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.
34 use Fcntl qw( :DEFAULT :flock :seek );
38 use vars qw( $VERSION );
42 # Set to 4 and 'N' for 32-bit offset tags (default). Theoretical limit of 4 GB per file.
43 # (Perl must be compiled with largefile support for files > 2 GB)
45 # Set to 8 and 'Q' for 64-bit offsets. Theoretical limit of 16 XB per file.
46 # (Perl must be compiled with largefile and 64-bit long support)
52 # Set to 4 and 'N' for 32-bit data length prefixes. Limit of 4 GB for each key/value.
53 # Upgrading this is possible (see above) but probably not necessary. If you need
54 # more than 4 GB for a single key or value, this module is really not for you :-)
56 #my $DATA_LENGTH_SIZE = 4;
57 #my $DATA_LENGTH_PACK = 'N';
58 our ($LONG_SIZE, $LONG_PACK, $DATA_LENGTH_SIZE, $DATA_LENGTH_PACK);
61 # Maximum number of buckets per list before another level of indexing is done.
62 # Increase this value for slightly greater speed, but larger database files.
63 # DO NOT decrease this value below 16, due to risk of recursive reindex overrun.
68 # Better not adjust anything below here, unless you're me :-)
72 # Setup digest function for keys
74 our ($DIGEST_FUNC, $HASH_SIZE);
75 #my $DIGEST_FUNC = \&Digest::MD5::md5;
78 # Precalculate index and bucket sizes based on values above.
81 my ($INDEX_SIZE, $BUCKET_SIZE, $BUCKET_LIST_SIZE);
88 # Setup file and tag signatures. These should never change.
90 sub SIG_FILE () { 'DPDB' }
91 sub SIG_HASH () { 'H' }
92 sub SIG_ARRAY () { 'A' }
93 sub SIG_SCALAR () { 'S' }
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; }
105 sub TYPE_SCALAR () { return SIG_SCALAR; }
111 if (scalar(@_) > 1) {
113 $proto->_throw_error( "Odd number of parameters to " . (caller(1))[2] );
117 elsif ( my $type = Scalar::Util::reftype($_[0]) ) {
118 if ( $type ne 'HASH' ) {
119 $proto->_throw_error( "Not a hashref in args to " . (caller(1))[2] );
124 $args = { file => shift };
132 # Class constructor method for Perl OO interface.
133 # Calls tie() and returns blessed reference to tied hash or array,
134 # providing a hybrid OO/tie interface.
137 my $args = $class->_get_args( @_ );
140 # Check if we want a tied hash or array.
143 if (defined($args->{type}) && $args->{type} eq TYPE_ARRAY) {
144 $class = 'DBM::Deep::Array';
145 require DBM::Deep::Array;
146 tie @$self, $class, %$args;
149 $class = 'DBM::Deep::Hash';
150 require DBM::Deep::Hash;
151 tie %$self, $class, %$args;
154 return bless $self, $class;
159 # Setup $self and bless into this class.
164 # These are the defaults to be optionally overridden below
167 base_offset => length(SIG_FILE),
170 foreach my $param ( keys %$self ) {
171 next unless exists $args->{$param};
172 $self->{$param} = delete $args->{$param}
175 $self->{root} = exists $args->{root}
177 : DBM::Deep::_::Root->new( $args );
179 if (!defined($self->fh)) { $self->_open(); }
186 require DBM::Deep::Hash;
187 return DBM::Deep::Hash->TIEHASH( @_ );
192 require DBM::Deep::Array;
193 return DBM::Deep::Array->TIEARRAY( @_ );
196 #XXX Unneeded now ...
202 # Open a FileHandle to the database, create if nonexistent.
203 # Make sure file signature matches DeepDB spec.
205 my $self = $_[0]->_get_self;
207 if (defined($self->fh)) { $self->_close(); }
210 # Theoretically, adding O_BINARY should remove the need for the binmode
211 # Of course, testing it is going to be ... interesting.
212 my $flags = O_RDWR | O_CREAT | O_BINARY;
214 #XXX Can the mode be anything but r+, w+, or a+??
215 #XXX ie, it has to be in read-write mode
216 #XXX So, should we verify that the mode is legitimate?
218 #XXX Maybe the mode thingy should just go away. There's no good
219 #XXX reason for it ...
220 if ( $self->root->{mode} eq 'w+' ) {
225 sysopen( $fh, $self->root->{file}, $flags )
227 $self->root->{fh} = $fh;
228 }; if ($@ ) { $self->_throw_error( "Received error: $@\n" ); }
229 if (! defined($self->fh)) {
230 return $self->_throw_error("Cannot sysopen file: " . $self->root->{file} . ": $!");
235 #XXX Can we remove this by using the right sysopen() flags?
