7 # Multi-level database module for storing hash trees, arrays and simple
8 # key/value pairs into FTP-able, cross-platform binary database files.
10 # Type `perldoc DBM::Deep` for complete documentation.
14 # tie %db, 'DBM::Deep', 'my_database.db'; # standard tie() method
16 # my $db = new DBM::Deep( 'my_database.db' ); # preferred OO method
18 # $db->{my_scalar} = 'hello world';
19 # $db->{my_hash} = { larry => 'genius', hashes => 'fast' };
20 # $db->{my_array} = [ 1, 2, 3, time() ];
21 # $db->{my_complex} = [ 'hello', { perl => 'rules' }, 42, 99 ];
22 # push @{$db->{my_array}}, 'another value';
23 # my @key_list = keys %{$db->{my_hash}};
24 # print "This module " . $db->{my_complex}->[1]->{perl} . "!\n";
27 # (c) 2002-2005 Joseph Huckaby. All Rights Reserved.
28 # This program is free software; you can redistribute it and/or
29 # modify it under the same terms as Perl itself.
38 use vars qw/$VERSION/;
43 # Set to 4 and 'N' for 32-bit offset tags (default). Theoretical limit of 4 GB per file.
44 # (Perl must be compiled with largefile support for files > 2 GB)
46 # Set to 8 and 'Q' for 64-bit offsets. Theoretical limit of 16 XB per file.
47 # (Perl must be compiled with largefile and 64-bit long support)
53 # Set to 4 and 'N' for 32-bit data length prefixes. Limit of 4 GB for each key/value.
54 # Upgrading this is possible (see above) but probably not necessary. If you need
55 # more than 4 GB for a single key or value, this module is really not for you :-)
57 #my $DATA_LENGTH_SIZE = 4;
58 #my $DATA_LENGTH_PACK = 'N';
59 my ($LONG_SIZE, $LONG_PACK, $DATA_LENGTH_SIZE, $DATA_LENGTH_PACK);
62 # Maximum number of buckets per list before another level of indexing is done.
63 # Increase this value for slightly greater speed, but larger database files.
64 # DO NOT decrease this value below 16, due to risk of recursive reindex overrun.
69 # Better not adjust anything below here, unless you're me :-)
73 # Setup digest function for keys
75 my ($DIGEST_FUNC, $HASH_SIZE);
76 #my $DIGEST_FUNC = \&Digest::MD5::md5;
79 # Precalculate index and bucket sizes based on values above.
82 my ($INDEX_SIZE, $BUCKET_SIZE, $BUCKET_LIST_SIZE);
89 # Setup file and tag signatures. These should never change.
91 sub SIG_FILE () { 'DPDB' }
92 sub SIG_HASH () { 'H' }
93 sub SIG_ARRAY () { 'A' }
94 sub SIG_NULL () { 'N' }
95 sub SIG_DATA () { 'D' }
96 sub SIG_INDEX () { 'I' }
97 sub SIG_BLIST () { 'B' }
101 # Setup constants for users to pass to new()
103 sub TYPE_HASH () { return SIG_HASH; }
104 sub TYPE_ARRAY () { return SIG_ARRAY; }
108 # Class constructor method for Perl OO interface.
109 # Calls tie() and returns blessed reference to tied hash or array,
110 # providing a hybrid OO/tie interface.
114 if (scalar(@_) > 1) { $args = {@_}; }
115 else { $args = { file => shift }; }
118 # Check if we want a tied hash or array.
121 if (defined($args->{type}) && $args->{type} eq TYPE_ARRAY) {
122 tie @$self, $class, %$args;
125 tie %$self, $class, %$args;
128 return bless $self, $class;
132 my @outer_params = qw( type base_offset );
135 # Setup $self and bless into this class.
142 base_offset => length(SIG_FILE),
153 filter_store_key => undef,
154 filter_store_value => undef,
155 filter_fetch_key => undef,
156 filter_fetch_value => undef,
165 foreach my $outer_parm ( @outer_params ) {
166 next unless exists $args->{$outer_parm};
167 $self->{$outer_parm} = $args->{$outer_parm}
170 if ( exists $args->{root} ) {
171 $self->{root} = $args->{root};
174 # This is cleanup based on the fact that the $args
175 # coming in is for both the root and non-root items
176 delete $self->root->{$_} for @outer_params;
178 $self->root->{links}++;
180 if (!defined($self->fh)) { $self->_open(); }
186 sub _get_self { tied( %{$_[0]} ) || $_[0] }
190 # Tied hash constructor method, called by Perl's tie() function.
194 if (scalar(@_) > 1) { $args = {@_}; }
195 #XXX This use of ref() is bad and is a bug
196 elsif (ref($_[0])) { $args = $_[0]; }
197 else { $args = { file => shift }; }
199 $args->{type} = TYPE_HASH;
201 return $class->_init($args);
206 # Tied array constructor method, called by Perl's tie() function.
210 if (scalar(@_) > 1) { $args = {@_}; }
211 #XXX This use of ref() is bad and is a bug
212 elsif (ref($_[0])) { $args = $_[0]; }
213 else { $args = { file => shift }; }
215 $args->{type} = TYPE_ARRAY;
217 return $class->_init($args);
222 # Class deconstructor. Close file handle if there are no more refs.
224 my $self = _get_self($_[0]);
227 $self->root->{links}--;
229 if (!$self->root->{links}) {
236 # Open a FileHandle to the database, create if nonexistent.
237 # Make sure file signature matches DeepDB spec.
239 my $self = _get_self($_[0]);
241 if (defined($self->fh)) { $self->_close(); }
243 if (!(-e $self->root->{file}) && $self->root->{mode} eq 'r+') {
244 my $temp = FileHandle->new( $self->root->{file}, 'w' );
248 #XXX Convert to set_fh()
249 $self->root->{fh} = FileHandle->new( $self->root->{file}, $self->root->{mode} );
250 if (! defined($self->fh)) {
251 return $self->_throw_error("Cannot open file: " . $self->root->{file} . ": $!");
254 binmode $self->fh; # for win32
255 if ($self->root->{autoflush}) {
256 $self->fh->autoflush();
260 seek($self->fh, 0, 0);
261 my $bytes_read = $self->fh->read($signature, length(SIG_FILE));
264 # File is empty -- write signature and master index
267 seek($self->fh, 0, 0);
268 $self->fh->print(SIG_FILE);
269 $self->root->{end} = length(SIG_FILE);
270 $self->_create_tag($self->base_offset, $self->type, chr(0) x $INDEX_SIZE);
272 my $plain_key = "[base]";
273 $self->fh->print( pack($DATA_LENGTH_PACK, length($plain_key)) . $plain_key );
274 $self->root->{end} += $DATA_LENGTH_SIZE + length($plain_key);
281 # Check signature was valid
283 unless ($signature eq SIG_FILE) {
285 return $self->_throw_error("Signature not found -- file is not a Deep DB");
288 $self->root->{end} = (stat($self->fh))[7];
291 # Get our type from master index signature
293 my $tag = $self->_load_tag($self->base_offset);
294 #XXX We probably also want to store the hash algorithm name and not assume anything
296 return $self->_throw_error("Corrupted file, no master index record");
298 if ($self->{type} ne $tag->{signature}) {
299 return $self->_throw_error("File type mismatch");
307 # Close database FileHandle
309 my $self = _get_self($_[0]);
310 undef $self->root->{fh};
315 # Given offset, signature and content, create tag and write to disk
317 my ($self, $offset, $sig, $content) = @_;
318 my $size = length($content);
320 seek($self->fh, $offset, 0);
321 $self->fh->print( $sig . pack($DATA_LENGTH_PACK, $size) . $content );
323 if ($offset == $self->root->{end}) {
324 $self->root->{end} += SIG_SIZE + $DATA_LENGTH_SIZE + $size;
330 offset => $offset + SIG_SIZE + $DATA_LENGTH_SIZE,
337 # Given offset, load single tag and return signature, size and data
342 seek($self->fh, $offset, 0);
343 if ($self->fh->eof()) { return undef; }
346 $self->fh->read($sig, SIG_SIZE);
349 $self->fh->read($size, $DATA_LENGTH_SIZE);
350 $size = unpack($DATA_LENGTH_PACK, $size);
353 $self->fh->read($buffer, $size);
358 offset => $offset + SIG_SIZE + $DATA_LENGTH_SIZE,
365 # Given index tag, lookup single entry in index and return .
368 my ($tag, $index) = @_;
370 my $location = unpack($LONG_PACK, substr($tag->{content}, $index * $LONG_SIZE, $LONG_SIZE) );
371 if (!$location) { return; }
373 return $self->_load_tag( $location );
378 # Adds one key/value pair to bucket list, given offset, MD5 digest of key,
379 # plain (undigested) key and value.
382 my ($tag, $md5, $plain_key, $value) = @_;
383 my $keys = $tag->{content};
387 my $is_dbm_deep = eval { $value->isa( 'DBM::Deep' ) };
388 my $internal_ref = $is_dbm_deep && ($value->root eq $self->root);
391 # Iterate through buckets, seeing if this is a new entry or a replace.
393 for (my $i=0; $i<$MAX_BUCKETS; $i++) {
394 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
395 my $subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
398 # Found empty bucket (end of list). Populate and exit loop.
402 $location = $internal_ref
403 ? $value->base_offset
404 : $self->root->{end};
406 seek($self->fh, $tag->{offset} + ($i * $BUCKET_SIZE), 0);
407 $self->fh->print( $md5 . pack($LONG_PACK, $location) );
410 elsif ($md5 eq $key) {
412 # Found existing bucket with same key. Replace with new value.
417 $location = $value->base_offset;
418 seek($self->fh, $tag->{offset} + ($i * $BUCKET_SIZE), 0);
419 $self->fh->print( $md5 . pack($LONG_PACK, $location) );
422 seek($self->fh, $subloc + SIG_SIZE, 0);
424 $self->fh->read($size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
427 # If value is a hash, array, or raw value with equal or less size, we can
428 # reuse the same content area of the database. Otherwise, we have to create
429 # a new content area at the EOF.
432 my $r = Scalar::Util::reftype( $value ) || '';
433 if ( $r eq 'HASH' || $r eq 'ARRAY' ) { $actual_length = $INDEX_SIZE; }
434 else { $actual_length = length($value); }
436 if ($actual_length <= $size) {
440 $location = $self->root->{end};
441 seek($self->fh, $tag->{offset} + ($i * $BUCKET_SIZE) + $HASH_SIZE, 0);
442 $self->fh->print( pack($LONG_PACK, $location) );
450 # If this is an internal reference, return now.
