7 # Multi-level database module for storing hash trees, arrays and simple
8 # key/value pairs into FTP-able, cross-platform binary database files.
10 # Type `perldoc DBM::Deep` for complete documentation.
14 # tie %db, 'DBM::Deep', 'my_database.db'; # standard tie() method
16 # my $db = new DBM::Deep( 'my_database.db' ); # preferred OO method
18 # $db->{my_scalar} = 'hello world';
19 # $db->{my_hash} = { larry => 'genius', hashes => 'fast' };
20 # $db->{my_array} = [ 1, 2, 3, time() ];
21 # $db->{my_complex} = [ 'hello', { perl => 'rules' }, 42, 99 ];
22 # push @{$db->{my_array}}, 'another value';
23 # my @key_list = keys %{$db->{my_hash}};
24 # print "This module " . $db->{my_complex}->[1]->{perl} . "!\n";
27 # (c) 2002-2005 Joseph Huckaby. All Rights Reserved.
28 # This program is free software; you can redistribute it and/or
29 # modify it under the same terms as Perl itself.
38 use vars qw/$VERSION/;
43 # Set to 4 and 'N' for 32-bit offset tags (default). Theoretical limit of 4 GB per file.
44 # (Perl must be compiled with largefile support for files > 2 GB)
46 # Set to 8 and 'Q' for 64-bit offsets. Theoretical limit of 16 XB per file.
47 # (Perl must be compiled with largefile and 64-bit long support)
53 # Set to 4 and 'N' for 32-bit data length prefixes. Limit of 4 GB for each key/value.
54 # Upgrading this is possible (see above) but probably not necessary. If you need
55 # more than 4 GB for a single key or value, this module is really not for you :-)
57 #my $DATA_LENGTH_SIZE = 4;
58 #my $DATA_LENGTH_PACK = 'N';
59 my ($LONG_SIZE, $LONG_PACK, $DATA_LENGTH_SIZE, $DATA_LENGTH_PACK);
62 # Maximum number of buckets per list before another level of indexing is done.
63 # Increase this value for slightly greater speed, but larger database files.
64 # DO NOT decrease this value below 16, due to risk of recursive reindex overrun.
69 # Better not adjust anything below here, unless you're me :-)
73 # Setup digest function for keys
75 my ($DIGEST_FUNC, $HASH_SIZE);
76 #my $DIGEST_FUNC = \&Digest::MD5::md5;
79 # Precalculate index and bucket sizes based on values above.
82 my ($INDEX_SIZE, $BUCKET_SIZE, $BUCKET_LIST_SIZE);
89 # Setup file and tag signatures. These should never change.
91 sub SIG_FILE () { 'DPDB' }
92 sub SIG_HASH () { 'H' }
93 sub SIG_ARRAY () { 'A' }
94 sub SIG_NULL () { 'N' }
95 sub SIG_DATA () { 'D' }
96 sub SIG_INDEX () { 'I' }
97 sub SIG_BLIST () { 'B' }
101 # Setup constants for users to pass to new()
103 sub TYPE_HASH () { return SIG_HASH; }
104 sub TYPE_ARRAY () { return SIG_ARRAY; }
108 # Class constructor method for Perl OO interface.
109 # Calls tie() and returns blessed reference to tied hash or array,
110 # providing a hybrid OO/tie interface.
114 if (scalar(@_) > 1) { $args = {@_}; }
115 else { $args = { file => shift }; }
118 # Check if we want a tied hash or array.
121 if (defined($args->{type}) && $args->{type} eq TYPE_ARRAY) {
122 tie @$self, $class, %$args;
125 tie %$self, $class, %$args;
128 return bless $self, $class;
132 my @outer_params = qw( type base_offset );
135 # Setup $self and bless into this class.
142 base_offset => length(SIG_FILE),
153 filter_store_key => undef,
154 filter_store_value => undef,
155 filter_fetch_key => undef,
156 filter_fetch_value => undef,
165 foreach my $outer_parm ( @outer_params ) {
166 next unless exists $args->{$outer_parm};
167 $self->{$outer_parm} = $args->{$outer_parm}
170 if ( exists $args->{root} ) {
171 $self->{root} = $args->{root};
174 # This is cleanup based on the fact that the $args
175 # coming in is for both the root and non-root items
176 delete $self->root->{$_} for @outer_params;
178 $self->root->{links}++;
180 if (!defined($self->fh)) { $self->_open(); }
186 sub _get_self { tied( %{$_[0]} ) || $_[0] }
190 # Tied hash constructor method, called by Perl's tie() function.
194 if (scalar(@_) > 1) { $args = {@_}; }
195 #XXX This use of ref() is bad and is a bug
196 elsif (ref($_[0])) { $args = $_[0]; }
197 else { $args = { file => shift }; }
199 $args->{type} = TYPE_HASH;
201 return $class->_init($args);
206 # Tied array constructor method, called by Perl's tie() function.
210 if (scalar(@_) > 1) { $args = {@_}; }
211 #XXX This use of ref() is bad and is a bug
212 elsif (ref($_[0])) { $args = $_[0]; }
213 else { $args = { file => shift }; }
215 $args->{type} = TYPE_ARRAY;
217 return $class->_init($args);
222 # Class deconstructor. Close file handle if there are no more refs.
224 my $self = _get_self($_[0]);
227 $self->root->{links}--;
229 if (!$self->root->{links}) {
235 return ! eval { eval("#" . substr(join("", @_), 0, 0)); 1 };
239 # Open a FileHandle to the database, create if nonexistent.
240 # Make sure file signature matches DeepDB spec.
242 my $self = _get_self($_[0]);
244 if (defined($self->fh)) { $self->_close(); }
247 if (!(-e $self->root->{file}) && $self->root->{mode} eq 'r+') {
248 my $temp = FileHandle->new( $self->root->{file}, 'w' );
251 #XXX Convert to set_fh()
252 $self->root->{fh} = FileHandle->new( $self->root->{file}, $self->root->{mode} );
253 # }; if ($@ ) { $self->_throw_error( "Received error: $@\n" ); }
254 if (! defined($self->fh)) {
255 return $self->_throw_error("Cannot open file: " . $self->root->{file} . ": $!");
258 binmode $self->fh; # for win32
259 if ($self->root->{autoflush}) {
260 $self->fh->autoflush();
264 seek($self->fh, 0, 0);
265 my $bytes_read = $self->fh->read($signature, length(SIG_FILE));
268 # File is empty -- write signature and master index
271 seek($self->fh, 0, 0);
272 $self->fh->print(SIG_FILE);
273 $self->root->{end} = length(SIG_FILE);
274 $self->_create_tag($self->base_offset, $self->type, chr(0) x $INDEX_SIZE);
276 my $plain_key = "[base]";
277 $self->fh->print( pack($DATA_LENGTH_PACK, length($plain_key)) . $plain_key );
278 $self->root->{end} += $DATA_LENGTH_SIZE + length($plain_key);
285 # Check signature was valid
287 unless ($signature eq SIG_FILE) {
289 return $self->_throw_error("Signature not found -- file is not a Deep DB");
292 $self->root->{end} = (stat($self->fh))[7];
295 # Get our type from master index signature
297 my $tag = $self->_load_tag($self->base_offset);
298 #XXX We probably also want to store the hash algorithm name and not assume anything
300 return $self->_throw_error("Corrupted file, no master index record");
302 if ($self->{type} ne $tag->{signature}) {
303 return $self->_throw_error("File type mismatch");
311 # Close database FileHandle
313 my $self = _get_self($_[0]);
314 undef $self->root->{fh};
319 # Given offset, signature and content, create tag and write to disk
321 my ($self, $offset, $sig, $content) = @_;
322 my $size = length($content);
324 seek($self->fh, $offset, 0);
325 $self->fh->print( $sig . pack($DATA_LENGTH_PACK, $size) . $content );
327 if ($offset == $self->root->{end}) {
328 $self->root->{end} += SIG_SIZE + $DATA_LENGTH_SIZE + $size;
334 offset => $offset + SIG_SIZE + $DATA_LENGTH_SIZE,
341 # Given offset, load single tag and return signature, size and data
346 seek($self->fh, $offset, 0);
347 if ($self->fh->eof()) { return undef; }
350 $self->fh->read($sig, SIG_SIZE);
353 $self->fh->read($size, $DATA_LENGTH_SIZE);
354 $size = unpack($DATA_LENGTH_PACK, $size);
357 $self->fh->read($buffer, $size);
362 offset => $offset + SIG_SIZE + $DATA_LENGTH_SIZE,
369 # Given index tag, lookup single entry in index and return .
372 my ($tag, $index) = @_;
374 my $location = unpack($LONG_PACK, substr($tag->{content}, $index * $LONG_SIZE, $LONG_SIZE) );
375 if (!$location) { return; }
377 return $self->_load_tag( $location );
382 # Adds one key/value pair to bucket list, given offset, MD5 digest of key,
383 # plain (undigested) key and value.
386 my ($tag, $md5, $plain_key, $value) = @_;
387 my $keys = $tag->{content};
391 my $is_dbm_deep = eval { $value->isa( 'DBM::Deep' ) };
392 my $internal_ref = $is_dbm_deep && ($value->root eq $self->root);
395 # Iterate through buckets, seeing if this is a new entry or a replace.
397 for (my $i=0; $i<$MAX_BUCKETS; $i++) {
398 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
399 my $subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
402 # Found empty bucket (end of list). Populate and exit loop.
406 $location = $internal_ref
407 ? $value->base_offset
408 : $self->root->{end};
410 seek($self->fh, $tag->{offset} + ($i * $BUCKET_SIZE), 0);
411 $self->fh->print( $md5 . pack($LONG_PACK, $location) );
414 elsif ($md5 eq $key) {
416 # Found existing bucket with same key. Replace with new value.
421 $location = $value->base_offset;
422 seek($self->fh, $tag->{offset} + ($i * $BUCKET_SIZE), 0);
423 $self->fh->print( $md5 . pack($LONG_PACK, $location) );
426 seek($self->fh, $subloc + SIG_SIZE, 0);
428 $self->fh->read($size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
431 # If value is a hash, array, or raw value with equal or less size, we can
432 # reuse the same content area of the database. Otherwise, we have to create
433 # a new content area at the EOF.