236 # Maybe ... q.v. above
237 binmode $fh; # for win32
239 if ($self->root->{autoflush}) {
240 my $old = select $fh;
246 seek($fh, 0, SEEK_SET);
249 my $bytes_read = read( $fh, $signature, length(SIG_FILE));
252 # File is empty -- write signature and master index
255 seek($fh, 0, SEEK_SET);
257 $self->_create_tag($self->base_offset, $self->type, chr(0) x $INDEX_SIZE);
259 my $plain_key = "[base]";
260 print($fh pack($DATA_LENGTH_PACK, length($plain_key)) . $plain_key );
262 # Flush the filehandle
263 my $old_fh = select $fh;
269 my @stats = stat($fh);
270 $self->root->{inode} = $stats[1];
271 $self->root->{end} = $stats[7];
277 # Check signature was valid
279 unless ($signature eq SIG_FILE) {
281 return $self->_throw_error("Signature not found -- file is not a Deep DB");
284 my @stats = stat($fh);
285 $self->root->{inode} = $stats[1];
286 $self->root->{end} = $stats[7];
289 # Get our type from master index signature
291 my $tag = $self->_load_tag($self->base_offset);
293 #XXX We probably also want to store the hash algorithm name and not assume anything
294 #XXX The cool thing would be to allow a different hashing algorithm at every level
297 return $self->_throw_error("Corrupted file, no master index record");
299 if ($self->{type} ne $tag->{signature}) {
300 return $self->_throw_error("File type mismatch");
308 # Close database FileHandle
310 my $self = $_[0]->_get_self;
311 close $self->root->{fh};
312 $self->root->{fh} = undef;
317 # Given offset, signature and content, create tag and write to disk
319 my ($self, $offset, $sig, $content) = @_;
320 my $size = length($content);
324 seek($fh, $offset, SEEK_SET);
325 print($fh $sig . pack($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 + $DATA_LENGTH_SIZE,
341 # Given offset, load single tag and return signature, size and data
348 seek($fh, $offset, SEEK_SET);
349 if (eof $fh) { return undef; }
352 read( $fh, $sig, SIG_SIZE);
355 read( $fh, $size, $DATA_LENGTH_SIZE);
356 $size = unpack($DATA_LENGTH_PACK, $size);
359 read( $fh, $buffer, $size);
364 offset => $offset + SIG_SIZE + $DATA_LENGTH_SIZE,
371 # Given index tag, lookup single entry in index and return .
374 my ($tag, $index) = @_;
376 my $location = unpack($LONG_PACK, substr($tag->{content}, $index * $LONG_SIZE, $LONG_SIZE) );
377 if (!$location) { return; }
379 return $self->_load_tag( $location );
384 # Adds one key/value pair to bucket list, given offset, MD5 digest of key,
385 # plain (undigested) key and value.
388 my ($tag, $md5, $plain_key, $value) = @_;
389 my $keys = $tag->{content};
393 # added ref() check first to avoid eval and runtime exception for every
394 # scalar value being stored. performance tweak.
395 my $is_dbm_deep = ref($value) && eval { $value->isa( 'DBM::Deep' ) };
397 my $internal_ref = $is_dbm_deep && ($value->root eq $self->root);
402 # Iterate through buckets, seeing if this is a new entry or a replace.
404 for (my $i=0; $i<$MAX_BUCKETS; $i++) {
405 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
406 my $subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
409 # Found empty bucket (end of list). Populate and exit loop.
413 $location = $internal_ref
414 ? $value->base_offset
415 : $self->root->{end};
417 seek($fh, $tag->{offset} + ($i * $BUCKET_SIZE), SEEK_SET);
418 print($fh $md5 . pack($LONG_PACK, $location) );
421 elsif ($md5 eq $key) {
423 # Found existing bucket with same key. Replace with new value.
428 $location = $value->base_offset;
429 seek($fh, $tag->{offset} + ($i * $BUCKET_SIZE), SEEK_SET);
430 print($fh $md5 . pack($LONG_PACK, $location) );
433 seek($fh, $subloc + SIG_SIZE, SEEK_SET);
435 read( $fh, $size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
438 # If value is a hash, array, or raw value with equal or less size, we can
439 # reuse the same content area of the database. Otherwise, we have to create
440 # a new content area at the EOF.
443 my $r = Scalar::Util::reftype( $value ) || '';
444 if ( $r eq 'HASH' || $r eq 'ARRAY' ) {
445 $actual_length = $INDEX_SIZE;
447 # if autobless is enabled, must also take into consideration
448 # the class name, as it is stored along with key/value.
449 if ( $self->root->{autobless} ) {
450 my $value_class = Scalar::Util::blessed($value);
451 if ( defined $value_class && $value_class ne 'DBM::Deep' ) {
452 $actual_length += length($value_class);
456 else { $actual_length = length($value); }
458 if ($actual_length <= $size) {
462 $location = $self->root->{end};
463 seek($fh, $tag->{offset} + ($i * $BUCKET_SIZE) + $HASH_SIZE, SEEK_SET);
464 print($fh pack($LONG_PACK, $location) );
472 # If this is an internal reference, return now.