451 # No need to write value or plain key
458 # If bucket didn't fit into list, split into a new index level
461 seek($self->fh, $tag->{ref_loc}, 0);
462 $self->fh->print( pack($LONG_PACK, $self->root->{end}) );
464 my $index_tag = $self->_create_tag($self->root->{end}, SIG_INDEX, chr(0) x $INDEX_SIZE);
467 $keys .= $md5 . pack($LONG_PACK, 0);
469 for (my $i=0; $i<=$MAX_BUCKETS; $i++) {
470 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
472 my $old_subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
473 my $num = ord(substr($key, $tag->{ch} + 1, 1));
475 if ($offsets[$num]) {
476 my $offset = $offsets[$num] + SIG_SIZE + $DATA_LENGTH_SIZE;
477 seek($self->fh, $offset, 0);
479 $self->fh->read($subkeys, $BUCKET_LIST_SIZE);
481 for (my $k=0; $k<$MAX_BUCKETS; $k++) {
482 my $subloc = unpack($LONG_PACK, substr($subkeys, ($k * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
484 seek($self->fh, $offset + ($k * $BUCKET_SIZE), 0);
485 $self->fh->print( $key . pack($LONG_PACK, $old_subloc || $self->root->{end}) );
491 $offsets[$num] = $self->root->{end};
492 seek($self->fh, $index_tag->{offset} + ($num * $LONG_SIZE), 0);
493 $self->fh->print( pack($LONG_PACK, $self->root->{end}) );
495 my $blist_tag = $self->_create_tag($self->root->{end}, SIG_BLIST, chr(0) x $BUCKET_LIST_SIZE);
497 seek($self->fh, $blist_tag->{offset}, 0);
498 $self->fh->print( $key . pack($LONG_PACK, $old_subloc || $self->root->{end}) );
503 $location ||= $self->root->{end};
504 } # re-index bucket list
507 # Seek to content area and store signature, value and plaintext key
511 seek($self->fh, $location, 0);
514 # Write signature based on content type, set content length and write actual value.
516 my $r = Scalar::Util::reftype($value) || '';
518 $self->fh->print( TYPE_HASH );
519 $self->fh->print( pack($DATA_LENGTH_PACK, $INDEX_SIZE) . chr(0) x $INDEX_SIZE );
520 $content_length = $INDEX_SIZE;
522 elsif ($r eq 'ARRAY') {
523 $self->fh->print( TYPE_ARRAY );
524 $self->fh->print( pack($DATA_LENGTH_PACK, $INDEX_SIZE) . chr(0) x $INDEX_SIZE );
525 $content_length = $INDEX_SIZE;
527 elsif (!defined($value)) {
528 $self->fh->print( SIG_NULL );
529 $self->fh->print( pack($DATA_LENGTH_PACK, 0) );
533 $self->fh->print( SIG_DATA );
534 $self->fh->print( pack($DATA_LENGTH_PACK, length($value)) . $value );
535 $content_length = length($value);
539 # Plain key is stored AFTER value, as keys are typically fetched less often.
541 $self->fh->print( pack($DATA_LENGTH_PACK, length($plain_key)) . $plain_key );
544 # If value is blessed, preserve class name
546 if ( $self->root->{autobless} ) {
547 my $value_class = Scalar::Util::blessed($value);
548 if ( defined $value_class && $value_class ne 'DBM::Deep' ) {
550 # Blessed ref -- will restore later
552 $self->fh->print( chr(1) );
553 $self->fh->print( pack($DATA_LENGTH_PACK, length($value_class)) . $value_class );
554 $content_length += 1;
555 $content_length += $DATA_LENGTH_SIZE + length($value_class);
558 $self->fh->print( chr(0) );
559 $content_length += 1;
564 # If this is a new content area, advance EOF counter
566 if ($location == $self->root->{end}) {
567 $self->root->{end} += SIG_SIZE;
568 $self->root->{end} += $DATA_LENGTH_SIZE + $content_length;
569 $self->root->{end} += $DATA_LENGTH_SIZE + length($plain_key);
573 # If content is a hash or array, create new child DeepDB object and
574 # pass each key or element to it.
577 my $branch = DBM::Deep->new(
579 base_offset => $location,
582 foreach my $key (keys %{$value}) {
583 $branch->{$key} = $value->{$key};
586 elsif ($r eq 'ARRAY') {
587 my $branch = DBM::Deep->new(
589 base_offset => $location,
593 foreach my $element (@{$value}) {
594 $branch->[$index] = $element;
602 return $self->_throw_error("Fatal error: indexing failed -- possibly due to corruption in file");
605 sub _get_bucket_value {
607 # Fetch single value given tag and MD5 digested key.
610 my ($tag, $md5) = @_;
611 my $keys = $tag->{content};
614 # Iterate through buckets, looking for a key match
617 for (my $i=0; $i<$MAX_BUCKETS; $i++) {
618 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
619 my $subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
623 # Hit end of list, no match
628 if ( $md5 ne $key ) {
633 # Found match -- seek to offset and read signature
636 seek($self->fh, $subloc, 0);
637 $self->fh->read($signature, SIG_SIZE);
640 # If value is a hash or array, return new DeepDB object with correct offset
642 if (($signature eq TYPE_HASH) || ($signature eq TYPE_ARRAY)) {
643 my $obj = DBM::Deep->new(
645 base_offset => $subloc,
649 if ($self->root->{autobless}) {
651 # Skip over value and plain key to see if object needs
654 seek($self->fh, $DATA_LENGTH_SIZE + $INDEX_SIZE, 1);
657 $self->fh->read($size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
658 if ($size) { seek($self->fh, $size, 1); }
661 $self->fh->read($bless_bit, 1);
662 if (ord($bless_bit)) {
664 # Yes, object needs to be re-blessed
667 $self->fh->read($size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
668 if ($size) { $self->fh->read($class_name, $size); }
669 if ($class_name) { $obj = bless( $obj, $class_name ); }
677 # Otherwise return actual value
679 elsif ($signature eq SIG_DATA) {
682 $self->fh->read($size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
683 if ($size) { $self->fh->read($value, $size); }
688 # Key exists, but content is null
698 # Delete single key/value pair given tag and MD5 digested key.
701 my ($tag, $md5) = @_;
702 my $keys = $tag->{content};
705 # Iterate through buckets, looking for a key match
708 for (my $i=0; $i<$MAX_BUCKETS; $i++) {
709 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
710 my $subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
714 # Hit end of list, no match
719 if ( $md5 ne $key ) {
724 # Matched key -- delete bucket and return
726 seek($self->fh, $tag->{offset} + ($i * $BUCKET_SIZE), 0);
727 $self->fh->print( substr($keys, ($i+1) * $BUCKET_SIZE ) );
728 $self->fh->print( chr(0) x $BUCKET_SIZE );
738 # Check existence of single key given tag and MD5 digested key.
741 my ($tag, $md5) = @_;
742 my $keys = $tag->{content};
745 # Iterate through buckets, looking for a key match
748 for (my $i=0; $i<$MAX_BUCKETS; $i++) {
749 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
750 my $subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
754 # Hit end of list, no match
759 if ( $md5 ne $key ) {
764 # Matched key -- return true
772 sub _find_bucket_list {
774 # Locate offset for bucket list, given digested key
780 # Locate offset for bucket list using digest index system
783 my $tag = $self->_load_tag($self->base_offset);
784 if (!$tag) { return; }
786 while ($tag->{signature} ne SIG_BLIST) {
787 $tag = $self->_index_lookup($tag, ord(substr($md5, $ch, 1)));
788 if (!$tag) { return; }
795 sub _traverse_index {
797 # Scan index and recursively step into deeper levels, looking for next key.
799 my ($self, $offset, $ch, $force_return_next) = @_;
800 $force_return_next = undef unless $force_return_next;
802 my $tag = $self->_load_tag( $offset );
804 if ($tag->{signature} ne SIG_BLIST) {
805 my $content = $tag->{content};
807 if ($self->{return_next}) { $start = 0; }
808 else { $start = ord(substr($self->{prev_md5}, $ch, 1)); }
810 for (my $index = $start; $index < 256; $index++) {
811 my $subloc = unpack($LONG_PACK, substr($content, $index * $LONG_SIZE, $LONG_SIZE) );
813 my $result = $self->_traverse_index( $subloc, $ch + 1, $force_return_next );
814 if (defined($result)) { return $result; }
818 $self->{return_next} = 1;
821 elsif ($tag->{signature} eq SIG_BLIST) {
822 my $keys = $tag->{content};
823 if ($force_return_next) { $self->{return_next} = 1; }
826 # Iterate through buckets, looking for a key match
828 for (my $i=0; $i<$MAX_BUCKETS; $i++) {
829 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
830 my $subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
834 # End of bucket list -- return to outer loop
836 $self->{return_next} = 1;
839 elsif ($key eq $self->{prev_md5}) {
841 # Located previous key -- return next one found
843 $self->{return_next} = 1;
846 elsif ($self->{return_next}) {
848 # Seek to bucket location and skip over signature
850 seek($self->fh, $subloc + SIG_SIZE, 0);
853 # Skip over value to get to plain key
856 $self->fh->read($size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
857 if ($size) { seek($self->fh, $size, 1); }
860 # Read in plain key and return as scalar
863 $self->fh->read($size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
864 if ($size) { $self->fh->read($plain_key, $size); }
870 $self->{return_next} = 1;
871 } # tag is a bucket list
878 # Locate next key, given digested previous one
880 my $self = _get_self($_[0]);
882 $self->{prev_md5} = $_[1] ? $_[1] : undef;
883 $self->{return_next} = 0;
886 # If the previous key was not specifed, start at the top and
887 # return the first one found.
889 if (!$self->{prev_md5}) {
890 $self->{prev_md5} = chr(0) x $HASH_SIZE;
891 $self->{return_next} = 1;
894 return $self->_traverse_index( $self->base_offset, 0 );
899 # If db locking is set, flock() the db file. If called multiple
900 # times before unlock(), then the same number of unlocks() must
901 # be called before the lock is released.