436 my $r = Scalar::Util::reftype( $value ) || '';
437 if ( $r eq 'HASH' || $r eq 'ARRAY' ) { $actual_length = $INDEX_SIZE; }
438 else { $actual_length = length($value); }
440 if ($actual_length <= $size) {
444 $location = $self->root->{end};
445 seek($self->fh, $tag->{offset} + ($i * $BUCKET_SIZE) + $HASH_SIZE, 0);
446 $self->fh->print( pack($LONG_PACK, $location) );
454 # If this is an internal reference, return now.
455 # No need to write value or plain key
462 # If bucket didn't fit into list, split into a new index level
465 seek($self->fh, $tag->{ref_loc}, 0);
466 $self->fh->print( pack($LONG_PACK, $self->root->{end}) );
468 my $index_tag = $self->_create_tag($self->root->{end}, SIG_INDEX, chr(0) x $INDEX_SIZE);
471 $keys .= $md5 . pack($LONG_PACK, 0);
473 for (my $i=0; $i<=$MAX_BUCKETS; $i++) {
474 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
476 my $old_subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
477 my $num = ord(substr($key, $tag->{ch} + 1, 1));
479 if ($offsets[$num]) {
480 my $offset = $offsets[$num] + SIG_SIZE + $DATA_LENGTH_SIZE;
481 seek($self->fh, $offset, 0);
483 $self->fh->read($subkeys, $BUCKET_LIST_SIZE);
485 for (my $k=0; $k<$MAX_BUCKETS; $k++) {
486 my $subloc = unpack($LONG_PACK, substr($subkeys, ($k * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
488 seek($self->fh, $offset + ($k * $BUCKET_SIZE), 0);
489 $self->fh->print( $key . pack($LONG_PACK, $old_subloc || $self->root->{end}) );
495 $offsets[$num] = $self->root->{end};
496 seek($self->fh, $index_tag->{offset} + ($num * $LONG_SIZE), 0);
497 $self->fh->print( pack($LONG_PACK, $self->root->{end}) );
499 my $blist_tag = $self->_create_tag($self->root->{end}, SIG_BLIST, chr(0) x $BUCKET_LIST_SIZE);
501 seek($self->fh, $blist_tag->{offset}, 0);
502 $self->fh->print( $key . pack($LONG_PACK, $old_subloc || $self->root->{end}) );
507 $location ||= $self->root->{end};
508 } # re-index bucket list
511 # Seek to content area and store signature, value and plaintext key
515 seek($self->fh, $location, 0);
518 # Write signature based on content type, set content length and write actual value.
520 my $r = Scalar::Util::reftype($value) || '';
522 $self->fh->print( TYPE_HASH );
523 $self->fh->print( pack($DATA_LENGTH_PACK, $INDEX_SIZE) . chr(0) x $INDEX_SIZE );
524 $content_length = $INDEX_SIZE;
526 elsif ($r eq 'ARRAY') {
527 $self->fh->print( TYPE_ARRAY );
528 $self->fh->print( pack($DATA_LENGTH_PACK, $INDEX_SIZE) . chr(0) x $INDEX_SIZE );
529 $content_length = $INDEX_SIZE;
531 elsif (!defined($value)) {
532 $self->fh->print( SIG_NULL );
533 $self->fh->print( pack($DATA_LENGTH_PACK, 0) );
537 $self->fh->print( SIG_DATA );
538 $self->fh->print( pack($DATA_LENGTH_PACK, length($value)) . $value );
539 $content_length = length($value);
543 # Plain key is stored AFTER value, as keys are typically fetched less often.
545 $self->fh->print( pack($DATA_LENGTH_PACK, length($plain_key)) . $plain_key );
548 # If value is blessed, preserve class name
550 if ( $self->root->{autobless} ) {
551 my $value_class = Scalar::Util::blessed($value);
552 if ( defined $value_class && $value_class ne 'DBM::Deep' ) {
554 # Blessed ref -- will restore later
556 $self->fh->print( chr(1) );
557 $self->fh->print( pack($DATA_LENGTH_PACK, length($value_class)) . $value_class );
558 $content_length += 1;
559 $content_length += $DATA_LENGTH_SIZE + length($value_class);
562 $self->fh->print( chr(0) );
563 $content_length += 1;
568 # If this is a new content area, advance EOF counter
570 if ($location == $self->root->{end}) {
571 $self->root->{end} += SIG_SIZE;
572 $self->root->{end} += $DATA_LENGTH_SIZE + $content_length;
573 $self->root->{end} += $DATA_LENGTH_SIZE + length($plain_key);
577 # If content is a hash or array, create new child DeepDB object and
578 # pass each key or element to it.
581 my $branch = DBM::Deep->new(
583 base_offset => $location,
586 foreach my $key (keys %{$value}) {
587 $branch->{$key} = $value->{$key};
590 elsif ($r eq 'ARRAY') {
591 my $branch = DBM::Deep->new(
593 base_offset => $location,
597 foreach my $element (@{$value}) {
598 $branch->[$index] = $element;
606 return $self->_throw_error("Fatal error: indexing failed -- possibly due to corruption in file");
609 sub _get_bucket_value {
611 # Fetch single value given tag and MD5 digested key.
614 my ($tag, $md5) = @_;
615 my $keys = $tag->{content};
618 # Iterate through buckets, looking for a key match
621 for (my $i=0; $i<$MAX_BUCKETS; $i++) {
622 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
623 my $subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
627 # Hit end of list, no match
632 if ( $md5 ne $key ) {
637 # Found match -- seek to offset and read signature
640 seek($self->fh, $subloc, 0);
641 $self->fh->read($signature, SIG_SIZE);
644 # If value is a hash or array, return new DeepDB object with correct offset
646 if (($signature eq TYPE_HASH) || ($signature eq TYPE_ARRAY)) {
647 my $obj = DBM::Deep->new(
649 base_offset => $subloc,
653 if ($self->root->{autobless}) {
655 # Skip over value and plain key to see if object needs
658 seek($self->fh, $DATA_LENGTH_SIZE + $INDEX_SIZE, 1);
661 $self->fh->read($size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
662 if ($size) { seek($self->fh, $size, 1); }
665 $self->fh->read($bless_bit, 1);
666 if (ord($bless_bit)) {
668 # Yes, object needs to be re-blessed
671 $self->fh->read($size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
672 if ($size) { $self->fh->read($class_name, $size); }
673 if ($class_name) { $obj = bless( $obj, $class_name ); }
681 # Otherwise return actual value
683 elsif ($signature eq SIG_DATA) {
686 $self->fh->read($size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
687 if ($size) { $self->fh->read($value, $size); }
692 # Key exists, but content is null
702 # Delete single key/value pair given tag and MD5 digested key.
705 my ($tag, $md5) = @_;
706 my $keys = $tag->{content};
709 # Iterate through buckets, looking for a key match
712 for (my $i=0; $i<$MAX_BUCKETS; $i++) {
713 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
714 my $subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
718 # Hit end of list, no match
723 if ( $md5 ne $key ) {
728 # Matched key -- delete bucket and return
730 seek($self->fh, $tag->{offset} + ($i * $BUCKET_SIZE), 0);
731 $self->fh->print( substr($keys, ($i+1) * $BUCKET_SIZE ) );
732 $self->fh->print( chr(0) x $BUCKET_SIZE );
742 # Check existence of single key given tag and MD5 digested key.
745 my ($tag, $md5) = @_;
746 my $keys = $tag->{content};
749 # Iterate through buckets, looking for a key match
752 for (my $i=0; $i<$MAX_BUCKETS; $i++) {
753 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
754 my $subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
758 # Hit end of list, no match
763 if ( $md5 ne $key ) {
768 # Matched key -- return true
776 sub _find_bucket_list {
778 # Locate offset for bucket list, given digested key
784 # Locate offset for bucket list using digest index system
787 my $tag = $self->_load_tag($self->base_offset);
788 if (!$tag) { return; }
790 while ($tag->{signature} ne SIG_BLIST) {
791 $tag = $self->_index_lookup($tag, ord(substr($md5, $ch, 1)));
792 if (!$tag) { return; }
799 sub _traverse_index {
801 # Scan index and recursively step into deeper levels, looking for next key.
803 my ($self, $offset, $ch, $force_return_next) = @_;
804 $force_return_next = undef unless $force_return_next;
806 my $tag = $self->_load_tag( $offset );
808 if ($tag->{signature} ne SIG_BLIST) {
809 my $content = $tag->{content};
811 if ($self->{return_next}) { $start = 0; }
812 else { $start = ord(substr($self->{prev_md5}, $ch, 1)); }
814 for (my $index = $start; $index < 256; $index++) {
815 my $subloc = unpack($LONG_PACK, substr($content, $index * $LONG_SIZE, $LONG_SIZE) );
817 my $result = $self->_traverse_index( $subloc, $ch + 1, $force_return_next );
818 if (defined($result)) { return $result; }
822 $self->{return_next} = 1;
825 elsif ($tag->{signature} eq SIG_BLIST) {
826 my $keys = $tag->{content};
827 if ($force_return_next) { $self->{return_next} = 1; }
830 # Iterate through buckets, looking for a key match
832 for (my $i=0; $i<$MAX_BUCKETS; $i++) {
833 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
834 my $subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
838 # End of bucket list -- return to outer loop
840 $self->{return_next} = 1;
843 elsif ($key eq $self->{prev_md5}) {
845 # Located previous key -- return next one found
847 $self->{return_next} = 1;
850 elsif ($self->{return_next}) {
852 # Seek to bucket location and skip over signature
854 seek($self->fh, $subloc + SIG_SIZE, 0);
857 # Skip over value to get to plain key
860 $self->fh->read($size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
861 if ($size) { seek($self->fh, $size, 1); }
864 # Read in plain key and return as scalar
867 $self->fh->read($size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
868 if ($size) { $self->fh->read($plain_key, $size); }
874 $self->{return_next} = 1;
875 } # tag is a bucket list
882 # Locate next key, given digested previous one
884 my $self = _get_self($_[0]);
886 $self->{prev_md5} = $_[1] ? $_[1] : undef;
887 $self->{return_next} = 0;
890 # If the previous key was not specifed, start at the top and
891 # return the first one found.
893 if (!$self->{prev_md5}) {
894 $self->{prev_md5} = chr(0) x $HASH_SIZE;
895 $self->{return_next} = 1;
898 return $self->_traverse_index( $self->base_offset, 0 );
903 # If db locking is set, flock() the db file. If called multiple
904 # times before unlock(), then the same number of unlocks() must
905 # be called before the lock is released.