473 # No need to write value or plain key
480 # If bucket didn't fit into list, split into a new index level
483 seek($fh, $tag->{ref_loc}, SEEK_SET);
484 print($fh pack($LONG_PACK, $self->root->{end}) );
486 my $index_tag = $self->_create_tag($self->root->{end}, SIG_INDEX, chr(0) x $INDEX_SIZE);
489 $keys .= $md5 . pack($LONG_PACK, 0);
491 for (my $i=0; $i<=$MAX_BUCKETS; $i++) {
492 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
494 my $old_subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
495 my $num = ord(substr($key, $tag->{ch} + 1, 1));
497 if ($offsets[$num]) {
498 my $offset = $offsets[$num] + SIG_SIZE + $DATA_LENGTH_SIZE;
499 seek($fh, $offset, SEEK_SET);
501 read( $fh, $subkeys, $BUCKET_LIST_SIZE);
503 for (my $k=0; $k<$MAX_BUCKETS; $k++) {
504 my $subloc = unpack($LONG_PACK, substr($subkeys, ($k * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
506 seek($fh, $offset + ($k * $BUCKET_SIZE), SEEK_SET);
507 print($fh $key . pack($LONG_PACK, $old_subloc || $self->root->{end}) );
513 $offsets[$num] = $self->root->{end};
514 seek($fh, $index_tag->{offset} + ($num * $LONG_SIZE), SEEK_SET);
515 print($fh pack($LONG_PACK, $self->root->{end}) );
517 my $blist_tag = $self->_create_tag($self->root->{end}, SIG_BLIST, chr(0) x $BUCKET_LIST_SIZE);
519 seek($fh, $blist_tag->{offset}, SEEK_SET);
520 print($fh $key . pack($LONG_PACK, $old_subloc || $self->root->{end}) );
525 $location ||= $self->root->{end};
526 } # re-index bucket list
529 # Seek to content area and store signature, value and plaintext key
533 seek($fh, $location, SEEK_SET);
536 # Write signature based on content type, set content length and write actual value.
538 my $r = Scalar::Util::reftype($value) || '';
540 print($fh TYPE_HASH );
541 print($fh pack($DATA_LENGTH_PACK, $INDEX_SIZE) . chr(0) x $INDEX_SIZE );
542 $content_length = $INDEX_SIZE;
544 elsif ($r eq 'ARRAY') {
545 print($fh TYPE_ARRAY );
546 print($fh pack($DATA_LENGTH_PACK, $INDEX_SIZE) . chr(0) x $INDEX_SIZE );
547 $content_length = $INDEX_SIZE;
549 elsif (!defined($value)) {
550 print($fh SIG_NULL );
551 print($fh pack($DATA_LENGTH_PACK, 0) );
555 print($fh SIG_DATA );
556 print($fh pack($DATA_LENGTH_PACK, length($value)) . $value );
557 $content_length = length($value);
561 # Plain key is stored AFTER value, as keys are typically fetched less often.
563 print($fh pack($DATA_LENGTH_PACK, length($plain_key)) . $plain_key );
566 # If value is blessed, preserve class name
568 if ( $self->root->{autobless} ) {
569 my $value_class = Scalar::Util::blessed($value);
570 if ( defined $value_class && $value_class ne 'DBM::Deep' ) {
572 # Blessed ref -- will restore later
575 print($fh pack($DATA_LENGTH_PACK, length($value_class)) . $value_class );
576 $content_length += 1;
577 $content_length += $DATA_LENGTH_SIZE + length($value_class);
581 $content_length += 1;
586 # If this is a new content area, advance EOF counter
588 if ($location == $self->root->{end}) {
589 $self->root->{end} += SIG_SIZE;
590 $self->root->{end} += $DATA_LENGTH_SIZE + $content_length;
591 $self->root->{end} += $DATA_LENGTH_SIZE + length($plain_key);
595 # If content is a hash or array, create new child DeepDB object and
596 # pass each key or element to it.
599 my $branch = DBM::Deep->new(
601 base_offset => $location,
604 foreach my $key (keys %{$value}) {
605 #$branch->{$key} = $value->{$key};
606 $branch->STORE( $key, $value->{$key} );
609 elsif ($r eq 'ARRAY') {
610 my $branch = DBM::Deep->new(
612 base_offset => $location,
616 foreach my $element (@{$value}) {
617 #$branch->[$index] = $element;
618 $branch->STORE( $index, $element );
626 return $self->_throw_error("Fatal error: indexing failed -- possibly due to corruption in file");
629 sub _get_bucket_value {
631 # Fetch single value given tag and MD5 digested key.
634 my ($tag, $md5) = @_;
635 my $keys = $tag->{content};
640 # Iterate through buckets, looking for a key match
643 for (my $i=0; $i<$MAX_BUCKETS; $i++) {
644 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
645 my $subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
649 # Hit end of list, no match
654 if ( $md5 ne $key ) {
659 # Found match -- seek to offset and read signature
662 seek($fh, $subloc, SEEK_SET);
663 read( $fh, $signature, SIG_SIZE);
666 # If value is a hash or array, return new DeepDB object with correct offset
668 if (($signature eq TYPE_HASH) || ($signature eq TYPE_ARRAY)) {
669 my $obj = DBM::Deep->new(
671 base_offset => $subloc,
675 if ($self->root->{autobless}) {
677 # Skip over value and plain key to see if object needs
680 seek($fh, $DATA_LENGTH_SIZE + $INDEX_SIZE, SEEK_CUR);
683 read( $fh, $size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
684 if ($size) { seek($fh, $size, SEEK_CUR); }
687 read( $fh, $bless_bit, 1);
688 if (ord($bless_bit)) {
690 # Yes, object needs to be re-blessed
693 read( $fh, $size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
694 if ($size) { read( $fh, $class_name, $size); }
695 if ($class_name) { $obj = bless( $obj, $class_name ); }
703 # Otherwise return actual value
705 elsif ($signature eq SIG_DATA) {
708 read( $fh, $size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
709 if ($size) { read( $fh, $value, $size); }
714 # Key exists, but content is null
724 # Delete single key/value pair given tag and MD5 digested key.