903 my $self = _get_self($_[0]);
905 $type = LOCK_EX unless defined $type;
907 if ($self->root->{locking}) {
908 if (!$self->root->{locked}) { flock($self->fh, $type); }
909 $self->root->{locked}++;
915 # If db locking is set, unlock the db file. See note in lock()
916 # regarding calling lock() multiple times.
918 my $self = _get_self($_[0]);
920 if ($self->root->{locking} && $self->root->{locked} > 0) {
921 $self->root->{locked}--;
922 if (!$self->root->{locked}) { flock($self->fh, LOCK_UN); }
926 #XXX These uses of ref() need verified
929 # Copy single level of keys or elements to new DB handle.
930 # Recurse for nested structures
932 my $self = _get_self($_[0]);
935 if ($self->type eq TYPE_HASH) {
936 my $key = $self->first_key();
938 my $value = $self->get($key);
939 #XXX This doesn't work with autobless
940 if (!ref($value)) { $db_temp->{$key} = $value; }
942 my $type = $value->type;
943 if ($type eq TYPE_HASH) { $db_temp->{$key} = {}; }
944 else { $db_temp->{$key} = []; }
945 $value->_copy_node( $db_temp->{$key} );
947 $key = $self->next_key($key);
951 my $length = $self->length();
952 for (my $index = 0; $index < $length; $index++) {
953 my $value = $self->get($index);
954 if (!ref($value)) { $db_temp->[$index] = $value; }
955 #XXX NO tests for this code
957 my $type = $value->type;
958 if ($type eq TYPE_HASH) { $db_temp->[$index] = {}; }
959 else { $db_temp->[$index] = []; }
960 $value->_copy_node( $db_temp->[$index] );
968 # Recursively export into standard Perl hashes and arrays.
970 my $self = _get_self($_[0]);
973 if ($self->type eq TYPE_HASH) { $temp = {}; }
974 elsif ($self->type eq TYPE_ARRAY) { $temp = []; }
977 $self->_copy_node( $temp );
985 # Recursively import Perl hash/array structure
987 #XXX This use of ref() seems to be ok
988 if (!ref($_[0])) { return; } # Perl calls import() on use -- ignore
990 my $self = _get_self($_[0]);
993 #XXX This use of ref() seems to be ok
996 # struct is not a reference, so just import based on our type
1000 if ($self->type eq TYPE_HASH) { $struct = {@_}; }
1001 elsif ($self->type eq TYPE_ARRAY) { $struct = [@_]; }
1004 my $r = Scalar::Util::reftype($struct) || '';
1005 if ($r eq "HASH" && $self->type eq TYPE_HASH) {
1006 foreach my $key (keys %$struct) { $self->put($key, $struct->{$key}); }
1008 elsif ($r eq "ARRAY" && $self->type eq TYPE_ARRAY) {
1009 $self->push( @$struct );
1012 return $self->_throw_error("Cannot import: type mismatch");
1020 # Rebuild entire database into new file, then move
1021 # it back on top of original.
1023 my $self = _get_self($_[0]);
1024 if ($self->root->{links} > 1) {
1025 return $self->_throw_error("Cannot optimize: reference count is greater than 1");
1028 my $db_temp = DBM::Deep->new(
1029 file => $self->root->{file} . '.tmp',
1033 return $self->_throw_error("Cannot optimize: failed to open temp file: $!");
1037 $self->_copy_node( $db_temp );
1041 # Attempt to copy user, group and permissions over to new file
1043 my @stats = stat($self->fh);
1044 my $perms = $stats[2] & 07777;
1045 my $uid = $stats[4];
1046 my $gid = $stats[5];
1047 chown( $uid, $gid, $self->root->{file} . '.tmp' );
1048 chmod( $perms, $self->root->{file} . '.tmp' );
1050 # q.v. perlport for more information on this variable
1051 if ( $^O eq 'MSWin32' ) {
1053 # Potential race condition when optmizing on Win32 with locking.
1054 # The Windows filesystem requires that the filehandle be closed
1055 # before it is overwritten with rename(). This could be redone
1062 if (!rename $self->root->{file} . '.tmp', $self->root->{file}) {
1063 unlink $self->root->{file} . '.tmp';
1065 return $self->_throw_error("Optimize failed: Cannot copy temp file over original: $!");
1077 # Make copy of object and return
1079 my $self = _get_self($_[0]);
1081 return DBM::Deep->new(
1082 type => $self->type,
1083 base_offset => $self->base_offset,
1089 my %is_legal_filter = map {
1092 store_key store_value
1093 fetch_key fetch_value
1098 # Setup filter function for storing or fetching the key or value
1100 my $self = _get_self($_[0]);
1101 my $type = lc $_[1];
1102 my $func = $_[2] ? $_[2] : undef;
1104 if ( $is_legal_filter{$type} ) {
1105 $self->root->{"filter_$type"} = $func;
1119 # Get access to the root structure
1121 my $self = _get_self($_[0]);
1122 return $self->{root};
1127 # Get access to the raw FileHandle
1129 #XXX It will be useful, though, when we split out HASH and ARRAY
1130 my $self = _get_self($_[0]);
1131 return $self->root->{fh};
1136 # Get type of current node (TYPE_HASH or TYPE_ARRAY)
1138 my $self = _get_self($_[0]);
1139 return $self->{type};
1144 # Get base_offset of current node (TYPE_HASH or TYPE_ARRAY)
1146 my $self = _get_self($_[0]);
1147 return $self->{base_offset};
1152 # Get last error string, or undef if no error
1155 ? ( _get_self($_[0])->{root}->{error} or undef )
1165 # Store error string in self
1167 my $self = _get_self($_[0]);
1168 my $error_text = $_[1];
1170 $self->root->{error} = $error_text;
1172 unless ($self->root->{debug}) {
1173 die "DBM::Deep: $error_text\n";
1176 warn "DBM::Deep: $error_text\n";
1184 my $self = _get_self($_[0]);
1186 undef $self->root->{error};
1191 # Precalculate index, bucket and bucket list sizes
1194 #XXX I don't like this ...
1195 set_pack() unless defined $LONG_SIZE;
1197 $INDEX_SIZE = 256 * $LONG_SIZE;
1198 $BUCKET_SIZE = $HASH_SIZE + $LONG_SIZE;
1199 $BUCKET_LIST_SIZE = $MAX_BUCKETS * $BUCKET_SIZE;
1204 # Set pack/unpack modes (see file header for more)
1206 my ($long_s, $long_p, $data_s, $data_p) = @_;
1208 $LONG_SIZE = $long_s ? $long_s : 4;
1209 $LONG_PACK = $long_p ? $long_p : 'N';
1211 $DATA_LENGTH_SIZE = $data_s ? $data_s : 4;
1212 $DATA_LENGTH_PACK = $data_p ? $data_p : 'N';
1219 # Set key digest function (default is MD5)
1221 my ($digest_func, $hash_size) = @_;
1223 $DIGEST_FUNC = $digest_func ? $digest_func : \&Digest::MD5::md5;
1224 $HASH_SIZE = $hash_size ? $hash_size : 16;
1230 # tie() methods (hashes and arrays)
1235 # Store single hash key/value or array element in database.
1237 my $self = _get_self($_[0]);
1238 my $key = ($self->root->{filter_store_key} && $self->type eq TYPE_HASH) ? $self->root->{filter_store_key}->($_[1]) : $_[1];
1239 #XXX What is ref() checking here?
1240 #YYY User may be storing a hash, in which case we do not want it run
1241 #YYY through the filtering system
1242 my $value = ($self->root->{filter_store_value} && !ref($_[2])) ? $self->root->{filter_store_value}->($_[2]) : $_[2];
1244 my $unpacked_key = $key;
1245 if (($self->type eq TYPE_ARRAY) && ($key =~ /^\d+$/)) { $key = pack($LONG_PACK, $key); }
1246 my $md5 = $DIGEST_FUNC->($key);
1249 # Make sure file is open
1251 if (!defined($self->fh) && !$self->_open()) {
1256 # Request exclusive lock for writing
1258 $self->lock( LOCK_EX );
1261 # If locking is enabled, set 'end' parameter again, in case another
1262 # DB instance appended to our file while we were unlocked.
1264 if ($self->root->{locking} || $self->root->{volatile}) {
1265 $self->root->{end} = (stat($self->fh))[7];
1269 # Locate offset for bucket list using digest index system
1271 my $tag = $self->_load_tag($self->base_offset);
1273 $tag = $self->_create_tag($self->base_offset, SIG_INDEX, chr(0) x $INDEX_SIZE);
1277 while ($tag->{signature} ne SIG_BLIST) {
1278 my $num = ord(substr($md5, $ch, 1));
1279 my $new_tag = $self->_index_lookup($tag, $num);
1281 my $ref_loc = $tag->{offset} + ($num * $LONG_SIZE);
1282 seek($self->fh, $ref_loc, 0);
1283 $self->fh->print( pack($LONG_PACK, $self->root->{end}) );
1285 $tag = $self->_create_tag($self->root->{end}, SIG_BLIST, chr(0) x $BUCKET_LIST_SIZE);
1286 $tag->{ref_loc} = $ref_loc;
1291 my $ref_loc = $tag->{offset} + ($num * $LONG_SIZE);
1293 $tag->{ref_loc} = $ref_loc;
1300 # Add key/value to bucket list
1302 my $result = $self->_add_bucket( $tag, $md5, $key, $value );
1305 # If this object is an array, and bucket was not a replace, and key is numerical,
1306 # and index is equal or greater than current length, advance length variable.
1308 if (($result == 2) && ($self->type eq TYPE_ARRAY) && ($unpacked_key =~ /^\d+$/) && ($unpacked_key >= $self->FETCHSIZE())) {
1309 $self->STORESIZE( $unpacked_key + 1 );
1319 # Fetch single value or element given plain key or array index
1321 my $self = _get_self($_[0]);
1324 if ( $self->type eq TYPE_HASH ) {
1325 if ( my $filter = $self->root->{filter_store_key} ) {
1326 $key = $filter->( $key );
1329 elsif ( $self->type eq TYPE_ARRAY ) {
1330 if ( $key =~ /^\d+$/ ) {
1331 $key = pack($LONG_PACK, $key);
1335 my $md5 = $DIGEST_FUNC->($key);
1338 # Make sure file is open
1340 if (!defined($self->fh)) { $self->_open(); }
1343 # Request shared lock for reading
1345 $self->lock( LOCK_SH );
1347 my $tag = $self->_find_bucket_list( $md5 );
1354 # Get value from bucket list
1356 my $result = $self->_get_bucket_value( $tag, $md5 );
1360 #XXX What is ref() checking here?