907 my $self = _get_self($_[0]);
909 $type = LOCK_EX unless defined $type;
911 if ($self->root->{locking}) {
912 if (!$self->root->{locked}) { flock($self->fh, $type); }
913 $self->root->{locked}++;
919 # If db locking is set, unlock the db file. See note in lock()
920 # regarding calling lock() multiple times.
922 my $self = _get_self($_[0]);
924 if ($self->root->{locking} && $self->root->{locked} > 0) {
925 $self->root->{locked}--;
926 if (!$self->root->{locked}) { flock($self->fh, LOCK_UN); }
930 #XXX These uses of ref() need verified
933 # Copy single level of keys or elements to new DB handle.
934 # Recurse for nested structures
936 my $self = _get_self($_[0]);
939 if ($self->type eq TYPE_HASH) {
940 my $key = $self->first_key();
942 my $value = $self->get($key);
943 #XXX This doesn't work with autobless
944 if (!ref($value)) { $db_temp->{$key} = $value; }
946 my $type = $value->type;
947 if ($type eq TYPE_HASH) { $db_temp->{$key} = {}; }
948 else { $db_temp->{$key} = []; }
949 $value->_copy_node( $db_temp->{$key} );
951 $key = $self->next_key($key);
955 my $length = $self->length();
956 for (my $index = 0; $index < $length; $index++) {
957 my $value = $self->get($index);
958 if (!ref($value)) { $db_temp->[$index] = $value; }
959 #XXX NO tests for this code
961 my $type = $value->type;
962 if ($type eq TYPE_HASH) { $db_temp->[$index] = {}; }
963 else { $db_temp->[$index] = []; }
964 $value->_copy_node( $db_temp->[$index] );
972 # Recursively export into standard Perl hashes and arrays.
974 my $self = _get_self($_[0]);
977 if ($self->type eq TYPE_HASH) { $temp = {}; }
978 elsif ($self->type eq TYPE_ARRAY) { $temp = []; }
981 $self->_copy_node( $temp );
989 # Recursively import Perl hash/array structure
991 #XXX This use of ref() seems to be ok
992 if (!ref($_[0])) { return; } # Perl calls import() on use -- ignore
994 my $self = _get_self($_[0]);
997 #XXX This use of ref() seems to be ok
1000 # struct is not a reference, so just import based on our type
1004 if ($self->type eq TYPE_HASH) { $struct = {@_}; }
1005 elsif ($self->type eq TYPE_ARRAY) { $struct = [@_]; }
1008 my $r = Scalar::Util::reftype($struct) || '';
1009 if ($r eq "HASH" && $self->type eq TYPE_HASH) {
1010 foreach my $key (keys %$struct) { $self->put($key, $struct->{$key}); }
1012 elsif ($r eq "ARRAY" && $self->type eq TYPE_ARRAY) {
1013 $self->push( @$struct );
1016 return $self->_throw_error("Cannot import: type mismatch");
1024 # Rebuild entire database into new file, then move
1025 # it back on top of original.
1027 my $self = _get_self($_[0]);
1028 if ($self->root->{links} > 1) {
1029 return $self->_throw_error("Cannot optimize: reference count is greater than 1");
1032 my $db_temp = DBM::Deep->new(
1033 file => $self->root->{file} . '.tmp',
1037 return $self->_throw_error("Cannot optimize: failed to open temp file: $!");
1041 $self->_copy_node( $db_temp );
1045 # Attempt to copy user, group and permissions over to new file
1047 my @stats = stat($self->fh);
1048 my $perms = $stats[2] & 07777;
1049 my $uid = $stats[4];
1050 my $gid = $stats[5];
1051 chown( $uid, $gid, $self->root->{file} . '.tmp' );
1052 chmod( $perms, $self->root->{file} . '.tmp' );
1054 # q.v. perlport for more information on this variable
1055 if ( $^O eq 'MSWin32' ) {
1057 # Potential race condition when optmizing on Win32 with locking.
1058 # The Windows filesystem requires that the filehandle be closed
1059 # before it is overwritten with rename(). This could be redone
1066 if (!rename $self->root->{file} . '.tmp', $self->root->{file}) {
1067 unlink $self->root->{file} . '.tmp';
1069 return $self->_throw_error("Optimize failed: Cannot copy temp file over original: $!");
1081 # Make copy of object and return
1083 my $self = _get_self($_[0]);
1085 return DBM::Deep->new(
1086 type => $self->type,
1087 base_offset => $self->base_offset,
1093 my %is_legal_filter = map {
1096 store_key store_value
1097 fetch_key fetch_value
1102 # Setup filter function for storing or fetching the key or value
1104 my $self = _get_self($_[0]);
1105 my $type = lc $_[1];
1106 my $func = $_[2] ? $_[2] : undef;
1108 if ( $is_legal_filter{$type} ) {
1109 $self->root->{"filter_$type"} = $func;
1123 # Get access to the root structure
1125 my $self = _get_self($_[0]);
1126 return $self->{root};
1131 # Get access to the raw FileHandle
1133 #XXX It will be useful, though, when we split out HASH and ARRAY
1134 my $self = _get_self($_[0]);
1135 return $self->root->{fh};
1140 # Get type of current node (TYPE_HASH or TYPE_ARRAY)
1142 my $self = _get_self($_[0]);
1143 return $self->{type};
1148 # Get base_offset of current node (TYPE_HASH or TYPE_ARRAY)
1150 my $self = _get_self($_[0]);
1151 return $self->{base_offset};
1156 # Get last error string, or undef if no error
1159 ? ( _get_self($_[0])->{root}->{error} or undef )
1169 # Store error string in self
1171 my $self = _get_self($_[0]);
1172 my $error_text = $_[1];
1174 $self->root->{error} = $error_text;
1176 unless ($self->root->{debug}) {
1177 die "DBM::Deep: $error_text\n";
1180 warn "DBM::Deep: $error_text\n";
1188 my $self = _get_self($_[0]);
1190 undef $self->root->{error};
1195 # Precalculate index, bucket and bucket list sizes
1198 #XXX I don't like this ...
1199 set_pack() unless defined $LONG_SIZE;
1201 $INDEX_SIZE = 256 * $LONG_SIZE;
1202 $BUCKET_SIZE = $HASH_SIZE + $LONG_SIZE;
1203 $BUCKET_LIST_SIZE = $MAX_BUCKETS * $BUCKET_SIZE;
1208 # Set pack/unpack modes (see file header for more)
1210 my ($long_s, $long_p, $data_s, $data_p) = @_;
1212 $LONG_SIZE = $long_s ? $long_s : 4;
1213 $LONG_PACK = $long_p ? $long_p : 'N';
1215 $DATA_LENGTH_SIZE = $data_s ? $data_s : 4;
1216 $DATA_LENGTH_PACK = $data_p ? $data_p : 'N';
1223 # Set key digest function (default is MD5)
1225 my ($digest_func, $hash_size) = @_;
1227 $DIGEST_FUNC = $digest_func ? $digest_func : \&Digest::MD5::md5;
1228 $HASH_SIZE = $hash_size ? $hash_size : 16;
1234 # tie() methods (hashes and arrays)
1239 # Store single hash key/value or array element in database.
1241 my $self = _get_self($_[0]);
1242 my $key = ($self->root->{filter_store_key} && $self->type eq TYPE_HASH) ? $self->root->{filter_store_key}->($_[1]) : $_[1];
1243 #XXX What is ref() checking here?
1244 #YYY User may be storing a hash, in which case we do not want it run
1245 #YYY through the filtering system
1246 my $value = ($self->root->{filter_store_value} && !ref($_[2])) ? $self->root->{filter_store_value}->($_[2]) : $_[2];
1248 my $unpacked_key = $key;
1249 if (($self->type eq TYPE_ARRAY) && ($key =~ /^\d+$/)) { $key = pack($LONG_PACK, $key); }
1250 my $md5 = $DIGEST_FUNC->($key);
1253 # Make sure file is open
1255 if (!defined($self->fh) && !$self->_open()) {
1260 # Request exclusive lock for writing
1262 $self->lock( LOCK_EX );
1265 # If locking is enabled, set 'end' parameter again, in case another
1266 # DB instance appended to our file while we were unlocked.
1268 if ($self->root->{locking} || $self->root->{volatile}) {
1269 $self->root->{end} = (stat($self->fh))[7];
1273 # Locate offset for bucket list using digest index system
1275 my $tag = $self->_load_tag($self->base_offset);
1277 $tag = $self->_create_tag($self->base_offset, SIG_INDEX, chr(0) x $INDEX_SIZE);
1281 while ($tag->{signature} ne SIG_BLIST) {
1282 my $num = ord(substr($md5, $ch, 1));
1283 my $new_tag = $self->_index_lookup($tag, $num);
1285 my $ref_loc = $tag->{offset} + ($num * $LONG_SIZE);
1286 seek($self->fh, $ref_loc, 0);
1287 $self->fh->print( pack($LONG_PACK, $self->root->{end}) );
1289 $tag = $self->_create_tag($self->root->{end}, SIG_BLIST, chr(0) x $BUCKET_LIST_SIZE);
1290 $tag->{ref_loc} = $ref_loc;
1295 my $ref_loc = $tag->{offset} + ($num * $LONG_SIZE);
1297 $tag->{ref_loc} = $ref_loc;
1304 # Add key/value to bucket list
1306 my $result = $self->_add_bucket( $tag, $md5, $key, $value );
1309 # If this object is an array, and bucket was not a replace, and key is numerical,
1310 # and index is equal or greater than current length, advance length variable.
1312 if (($result == 2) && ($self->type eq TYPE_ARRAY) && ($unpacked_key =~ /^\d+$/) && ($unpacked_key >= $self->FETCHSIZE())) {
1313 $self->STORESIZE( $unpacked_key + 1 );
1323 # Fetch single value or element given plain key or array index
1325 my $self = _get_self($_[0]);
1328 if ( $self->type eq TYPE_HASH ) {
1329 if ( my $filter = $self->root->{filter_store_key} ) {
1330 $key = $filter->( $key );
1333 elsif ( $self->type eq TYPE_ARRAY ) {
1334 if ( $key =~ /^\d+$/ ) {
1335 $key = pack($LONG_PACK, $key);
1339 my $md5 = $DIGEST_FUNC->($key);
1342 # Make sure file is open
1344 if (!defined($self->fh)) { $self->_open(); }
1347 # Request shared lock for reading
1349 $self->lock( LOCK_SH );
1351 my $tag = $self->_find_bucket_list( $md5 );
1358 # Get value from bucket list
1360 my $result = $self->_get_bucket_value( $tag, $md5 );
1364 #XXX What is ref() checking here?