727 my ($tag, $md5) = @_;
728 my $keys = $tag->{content};
733 # Iterate through buckets, looking for a key match
736 for (my $i=0; $i<$MAX_BUCKETS; $i++) {
737 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
738 my $subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
742 # Hit end of list, no match
747 if ( $md5 ne $key ) {
752 # Matched key -- delete bucket and return
754 seek($fh, $tag->{offset} + ($i * $BUCKET_SIZE), SEEK_SET);
755 print($fh substr($keys, ($i+1) * $BUCKET_SIZE ) );
756 print($fh chr(0) x $BUCKET_SIZE );
766 # Check existence of single key given tag and MD5 digested key.
769 my ($tag, $md5) = @_;
770 my $keys = $tag->{content};
773 # Iterate through buckets, looking for a key match
776 for (my $i=0; $i<$MAX_BUCKETS; $i++) {
777 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
778 my $subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
782 # Hit end of list, no match
787 if ( $md5 ne $key ) {
792 # Matched key -- return true
800 sub _find_bucket_list {
802 # Locate offset for bucket list, given digested key
808 # Locate offset for bucket list using digest index system
811 my $tag = $self->_load_tag($self->base_offset);
812 if (!$tag) { return; }
814 while ($tag->{signature} ne SIG_BLIST) {
815 $tag = $self->_index_lookup($tag, ord(substr($md5, $ch, 1)));
816 if (!$tag) { return; }
823 sub _traverse_index {
825 # Scan index and recursively step into deeper levels, looking for next key.
827 my ($self, $offset, $ch, $force_return_next) = @_;
828 $force_return_next = undef unless $force_return_next;
830 my $tag = $self->_load_tag( $offset );
834 if ($tag->{signature} ne SIG_BLIST) {
835 my $content = $tag->{content};
837 if ($self->{return_next}) { $start = 0; }
838 else { $start = ord(substr($self->{prev_md5}, $ch, 1)); }
840 for (my $index = $start; $index < 256; $index++) {
841 my $subloc = unpack($LONG_PACK, substr($content, $index * $LONG_SIZE, $LONG_SIZE) );
843 my $result = $self->_traverse_index( $subloc, $ch + 1, $force_return_next );
844 if (defined($result)) { return $result; }
848 $self->{return_next} = 1;
851 elsif ($tag->{signature} eq SIG_BLIST) {
852 my $keys = $tag->{content};
853 if ($force_return_next) { $self->{return_next} = 1; }
856 # Iterate through buckets, looking for a key match
858 for (my $i=0; $i<$MAX_BUCKETS; $i++) {
859 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
860 my $subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
864 # End of bucket list -- return to outer loop
866 $self->{return_next} = 1;
869 elsif ($key eq $self->{prev_md5}) {
871 # Located previous key -- return next one found
873 $self->{return_next} = 1;
876 elsif ($self->{return_next}) {
878 # Seek to bucket location and skip over signature
880 seek($fh, $subloc + SIG_SIZE, SEEK_SET);
883 # Skip over value to get to plain key
886 read( $fh, $size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
887 if ($size) { seek($fh, $size, SEEK_CUR); }
890 # Read in plain key and return as scalar
893 read( $fh, $size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
894 if ($size) { read( $fh, $plain_key, $size); }
900 $self->{return_next} = 1;
901 } # tag is a bucket list
908 # Locate next key, given digested previous one
910 my $self = $_[0]->_get_self;
912 $self->{prev_md5} = $_[1] ? $_[1] : undef;
913 $self->{return_next} = 0;
916 # If the previous key was not specifed, start at the top and
917 # return the first one found.
919 if (!$self->{prev_md5}) {
920 $self->{prev_md5} = chr(0) x $HASH_SIZE;
921 $self->{return_next} = 1;
924 return $self->_traverse_index( $self->base_offset, 0 );
929 # If db locking is set, flock() the db file. If called multiple
930 # times before unlock(), then the same number of unlocks() must
931 # be called before the lock is released.
933 my $self = $_[0]->_get_self;
935 $type = LOCK_EX unless defined $type;
937 if (!defined($self->fh)) { return; }
939 if ($self->root->{locking}) {
940 if (!$self->root->{locked}) {
941 flock($self->fh, $type);
943 # double-check file inode, in case another process
944 # has optimize()d our file while we were waiting.
945 if ((stat($self->root->{file}))[1] != $self->root->{inode}) {
946 $self->_open(); # re-open
947 flock($self->fh, $type); # re-lock
950 $self->root->{locked}++;
960 # If db locking is set, unlock the db file. See note in lock()
961 # regarding calling lock() multiple times.