1361 return ($result && !ref($result) && $self->root->{filter_fetch_value}) ? $self->root->{filter_fetch_value}->($result) : $result;
1366 # Delete single key/value pair or element given plain key or array index
1368 my $self = _get_self($_[0]);
1369 my $key = ($self->root->{filter_store_key} && $self->type eq TYPE_HASH) ? $self->root->{filter_store_key}->($_[1]) : $_[1];
1371 my $unpacked_key = $key;
1372 if (($self->type eq TYPE_ARRAY) && ($key =~ /^\d+$/)) { $key = pack($LONG_PACK, $key); }
1373 my $md5 = $DIGEST_FUNC->($key);
1376 # Make sure file is open
1378 if (!defined($self->fh)) { $self->_open(); }
1381 # Request exclusive lock for writing
1383 $self->lock( LOCK_EX );
1385 my $tag = $self->_find_bucket_list( $md5 );
1394 my $result = $self->_delete_bucket( $tag, $md5 );
1397 # If this object is an array and the key deleted was on the end of the stack,
1398 # decrement the length variable.
1400 if ($result && ($self->type eq TYPE_ARRAY) && ($unpacked_key == $self->FETCHSIZE() - 1)) {
1401 $self->STORESIZE( $unpacked_key );
1411 # Check if a single key or element exists given plain key or array index
1413 my $self = _get_self($_[0]);
1414 my $key = ($self->root->{filter_store_key} && $self->type eq TYPE_HASH) ? $self->root->{filter_store_key}->($_[1]) : $_[1];
1416 if (($self->type eq TYPE_ARRAY) && ($key =~ /^\d+$/)) { $key = pack($LONG_PACK, $key); }
1417 my $md5 = $DIGEST_FUNC->($key);
1420 # Make sure file is open
1422 if (!defined($self->fh)) { $self->_open(); }
1425 # Request shared lock for reading
1427 $self->lock( LOCK_SH );
1429 my $tag = $self->_find_bucket_list( $md5 );
1432 # For some reason, the built-in exists() function returns '' for false
1440 # Check if bucket exists and return 1 or ''
1442 my $result = $self->_bucket_exists( $tag, $md5 ) || '';
1451 # Clear all keys from hash, or all elements from array.
1453 my $self = _get_self($_[0]);
1456 # Make sure file is open
1458 if (!defined($self->fh)) { $self->_open(); }
1461 # Request exclusive lock for writing
1463 $self->lock( LOCK_EX );
1465 seek($self->fh, $self->base_offset, 0);
1466 if ($self->fh->eof()) {
1471 $self->_create_tag($self->base_offset, $self->type, chr(0) x $INDEX_SIZE);
1480 # Locate and return first key (in no particular order)
1482 my $self = _get_self($_[0]);
1483 if ($self->type ne TYPE_HASH) {
1484 return $self->_throw_error("FIRSTKEY method only supported for hashes");
1488 # Make sure file is open
1490 if (!defined($self->fh)) { $self->_open(); }
1493 # Request shared lock for reading
1495 $self->lock( LOCK_SH );
1497 my $result = $self->_get_next_key();
1501 return ($result && $self->root->{filter_fetch_key}) ? $self->root->{filter_fetch_key}->($result) : $result;
1506 # Return next key (in no particular order), given previous one
1508 my $self = _get_self($_[0]);
1509 if ($self->type ne TYPE_HASH) {
1510 return $self->_throw_error("NEXTKEY method only supported for hashes");
1512 my $prev_key = ($self->root->{filter_store_key} && $self->type eq TYPE_HASH) ? $self->root->{filter_store_key}->($_[1]) : $_[1];
1513 my $prev_md5 = $DIGEST_FUNC->($prev_key);
1516 # Make sure file is open
1518 if (!defined($self->fh)) { $self->_open(); }
1521 # Request shared lock for reading
1523 $self->lock( LOCK_SH );
1525 my $result = $self->_get_next_key( $prev_md5 );
1529 return ($result && $self->root->{filter_fetch_key}) ? $self->root->{filter_fetch_key}->($result) : $result;
1533 # The following methods are for arrays only
1538 # Return the length of the array
1540 my $self = _get_self($_[0]);
1541 if ($self->type ne TYPE_ARRAY) {
1542 return $self->_throw_error("FETCHSIZE method only supported for arrays");
1545 my $SAVE_FILTER = $self->root->{filter_fetch_value};
1546 $self->root->{filter_fetch_value} = undef;
1548 my $packed_size = $self->FETCH('length');
1550 $self->root->{filter_fetch_value} = $SAVE_FILTER;
1552 if ($packed_size) { return int(unpack($LONG_PACK, $packed_size)); }
1558 # Set the length of the array
1560 my $self = _get_self($_[0]);
1561 if ($self->type ne TYPE_ARRAY) {
1562 return $self->_throw_error("STORESIZE method only supported for arrays");
1564 my $new_length = $_[1];
1566 my $SAVE_FILTER = $self->root->{filter_store_value};
1567 $self->root->{filter_store_value} = undef;
1569 my $result = $self->STORE('length', pack($LONG_PACK, $new_length));
1571 $self->root->{filter_store_value} = $SAVE_FILTER;
1578 # Remove and return the last element on the array
1580 my $self = _get_self($_[0]);
1581 if ($self->type ne TYPE_ARRAY) {
1582 return $self->_throw_error("POP method only supported for arrays");
1584 my $length = $self->FETCHSIZE();
1587 my $content = $self->FETCH( $length - 1 );
1588 $self->DELETE( $length - 1 );
1598 # Add new element(s) to the end of the array
1600 my $self = _get_self(shift);
1601 if ($self->type ne TYPE_ARRAY) {
1602 return $self->_throw_error("PUSH method only supported for arrays");
1604 my $length = $self->FETCHSIZE();
1606 while (my $content = shift @_) {
1607 $self->STORE( $length, $content );
1614 # Remove and return first element on the array.
1615 # Shift over remaining elements to take up space.
1617 my $self = _get_self($_[0]);
1618 if ($self->type ne TYPE_ARRAY) {
1619 return $self->_throw_error("SHIFT method only supported for arrays");
1621 my $length = $self->FETCHSIZE();
1624 my $content = $self->FETCH( 0 );
1627 # Shift elements over and remove last one.
1629 for (my $i = 0; $i < $length - 1; $i++) {
1630 $self->STORE( $i, $self->FETCH($i + 1) );
1632 $self->DELETE( $length - 1 );
1643 # Insert new element(s) at beginning of array.
1644 # Shift over other elements to make space.
1646 my $self = _get_self($_[0]);shift @_;
1647 if ($self->type ne TYPE_ARRAY) {
1648 return $self->_throw_error("UNSHIFT method only supported for arrays");
1650 my @new_elements = @_;
1651 my $length = $self->FETCHSIZE();
1652 my $new_size = scalar @new_elements;
1655 for (my $i = $length - 1; $i >= 0; $i--) {
1656 $self->STORE( $i + $new_size, $self->FETCH($i) );
1660 for (my $i = 0; $i < $new_size; $i++) {
1661 $self->STORE( $i, $new_elements[$i] );
1667 # Splices section of array with optional new section.
1668 # Returns deleted section, or last element deleted in scalar context.
1670 my $self = _get_self($_[0]);shift @_;
1671 if ($self->type ne TYPE_ARRAY) {
1672 return $self->_throw_error("SPLICE method only supported for arrays");
1674 my $length = $self->FETCHSIZE();
1677 # Calculate offset and length of splice
1679 my $offset = shift || 0;
1680 if ($offset < 0) { $offset += $length; }
1683 if (scalar @_) { $splice_length = shift; }
1684 else { $splice_length = $length - $offset; }
1685 if ($splice_length < 0) { $splice_length += ($length - $offset); }
1688 # Setup array with new elements, and copy out old elements for return
1690 my @new_elements = @_;
1691 my $new_size = scalar @new_elements;
1693 my @old_elements = ();
1694 for (my $i = $offset; $i < $offset + $splice_length; $i++) {
1695 push @old_elements, $self->FETCH( $i );
1699 # Adjust array length, and shift elements to accomodate new section.
1701 if ( $new_size != $splice_length ) {
1702 if ($new_size > $splice_length) {
1703 for (my $i = $length - 1; $i >= $offset + $splice_length; $i--) {
1704 $self->STORE( $i + ($new_size - $splice_length), $self->FETCH($i) );
1708 for (my $i = $offset + $splice_length; $i < $length; $i++) {
1709 $self->STORE( $i + ($new_size - $splice_length), $self->FETCH($i) );
1711 for (my $i = 0; $i < $splice_length - $new_size; $i++) {
1712 $self->DELETE( $length - 1 );
1719 # Insert new elements into array
1721 for (my $i = $offset; $i < $offset + $new_size; $i++) {
1722 $self->STORE( $i, shift @new_elements );
1726 # Return deleted section, or last element in scalar context.
1728 return wantarray ? @old_elements : $old_elements[-1];
1731 #XXX We don't need to define it.
1732 #XXX It will be useful, though, when we split out HASH and ARRAY
1735 # Perl will call EXTEND() when the array is likely to grow.
1736 # We don't care, but include it for compatibility.
1741 # Public method aliases
1743 *put = *store = *STORE;
1744 *get = *fetch = *FETCH;
1748 *first_key = *FIRSTKEY;
1749 *next_key = *NEXTKEY;
1750 *length = *FETCHSIZE;
1754 *unshift = *UNSHIFT;
1763 DBM::Deep - A pure perl multi-level hash/array DBM
1768 my $db = DBM::Deep->new( "foo.db" );
1770 $db->{key} = 'value'; # tie() style
1773 $db->put('key', 'value'); # OO style
1774 print $db->get('key');
1776 # true multi-level support
1777 $db->{my_complex} = [
1778 'hello', { perl => 'rules' },
1783 A unique flat-file database module, written in pure perl. True
1784 multi-level hash/array support (unlike MLDBM, which is faked), hybrid
1785 OO / tie() interface, cross-platform FTPable files, and quite fast. Can
1786 handle millions of keys and unlimited hash levels without significant
1787 slow-down. Written from the ground-up in pure perl -- this is NOT a
1788 wrapper around a C-based DBM. Out-of-the-box compatibility with Unix,
1789 Mac OS X and Windows.
1793 Hopefully you are using CPAN's excellent Perl module, which will download
1794 and install the module for you. If not, get the tarball, and run these
1806 Construction can be done OO-style (which is the recommended way), or using
1807 Perl's tie() function. Both are examined here.
1809 =head2 OO CONSTRUCTION
1811 The recommended way to construct a DBM::Deep object is to use the new()
1812 method, which gets you a blessed, tied hash or array reference.
