end of a file, or get
2728 stuck in an infinite loop depending on the level of corruption. File write
2729 operations are not checked for failure (for speed), so if you happen to run
2730 out of disk space, DBM::Deep will probably fail in a bad way. These things will
2731 be addressed in a later version of DBM::Deep.
2735 Beware of using DB files over NFS. DBM::Deep uses flock(), which works well on local
2736 filesystems, but will NOT protect you from file corruption over NFS. I've heard
2737 about setting up your NFS server with a locking daemon, then using lockf() to
2738 lock your files, but your milage may vary there as well. From what I
2739 understand, there is no real way to do it. However, if you need access to the
2740 underlying FileHandle in DBM::Deep for using some other kind of locking scheme like
2741 lockf(), see the L<LOW-LEVEL ACCESS> section above.
2743 =head2 COPYING OBJECTS
2745 Beware of copying tied objects in Perl. Very strange things can happen.
2746 Instead, use DBM::Deep's C<clone()> method which safely copies the object and
2747 returns a new, blessed, tied hash or array to the same level in the DB.
2749 my $copy = $db->clone();
2753 Beware of using C<shift()>, C<unshift()> or C<splice()> with large arrays.
2754 These functions cause every element in the array to move, which can be murder
2755 on DBM::Deep, as every element has to be fetched from disk, then stored again in
2756 a different location. This may be addressed in a later version.
2760 This section discusses DBM::Deep's speed and memory usage.
2764 Obviously, DBM::Deep isn't going to be as fast as some C-based DBMs, such as
2765 the almighty I<BerkeleyDB>. But it makes up for it in features like true
2766 multi-level hash/array support, and cross-platform FTPable files. Even so,
2767 DBM::Deep is still pretty fast, and the speed stays fairly consistent, even
2768 with huge databases. Here is some test data:
2770 Adding 1,000,000 keys to new DB file...
2772 At 100 keys, avg. speed is 2,703 keys/sec
2773 At 200 keys, avg. speed is 2,642 keys/sec
2774 At 300 keys, avg. speed is 2,598 keys/sec
2775 At 400 keys, avg. speed is 2,578 keys/sec
2776 At 500 keys, avg. speed is 2,722 keys/sec
2777 At 600 keys, avg. speed is 2,628 keys/sec
2778 At 700 keys, avg. speed is 2,700 keys/sec
2779 At 800 keys, avg. speed is 2,607 keys/sec
2780 At 900 keys, avg. speed is 2,190 keys/sec
2781 At 1,000 keys, avg. speed is 2,570 keys/sec
2782 At 2,000 keys, avg. speed is 2,417 keys/sec
2783 At 3,000 keys, avg. speed is 1,982 keys/sec
2784 At 4,000 keys, avg. speed is 1,568 keys/sec
2785 At 5,000 keys, avg. speed is 1,533 keys/sec
2786 At 6,000 keys, avg. speed is 1,787 keys/sec
2787 At 7,000 keys, avg. speed is 1,977 keys/sec
2788 At 8,000 keys, avg. speed is 2,028 keys/sec
2789 At 9,000 keys, avg. speed is 2,077 keys/sec
2790 At 10,000 keys, avg. speed is 2,031 keys/sec
2791 At 20,000 keys, avg. speed is 1,970 keys/sec
2792 At 30,000 keys, avg. speed is 2,050 keys/sec
2793 At 40,000 keys, avg. speed is 2,073 keys/sec
2794 At 50,000 keys, avg. speed is 1,973 keys/sec
2795 At 60,000 keys, avg. speed is 1,914 keys/sec
2796 At 70,000 keys, avg. speed is 2,091 keys/sec
2797 At 80,000 keys, avg. speed is 2,103 keys/sec
2798 At 90,000 keys, avg. speed is 1,886 keys/sec
2799 At 100,000 keys, avg. speed is 1,970 keys/sec
2800 At 200,000 keys, avg. speed is 2,053 keys/sec
2801 At 300,000 keys, avg. speed is 1,697 keys/sec
2802 At 400,000 keys, avg. speed is 1,838 keys/sec
2803 At 500,000 keys, avg. speed is 1,941 keys/sec
2804 At 600,000 keys, avg. speed is 1,930 keys/sec
2805 At 700,000 keys, avg. speed is 1,735 keys/sec
2806 At 800,000 keys, avg. speed is 1,795 keys/sec
2807 At 900,000 keys, avg. speed is 1,221 keys/sec
2808 At 1,000,000 keys, avg. speed is 1,077 keys/sec
2810 This test was performed on a PowerMac G4 1gHz running Mac OS X 10.3.2 & Perl
2811 5.8.1, with an 80GB Ultra ATA/100 HD spinning at 7200RPM. The hash keys and
2812 values were between 6 - 12 chars in length. The DB file ended up at 210MB.