963 my $self = $_[0]->_get_self;
965 if (!defined($self->fh)) { return; }
967 if ($self->root->{locking} && $self->root->{locked} > 0) {
968 $self->root->{locked}--;
969 if (!$self->root->{locked}) { flock($self->fh, LOCK_UN); }
977 #XXX These uses of ref() need verified
980 # Copy single level of keys or elements to new DB handle.
981 # Recurse for nested structures
983 my $self = $_[0]->_get_self;
986 if ($self->type eq TYPE_HASH) {
987 my $key = $self->first_key();
989 my $value = $self->get($key);
990 #XXX This doesn't work with autobless
991 if (!ref($value)) { $db_temp->{$key} = $value; }
993 my $type = $value->type;
994 if ($type eq TYPE_HASH) { $db_temp->{$key} = {}; }
995 else { $db_temp->{$key} = []; }
996 $value->_copy_node( $db_temp->{$key} );
998 $key = $self->next_key($key);
1002 my $length = $self->length();
1003 for (my $index = 0; $index < $length; $index++) {
1004 my $value = $self->get($index);
1005 if (!ref($value)) { $db_temp->[$index] = $value; }
1006 #XXX NO tests for this code
1008 my $type = $value->type;
1009 if ($type eq TYPE_HASH) { $db_temp->[$index] = {}; }
1010 else { $db_temp->[$index] = []; }
1011 $value->_copy_node( $db_temp->[$index] );
1019 # Recursively export into standard Perl hashes and arrays.
1021 my $self = $_[0]->_get_self;
1024 if ($self->type eq TYPE_HASH) { $temp = {}; }
1025 elsif ($self->type eq TYPE_ARRAY) { $temp = []; }
1028 $self->_copy_node( $temp );
1036 # Recursively import Perl hash/array structure
1038 #XXX This use of ref() seems to be ok
1039 if (!ref($_[0])) { return; } # Perl calls import() on use -- ignore
1041 my $self = $_[0]->_get_self;
1044 #XXX This use of ref() seems to be ok
1045 if (!ref($struct)) {
1047 # struct is not a reference, so just import based on our type
1051 if ($self->type eq TYPE_HASH) { $struct = {@_}; }
1052 elsif ($self->type eq TYPE_ARRAY) { $struct = [@_]; }
1055 my $r = Scalar::Util::reftype($struct) || '';
1056 if ($r eq "HASH" && $self->type eq TYPE_HASH) {
1057 foreach my $key (keys %$struct) { $self->put($key, $struct->{$key}); }
1059 elsif ($r eq "ARRAY" && $self->type eq TYPE_ARRAY) {
1060 $self->push( @$struct );
1063 return $self->_throw_error("Cannot import: type mismatch");
1071 # Rebuild entire database into new file, then move
1072 # it back on top of original.
1074 my $self = $_[0]->_get_self;
1076 #XXX Need to create a new test for this
1077 # if ($self->root->{links} > 1) {
1078 # return $self->_throw_error("Cannot optimize: reference count is greater than 1");
1081 my $db_temp = DBM::Deep->new(
1082 file => $self->root->{file} . '.tmp',
1086 return $self->_throw_error("Cannot optimize: failed to open temp file: $!");
1090 $self->_copy_node( $db_temp );
1094 # Attempt to copy user, group and permissions over to new file
1096 my @stats = stat($self->fh);
1097 my $perms = $stats[2] & 07777;
1098 my $uid = $stats[4];
1099 my $gid = $stats[5];
1100 chown( $uid, $gid, $self->root->{file} . '.tmp' );
1101 chmod( $perms, $self->root->{file} . '.tmp' );
1103 # q.v. perlport for more information on this variable
1104 if ( $^O eq 'MSWin32' ) {
1106 # Potential race condition when optmizing on Win32 with locking.
1107 # The Windows filesystem requires that the filehandle be closed
1108 # before it is overwritten with rename(). This could be redone
1115 if (!rename $self->root->{file} . '.tmp', $self->root->{file}) {
1116 unlink $self->root->{file} . '.tmp';
1118 return $self->_throw_error("Optimize failed: Cannot copy temp file over original: $!");
1130 # Make copy of object and return
1132 my $self = $_[0]->_get_self;
1134 return DBM::Deep->new(
1135 type => $self->type,
1136 base_offset => $self->base_offset,
1142 my %is_legal_filter = map {
1145 store_key store_value
1146 fetch_key fetch_value
1151 # Setup filter function for storing or fetching the key or value
1153 my $self = $_[0]->_get_self;
1154 my $type = lc $_[1];
1155 my $func = $_[2] ? $_[2] : undef;
1157 if ( $is_legal_filter{$type} ) {
1158 $self->root->{"filter_$type"} = $func;
1172 # Get access to the root structure
1174 my $self = $_[0]->_get_self;
1175 return $self->{root};
1180 # Get access to the raw FileHandle
1182 #XXX It will be useful, though, when we split out HASH and ARRAY
1183 my $self = $_[0]->_get_self;
1184 return $self->root->{fh};
1189 # Get type of current node (TYPE_HASH or TYPE_ARRAY)
1191 my $self = $_[0]->_get_self;
1192 return $self->{type};
1197 # Get base_offset of current node (TYPE_HASH or TYPE_ARRAY)
1199 my $self = $_[0]->_get_self;
1200 return $self->{base_offset};
1205 # Get last error string, or undef if no error
1208 #? ( _get_self($_[0])->{root}->{error} or undef )
1209 ? ( $_[0]->_get_self->{root}->{error} or undef )
1219 # Store error string in self
1221 my $self = $_[0]->_get_self;
1222 my $error_text = $_[1];
1224 if ( Scalar::Util::blessed $self ) {
1225 $self->root->{error} = $error_text;
1227 unless ($self->root->{debug}) {
1228 die "DBM::Deep: $error_text\n";
1231 warn "DBM::Deep: $error_text\n";
1235 die "DBM::Deep: $error_text\n";
1243 my $self = $_[0]->_get_self;
1245 undef $self->root->{error};
1250 # Precalculate index, bucket and bucket list sizes
1253 #XXX I don't like this ...