1814 my $db = DBM::Deep->new( "foo.db" );
1816 This opens a new database handle, mapped to the file "foo.db". If this
1817 file does not exist, it will automatically be created. DB files are
1818 opened in "r+" (read/write) mode, and the type of object returned is a
1819 hash, unless otherwise specified (see L<OPTIONS> below).
1823 You can pass a number of options to the constructor to specify things like
1824 locking, autoflush, etc. This is done by passing an inline hash:
1826 my $db = DBM::Deep->new(
1832 Notice that the filename is now specified I<inside> the hash with
1833 the "file" parameter, as opposed to being the sole argument to the
1834 constructor. This is required if any options are specified.
1835 See L<OPTIONS> below for the complete list.
1839 You can also start with an array instead of a hash. For this, you must
1840 specify the C<type> parameter:
1842 my $db = DBM::Deep->new(
1844 type => DBM::Deep->TYPE_ARRAY
1847 B<Note:> Specifing the C<type> parameter only takes effect when beginning
1848 a new DB file. If you create a DBM::Deep object with an existing file, the
1849 C<type> will be loaded from the file header, and ignored if it is passed
1852 =head2 TIE CONSTRUCTION
1854 Alternatively, you can create a DBM::Deep handle by using Perl's built-in
1855 tie() function. This is not ideal, because you get only a basic, tied hash
1856 (or array) which is not blessed, so you can't call any functions on it.
1859 tie %hash, "DBM::Deep", "foo.db";
1862 tie @array, "DBM::Deep", "bar.db";
1864 As with the OO constructor, you can replace the DB filename parameter with
1865 a hash containing one or more options (see L<OPTIONS> just below for the
1868 tie %hash, "DBM::Deep", {
1876 There are a number of options that can be passed in when constructing your
1877 DBM::Deep objects. These apply to both the OO- and tie- based approaches.
1883 Filename of the DB file to link the handle to. You can pass a full absolute
1884 filesystem path, partial path, or a plain filename if the file is in the
1885 current working directory. This is a required parameter.
1889 File open mode (read-only, read-write, etc.) string passed to Perl's FileHandle
1890 module. This is an optional parameter, and defaults to "r+" (read/write).
1891 B<Note:> If the default (r+) mode is selected, the file will also be auto-
1892 created if it doesn't exist.
1896 This parameter specifies what type of object to create, a hash or array. Use
1897 one of these two constants: C<DBM::Deep-E<gt>TYPE_HASH> or C<DBM::Deep-E<gt>TYPE_ARRAY>.
1898 This only takes effect when beginning a new file. This is an optional
1899 parameter, and defaults to C<DBM::Deep-E<gt>TYPE_HASH>.
1903 Specifies whether locking is to be enabled. DBM::Deep uses Perl's Fnctl flock()
1904 function to lock the database in exclusive mode for writes, and shared mode for
1905 reads. Pass any true value to enable. This affects the base DB handle I<and
1906 any child hashes or arrays> that use the same DB file. This is an optional
1907 parameter, and defaults to 0 (disabled). See L<LOCKING> below for more.
1911 Specifies whether autoflush is to be enabled on the underlying FileHandle.
1912 This obviously slows down write operations, but is required if you may have
1913 multiple processes accessing the same DB file (also consider enable I<locking>
1914 or at least I<volatile>). Pass any true value to enable. This is an optional
1915 parameter, and defaults to 0 (disabled).
1919 If I<volatile> mode is enabled, DBM::Deep will stat() the DB file before each
1920 STORE() operation. This is required if an outside force may change the size of
1921 the file between transactions. Locking also implicitly enables volatile. This
1922 is useful if you want to use a different locking system or write your own. Pass
1923 any true value to enable. This is an optional parameter, and defaults to 0
1928 If I<autobless> mode is enabled, DBM::Deep will preserve blessed hashes, and
1929 restore them when fetched. This is an B<experimental> feature, and does have
1930 side-effects. Basically, when hashes are re-blessed into their original
1931 classes, they are no longer blessed into the DBM::Deep class! So you won't be
1932 able to call any DBM::Deep methods on them. You have been warned.
1933 This is an optional parameter, and defaults to 0 (disabled).
1937 See L<FILTERS> below.
1941 Setting I<debug> mode will make all errors non-fatal, dump them out to
1942 STDERR, and continue on. This is for debugging purposes only, and probably
1943 not what you want. This is an optional parameter, and defaults to 0 (disabled).
1947 Instead of passing a file path, you can instead pass a handle to an pre-opened
1948 filehandle. Note: Beware of using the magick *DATA handle, as this actually
1949 contains your entire Perl script, as well as the data following the __DATA__
1950 marker. This will not work, because DBM::Deep uses absolute seek()s into the
1951 file. Instead, consider reading *DATA into an IO::Scalar handle, then passing
1956 =head1 TIE INTERFACE
1958 With DBM::Deep you can access your databases using Perl's standard hash/array
1959 syntax. Because all DBM::Deep objects are I<tied> to hashes or arrays, you can treat
1960 them as such. DBM::Deep will intercept all reads/writes and direct them to the right
1961 place -- the DB file. This has nothing to do with the L<TIE CONSTRUCTION>
1962 section above. This simply tells you how to use DBM::Deep using regular hashes
1963 and arrays, rather than calling functions like C<get()> and C<put()> (although those
1964 work too). It is entirely up to you how to want to access your databases.
1968 You can treat any DBM::Deep object like a normal Perl hash reference. Add keys,
1969 or even nested hashes (or arrays) using standard Perl syntax:
1971 my $db = DBM::Deep->new( "foo.db" );
1973 $db->{mykey} = "myvalue";
1975 $db->{myhash}->{subkey} = "subvalue";
1977 print $db->{myhash}->{subkey} . "\n";
1979 You can even step through hash keys using the normal Perl C<keys()> function:
1981 foreach my $key (keys %$db) {
1982 print "$key: " . $db->{$key} . "\n";
1985 Remember that Perl's C<keys()> function extracts I<every> key from the hash and
1986 pushes them onto an array, all before the loop even begins. If you have an
1987 extra large hash, this may exhaust Perl's memory. Instead, consider using
1988 Perl's C<each()> function, which pulls keys/values one at a time, using very
1991 while (my ($key, $value) = each %$db) {
1992 print "$key: $value\n";
1995 Please note that when using C<each()>, you should always pass a direct
1996 hash reference, not a lookup. Meaning, you should B<never> do this:
1999 while (my ($key, $value) = each %{$db->{foo}}) { # BAD
2001 This causes an infinite loop, because for each iteration, Perl is calling
2002 FETCH() on the $db handle, resulting in a "new" hash for foo every time, so
2003 it effectively keeps returning the first key over and over again. Instead,
2004 assign a temporary variable to C<$db->{foo}>, then pass that to each().
2008 As with hashes, you can treat any DBM::Deep object like a normal Perl array
2009 reference. This includes inserting, removing and manipulating elements,
2010 and the C<push()>, C<pop()>, C<shift()>, C<unshift()> and C<splice()> functions.
2011 The object must have first been created using type C<DBM::Deep-E<gt>TYPE_ARRAY>,
2012 or simply be a nested array reference inside a hash. Example:
2014 my $db = DBM::Deep->new(
2015 file => "foo-array.db",
2016 type => DBM::Deep->TYPE_ARRAY
2020 push @$db, "bar", "baz";
2021 unshift @$db, "bah";
2023 my $last_elem = pop @$db; # baz
2024 my $first_elem = shift @$db; # bah
2025 my $second_elem = $db->[1]; # bar
2027 my $num_elements = scalar @$db;
2031 In addition to the I<tie()> interface, you can also use a standard OO interface
2032 to manipulate all aspects of DBM::Deep databases. Each type of object (hash or
2033 array) has its own methods, but both types share the following common methods:
2034 C<put()>, C<get()>, C<exists()>, C<delete()> and C<clear()>.
2040 Stores a new hash key/value pair, or sets an array element value. Takes two
2041 arguments, the hash key or array index, and the new value. The value can be
2042 a scalar, hash ref or array ref. Returns true on success, false on failure.
2044 $db->put("foo", "bar"); # for hashes
2045 $db->put(1, "bar"); # for arrays
2049 Fetches the value of a hash key or array element. Takes one argument: the hash
2050 key or array index. Returns a scalar, hash ref or array ref, depending on the
2053 my $value = $db->get("foo"); # for hashes
2054 my $value = $db->get(1); # for arrays
2058 Checks if a hash key or array index exists. Takes one argument: the hash key
2059 or array index. Returns true if it exists, false if not.
2061 if ($db->exists("foo")) { print "yay!\n"; } # for hashes
2062 if ($db->exists(1)) { print "yay!\n"; } # for arrays
2066 Deletes one hash key/value pair or array element. Takes one argument: the hash
2067 key or array index. Returns true on success, false if not found. For arrays,
2068 the remaining elements located after the deleted element are NOT moved over.
2069 The deleted element is essentially just undefined, which is exactly how Perl's
2070 internal arrays work. Please note that the space occupied by the deleted
2071 key/value or element is B<not> reused again -- see L<UNUSED SPACE RECOVERY>
2072 below for details and workarounds.
2074 $db->delete("foo"); # for hashes
2075 $db->delete(1); # for arrays
2079 Deletes B<all> hash keys or array elements. Takes no arguments. No return
2080 value. Please note that the space occupied by the deleted keys/values or
2081 elements is B<not> reused again -- see L<UNUSED SPACE RECOVERY> below for
2082 details and workarounds.
2084 $db->clear(); # hashes or arrays
2090 For hashes, DBM::Deep supports all the common methods described above, and the
2091 following additional methods: C<first_key()> and C<next_key()>.
2097 Returns the "first" key in the hash. As with built-in Perl hashes, keys are
2098 fetched in an undefined order (which appears random). Takes no arguments,
2099 returns the key as a scalar value.