1365 return ($result && !ref($result) && $self->root->{filter_fetch_value}) ? $self->root->{filter_fetch_value}->($result) : $result;
1370 # Delete single key/value pair or element given plain key or array index
1372 my $self = _get_self($_[0]);
1373 my $key = ($self->root->{filter_store_key} && $self->type eq TYPE_HASH) ? $self->root->{filter_store_key}->($_[1]) : $_[1];
1375 my $unpacked_key = $key;
1376 if (($self->type eq TYPE_ARRAY) && ($key =~ /^\d+$/)) { $key = pack($LONG_PACK, $key); }
1377 my $md5 = $DIGEST_FUNC->($key);
1380 # Make sure file is open
1382 if (!defined($self->fh)) { $self->_open(); }
1385 # Request exclusive lock for writing
1387 $self->lock( LOCK_EX );
1389 my $tag = $self->_find_bucket_list( $md5 );
1398 my $result = $self->_delete_bucket( $tag, $md5 );
1401 # If this object is an array and the key deleted was on the end of the stack,
1402 # decrement the length variable.
1404 if ($result && ($self->type eq TYPE_ARRAY) && ($unpacked_key == $self->FETCHSIZE() - 1)) {
1405 $self->STORESIZE( $unpacked_key );
1415 # Check if a single key or element exists given plain key or array index
1417 my $self = _get_self($_[0]);
1418 my $key = ($self->root->{filter_store_key} && $self->type eq TYPE_HASH) ? $self->root->{filter_store_key}->($_[1]) : $_[1];
1420 if (($self->type eq TYPE_ARRAY) && ($key =~ /^\d+$/)) { $key = pack($LONG_PACK, $key); }
1421 my $md5 = $DIGEST_FUNC->($key);
1424 # Make sure file is open
1426 if (!defined($self->fh)) { $self->_open(); }
1429 # Request shared lock for reading
1431 $self->lock( LOCK_SH );
1433 my $tag = $self->_find_bucket_list( $md5 );
1436 # For some reason, the built-in exists() function returns '' for false
1444 # Check if bucket exists and return 1 or ''
1446 my $result = $self->_bucket_exists( $tag, $md5 ) || '';
1455 # Clear all keys from hash, or all elements from array.
1457 my $self = _get_self($_[0]);
1460 # Make sure file is open
1462 if (!defined($self->fh)) { $self->_open(); }
1465 # Request exclusive lock for writing
1467 $self->lock( LOCK_EX );
1469 seek($self->fh, $self->base_offset, 0);
1470 if ($self->fh->eof()) {
1475 $self->_create_tag($self->base_offset, $self->type, chr(0) x $INDEX_SIZE);
1484 # Locate and return first key (in no particular order)
1486 my $self = _get_self($_[0]);
1487 if ($self->type ne TYPE_HASH) {
1488 return $self->_throw_error("FIRSTKEY method only supported for hashes");
1492 # Make sure file is open
1494 if (!defined($self->fh)) { $self->_open(); }
1497 # Request shared lock for reading
1499 $self->lock( LOCK_SH );
1501 my $result = $self->_get_next_key();
1505 return ($result && $self->root->{filter_fetch_key}) ? $self->root->{filter_fetch_key}->($result) : $result;
1510 # Return next key (in no particular order), given previous one
1512 my $self = _get_self($_[0]);
1513 if ($self->type ne TYPE_HASH) {
1514 return $self->_throw_error("NEXTKEY method only supported for hashes");
1516 my $prev_key = ($self->root->{filter_store_key} && $self->type eq TYPE_HASH) ? $self->root->{filter_store_key}->($_[1]) : $_[1];
1517 my $prev_md5 = $DIGEST_FUNC->($prev_key);
1520 # Make sure file is open
1522 if (!defined($self->fh)) { $self->_open(); }
1525 # Request shared lock for reading
1527 $self->lock( LOCK_SH );
1529 my $result = $self->_get_next_key( $prev_md5 );
1533 return ($result && $self->root->{filter_fetch_key}) ? $self->root->{filter_fetch_key}->($result) : $result;
1537 # The following methods are for arrays only
1542 # Return the length of the array
1544 my $self = _get_self($_[0]);
1545 if ($self->type ne TYPE_ARRAY) {
1546 return $self->_throw_error("FETCHSIZE method only supported for arrays");
1549 my $SAVE_FILTER = $self->root->{filter_fetch_value};
1550 $self->root->{filter_fetch_value} = undef;
1552 my $packed_size = $self->FETCH('length');
1554 $self->root->{filter_fetch_value} = $SAVE_FILTER;
1556 if ($packed_size) { return int(unpack($LONG_PACK, $packed_size)); }
1562 # Set the length of the array
1564 my $self = _get_self($_[0]);
1565 if ($self->type ne TYPE_ARRAY) {
1566 return $self->_throw_error("STORESIZE method only supported for arrays");
1568 my $new_length = $_[1];
1570 my $SAVE_FILTER = $self->root->{filter_store_value};
1571 $self->root->{filter_store_value} = undef;
1573 my $result = $self->STORE('length', pack($LONG_PACK, $new_length));
1575 $self->root->{filter_store_value} = $SAVE_FILTER;
1582 # Remove and return the last element on the array
1584 my $self = _get_self($_[0]);
1585 if ($self->type ne TYPE_ARRAY) {
1586 return $self->_throw_error("POP method only supported for arrays");
1588 my $length = $self->FETCHSIZE();
1591 my $content = $self->FETCH( $length - 1 );
1592 $self->DELETE( $length - 1 );
1602 # Add new element(s) to the end of the array
1604 my $self = _get_self(shift);
1605 if ($self->type ne TYPE_ARRAY) {
1606 return $self->_throw_error("PUSH method only supported for arrays");
1608 my $length = $self->FETCHSIZE();
1610 while (my $content = shift @_) {
1611 $self->STORE( $length, $content );
1618 # Remove and return first element on the array.
1619 # Shift over remaining elements to take up space.
1621 my $self = _get_self($_[0]);
1622 if ($self->type ne TYPE_ARRAY) {
1623 return $self->_throw_error("SHIFT method only supported for arrays");
1625 my $length = $self->FETCHSIZE();
1628 my $content = $self->FETCH( 0 );
1631 # Shift elements over and remove last one.
1633 for (my $i = 0; $i < $length - 1; $i++) {
1634 $self->STORE( $i, $self->FETCH($i + 1) );
1636 $self->DELETE( $length - 1 );
1647 # Insert new element(s) at beginning of array.
1648 # Shift over other elements to make space.
1650 my $self = _get_self($_[0]);shift @_;
1651 if ($self->type ne TYPE_ARRAY) {
1652 return $self->_throw_error("UNSHIFT method only supported for arrays");
1654 my @new_elements = @_;
1655 my $length = $self->FETCHSIZE();
1656 my $new_size = scalar @new_elements;
1659 for (my $i = $length - 1; $i >= 0; $i--) {
1660 $self->STORE( $i + $new_size, $self->FETCH($i) );
1664 for (my $i = 0; $i < $new_size; $i++) {
1665 $self->STORE( $i, $new_elements[$i] );
1671 # Splices section of array with optional new section.
1672 # Returns deleted section, or last element deleted in scalar context.
1674 my $self = _get_self($_[0]);shift @_;
1675 if ($self->type ne TYPE_ARRAY) {
1676 return $self->_throw_error("SPLICE method only supported for arrays");
1678 my $length = $self->FETCHSIZE();
1681 # Calculate offset and length of splice
1683 my $offset = shift || 0;
1684 if ($offset < 0) { $offset += $length; }
1687 if (scalar @_) { $splice_length = shift; }
1688 else { $splice_length = $length - $offset; }
1689 if ($splice_length < 0) { $splice_length += ($length - $offset); }
1692 # Setup array with new elements, and copy out old elements for return
1694 my @new_elements = @_;
1695 my $new_size = scalar @new_elements;
1697 my @old_elements = ();
1698 for (my $i = $offset; $i < $offset + $splice_length; $i++) {
1699 push @old_elements, $self->FETCH( $i );
1703 # Adjust array length, and shift elements to accomodate new section.
1705 if ( $new_size != $splice_length ) {
1706 if ($new_size > $splice_length) {
1707 for (my $i = $length - 1; $i >= $offset + $splice_length; $i--) {
1708 $self->STORE( $i + ($new_size - $splice_length), $self->FETCH($i) );
1712 for (my $i = $offset + $splice_length; $i < $length; $i++) {
1713 $self->STORE( $i + ($new_size - $splice_length), $self->FETCH($i) );
1715 for (my $i = 0; $i < $splice_length - $new_size; $i++) {
1716 $self->DELETE( $length - 1 );
1723 # Insert new elements into array
1725 for (my $i = $offset; $i < $offset + $new_size; $i++) {
1726 $self->STORE( $i, shift @new_elements );
1730 # Return deleted section, or last element in scalar context.
1732 return wantarray ? @old_elements : $old_elements[-1];
1735 #XXX We don't need to define it.
1736 #XXX It will be useful, though, when we split out HASH and ARRAY
1739 # Perl will call EXTEND() when the array is likely to grow.
1740 # We don't care, but include it for compatibility.
1745 # Public method aliases
1747 *put = *store = *STORE;
1748 *get = *fetch = *FETCH;
1752 *first_key = *FIRSTKEY;
1753 *next_key = *NEXTKEY;
1754 *length = *FETCHSIZE;
1758 *unshift = *UNSHIFT;
1767 DBM::Deep - A pure perl multi-level hash/array DBM
1772 my $db = DBM::Deep->new( "foo.db" );
1774 $db->{key} = 'value'; # tie() style
1777 $db->put('key', 'value'); # OO style
1778 print $db->get('key');
1780 # true multi-level support
1781 $db->{my_complex} = [
1782 'hello', { perl => 'rules' },
1787 A unique flat-file database module, written in pure perl. True
1788 multi-level hash/array support (unlike MLDBM, which is faked), hybrid
1789 OO / tie() interface, cross-platform FTPable files, and quite fast. Can
1790 handle millions of keys and unlimited hash levels without significant
1791 slow-down. Written from the ground-up in pure perl -- this is NOT a
1792 wrapper around a C-based DBM. Out-of-the-box compatibility with Unix,
1793 Mac OS X and Windows.
1797 Hopefully you are using CPAN's excellent Perl module, which will download
1798 and install the module for you. If not, get the tarball, and run these
1810 Construction can be done OO-style (which is the recommended way), or using
1811 Perl's tie() function. Both are examined here.
1813 =head2 OO CONSTRUCTION
1815 The recommended way to construct a DBM::Deep object is to use the new()
1816 method, which gets you a blessed, tied hash or array reference.