1254 set_pack() unless defined $LONG_SIZE;
1256 $INDEX_SIZE = 256 * $LONG_SIZE;
1257 $BUCKET_SIZE = $HASH_SIZE + $LONG_SIZE;
1258 $BUCKET_LIST_SIZE = $MAX_BUCKETS * $BUCKET_SIZE;
1263 # Set pack/unpack modes (see file header for more)
1265 my ($long_s, $long_p, $data_s, $data_p) = @_;
1267 $LONG_SIZE = $long_s ? $long_s : 4;
1268 $LONG_PACK = $long_p ? $long_p : 'N';
1270 $DATA_LENGTH_SIZE = $data_s ? $data_s : 4;
1271 $DATA_LENGTH_PACK = $data_p ? $data_p : 'N';
1278 # Set key digest function (default is MD5)
1280 my ($digest_func, $hash_size) = @_;
1282 $DIGEST_FUNC = $digest_func ? $digest_func : \&Digest::MD5::md5;
1283 $HASH_SIZE = $hash_size ? $hash_size : 16;
1289 # tie() methods (hashes and arrays)
1294 # Store single hash key/value or array element in database.
1296 my $self = $_[0]->_get_self;
1299 # User may be storing a hash, in which case we do not want it run
1300 # through the filtering system
1301 my $value = ($self->root->{filter_store_value} && !ref($_[2]))
1302 ? $self->root->{filter_store_value}->($_[2])
1305 my $md5 = $DIGEST_FUNC->($key);
1308 # Make sure file is open
1310 if (!defined($self->fh) && !$self->_open()) {
1316 # Request exclusive lock for writing
1318 $self->lock( LOCK_EX );
1323 # If locking is enabled, set 'end' parameter again, in case another
1324 # DB instance appended to our file while we were unlocked.
1326 if ($self->root->{locking} || $self->root->{volatile}) {
1327 $self->root->{end} = (stat($fh))[7];
1331 # Locate offset for bucket list using digest index system
1333 my $tag = $self->_load_tag($self->base_offset);
1335 $tag = $self->_create_tag($self->base_offset, SIG_INDEX, chr(0) x $INDEX_SIZE);
1339 while ($tag->{signature} ne SIG_BLIST) {
1340 my $num = ord(substr($md5, $ch, 1));
1341 my $new_tag = $self->_index_lookup($tag, $num);
1343 my $ref_loc = $tag->{offset} + ($num * $LONG_SIZE);
1344 seek($fh, $ref_loc, SEEK_SET);
1345 print($fh pack($LONG_PACK, $self->root->{end}) );
1347 $tag = $self->_create_tag($self->root->{end}, SIG_BLIST, chr(0) x $BUCKET_LIST_SIZE);
1348 $tag->{ref_loc} = $ref_loc;
1353 my $ref_loc = $tag->{offset} + ($num * $LONG_SIZE);
1355 $tag->{ref_loc} = $ref_loc;
1362 # Add key/value to bucket list
1364 my $result = $self->_add_bucket( $tag, $md5, $key, $value );
1373 # Fetch single value or element given plain key or array index
1375 my $self = shift->_get_self;
1379 # Make sure file is open
1381 if (!defined($self->fh)) { $self->_open(); }
1383 my $md5 = $DIGEST_FUNC->($key);
1386 # Request shared lock for reading
1388 $self->lock( LOCK_SH );
1390 my $tag = $self->_find_bucket_list( $md5 );
1397 # Get value from bucket list
1399 my $result = $self->_get_bucket_value( $tag, $md5 );
1403 #XXX What is ref() checking here?
1404 #YYY Filters only apply on scalar values, so the ref check is making
1405 #YYY sure the fetched bucket is a scalar, not a child hash or array.
1406 return ($result && !ref($result) && $self->root->{filter_fetch_value})
1407 ? $self->root->{filter_fetch_value}->($result)
1413 # Delete single key/value pair or element given plain key or array index
1415 my $self = $_[0]->_get_self;
1418 my $md5 = $DIGEST_FUNC->($key);
1421 # Make sure file is open
1423 if (!defined($self->fh)) { $self->_open(); }
1426 # Request exclusive lock for writing
1428 $self->lock( LOCK_EX );
1430 my $tag = $self->_find_bucket_list( $md5 );
1439 my $value = $self->_get_bucket_value( $tag, $md5 );
1440 if ($value && !ref($value) && $self->root->{filter_fetch_value}) {
1441 $value = $self->root->{filter_fetch_value}->($value);
1444 my $result = $self->_delete_bucket( $tag, $md5 );
1447 # If this object is an array and the key deleted was on the end of the stack,
1448 # decrement the length variable.