2813 Run time was 12 min 3 sec.
2817 One of the great things about DBM::Deep is that it uses very little memory.
2818 Even with huge databases (1,000,000+ keys) you will not see much increased
2819 memory on your process. DBM::Deep relies solely on the filesystem for storing
2820 and fetching data. Here is output from I</usr/bin/top> before even opening a
2823 PID USER PRI NI SIZE RSS SHARE STAT %CPU %MEM TIME COMMAND
2824 22831 root 11 0 2716 2716 1296 R 0.0 0.2 0:07 perl
2826 Basically the process is taking 2,716K of memory. And here is the same
2827 process after storing and fetching 1,000,000 keys:
2829 PID USER PRI NI SIZE RSS SHARE STAT %CPU %MEM TIME COMMAND
2830 22831 root 14 0 2772 2772 1328 R 0.0 0.2 13:32 perl
2832 Notice the memory usage increased by only 56K. Test was performed on a 700mHz
2833 x86 box running Linux RedHat 7.2 & Perl 5.6.1.
2835 =head1 DB FILE FORMAT
2837 In case you were interested in the underlying DB file format, it is documented
2838 here in this section. You don't need to know this to use the module, it's just
2839 included for reference.
2843 DBM::Deep files always start with a 32-bit signature to identify the file type.
2844 This is at offset 0. The signature is "DPDB" in network byte order. This is
2845 checked when the file is opened.
2849 The DBM::Deep file is in a I<tagged format>, meaning each section of the file
2850 has a standard header containing the type of data, the length of data, and then
2851 the data itself. The type is a single character (1 byte), the length is a
2852 32-bit unsigned long in network byte order, and the data is, well, the data.
2853 Here is how it unfolds:
2857 Immediately after the 32-bit file signature is the I<Master Index> record.
2858 This is a standard tag header followed by 1024 bytes (in 32-bit mode) or 2048
2859 bytes (in 64-bit mode) of data. The type is I<H> for hash or I<A> for array,
2860 depending on how the DBM::Deep object was constructed.
2864 The index works by looking at a I<MD5 Hash> of the hash key (or array index
2865 number). The first 8-bit char of the MD5 signature is the offset into the
2866 index, multipled by 4 in 32-bit mode, or 8 in 64-bit mode. The value of the
2867 index element is a file offset of the next tag for the key/element in question,
2868 which is usually a I<Bucket List> tag (see below).
2872 The next tag I<could> be another index, depending on how many keys/elements
2873 exist. See L<RE-INDEXING> below for details.
2877 A I<Bucket List> is a collection of 16 MD5 hashes for keys/elements, plus
2878 file offsets to where the actual data is stored. It starts with a standard
2879 tag header, with type I<B>, and a data size of 320 bytes in 32-bit mode, or
2880 384 bytes in 64-bit mode. Each MD5 hash is stored in full (16 bytes), plus
2881 the 32-bit or 64-bit file offset for the I<Bucket> containing the actual data.
2882 When the list fills up, a I<Re-Index> operation is performed (See
2883 L<RE-INDEXING> below).
2887 A I<Bucket> is a tag containing a key/value pair (in hash mode), or a
2888 index/value pair (in array mode). It starts with a standard tag header with
2889 type I<D> for scalar data (string, binary, etc.), or it could be a nested
2890 hash (type I<H>) or array (type I<A>). The value comes just after the tag
2891 header. The size reported in the tag header is only for the value, but then,
2892 just after the value is another size (32-bit unsigned long) and then the plain
2893 key itself. Since the value is likely to be fetched more often than the plain
2894 key, I figured it would be I<slightly> faster to store the value first.
2898 If the type is I<H> (hash) or I<A> (array), the value is another I<Master Index>
2899 record for the nested structure, where the process begins all over again.