1818 my $db = DBM::Deep->new( "foo.db" );
1820 This opens a new database handle, mapped to the file "foo.db". If this
1821 file does not exist, it will automatically be created. DB files are
1822 opened in "r+" (read/write) mode, and the type of object returned is a
1823 hash, unless otherwise specified (see L<OPTIONS> below).
1827 You can pass a number of options to the constructor to specify things like
1828 locking, autoflush, etc. This is done by passing an inline hash:
1830 my $db = DBM::Deep->new(
1836 Notice that the filename is now specified I<inside> the hash with
1837 the "file" parameter, as opposed to being the sole argument to the
1838 constructor. This is required if any options are specified.
1839 See L<OPTIONS> below for the complete list.
1843 You can also start with an array instead of a hash. For this, you must
1844 specify the C<type> parameter:
1846 my $db = DBM::Deep->new(
1848 type => DBM::Deep->TYPE_ARRAY
1851 B<Note:> Specifing the C<type> parameter only takes effect when beginning
1852 a new DB file. If you create a DBM::Deep object with an existing file, the
1853 C<type> will be loaded from the file header, and ignored if it is passed
1856 =head2 TIE CONSTRUCTION
1858 Alternatively, you can create a DBM::Deep handle by using Perl's built-in
1859 tie() function. This is not ideal, because you get only a basic, tied hash
1860 (or array) which is not blessed, so you can't call any functions on it.
1863 tie %hash, "DBM::Deep", "foo.db";
1866 tie @array, "DBM::Deep", "bar.db";
1868 As with the OO constructor, you can replace the DB filename parameter with
1869 a hash containing one or more options (see L<OPTIONS> just below for the
1872 tie %hash, "DBM::Deep", {
1880 There are a number of options that can be passed in when constructing your
1881 DBM::Deep objects. These apply to both the OO- and tie- based approaches.
1887 Filename of the DB file to link the handle to. You can pass a full absolute
1888 filesystem path, partial path, or a plain filename if the file is in the
1889 current working directory. This is a required parameter.
1893 File open mode (read-only, read-write, etc.) string passed to Perl's FileHandle
1894 module. This is an optional parameter, and defaults to "r+" (read/write).
1895 B<Note:> If the default (r+) mode is selected, the file will also be auto-
1896 created if it doesn't exist.
1900 This parameter specifies what type of object to create, a hash or array. Use
1901 one of these two constants: C<DBM::Deep-E<gt>TYPE_HASH> or C<DBM::Deep-E<gt>TYPE_ARRAY>.
1902 This only takes effect when beginning a new file. This is an optional
1903 parameter, and defaults to C<DBM::Deep-E<gt>TYPE_HASH>.
1907 Specifies whether locking is to be enabled. DBM::Deep uses Perl's Fnctl flock()
1908 function to lock the database in exclusive mode for writes, and shared mode for
1909 reads. Pass any true value to enable. This affects the base DB handle I<and
1910 any child hashes or arrays> that use the same DB file. This is an optional
1911 parameter, and defaults to 0 (disabled). See L<LOCKING> below for more.
1915 Specifies whether autoflush is to be enabled on the underlying FileHandle.
1916 This obviously slows down write operations, but is required if you may have
1917 multiple processes accessing the same DB file (also consider enable I<locking>
1918 or at least I<volatile>). Pass any true value to enable. This is an optional
1919 parameter, and defaults to 0 (disabled).
1923 If I<volatile> mode is enabled, DBM::Deep will stat() the DB file before each
1924 STORE() operation. This is required if an outside force may change the size of
1925 the file between transactions. Locking also implicitly enables volatile. This
1926 is useful if you want to use a different locking system or write your own. Pass
1927 any true value to enable. This is an optional parameter, and defaults to 0
1932 If I<autobless> mode is enabled, DBM::Deep will preserve blessed hashes, and
1933 restore them when fetched. This is an B<experimental> feature, and does have
1934 side-effects. Basically, when hashes are re-blessed into their original
1935 classes, they are no longer blessed into the DBM::Deep class! So you won't be
1936 able to call any DBM::Deep methods on them. You have been warned.
1937 This is an optional parameter, and defaults to 0 (disabled).
1941 See L<FILTERS> below.
1945 Setting I<debug> mode will make all errors non-fatal, dump them out to
1946 STDERR, and continue on. This is for debugging purposes only, and probably
1947 not what you want. This is an optional parameter, and defaults to 0 (disabled).
1951 Instead of passing a file path, you can instead pass a handle to an pre-opened
1952 filehandle. Note: Beware of using the magick *DATA handle, as this actually
1953 contains your entire Perl script, as well as the data following the __DATA__
1954 marker. This will not work, because DBM::Deep uses absolute seek()s into the
1955 file. Instead, consider reading *DATA into an IO::Scalar handle, then passing
1960 =head1 TIE INTERFACE
1962 With DBM::Deep you can access your databases using Perl's standard hash/array
1963 syntax. Because all DBM::Deep objects are I<tied> to hashes or arrays, you can treat
1964 them as such. DBM::Deep will intercept all reads/writes and direct them to the right
1965 place -- the DB file. This has nothing to do with the L<TIE CONSTRUCTION>
1966 section above. This simply tells you how to use DBM::Deep using regular hashes
1967 and arrays, rather than calling functions like C<get()> and C<put()> (although those
1968 work too). It is entirely up to you how to want to access your databases.
1972 You can treat any DBM::Deep object like a normal Perl hash reference. Add keys,
1973 or even nested hashes (or arrays) using standard Perl syntax:
1975 my $db = DBM::Deep->new( "foo.db" );
1977 $db->{mykey} = "myvalue";
1979 $db->{myhash}->{subkey} = "subvalue";
1981 print $db->{myhash}->{subkey} . "\n";
1983 You can even step through hash keys using the normal Perl C<keys()> function:
1985 foreach my $key (keys %$db) {
1986 print "$key: " . $db->{$key} . "\n";
1989 Remember that Perl's C<keys()> function extracts I<every> key from the hash and
1990 pushes them onto an array, all before the loop even begins. If you have an
1991 extra large hash, this may exhaust Perl's memory. Instead, consider using
1992 Perl's C<each()> function, which pulls keys/values one at a time, using very
1995 while (my ($key, $value) = each %$db) {
1996 print "$key: $value\n";
1999 Please note that when using C<each()>, you should always pass a direct
2000 hash reference, not a lookup. Meaning, you should B<never> do this:
2003 while (my ($key, $value) = each %{$db->{foo}}) { # BAD
2005 This causes an infinite loop, because for each iteration, Perl is calling
2006 FETCH() on the $db handle, resulting in a "new" hash for foo every time, so
2007 it effectively keeps returning the first key over and over again. Instead,
2008 assign a temporary variable to C<$db->{foo}>, then pass that to each().
2012 As with hashes, you can treat any DBM::Deep object like a normal Perl array
2013 reference. This includes inserting, removing and manipulating elements,
2014 and the C<push()>, C<pop()>, C<shift()>, C<unshift()> and C<splice()> functions.
2015 The object must have first been created using type C<DBM::Deep-E<gt>TYPE_ARRAY>,
2016 or simply be a nested array reference inside a hash. Example:
2018 my $db = DBM::Deep->new(
2019 file => "foo-array.db",
2020 type => DBM::Deep->TYPE_ARRAY
2024 push @$db, "bar", "baz";
2025 unshift @$db, "bah";
2027 my $last_elem = pop @$db; # baz
2028 my $first_elem = shift @$db; # bah
2029 my $second_elem = $db->[1]; # bar
2031 my $num_elements = scalar @$db;
2035 In addition to the I<tie()> interface, you can also use a standard OO interface
2036 to manipulate all aspects of DBM::Deep databases. Each type of object (hash or
2037 array) has its own methods, but both types share the following common methods:
2038 C<put()>, C<get()>, C<exists()>, C<delete()> and C<clear()>.
2044 Stores a new hash key/value pair, or sets an array element value. Takes two
2045 arguments, the hash key or array index, and the new value. The value can be
2046 a scalar, hash ref or array ref. Returns true on success, false on failure.
2048 $db->put("foo", "bar"); # for hashes
2049 $db->put(1, "bar"); # for arrays
2053 Fetches the value of a hash key or array element. Takes one argument: the hash
2054 key or array index. Returns a scalar, hash ref or array ref, depending on the
2057 my $value = $db->get("foo"); # for hashes
2058 my $value = $db->get(1); # for arrays
2062 Checks if a hash key or array index exists. Takes one argument: the hash key
2063 or array index. Returns true if it exists, false if not.
2065 if ($db->exists("foo")) { print "yay!\n"; } # for hashes
2066 if ($db->exists(1)) { print "yay!\n"; } # for arrays
2070 Deletes one hash key/value pair or array element. Takes one argument: the hash
2071 key or array index. Returns true on success, false if not found. For arrays,
2072 the remaining elements located after the deleted element are NOT moved over.
2073 The deleted element is essentially just undefined, which is exactly how Perl's
2074 internal arrays work. Please note that the space occupied by the deleted
2075 key/value or element is B<not> reused again -- see L<UNUSED SPACE RECOVERY>
2076 below for details and workarounds.
2078 $db->delete("foo"); # for hashes
2079 $db->delete(1); # for arrays
2083 Deletes B<all> hash keys or array elements. Takes no arguments. No return
2084 value. Please note that the space occupied by the deleted keys/values or
2085 elements is B<not> reused again -- see L<UNUSED SPACE RECOVERY> below for
2086 details and workarounds.
2088 $db->clear(); # hashes or arrays
2094 For hashes, DBM::Deep supports all the common methods described above, and the
2095 following additional methods: C<first_key()> and C<next_key()>.
2101 Returns the "first" key in the hash. As with built-in Perl hashes, keys are
2102 fetched in an undefined order (which appears random). Takes no arguments,
2103 returns the key as a scalar value.