1458 # Check if a single key or element exists given plain key or array index
1460 my $self = $_[0]->_get_self;
1463 my $md5 = $DIGEST_FUNC->($key);
1466 # Make sure file is open
1468 if (!defined($self->fh)) { $self->_open(); }
1471 # Request shared lock for reading
1473 $self->lock( LOCK_SH );
1475 my $tag = $self->_find_bucket_list( $md5 );
1478 # For some reason, the built-in exists() function returns '' for false
1486 # Check if bucket exists and return 1 or ''
1488 my $result = $self->_bucket_exists( $tag, $md5 ) || '';
1497 # Clear all keys from hash, or all elements from array.
1499 my $self = $_[0]->_get_self;
1502 # Make sure file is open
1504 if (!defined($self->fh)) { $self->_open(); }
1507 # Request exclusive lock for writing
1509 $self->lock( LOCK_EX );
1513 seek($fh, $self->base_offset, SEEK_SET);
1519 $self->_create_tag($self->base_offset, $self->type, chr(0) x $INDEX_SIZE);
1527 # Public method aliases
1529 sub put { (shift)->STORE( @_ ) }
1530 sub store { (shift)->STORE( @_ ) }
1531 sub get { (shift)->FETCH( @_ ) }
1532 sub fetch { (shift)->FETCH( @_ ) }
1533 sub delete { (shift)->DELETE( @_ ) }
1534 sub exists { (shift)->EXISTS( @_ ) }
1535 sub clear { (shift)->CLEAR( @_ ) }
1537 package DBM::Deep::_::Root;
1552 filter_store_key => undef,
1553 filter_store_value => undef,
1554 filter_fetch_key => undef,
1555 filter_fetch_value => undef,
1566 return unless $self;
1568 close $self->{fh} if $self->{fh};
1579 DBM::Deep - A pure perl multi-level hash/array DBM
1584 my $db = DBM::Deep->new( "foo.db" );
1586 $db->{key} = 'value'; # tie() style
1589 $db->put('key', 'value'); # OO style
1590 print $db->get('key');
1592 # true multi-level support
1593 $db->{my_complex} = [
1594 'hello', { perl => 'rules' },
1599 A unique flat-file database module, written in pure perl. True
1600 multi-level hash/array support (unlike MLDBM, which is faked), hybrid
1601 OO / tie() interface, cross-platform FTPable files, and quite fast. Can
1602 handle millions of keys and unlimited hash levels without significant
1603 slow-down. Written from the ground-up in pure perl -- this is NOT a
1604 wrapper around a C-based DBM. Out-of-the-box compatibility with Unix,
1605 Mac OS X and Windows.
1609 Hopefully you are using CPAN's excellent Perl module, which will download
1610 and install the module for you. If not, get the tarball, and run these
1622 Construction can be done OO-style (which is the recommended way), or using
1623 Perl's tie() function. Both are examined here.
1625 =head2 OO CONSTRUCTION
1627 The recommended way to construct a DBM::Deep object is to use the new()
1628 method, which gets you a blessed, tied hash or array reference.
1630 my $db = DBM::Deep->new( "foo.db" );
1632 This opens a new database handle, mapped to the file "foo.db". If this
1633 file does not exist, it will automatically be created. DB files are
1634 opened in "r+" (read/write) mode, and the type of object returned is a
1635 hash, unless otherwise specified (see L<OPTIONS> below).
1639 You can pass a number of options to the constructor to specify things like
1640 locking, autoflush, etc. This is done by passing an inline hash:
1642 my $db = DBM::Deep->new(
1648 Notice that the filename is now specified I<inside> the hash with
1649 the "file" parameter, as opposed to being the sole argument to the
1650 constructor. This is required if any options are specified.
1651 See L<OPTIONS> below for the complete list.
1655 You can also start with an array instead of a hash. For this, you must
1656 specify the C<type> parameter:
1658 my $db = DBM::Deep->new(
1660 type => DBM::Deep->TYPE_ARRAY
1663 B<Note:> Specifing the C<type> parameter only takes effect when beginning
1664 a new DB file. If you create a DBM::Deep object with an existing file, the
1665 C<type> will be loaded from the file header, and ignored if it is passed
1668 =head2 TIE CONSTRUCTION
1670 Alternatively, you can create a DBM::Deep handle by using Perl's built-in
1671 tie() function. This is not ideal, because you get only a basic, tied hash
1672 (or array) which is not blessed, so you can't call any functions on it.
1675 tie %hash, "DBM::Deep", "foo.db";
1678 tie @array, "DBM::Deep", "bar.db";
1680 As with the OO constructor, you can replace the DB filename parameter with
1681 a hash containing one or more options (see L<OPTIONS> just below for the
1684 tie %hash, "DBM::Deep", {
1692 There are a number of options that can be passed in when constructing your
1693 DBM::Deep objects. These apply to both the OO- and tie- based approaches.
1699 Filename of the DB file to link the handle to. You can pass a full absolute
1700 filesystem path, partial path, or a plain filename if the file is in the
1701 current working directory. This is a required parameter.