2101 my $key = $db->first_key();
2105 Returns the "next" key in the hash, given the previous one as the sole argument.
2106 Returns undef if there are no more keys to be fetched.
2108 $key = $db->next_key($key);
2112 Here are some examples of using hashes:
2114 my $db = DBM::Deep->new( "foo.db" );
2116 $db->put("foo", "bar");
2117 print "foo: " . $db->get("foo") . "\n";
2119 $db->put("baz", {}); # new child hash ref
2120 $db->get("baz")->put("buz", "biz");
2121 print "buz: " . $db->get("baz")->get("buz") . "\n";
2123 my $key = $db->first_key();
2125 print "$key: " . $db->get($key) . "\n";
2126 $key = $db->next_key($key);
2129 if ($db->exists("foo")) { $db->delete("foo"); }
2133 For arrays, DBM::Deep supports all the common methods described above, and the
2134 following additional methods: C<length()>, C<push()>, C<pop()>, C<shift()>,
2135 C<unshift()> and C<splice()>.
2141 Returns the number of elements in the array. Takes no arguments.
2143 my $len = $db->length();
2147 Adds one or more elements onto the end of the array. Accepts scalars, hash
2148 refs or array refs. No return value.
2150 $db->push("foo", "bar", {});
2154 Fetches the last element in the array, and deletes it. Takes no arguments.
2155 Returns undef if array is empty. Returns the element value.
2157 my $elem = $db->pop();
2161 Fetches the first element in the array, deletes it, then shifts all the
2162 remaining elements over to take up the space. Returns the element value. This
2163 method is not recommended with large arrays -- see L<LARGE ARRAYS> below for
2166 my $elem = $db->shift();
2170 Inserts one or more elements onto the beginning of the array, shifting all
2171 existing elements over to make room. Accepts scalars, hash refs or array refs.
2172 No return value. This method is not recommended with large arrays -- see
2173 <LARGE ARRAYS> below for details.
2175 $db->unshift("foo", "bar", {});
2179 Performs exactly like Perl's built-in function of the same name. See L<perldoc
2180 -f splice> for usage -- it is too complicated to document here. This method is
2181 not recommended with large arrays -- see L<LARGE ARRAYS> below for details.
2185 Here are some examples of using arrays:
2187 my $db = DBM::Deep->new(
2189 type => DBM::Deep->TYPE_ARRAY
2192 $db->push("bar", "baz");
2193 $db->unshift("foo");
2196 my $len = $db->length();
2197 print "length: $len\n"; # 4
2199 for (my $k=0; $k<$len; $k++) {
2200 print "$k: " . $db->get($k) . "\n";
2203 $db->splice(1, 2, "biz", "baf");
2205 while (my $elem = shift @$db) {
2206 print "shifted: $elem\n";
2211 Enable automatic file locking by passing a true value to the C<locking>
2212 parameter when constructing your DBM::Deep object (see L<SETUP> above).
2214 my $db = DBM::Deep->new(
2219 This causes DBM::Deep to C<flock()> the underlying FileHandle object with exclusive
2220 mode for writes, and shared mode for reads. This is required if you have
2221 multiple processes accessing the same database file, to avoid file corruption.
2222 Please note that C<flock()> does NOT work for files over NFS. See L<DB OVER
2223 NFS> below for more.
2225 =head2 EXPLICIT LOCKING
2227 You can explicitly lock a database, so it remains locked for multiple
2228 transactions. This is done by calling the C<lock()> method, and passing an
2229 optional lock mode argument (defaults to exclusive mode). This is particularly
2230 useful for things like counters, where the current value needs to be fetched,
2231 then incremented, then stored again.
2234 my $counter = $db->get("counter");
2236 $db->put("counter", $counter);
2245 You can pass C<lock()> an optional argument, which specifies which mode to use
2246 (exclusive or shared). Use one of these two constants: C<DBM::Deep-E<gt>LOCK_EX>
2247 or C<DBM::Deep-E<gt>LOCK_SH>. These are passed directly to C<flock()>, and are the
2248 same as the constants defined in Perl's C<Fcntl> module.
2250 $db->lock( DBM::Deep->LOCK_SH );
2254 If you want to implement your own file locking scheme, be sure to create your
2255 DBM::Deep objects setting the C<volatile> option to true. This hints to DBM::Deep
2256 that the DB file may change between transactions. See L<LOW-LEVEL ACCESS>
2259 =head1 IMPORTING/EXPORTING
2261 You can import existing complex structures by calling the C<import()> method,
2262 and export an entire database into an in-memory structure using the C<export()>
2263 method. Both are examined here.
2267 Say you have an existing hash with nested hashes/arrays inside it. Instead of
2268 walking the structure and adding keys/elements to the database as you go,
2269 simply pass a reference to the C<import()> method. This recursively adds
2270 everything to an existing DBM::Deep object for you. Here is an example:
2275 array1 => [ "elem0", "elem1", "elem2" ],
2277 subkey1 => "subvalue1",
2278 subkey2 => "subvalue2"
2282 my $db = DBM::Deep->new( "foo.db" );
2283 $db->import( $struct );
2285 print $db->{key1} . "\n"; # prints "value1"
2287 This recursively imports the entire C<$struct> object into C<$db>, including
2288 all nested hashes and arrays. If the DBM::Deep object contains exsiting data,
2289 keys are merged with the existing ones, replacing if they already exist.
2290 The C<import()> method can be called on any database level (not just the base
2291 level), and works with both hash and array DB types.
2295 B<Note:> Make sure your existing structure has no circular references in it.
2296 These will cause an infinite loop when importing.
2300 Calling the C<export()> method on an existing DBM::Deep object will return
2301 a reference to a new in-memory copy of the database. The export is done
2302 recursively, so all nested hashes/arrays are all exported to standard Perl
2303 objects. Here is an example:
2305 my $db = DBM::Deep->new( "foo.db" );
2307 $db->{key1} = "value1";
2308 $db->{key2} = "value2";
2310 $db->{hash1}->{subkey1} = "subvalue1";
2311 $db->{hash1}->{subkey2} = "subvalue2";
2313 my $struct = $db->export();
2315 print $struct->{key1} . "\n"; # prints "value1"
2317 This makes a complete copy of the database in memory, and returns a reference
2318 to it. The C<export()> method can be called on any database level (not just
2319 the base level), and works with both hash and array DB types. Be careful of
2320 large databases -- you can store a lot more data in a DBM::Deep object than an
2321 in-memory Perl structure.
2325 B<Note:> Make sure your database has no circular references in it.
2326 These will cause an infinite loop when exporting.
2330 DBM::Deep has a number of hooks where you can specify your own Perl function
2331 to perform filtering on incoming or outgoing data. This is a perfect
2332 way to extend the engine, and implement things like real-time compression or
2333 encryption. Filtering applies to the base DB level, and all child hashes /
2334 arrays. Filter hooks can be specified when your DBM::Deep object is first
2335 constructed, or by calling the C<set_filter()> method at any time. There are
2336 four available filter hooks, described below:
2340 =item * filter_store_key
2342 This filter is called whenever a hash key is stored. It
2343 is passed the incoming key, and expected to return a transformed key.
2345 =item * filter_store_value
2347 This filter is called whenever a hash key or array element is stored. It
2348 is passed the incoming value, and expected to return a transformed value.
2350 =item * filter_fetch_key
2352 This filter is called whenever a hash key is fetched (i.e. via
2353 C<first_key()> or C<next_key()>). It is passed the transformed key,
2354 and expected to return the plain key.
2356 =item * filter_fetch_value
2358 This filter is called whenever a hash key or array element is fetched.
2359 It is passed the transformed value, and expected to return the plain value.
2363 Here are the two ways to setup a filter hook:
2365 my $db = DBM::Deep->new(
2367 filter_store_value => \&my_filter_store,
2368 filter_fetch_value => \&my_filter_fetch
2373 $db->set_filter( "filter_store_value", \&my_filter_store );
2374 $db->set_filter( "filter_fetch_value", \&my_filter_fetch );
2376 Your filter function will be called only when dealing with SCALAR keys or
2377 values. When nested hashes and arrays are being stored/fetched, filtering
2378 is bypassed. Filters are called as static functions, passed a single SCALAR
2379 argument, and expected to return a single SCALAR value. If you want to
2380 remove a filter, set the function reference to C<undef>:
2382 $db->set_filter( "filter_store_value", undef );
2384 =head2 REAL-TIME ENCRYPTION EXAMPLE
2386 Here is a working example that uses the I<Crypt::Blowfish> module to
2387 do real-time encryption / decryption of keys & values with DBM::Deep Filters.
2388 Please visit L<http://search.cpan.org/search?module=Crypt::Blowfish> for more
2389 on I<Crypt::Blowfish>. You'll also need the I<Crypt::CBC> module.
2392 use Crypt::Blowfish;
2395 my $cipher = Crypt::CBC->new({
2396 'key' => 'my secret key',
2397 'cipher' => 'Blowfish',
2399 'regenerate_key' => 0,
2400 'padding' => 'space',
2404 my $db = DBM::Deep->new(
2405 file => "foo-encrypt.db",
2406 filter_store_key => \&my_encrypt,
2407 filter_store_value => \&my_encrypt,
2408 filter_fetch_key => \&my_decrypt,
2409 filter_fetch_value => \&my_decrypt,
2412 $db->{key1} = "value1";
2413 $db->{key2} = "value2";
2414 print "key1: " . $db->{key1} . "\n";
2415 print "key2: " . $db->{key2} . "\n";
2421 return $cipher->encrypt( $_[0] );
2424 return $cipher->decrypt( $_[0] );
2427 =head2 REAL-TIME COMPRESSION EXAMPLE
2429 Here is a working example that uses the I<Compress::Zlib> module to do real-time
2430 compression / decompression of keys & values with DBM::Deep Filters.
2431 Please visit L<http://search.cpan.org/search?module=Compress::Zlib> for
2432 more on I<Compress::Zlib>.
2437 my $db = DBM::Deep->new(
2438 file => "foo-compress.db",
2439 filter_store_key => \&my_compress,
2440 filter_store_value => \&my_compress,
2441 filter_fetch_key => \&my_decompress,
2442 filter_fetch_value => \&my_decompress,
2445 $db->{key1} = "value1";
2446 $db->{key2} = "value2";
2447 print "key1: " . $db->{key1} . "\n";
2448 print "key2: " . $db->{key2} . "\n";
2454 return Compress::Zlib::memGzip( $_[0] ) ;
2457 return Compress::Zlib::memGunzip( $_[0] ) ;
2460 B<Note:> Filtering of keys only applies to hashes. Array "keys" are
2461 actually numerical index numbers, and are not filtered.