2105 my $key = $db->first_key();
2109 Returns the "next" key in the hash, given the previous one as the sole argument.
2110 Returns undef if there are no more keys to be fetched.
2112 $key = $db->next_key($key);
2116 Here are some examples of using hashes:
2118 my $db = DBM::Deep->new( "foo.db" );
2120 $db->put("foo", "bar");
2121 print "foo: " . $db->get("foo") . "\n";
2123 $db->put("baz", {}); # new child hash ref
2124 $db->get("baz")->put("buz", "biz");
2125 print "buz: " . $db->get("baz")->get("buz") . "\n";
2127 my $key = $db->first_key();
2129 print "$key: " . $db->get($key) . "\n";
2130 $key = $db->next_key($key);
2133 if ($db->exists("foo")) { $db->delete("foo"); }
2137 For arrays, DBM::Deep supports all the common methods described above, and the
2138 following additional methods: C<length()>, C<push()>, C<pop()>, C<shift()>,
2139 C<unshift()> and C<splice()>.
2145 Returns the number of elements in the array. Takes no arguments.
2147 my $len = $db->length();
2151 Adds one or more elements onto the end of the array. Accepts scalars, hash
2152 refs or array refs. No return value.
2154 $db->push("foo", "bar", {});
2158 Fetches the last element in the array, and deletes it. Takes no arguments.
2159 Returns undef if array is empty. Returns the element value.
2161 my $elem = $db->pop();
2165 Fetches the first element in the array, deletes it, then shifts all the
2166 remaining elements over to take up the space. Returns the element value. This
2167 method is not recommended with large arrays -- see L<LARGE ARRAYS> below for
2170 my $elem = $db->shift();
2174 Inserts one or more elements onto the beginning of the array, shifting all
2175 existing elements over to make room. Accepts scalars, hash refs or array refs.
2176 No return value. This method is not recommended with large arrays -- see
2177 <LARGE ARRAYS> below for details.
2179 $db->unshift("foo", "bar", {});
2183 Performs exactly like Perl's built-in function of the same name. See L<perldoc
2184 -f splice> for usage -- it is too complicated to document here. This method is
2185 not recommended with large arrays -- see L<LARGE ARRAYS> below for details.
2189 Here are some examples of using arrays:
2191 my $db = DBM::Deep->new(
2193 type => DBM::Deep->TYPE_ARRAY
2196 $db->push("bar", "baz");
2197 $db->unshift("foo");
2200 my $len = $db->length();
2201 print "length: $len\n"; # 4
2203 for (my $k=0; $k<$len; $k++) {
2204 print "$k: " . $db->get($k) . "\n";
2207 $db->splice(1, 2, "biz", "baf");
2209 while (my $elem = shift @$db) {
2210 print "shifted: $elem\n";
2215 Enable automatic file locking by passing a true value to the C<locking>
2216 parameter when constructing your DBM::Deep object (see L<SETUP> above).
2218 my $db = DBM::Deep->new(
2223 This causes DBM::Deep to C<flock()> the underlying FileHandle object with exclusive
2224 mode for writes, and shared mode for reads. This is required if you have
2225 multiple processes accessing the same database file, to avoid file corruption.
2226 Please note that C<flock()> does NOT work for files over NFS. See L<DB OVER
2227 NFS> below for more.
2229 =head2 EXPLICIT LOCKING
2231 You can explicitly lock a database, so it remains locked for multiple
2232 transactions. This is done by calling the C<lock()> method, and passing an
2233 optional lock mode argument (defaults to exclusive mode). This is particularly
2234 useful for things like counters, where the current value needs to be fetched,
2235 then incremented, then stored again.
2238 my $counter = $db->get("counter");
2240 $db->put("counter", $counter);
2249 You can pass C<lock()> an optional argument, which specifies which mode to use
2250 (exclusive or shared). Use one of these two constants: C<DBM::Deep-E<gt>LOCK_EX>
2251 or C<DBM::Deep-E<gt>LOCK_SH>. These are passed directly to C<flock()>, and are the
2252 same as the constants defined in Perl's C<Fcntl> module.
2254 $db->lock( DBM::Deep->LOCK_SH );
2258 If you want to implement your own file locking scheme, be sure to create your
2259 DBM::Deep objects setting the C<volatile> option to true. This hints to DBM::Deep
2260 that the DB file may change between transactions. See L<LOW-LEVEL ACCESS>
2263 =head1 IMPORTING/EXPORTING
2265 You can import existing complex structures by calling the C<import()> method,
2266 and export an entire database into an in-memory structure using the C<export()>
2267 method. Both are examined here.
2271 Say you have an existing hash with nested hashes/arrays inside it. Instead of
2272 walking the structure and adding keys/elements to the database as you go,
2273 simply pass a reference to the C<import()> method. This recursively adds
2274 everything to an existing DBM::Deep object for you. Here is an example:
2279 array1 => [ "elem0", "elem1", "elem2" ],
2281 subkey1 => "subvalue1",
2282 subkey2 => "subvalue2"
2286 my $db = DBM::Deep->new( "foo.db" );
2287 $db->import( $struct );
2289 print $db->{key1} . "\n"; # prints "value1"
2291 This recursively imports the entire C<$struct> object into C<$db>, including
2292 all nested hashes and arrays. If the DBM::Deep object contains exsiting data,
2293 keys are merged with the existing ones, replacing if they already exist.
2294 The C<import()> method can be called on any database level (not just the base
2295 level), and works with both hash and array DB types.
2299 B<Note:> Make sure your existing structure has no circular references in it.
2300 These will cause an infinite loop when importing.
2304 Calling the C<export()> method on an existing DBM::Deep object will return
2305 a reference to a new in-memory copy of the database. The export is done
2306 recursively, so all nested hashes/arrays are all exported to standard Perl
2307 objects. Here is an example:
2309 my $db = DBM::Deep->new( "foo.db" );
2311 $db->{key1} = "value1";
2312 $db->{key2} = "value2";
2314 $db->{hash1}->{subkey1} = "subvalue1";
2315 $db->{hash1}->{subkey2} = "subvalue2";
2317 my $struct = $db->export();
2319 print $struct->{key1} . "\n"; # prints "value1"
2321 This makes a complete copy of the database in memory, and returns a reference
2322 to it. The C<export()> method can be called on any database level (not just
2323 the base level), and works with both hash and array DB types. Be careful of
2324 large databases -- you can store a lot more data in a DBM::Deep object than an
2325 in-memory Perl structure.
2329 B<Note:> Make sure your database has no circular references in it.
2330 These will cause an infinite loop when exporting.
2334 DBM::Deep has a number of hooks where you can specify your own Perl function
2335 to perform filtering on incoming or outgoing data. This is a perfect
2336 way to extend the engine, and implement things like real-time compression or
2337 encryption. Filtering applies to the base DB level, and all child hashes /
2338 arrays. Filter hooks can be specified when your DBM::Deep object is first
2339 constructed, or by calling the C<set_filter()> method at any time. There are
2340 four available filter hooks, described below:
2344 =item * filter_store_key
2346 This filter is called whenever a hash key is stored. It
2347 is passed the incoming key, and expected to return a transformed key.
2349 =item * filter_store_value
2351 This filter is called whenever a hash key or array element is stored. It
2352 is passed the incoming value, and expected to return a transformed value.
2354 =item * filter_fetch_key
2356 This filter is called whenever a hash key is fetched (i.e. via
2357 C<first_key()> or C<next_key()>). It is passed the transformed key,
2358 and expected to return the plain key.
2360 =item * filter_fetch_value
2362 This filter is called whenever a hash key or array element is fetched.
2363 It is passed the transformed value, and expected to return the plain value.
2367 Here are the two ways to setup a filter hook:
2369 my $db = DBM::Deep->new(
2371 filter_store_value => \&my_filter_store,
2372 filter_fetch_value => \&my_filter_fetch
2377 $db->set_filter( "filter_store_value", \&my_filter_store );
2378 $db->set_filter( "filter_fetch_value", \&my_filter_fetch );
2380 Your filter function will be called only when dealing with SCALAR keys or
2381 values. When nested hashes and arrays are being stored/fetched, filtering
2382 is bypassed. Filters are called as static functions, passed a single SCALAR
2383 argument, and expected to return a single SCALAR value. If you want to
2384 remove a filter, set the function reference to C<undef>:
2386 $db->set_filter( "filter_store_value", undef );
2388 =head2 REAL-TIME ENCRYPTION EXAMPLE
2390 Here is a working example that uses the I<Crypt::Blowfish> module to
2391 do real-time encryption / decryption of keys & values with DBM::Deep Filters.
2392 Please visit L<http://search.cpan.org/search?module=Crypt::Blowfish> for more
2393 on I<Crypt::Blowfish>. You'll also need the I<Crypt::CBC> module.
2396 use Crypt::Blowfish;
2399 my $cipher = Crypt::CBC->new({
2400 'key' => 'my secret key',
2401 'cipher' => 'Blowfish',
2403 'regenerate_key' => 0,
2404 'padding' => 'space',
2408 my $db = DBM::Deep->new(
2409 file => "foo-encrypt.db",
2410 filter_store_key => \&my_encrypt,
2411 filter_store_value => \&my_encrypt,
2412 filter_fetch_key => \&my_decrypt,
2413 filter_fetch_value => \&my_decrypt,
2416 $db->{key1} = "value1";
2417 $db->{key2} = "value2";
2418 print "key1: " . $db->{key1} . "\n";
2419 print "key2: " . $db->{key2} . "\n";
2425 return $cipher->encrypt( $_[0] );
2428 return $cipher->decrypt( $_[0] );
2431 =head2 REAL-TIME COMPRESSION EXAMPLE
2433 Here is a working example that uses the I<Compress::Zlib> module to do real-time
2434 compression / decompression of keys & values with DBM::Deep Filters.
2435 Please visit L<http://search.cpan.org/search?module=Compress::Zlib> for
2436 more on I<Compress::Zlib>.
2441 my $db = DBM::Deep->new(
2442 file => "foo-compress.db",
2443 filter_store_key => \&my_compress,
2444 filter_store_value => \&my_compress,
2445 filter_fetch_key => \&my_decompress,
2446 filter_fetch_value => \&my_decompress,
2449 $db->{key1} = "value1";
2450 $db->{key2} = "value2";
2451 print "key1: " . $db->{key1} . "\n";
2452 print "key2: " . $db->{key2} . "\n";
2458 return Compress::Zlib::memGzip( $_[0] ) ;
2461 return Compress::Zlib::memGunzip( $_[0] ) ;
2464 B<Note:> Filtering of keys only applies to hashes. Array "keys" are
2465 actually numerical index numbers, and are not filtered.