2903 After a I<Bucket List> grows to 16 records, its allocated space in the file is
2904 exhausted. Then, when another key/element comes in, the list is converted to a
2905 new index record. However, this index will look at the next char in the MD5
2906 hash, and arrange new Bucket List pointers accordingly. This process is called
2907 I<Re-Indexing>. Basically, a new index tag is created at the file EOF, and all
2908 17 (16 + new one) keys/elements are removed from the old Bucket List and
2909 inserted into the new index. Several new Bucket Lists are created in the
2910 process, as a new MD5 char from the key is being examined (it is unlikely that
2911 the keys will all share the same next char of their MD5s).
2915 Because of the way the I<MD5> algorithm works, it is impossible to tell exactly
2916 when the Bucket Lists will turn into indexes, but the first round tends to
2917 happen right around 4,000 keys. You will see a I<slight> decrease in
2918 performance here, but it picks back up pretty quick (see L<SPEED> above). Then
2919 it takes B<a lot> more keys to exhaust the next level of Bucket Lists. It's
2920 right around 900,000 keys. This process can continue nearly indefinitely --
2921 right up until the point the I<MD5> signatures start colliding with each other,
2922 and this is B<EXTREMELY> rare -- like winning the lottery 5 times in a row AND
2923 getting struck by lightning while you are walking to cash in your tickets.
2924 Theoretically, since I<MD5> hashes are 128-bit values, you I<could> have up to
2925 340,282,366,921,000,000,000,000,000,000,000,000,000 keys/elements (I believe
2926 this is 340 unodecillion, but don't quote me).
2930 When a new key/element is stored, the key (or index number) is first ran through
2931 I<Digest::MD5> to get a 128-bit signature (example, in hex:
2932 b05783b0773d894396d475ced9d2f4f6). Then, the I<Master Index> record is checked
2933 for the first char of the signature (in this case I<b>). If it does not exist,
2934 a new I<Bucket List> is created for our key (and the next 15 future keys that
2935 happen to also have I<b> as their first MD5 char). The entire MD5 is written
2936 to the I<Bucket List> along with the offset of the new I<Bucket> record (EOF at
2937 this point, unless we are replacing an existing I<Bucket>), where the actual
2938 data will be stored.
2942 Fetching an existing key/element involves getting a I<Digest::MD5> of the key
2943 (or index number), then walking along the indexes. If there are enough
2944 keys/elements in this DB level, there might be nested indexes, each linked to
2945 a particular char of the MD5. Finally, a I<Bucket List> is pointed to, which
2946 contains up to 16 full MD5 hashes. Each is checked for equality to the key in
2947 question. If we found a match, the I<Bucket> tag is loaded, where the value and
2948 plain key are stored.
2952 Fetching the plain key occurs when calling the I<first_key()> and I<next_key()>
2953 methods. In this process the indexes are walked systematically, and each key
2954 fetched in increasing MD5 order (which is why it appears random). Once the
2955 I<Bucket> is found, the value is skipped the plain key returned instead.
2956 B<Note:> Do not count on keys being fetched as if the MD5 hashes were
2957 alphabetically sorted. This only happens on an index-level -- as soon as the
2958 I<Bucket Lists> are hit, the keys will come out in the order they went in --
2959 so it's pretty much undefined how the keys will come out -- just like Perl's
2962 =head1 CODE COVERAGE
2964 I use B<Devel::Cover> to test the code coverage of my tests, below is the B<Devel::Cover> report on this
2965 module's test suite.
2967 ---------------------------- ------ ------ ------ ------ ------ ------ ------
2968 File stmt bran cond sub pod time total
2969 ---------------------------- ------ ------ ------ ------ ------ ------ ------
2970 blib/lib/DBM/Deep.pm 94.9 84.5 77.8 100.0 11.1 100.0 89.7
2971 Total 94.9 84.5 77.8 100.0 11.1 100.0 89.7
2972 ---------------------------- ------ ------ ------ ------ ------ ------ ------
2976 Joseph Huckaby, L<jhuckaby@cpan.org>
2978 Special thanks to Adam Sah and Rich Gaushell! You know why :-)
2982 perltie(1), Tie::Hash(3), Digest::MD5(3), Fcntl(3), flock(2), lockf(3), nfs(5),
2983 Digest::SHA256(3), Crypt::Blowfish(3), Compress::Zlib(3)
2987 Copyright (c) 2002-2005 Joseph Huckaby. All Rights Reserved.
2988 This is free software, you may use it and distribute it under the
2989 same terms as Perl itself.