2463 =head1 ERROR HANDLING
2465 Most DBM::Deep methods return a true value for success, and call die() on
2466 failure. You can wrap calls in an eval block to catch the die. Also, the
2467 actual error message is stored in an internal scalar, which can be fetched by
2468 calling the C<error()> method.
2470 my $db = DBM::Deep->new( "foo.db" ); # create hash
2471 eval { $db->push("foo"); }; # ILLEGAL -- push is array-only call
2473 print $db->error(); # prints error message
2475 You can then call C<clear_error()> to clear the current error state.
2479 If you set the C<debug> option to true when creating your DBM::Deep object,
2480 all errors are considered NON-FATAL, and dumped to STDERR. This is only
2481 for debugging purposes.
2483 =head1 LARGEFILE SUPPORT
2485 If you have a 64-bit system, and your Perl is compiled with both LARGEFILE
2486 and 64-bit support, you I<may> be able to create databases larger than 2 GB.
2487 DBM::Deep by default uses 32-bit file offset tags, but these can be changed
2488 by calling the static C<set_pack()> method before you do anything else.
2490 DBM::Deep::set_pack(8, 'Q');
2492 This tells DBM::Deep to pack all file offsets with 8-byte (64-bit) quad words
2493 instead of 32-bit longs. After setting these values your DB files have a
2494 theoretical maximum size of 16 XB (exabytes).
2498 B<Note:> Changing these values will B<NOT> work for existing database files.
2499 Only change this for new files, and make sure it stays set consistently
2500 throughout the file's life. If you do set these values, you can no longer
2501 access 32-bit DB files. You can, however, call C<set_pack(4, 'N')> to change
2502 back to 32-bit mode.
2506 B<Note:> I have not personally tested files > 2 GB -- all my systems have
2507 only a 32-bit Perl. However, I have received user reports that this does
2510 =head1 LOW-LEVEL ACCESS
2512 If you require low-level access to the underlying FileHandle that DBM::Deep uses,
2513 you can call the C<fh()> method, which returns the handle:
2517 This method can be called on the root level of the datbase, or any child
2518 hashes or arrays. All levels share a I<root> structure, which contains things
2519 like the FileHandle, a reference counter, and all your options you specified
2520 when you created the object. You can get access to this root structure by
2521 calling the C<root()> method.
2523 my $root = $db->root();
2525 This is useful for changing options after the object has already been created,
2526 such as enabling/disabling locking, volatile or debug modes. You can also
2527 store your own temporary user data in this structure (be wary of name
2528 collision), which is then accessible from any child hash or array.
2530 =head1 CUSTOM DIGEST ALGORITHM
2532 DBM::Deep by default uses the I<Message Digest 5> (MD5) algorithm for hashing
2533 keys. However you can override this, and use another algorithm (such as SHA-256)
2534 or even write your own. But please note that DBM::Deep currently expects zero
2535 collisions, so your algorithm has to be I<perfect>, so to speak.
2536 Collision detection may be introduced in a later version.
2540 You can specify a custom digest algorithm by calling the static C<set_digest()>
2541 function, passing a reference to a subroutine, and the length of the algorithm's
2542 hashes (in bytes). This is a global static function, which affects ALL DBM::Deep
2543 objects. Here is a working example that uses a 256-bit hash from the
2544 I<Digest::SHA256> module. Please see
2545 L<http://search.cpan.org/search?module=Digest::SHA256> for more.
2550 my $context = Digest::SHA256::new(256);
2552 DBM::Deep::set_digest( \&my_digest, 32 );
2554 my $db = DBM::Deep->new( "foo-sha.db" );
2556 $db->{key1} = "value1";
2557 $db->{key2} = "value2";
2558 print "key1: " . $db->{key1} . "\n";
2559 print "key2: " . $db->{key2} . "\n";
2565 return substr( $context->hash($_[0]), 0, 32 );
2568 B<Note:> Your returned digest strings must be B<EXACTLY> the number
2569 of bytes you specify in the C<set_digest()> function (in this case 32).
2571 =head1 CIRCULAR REFERENCES
2573 DBM::Deep has B<experimental> support for circular references. Meaning you
2574 can have a nested hash key or array element that points to a parent object.
2575 This relationship is stored in the DB file, and is preserved between sessions.
2578 my $db = DBM::Deep->new( "foo.db" );
2581 $db->{circle} = $db; # ref to self
2583 print $db->{foo} . "\n"; # prints "foo"
2584 print $db->{circle}->{foo} . "\n"; # prints "foo" again
2586 One catch is, passing the object to a function that recursively walks the
2587 object tree (such as I<Data::Dumper> or even the built-in C<optimize()> or
2588 C<export()> methods) will result in an infinite loop. The other catch is,
2589 if you fetch the I<key> of a circular reference (i.e. using the C<first_key()>
2590 or C<next_key()> methods), you will get the I<target object's key>, not the
2591 ref's key. This gets even more interesting with the above example, where
2592 the I<circle> key points to the base DB object, which technically doesn't
2593 have a key. So I made DBM::Deep return "[base]" as the key name in that
2596 =head1 CAVEATS / ISSUES / BUGS
2598 This section describes all the known issues with DBM::Deep. It you have found
2599 something that is not listed here, please send e-mail to L<jhuckaby@cpan.org>.
2601 =head2 UNUSED SPACE RECOVERY
2603 One major caveat with DBM::Deep is that space occupied by existing keys and
2604 values is not recovered when they are deleted. Meaning if you keep deleting
2605 and adding new keys, your file will continuously grow. I am working on this,
2606 but in the meantime you can call the built-in C<optimize()> method from time to
2607 time (perhaps in a crontab or something) to recover all your unused space.
2609 $db->optimize(); # returns true on success
2611 This rebuilds the ENTIRE database into a new file, then moves it on top of
2612 the original. The new file will have no unused space, thus it will take up as
2613 little disk space as possible. Please note that this operation can take
2614 a long time for large files, and you need enough disk space to temporarily hold
2615 2 copies of your DB file. The temporary file is created in the same directory
2616 as the original, named with a ".tmp" extension, and is deleted when the
2617 operation completes. Oh, and if locking is enabled, the DB is automatically
2618 locked for the entire duration of the copy.
2622 B<WARNING:> Only call optimize() on the top-level node of the database, and
2623 make sure there are no child references lying around. DBM::Deep keeps a reference
2624 counter, and if it is greater than 1, optimize() will abort and return undef.
2626 =head2 AUTOVIVIFICATION
2628 Unfortunately, autovivification doesn't work with tied hashes. This appears to
2629 be a bug in Perl's tie() system, as I<Jakob Schmidt> encountered the very same
2630 issue with his I<DWH_FIle> module (see L<http://search.cpan.org/search?module=DWH_File>),
2631 and it is also mentioned in the BUGS section for the I<MLDBM> module <see
2632 L<http://search.cpan.org/search?module=MLDBM>). Basically, on a new db file,
2635 $db->{foo}->{bar} = "hello";
2637 Since "foo" doesn't exist, you cannot add "bar" to it. You end up with "foo"
2638 being an empty hash. Try this instead, which works fine:
2640 $db->{foo} = { bar => "hello" };
2642 As of Perl 5.8.7, this bug still exists. I have walked very carefully through
2643 the execution path, and Perl indeed passes an empty hash to the STORE() method.
2644 Probably a bug in Perl.
2646 =head2 FILE CORRUPTION
2648 The current level of error handling in DBM::Deep is minimal. Files I<are> checked
2649 for a 32-bit signature when opened, but other corruption in files can cause
2650 segmentation faults. DBM::Deep may try to seek() past the end of a file, or get
2651 stuck in an infinite loop depending on the level of corruption. File write
2652 operations are not checked for failure (for speed), so if you happen to run
2653 out of disk space, DBM::Deep will probably fail in a bad way. These things will
2654 be addressed in a later version of DBM::Deep.
2658 Beware of using DB files over NFS. DBM::Deep uses flock(), which works well on local
2659 filesystems, but will NOT protect you from file corruption over NFS. I've heard
2660 about setting up your NFS server with a locking daemon, then using lockf() to
2661 lock your files, but your milage may vary there as well. From what I
2662 understand, there is no real way to do it. However, if you need access to the
2663 underlying FileHandle in DBM::Deep for using some other kind of locking scheme like
2664 lockf(), see the L<LOW-LEVEL ACCESS> section above.
2666 =head2 COPYING OBJECTS
2668 Beware of copying tied objects in Perl. Very strange things can happen.
2669 Instead, use DBM::Deep's C<clone()> method which safely copies the object and
2670 returns a new, blessed, tied hash or array to the same level in the DB.
2672 my $copy = $db->clone();
2676 Beware of using C<shift()>, C<unshift()> or C<splice()> with large arrays.
2677 These functions cause every element in the array to move, which can be murder
2678 on DBM::Deep, as every element has to be fetched from disk, then stored again in
2679 a different location. This may be addressed in a later version.