2467 =head1 ERROR HANDLING
2469 Most DBM::Deep methods return a true value for success, and call die() on
2470 failure. You can wrap calls in an eval block to catch the die. Also, the
2471 actual error message is stored in an internal scalar, which can be fetched by
2472 calling the C<error()> method.
2474 my $db = DBM::Deep->new( "foo.db" ); # create hash
2475 eval { $db->push("foo"); }; # ILLEGAL -- push is array-only call
2477 print $db->error(); # prints error message
2479 You can then call C<clear_error()> to clear the current error state.
2483 If you set the C<debug> option to true when creating your DBM::Deep object,
2484 all errors are considered NON-FATAL, and dumped to STDERR. This is only
2485 for debugging purposes.
2487 =head1 LARGEFILE SUPPORT
2489 If you have a 64-bit system, and your Perl is compiled with both LARGEFILE
2490 and 64-bit support, you I<may> be able to create databases larger than 2 GB.
2491 DBM::Deep by default uses 32-bit file offset tags, but these can be changed
2492 by calling the static C<set_pack()> method before you do anything else.
2494 DBM::Deep::set_pack(8, 'Q');
2496 This tells DBM::Deep to pack all file offsets with 8-byte (64-bit) quad words
2497 instead of 32-bit longs. After setting these values your DB files have a
2498 theoretical maximum size of 16 XB (exabytes).
2502 B<Note:> Changing these values will B<NOT> work for existing database files.
2503 Only change this for new files, and make sure it stays set consistently
2504 throughout the file's life. If you do set these values, you can no longer
2505 access 32-bit DB files. You can, however, call C<set_pack(4, 'N')> to change
2506 back to 32-bit mode.
2510 B<Note:> I have not personally tested files > 2 GB -- all my systems have
2511 only a 32-bit Perl. However, I have received user reports that this does
2514 =head1 LOW-LEVEL ACCESS
2516 If you require low-level access to the underlying FileHandle that DBM::Deep uses,
2517 you can call the C<fh()> method, which returns the handle:
2521 This method can be called on the root level of the datbase, or any child
2522 hashes or arrays. All levels share a I<root> structure, which contains things
2523 like the FileHandle, a reference counter, and all your options you specified
2524 when you created the object. You can get access to this root structure by
2525 calling the C<root()> method.
2527 my $root = $db->root();
2529 This is useful for changing options after the object has already been created,
2530 such as enabling/disabling locking, volatile or debug modes. You can also
2531 store your own temporary user data in this structure (be wary of name
2532 collision), which is then accessible from any child hash or array.
2534 =head1 CUSTOM DIGEST ALGORITHM
2536 DBM::Deep by default uses the I<Message Digest 5> (MD5) algorithm for hashing
2537 keys. However you can override this, and use another algorithm (such as SHA-256)
2538 or even write your own. But please note that DBM::Deep currently expects zero
2539 collisions, so your algorithm has to be I<perfect>, so to speak.
2540 Collision detection may be introduced in a later version.
2544 You can specify a custom digest algorithm by calling the static C<set_digest()>
2545 function, passing a reference to a subroutine, and the length of the algorithm's
2546 hashes (in bytes). This is a global static function, which affects ALL DBM::Deep
2547 objects. Here is a working example that uses a 256-bit hash from the
2548 I<Digest::SHA256> module. Please see
2549 L<http://search.cpan.org/search?module=Digest::SHA256> for more.
2554 my $context = Digest::SHA256::new(256);
2556 DBM::Deep::set_digest( \&my_digest, 32 );
2558 my $db = DBM::Deep->new( "foo-sha.db" );
2560 $db->{key1} = "value1";
2561 $db->{key2} = "value2";
2562 print "key1: " . $db->{key1} . "\n";
2563 print "key2: " . $db->{key2} . "\n";
2569 return substr( $context->hash($_[0]), 0, 32 );
2572 B<Note:> Your returned digest strings must be B<EXACTLY> the number
2573 of bytes you specify in the C<set_digest()> function (in this case 32).
2575 =head1 CIRCULAR REFERENCES
2577 DBM::Deep has B<experimental> support for circular references. Meaning you
2578 can have a nested hash key or array element that points to a parent object.
2579 This relationship is stored in the DB file, and is preserved between sessions.
2582 my $db = DBM::Deep->new( "foo.db" );
2585 $db->{circle} = $db; # ref to self
2587 print $db->{foo} . "\n"; # prints "foo"
2588 print $db->{circle}->{foo} . "\n"; # prints "foo" again
2590 One catch is, passing the object to a function that recursively walks the
2591 object tree (such as I<Data::Dumper> or even the built-in C<optimize()> or
2592 C<export()> methods) will result in an infinite loop. The other catch is,
2593 if you fetch the I<key> of a circular reference (i.e. using the C<first_key()>
2594 or C<next_key()> methods), you will get the I<target object's key>, not the
2595 ref's key. This gets even more interesting with the above example, where
2596 the I<circle> key points to the base DB object, which technically doesn't
2597 have a key. So I made DBM::Deep return "[base]" as the key name in that
2600 =head1 CAVEATS / ISSUES / BUGS
2602 This section describes all the known issues with DBM::Deep. It you have found
2603 something that is not listed here, please send e-mail to L<jhuckaby@cpan.org>.
2605 =head2 UNUSED SPACE RECOVERY
2607 One major caveat with DBM::Deep is that space occupied by existing keys and
2608 values is not recovered when they are deleted. Meaning if you keep deleting
2609 and adding new keys, your file will continuously grow. I am working on this,
2610 but in the meantime you can call the built-in C<optimize()> method from time to
2611 time (perhaps in a crontab or something) to recover all your unused space.
2613 $db->optimize(); # returns true on success
2615 This rebuilds the ENTIRE database into a new file, then moves it on top of
2616 the original. The new file will have no unused space, thus it will take up as
2617 little disk space as possible. Please note that this operation can take
2618 a long time for large files, and you need enough disk space to temporarily hold
2619 2 copies of your DB file. The temporary file is created in the same directory
2620 as the original, named with a ".tmp" extension, and is deleted when the
2621 operation completes. Oh, and if locking is enabled, the DB is automatically
2622 locked for the entire duration of the copy.
2626 B<WARNING:> Only call optimize() on the top-level node of the database, and
2627 make sure there are no child references lying around. DBM::Deep keeps a reference
2628 counter, and if it is greater than 1, optimize() will abort and return undef.
2630 =head2 AUTOVIVIFICATION
2632 Unfortunately, autovivification doesn't work with tied hashes. This appears to
2633 be a bug in Perl's tie() system, as I<Jakob Schmidt> encountered the very same
2634 issue with his I<DWH_FIle> module (see L<http://search.cpan.org/search?module=DWH_File>),
2635 and it is also mentioned in the BUGS section for the I<MLDBM> module <see
2636 L<http://search.cpan.org/search?module=MLDBM>). Basically, on a new db file,
2639 $db->{foo}->{bar} = "hello";
2641 Since "foo" doesn't exist, you cannot add "bar" to it. You end up with "foo"
2642 being an empty hash. Try this instead, which works fine:
2644 $db->{foo} = { bar => "hello" };
2646 As of Perl 5.8.7, this bug still exists. I have walked very carefully through
2647 the execution path, and Perl indeed passes an empty hash to the STORE() method.
2648 Probably a bug in Perl.
2650 =head2 FILE CORRUPTION
2652 The current level of error handling in DBM::Deep is minimal. Files I<are> checked
2653 for a 32-bit signature when opened, but other corruption in files can cause
2654 segmentation faults. DBM::Deep may try to seek() past the end of a file, or get
2655 stuck in an infinite loop depending on the level of corruption. File write
2656 operations are not checked for failure (for speed), so if you happen to run
2657 out of disk space, DBM::Deep will probably fail in a bad way. These things will
2658 be addressed in a later version of DBM::Deep.
2662 Beware of using DB files over NFS. DBM::Deep uses flock(), which works well on local
2663 filesystems, but will NOT protect you from file corruption over NFS. I've heard
2664 about setting up your NFS server with a locking daemon, then using lockf() to
2665 lock your files, but your milage may vary there as well. From what I
2666 understand, there is no real way to do it. However, if you need access to the
2667 underlying FileHandle in DBM::Deep for using some other kind of locking scheme like
2668 lockf(), see the L<LOW-LEVEL ACCESS> section above.
2670 =head2 COPYING OBJECTS
2672 Beware of copying tied objects in Perl. Very strange things can happen.
2673 Instead, use DBM::Deep's C<clone()> method which safely copies the object and
2674 returns a new, blessed, tied hash or array to the same level in the DB.
2676 my $copy = $db->clone();
2680 Beware of using C<shift()>, C<unshift()> or C<splice()> with large arrays.
2681 These functions cause every element in the array to move, which can be murder
2682 on DBM::Deep, as every element has to be fetched from disk, then stored again in
2683 a different location. This may be addressed in a later version.
2687 This section discusses DBM::Deep's speed and memory usage.
2691 Obviously, DBM::Deep isn't going to be as fast as some C-based DBMs, such as
2692 the almighty I<BerkeleyDB>. But it makes up for it in features like true
2693 multi-level hash/array support, and cross-platform FTPable files. Even so,
2694 DBM::Deep is still pretty fast, and the speed stays fairly consistent, even
2695 with huge databases. Here is some test data:
2697 Adding 1,000,000 keys to new DB file...
2699 At 100 keys, avg. speed is 2,703 keys/sec
2700 At 200 keys, avg. speed is 2,642 keys/sec
2701 At 300 keys, avg. speed is 2,598 keys/sec
2702 At 400 keys, avg. speed is 2,578 keys/sec
2703 At 500 keys, avg. speed is 2,722 keys/sec
2704 At 600 keys, avg. speed is 2,628 keys/sec
2705 At 700 keys, avg. speed is 2,700 keys/sec
2706 At 800 keys, avg. speed is 2,607 keys/sec
2707 At 900 keys, avg. speed is 2,190 keys/sec
2708 At 1,000 keys, avg. speed is 2,570 keys/sec
2709 At 2,000 keys, avg. speed is 2,417 keys/sec
2710 At 3,000 keys, avg. speed is 1,982 keys/sec
2711 At 4,000 keys, avg. speed is 1,568 keys/sec
2712 At 5,000 keys, avg. speed is 1,533 keys/sec
2713 At 6,000 keys, avg. speed is 1,787 keys/sec
2714 At 7,000 keys, avg. speed is 1,977 keys/sec
2715 At 8,000 keys, avg. speed is 2,028 keys/sec
2716 At 9,000 keys, avg. speed is 2,077 keys/sec
2717 At 10,000 keys, avg. speed is 2,031 keys/sec
2718 At 20,000 keys, avg. speed is 1,970 keys/sec
2719 At 30,000 keys, avg. speed is 2,050 keys/sec
2720 At 40,000 keys, avg. speed is 2,073 keys/sec
2721 At 50,000 keys, avg. speed is 1,973 keys/sec
2722 At 60,000 keys, avg. speed is 1,914 keys/sec
2723 At 70,000 keys, avg. speed is 2,091 keys/sec
2724 At 80,000 keys, avg. speed is 2,103 keys/sec
2725 At 90,000 keys, avg. speed is 1,886 keys/sec
2726 At 100,000 keys, avg. speed is 1,970 keys/sec
2727 At 200,000 keys, avg. speed is 2,053 keys/sec
2728 At 300,000 keys, avg. speed is 1,697 keys/sec
2729 At 400,000 keys, avg. speed is 1,838 keys/sec
2730 At 500,000 keys, avg. speed is 1,941 keys/sec
2731 At 600,000 keys, avg. speed is 1,930 keys/sec
2732 At 700,000 keys, avg. speed is 1,735 keys/sec
2733 At 800,000 keys, avg. speed is 1,795 keys/sec
2734 At 900,000 keys, avg. speed is 1,221 keys/sec
2735 At 1,000,000 keys, avg. speed is 1,077 keys/sec
2737 This test was performed on a PowerMac G4 1gHz running Mac OS X 10.3.2 & Perl
2738 5.8.1, with an 80GB Ultra ATA/100 HD spinning at 7200RPM. The hash keys and
2739 values were between 6 - 12 chars in length. The DB file ended up at 210MB.
2740 Run time was 12 min 3 sec.
2744 One of the great things about DBM::Deep is that it uses very little memory.
2745 Even with huge databases (1,000,000+ keys) you will not see much increased
2746 memory on your process. DBM::Deep relies solely on the filesystem for storing
2747 and fetching data. Here is output from I</usr/bin/top> before even opening a
2750 PID USER PRI NI SIZE RSS SHARE STAT %CPU %MEM TIME COMMAND
2751 22831 root 11 0 2716 2716 1296 R 0.0 0.2 0:07 perl
2753 Basically the process is taking 2,716K of memory. And here is the same
2754 process after storing and fetching 1,000,000 keys:
2756 PID USER PRI NI SIZE RSS SHARE STAT %CPU %MEM TIME COMMAND
2757 22831 root 14 0 2772 2772 1328 R 0.0 0.2 13:32 perl
2759 Notice the memory usage increased by only 56K. Test was performed on a 700mHz
2760 x86 box running Linux RedHat 7.2 & Perl 5.6.1.
2762 =head1 DB FILE FORMAT
2764 In case you were interested in the underlying DB file format, it is documented
2765 here in this section. You don't need to know this to use the module, it's just
2766 included for reference.
2770 DBM::Deep files always start with a 32-bit signature to identify the file type.
2771 This is at offset 0. The signature is "DPDB" in network byte order. This is
2772 checked when the file is opened.
2776 The DBM::Deep file is in a I<tagged format>, meaning each section of the file
2777 has a standard header containing the type of data, the length of data, and then
2778 the data itself. The type is a single character (1 byte), the length is a
2779 32-bit unsigned long in network byte order, and the data is, well, the data.
2780 Here is how it unfolds:
2784 Immediately after the 32-bit file signature is the I<Master Index> record.
2785 This is a standard tag header followed by 1024 bytes (in 32-bit mode) or 2048
2786 bytes (in 64-bit mode) of data. The type is I<H> for hash or I<A> for array,
2787 depending on how the DBM::Deep object was constructed.
2791 The index works by looking at a I<MD5 Hash> of the hash key (or array index
2792 number). The first 8-bit char of the MD5 signature is the offset into the
2793 index, multipled by 4 in 32-bit mode, or 8 in 64-bit mode. The value of the
2794 index element is a file offset of the next tag for the key/element in question,
2795 which is usually a I<Bucket List> tag (see below).
2799 The next tag I<could> be another index, depending on how many keys/elements
2800 exist. See L<RE-INDEXING> below for details.
2804 A I<Bucket List> is a collection of 16 MD5 hashes for keys/elements, plus
2805 file offsets to where the actual data is stored. It starts with a standard
2806 tag header, with type I<B>, and a data size of 320 bytes in 32-bit mode, or
2807 384 bytes in 64-bit mode. Each MD5 hash is stored in full (16 bytes), plus
2808 the 32-bit or 64-bit file offset for the I<Bucket> containing the actual data.
2809 When the list fills up, a I<Re-Index> operation is performed (See
2810 L<RE-INDEXING> below).
2814 A I<Bucket> is a tag containing a key/value pair (in hash mode), or a
2815 index/value pair (in array mode). It starts with a standard tag header with
2816 type I<D> for scalar data (string, binary, etc.), or it could be a nested
2817 hash (type I<H>) or array (type I<A>). The value comes just after the tag
2818 header. The size reported in the tag header is only for the value, but then,
2819 just after the value is another size (32-bit unsigned long) and then the plain
2820 key itself. Since the value is likely to be fetched more often than the plain
2821 key, I figured it would be I<slightly> faster to store the value first.
2825 If the type is I<H> (hash) or I<A> (array), the value is another I<Master Index>
2826 record for the nested structure, where the process begins all over again.
2830 After a I<Bucket List> grows to 16 records, its allocated space in the file is
2831 exhausted. Then, when another key/element comes in, the list is converted to a
2832 new index record. However, this index will look at the next char in the MD5
2833 hash, and arrange new Bucket List pointers accordingly. This process is called
2834 I<Re-Indexing>. Basically, a new index tag is created at the file EOF, and all
2835 17 (16 + new one) keys/elements are removed from the old Bucket List and
2836 inserted into the new index. Several new Bucket Lists are created in the
2837 process, as a new MD5 char from the key is being examined (it is unlikely that
2838 the keys will all share the same next char of their MD5s).
2842 Because of the way the I<MD5> algorithm works, it is impossible to tell exactly
2843 when the Bucket Lists will turn into indexes, but the first round tends to
2844 happen right around 4,000 keys. You will see a I<slight> decrease in
2845 performance here, but it picks back up pretty quick (see L<SPEED> above). Then
2846 it takes B<a lot> more keys to exhaust the next level of Bucket Lists. It's
2847 right around 900,000 keys. This process can continue nearly indefinitely --
2848 right up until the point the I<MD5> signatures start colliding with each other,
2849 and this is B<EXTREMELY> rare -- like winning the lottery 5 times in a row AND
2850 getting struck by lightning while you are walking to cash in your tickets.
2851 Theoretically, since I<MD5> hashes are 128-bit values, you I<could> have up to
2852 340,282,366,921,000,000,000,000,000,000,000,000,000 keys/elements (I believe
2853 this is 340 unodecillion, but don't quote me).
2857 When a new key/element is stored, the key (or index number) is first ran through
2858 I<Digest::MD5> to get a 128-bit signature (example, in hex:
2859 b05783b0773d894396d475ced9d2f4f6). Then, the I<Master Index> record is checked
2860 for the first char of the signature (in this case I<b>). If it does not exist,
2861 a new I<Bucket List> is created for our key (and the next 15 future keys that
2862 happen to also have I<b> as their first MD5 char). The entire MD5 is written
2863 to the I<Bucket List> along with the offset of the new I<Bucket> record (EOF at
2864 this point, unless we are replacing an existing I<Bucket>), where the actual
2865 data will be stored.
2869 Fetching an existing key/element involves getting a I<Digest::MD5> of the key
2870 (or index number), then walking along the indexes. If there are enough
2871 keys/elements in this DB level, there might be nested indexes, each linked to
2872 a particular char of the MD5. Finally, a I<Bucket List> is pointed to, which
2873 contains up to 16 full MD5 hashes. Each is checked for equality to the key in
2874 question. If we found a match, the I<Bucket> tag is loaded, where the value and
2875 plain key are stored.
2879 Fetching the plain key occurs when calling the I<first_key()> and I<next_key()>
2880 methods. In this process the indexes are walked systematically, and each key
2881 fetched in increasing MD5 order (which is why it appears random). Once the
2882 I<Bucket> is found, the value is skipped the plain key returned instead.
2883 B<Note:> Do not count on keys being fetched as if the MD5 hashes were
2884 alphabetically sorted. This only happens on an index-level -- as soon as the
2885 I<Bucket Lists> are hit, the keys will come out in the order they went in --
2886 so it's pretty much undefined how the keys will come out -- just like Perl's
2889 =head1 CODE COVERAGE
2891 I use B<Devel::Cover> to test the code coverage of my tests, below is the B<Devel::Cover> report on this
2892 module's test suite.
2894 ---------------------------- ------ ------ ------ ------ ------ ------ ------
2895 File stmt bran cond sub pod time total
2896 ---------------------------- ------ ------ ------ ------ ------ ------ ------
2897 blib/lib/DBM/Deep.pm 94.9 84.5 77.8 100.0 11.1 100.0 89.7
2898 Total 94.9 84.5 77.8 100.0 11.1 100.0 89.7
2899 ---------------------------- ------ ------ ------ ------ ------ ------ ------
2903 Joseph Huckaby, L<jhuckaby@cpan.org>
2905 Special thanks to Adam Sah and Rich Gaushell! You know why :-)
2909 perltie(1), Tie::Hash(3), Digest::MD5(3), Fcntl(3), flock(2), lockf(3), nfs(5),
2910 Digest::SHA256(3), Crypt::Blowfish(3), Compress::Zlib(3)
2914 Copyright (c) 2002-2005 Joseph Huckaby. All Rights Reserved.
2915 This is free software, you may use it and distribute it under the
2916 same terms as Perl itself.