2683 This section discusses DBM::Deep's speed and memory usage.
2687 Obviously, DBM::Deep isn't going to be as fast as some C-based DBMs, such as
2688 the almighty I<BerkeleyDB>. But it makes up for it in features like true
2689 multi-level hash/array support, and cross-platform FTPable files. Even so,
2690 DBM::Deep is still pretty fast, and the speed stays fairly consistent, even
2691 with huge databases. Here is some test data:
2693 Adding 1,000,000 keys to new DB file...
2695 At 100 keys, avg. speed is 2,703 keys/sec
2696 At 200 keys, avg. speed is 2,642 keys/sec
2697 At 300 keys, avg. speed is 2,598 keys/sec
2698 At 400 keys, avg. speed is 2,578 keys/sec
2699 At 500 keys, avg. speed is 2,722 keys/sec
2700 At 600 keys, avg. speed is 2,628 keys/sec
2701 At 700 keys, avg. speed is 2,700 keys/sec
2702 At 800 keys, avg. speed is 2,607 keys/sec
2703 At 900 keys, avg. speed is 2,190 keys/sec
2704 At 1,000 keys, avg. speed is 2,570 keys/sec
2705 At 2,000 keys, avg. speed is 2,417 keys/sec
2706 At 3,000 keys, avg. speed is 1,982 keys/sec
2707 At 4,000 keys, avg. speed is 1,568 keys/sec
2708 At 5,000 keys, avg. speed is 1,533 keys/sec
2709 At 6,000 keys, avg. speed is 1,787 keys/sec
2710 At 7,000 keys, avg. speed is 1,977 keys/sec
2711 At 8,000 keys, avg. speed is 2,028 keys/sec
2712 At 9,000 keys, avg. speed is 2,077 keys/sec
2713 At 10,000 keys, avg. speed is 2,031 keys/sec
2714 At 20,000 keys, avg. speed is 1,970 keys/sec
2715 At 30,000 keys, avg. speed is 2,050 keys/sec
2716 At 40,000 keys, avg. speed is 2,073 keys/sec
2717 At 50,000 keys, avg. speed is 1,973 keys/sec
2718 At 60,000 keys, avg. speed is 1,914 keys/sec
2719 At 70,000 keys, avg. speed is 2,091 keys/sec
2720 At 80,000 keys, avg. speed is 2,103 keys/sec
2721 At 90,000 keys, avg. speed is 1,886 keys/sec
2722 At 100,000 keys, avg. speed is 1,970 keys/sec
2723 At 200,000 keys, avg. speed is 2,053 keys/sec
2724 At 300,000 keys, avg. speed is 1,697 keys/sec
2725 At 400,000 keys, avg. speed is 1,838 keys/sec
2726 At 500,000 keys, avg. speed is 1,941 keys/sec
2727 At 600,000 keys, avg. speed is 1,930 keys/sec
2728 At 700,000 keys, avg. speed is 1,735 keys/sec
2729 At 800,000 keys, avg. speed is 1,795 keys/sec
2730 At 900,000 keys, avg. speed is 1,221 keys/sec
2731 At 1,000,000 keys, avg. speed is 1,077 keys/sec
2733 This test was performed on a PowerMac G4 1gHz running Mac OS X 10.3.2 & Perl
2734 5.8.1, with an 80GB Ultra ATA/100 HD spinning at 7200RPM. The hash keys and
2735 values were between 6 - 12 chars in length. The DB file ended up at 210MB.
2736 Run time was 12 min 3 sec.
2740 One of the great things about DBM::Deep is that it uses very little memory.
2741 Even with huge databases (1,000,000+ keys) you will not see much increased
2742 memory on your process. DBM::Deep relies solely on the filesystem for storing
2743 and fetching data. Here is output from I</usr/bin/top> before even opening a
2746 PID USER PRI NI SIZE RSS SHARE STAT %CPU %MEM TIME COMMAND
2747 22831 root 11 0 2716 2716 1296 R 0.0 0.2 0:07 perl
2749 Basically the process is taking 2,716K of memory. And here is the same
2750 process after storing and fetching 1,000,000 keys:
2752 PID USER PRI NI SIZE RSS SHARE STAT %CPU %MEM TIME COMMAND
2753 22831 root 14 0 2772 2772 1328 R 0.0 0.2 13:32 perl
2755 Notice the memory usage increased by only 56K. Test was performed on a 700mHz
2756 x86 box running Linux RedHat 7.2 & Perl 5.6.1.
2758 =head1 DB FILE FORMAT
2760 In case you were interested in the underlying DB file format, it is documented
2761 here in this section. You don't need to know this to use the module, it's just
2762 included for reference.
2766 DBM::Deep files always start with a 32-bit signature to identify the file type.
2767 This is at offset 0. The signature is "DPDB" in network byte order. This is
2768 checked when the file is opened.
2772 The DBM::Deep file is in a I<tagged format>, meaning each section of the file
2773 has a standard header containing the type of data, the length of data, and then
2774 the data itself. The type is a single character (1 byte), the length is a
2775 32-bit unsigned long in network byte order, and the data is, well, the data.
2776 Here is how it unfolds:
2780 Immediately after the 32-bit file signature is the I<Master Index> record.
2781 This is a standard tag header followed by 1024 bytes (in 32-bit mode) or 2048
2782 bytes (in 64-bit mode) of data. The type is I<H> for hash or I<A> for array,
2783 depending on how the DBM::Deep object was constructed.
2787 The index works by looking at a I<MD5 Hash> of the hash key (or array index
2788 number). The first 8-bit char of the MD5 signature is the offset into the
2789 index, multipled by 4 in 32-bit mode, or 8 in 64-bit mode. The value of the
2790 index element is a file offset of the next tag for the key/element in question,
2791 which is usually a I<Bucket List> tag (see below).
2795 The next tag I<could> be another index, depending on how many keys/elements
2796 exist. See L<RE-INDEXING> below for details.
2800 A I<Bucket List> is a collection of 16 MD5 hashes for keys/elements, plus
2801 file offsets to where the actual data is stored. It starts with a standard
2802 tag header, with type I<B>, and a data size of 320 bytes in 32-bit mode, or
2803 384 bytes in 64-bit mode. Each MD5 hash is stored in full (16 bytes), plus
2804 the 32-bit or 64-bit file offset for the I<Bucket> containing the actual data.
2805 When the list fills up, a I<Re-Index> operation is performed (See
2806 L<RE-INDEXING> below).
2810 A I<Bucket> is a tag containing a key/value pair (in hash mode), or a
2811 index/value pair (in array mode). It starts with a standard tag header with
2812 type I<D> for scalar data (string, binary, etc.), or it could be a nested
2813 hash (type I<H>) or array (type I<A>). The value comes just after the tag
2814 header. The size reported in the tag header is only for the value, but then,
2815 just after the value is another size (32-bit unsigned long) and then the plain
2816 key itself. Since the value is likely to be fetched more often than the plain
2817 key, I figured it would be I<slightly> faster to store the value first.
2821 If the type is I<H> (hash) or I<A> (array), the value is another I<Master Index>
2822 record for the nested structure, where the process begins all over again.
2826 After a I<Bucket List> grows to 16 records, its allocated space in the file is
2827 exhausted. Then, when another key/element comes in, the list is converted to a
2828 new index record. However, this index will look at the next char in the MD5
2829 hash, and arrange new Bucket List pointers accordingly. This process is called
2830 I<Re-Indexing>. Basically, a new index tag is created at the file EOF, and all
2831 17 (16 + new one) keys/elements are removed from the old Bucket List and
2832 inserted into the new index. Several new Bucket Lists are created in the
2833 process, as a new MD5 char from the key is being examined (it is unlikely that
2834 the keys will all share the same next char of their MD5s).
2838 Because of the way the I<MD5> algorithm works, it is impossible to tell exactly
2839 when the Bucket Lists will turn into indexes, but the first round tends to
2840 happen right around 4,000 keys. You will see a I<slight> decrease in
2841 performance here, but it picks back up pretty quick (see L<SPEED> above). Then
2842 it takes B<a lot> more keys to exhaust the next level of Bucket Lists. It's
2843 right around 900,000 keys. This process can continue nearly indefinitely --
2844 right up until the point the I<MD5> signatures start colliding with each other,
2845 and this is B<EXTREMELY> rare -- like winning the lottery 5 times in a row AND
2846 getting struck by lightning while you are walking to cash in your tickets.
2847 Theoretically, since I<MD5> hashes are 128-bit values, you I<could> have up to
2848 340,282,366,921,000,000,000,000,000,000,000,000,000 keys/elements (I believe
2849 this is 340 unodecillion, but don't quote me).
2853 When a new key/element is stored, the key (or index number) is first ran through
2854 I<Digest::MD5> to get a 128-bit signature (example, in hex:
2855 b05783b0773d894396d475ced9d2f4f6). Then, the I<Master Index> record is checked
2856 for the first char of the signature (in this case I<b>). If it does not exist,
2857 a new I<Bucket List> is created for our key (and the next 15 future keys that
2858 happen to also have I<b> as their first MD5 char). The entire MD5 is written
2859 to the I<Bucket List> along with the offset of the new I<Bucket> record (EOF at
2860 this point, unless we are replacing an existing I<Bucket>), where the actual
2861 data will be stored.
2865 Fetching an existing key/element involves getting a I<Digest::MD5> of the key
2866 (or index number), then walking along the indexes. If there are enough
2867 keys/elements in this DB level, there might be nested indexes, each linked to
2868 a particular char of the MD5. Finally, a I<Bucket List> is pointed to, which
2869 contains up to 16 full MD5 hashes. Each is checked for equality to the key in
2870 question. If we found a match, the I<Bucket> tag is loaded, where the value and
2871 plain key are stored.
2875 Fetching the plain key occurs when calling the I<first_key()> and I<next_key()>
2876 methods. In this process the indexes are walked systematically, and each key
2877 fetched in increasing MD5 order (which is why it appears random). Once the
2878 I<Bucket> is found, the value is skipped the plain key returned instead.
2879 B<Note:> Do not count on keys being fetched as if the MD5 hashes were
2880 alphabetically sorted. This only happens on an index-level -- as soon as the
2881 I<Bucket Lists> are hit, the keys will come out in the order they went in --
2882 so it's pretty much undefined how the keys will come out -- just like Perl's
2885 =head1 CODE COVERAGE
2887 I use B<Devel::Cover> to test the code coverage of my tests, below is the B<Devel::Cover> report on this
2888 module's test suite.
2890 ---------------------------- ------ ------ ------ ------ ------ ------ ------
2891 File stmt bran cond sub pod time total
2892 ---------------------------- ------ ------ ------ ------ ------ ------ ------
2893 blib/lib/DBM/Deep.pm 94.9 84.5 77.8 100.0 11.1 100.0 89.7
2894 Total 94.9 84.5 77.8 100.0 11.1 100.0 89.7
2895 ---------------------------- ------ ------ ------ ------ ------ ------ ------
2899 Joseph Huckaby, L<jhuckaby@cpan.org>
2901 Special thanks to Adam Sah and Rich Gaushell! You know why :-)
2905 perltie(1), Tie::Hash(3), Digest::MD5(3), Fcntl(3), flock(2), lockf(3), nfs(5),
2906 Digest::SHA256(3), Crypt::Blowfish(3), Compress::Zlib(3)
2910 Copyright (c) 2002-2005 Joseph Huckaby. All Rights Reserved.
2911 This is free software, you may use it and distribute it under the
2912 same terms as Perl itself.