1705 File open mode (read-only, read-write, etc.) string passed to Perl's FileHandle
1706 module. This is an optional parameter, and defaults to "r+" (read/write).
1707 B<Note:> If the default (r+) mode is selected, the file will also be auto-
1708 created if it doesn't exist.
1712 This parameter specifies what type of object to create, a hash or array. Use
1713 one of these two constants: C<DBM::Deep-E<gt>TYPE_HASH> or C<DBM::Deep-E<gt>TYPE_ARRAY>.
1714 This only takes effect when beginning a new file. This is an optional
1715 parameter, and defaults to C<DBM::Deep-E<gt>TYPE_HASH>.
1719 Specifies whether locking is to be enabled. DBM::Deep uses Perl's Fnctl flock()
1720 function to lock the database in exclusive mode for writes, and shared mode for
1721 reads. Pass any true value to enable. This affects the base DB handle I<and
1722 any child hashes or arrays> that use the same DB file. This is an optional
1723 parameter, and defaults to 0 (disabled). See L<LOCKING> below for more.
1727 Specifies whether autoflush is to be enabled on the underlying FileHandle.
1728 This obviously slows down write operations, but is required if you may have
1729 multiple processes accessing the same DB file (also consider enable I<locking>
1730 or at least I<volatile>). Pass any true value to enable. This is an optional
1731 parameter, and defaults to 0 (disabled).
1735 If I<volatile> mode is enabled, DBM::Deep will stat() the DB file before each
1736 STORE() operation. This is required if an outside force may change the size of
1737 the file between transactions. Locking also implicitly enables volatile. This
1738 is useful if you want to use a different locking system or write your own. Pass
1739 any true value to enable. This is an optional parameter, and defaults to 0
1744 If I<autobless> mode is enabled, DBM::Deep will preserve blessed hashes, and
1745 restore them when fetched. This is an B<experimental> feature, and does have
1746 side-effects. Basically, when hashes are re-blessed into their original
1747 classes, they are no longer blessed into the DBM::Deep class! So you won't be
1748 able to call any DBM::Deep methods on them. You have been warned.
1749 This is an optional parameter, and defaults to 0 (disabled).
1753 See L<FILTERS> below.
1757 Setting I<debug> mode will make all errors non-fatal, dump them out to
1758 STDERR, and continue on. This is for debugging purposes only, and probably
1759 not what you want. This is an optional parameter, and defaults to 0 (disabled).
1763 Instead of passing a file path, you can instead pass a handle to an pre-opened
1764 filehandle. Note: Beware of using the magick *DATA handle, as this actually
1765 contains your entire Perl script, as well as the data following the __DATA__
1766 marker. This will not work, because DBM::Deep uses absolute seek()s into the
1767 file. Instead, consider reading *DATA into an IO::Scalar handle, then passing
1768 in that. Also please note optimize() will NOT work when passing in only a
1769 handle. Pass in a real filename in order to use optimize().
1773 =head1 TIE INTERFACE
1775 With DBM::Deep you can access your databases using Perl's standard hash/array
1776 syntax. Because all DBM::Deep objects are I<tied> to hashes or arrays, you can treat
1777 them as such. DBM::Deep will intercept all reads/writes and direct them to the right
1778 place -- the DB file. This has nothing to do with the L<TIE CONSTRUCTION>
1779 section above. This simply tells you how to use DBM::Deep using regular hashes
1780 and arrays, rather than calling functions like C<get()> and C<put()> (although those
1781 work too). It is entirely up to you how to want to access your databases.
1785 You can treat any DBM::Deep object like a normal Perl hash reference. Add keys,
1786 or even nested hashes (or arrays) using standard Perl syntax:
1788 my $db = DBM::Deep->new( "foo.db" );
1790 $db->{mykey} = "myvalue";
1792 $db->{myhash}->{subkey} = "subvalue";
1794 print $db->{myhash}->{subkey} . "\n";
1796 You can even step through hash keys using the normal Perl C<keys()> function:
1798 foreach my $key (keys %$db) {
1799 print "$key: " . $db->{$key} . "\n";
1802 Remember that Perl's C<keys()> function extracts I<every> key from the hash and
1803 pushes them onto an array, all before the loop even begins. If you have an
1804 extra large hash, this may exhaust Perl's memory. Instead, consider using
1805 Perl's C<each()> function, which pulls keys/values one at a time, using very
1808 while (my ($key, $value) = each %$db) {
1809 print "$key: $value\n";
1812 Please note that when using C<each()>, you should always pass a direct
1813 hash reference, not a lookup. Meaning, you should B<never> do this:
1816 while (my ($key, $value) = each %{$db->{foo}}) { # BAD
1818 This causes an infinite loop, because for each iteration, Perl is calling
1819 FETCH() on the $db handle, resulting in a "new" hash for foo every time, so
1820 it effectively keeps returning the first key over and over again. Instead,
1821 assign a temporary variable to C<$db->{foo}>, then pass that to each().
1825 As with hashes, you can treat any DBM::Deep object like a normal Perl array
1826 reference. This includes inserting, removing and manipulating elements,
1827 and the C<push()>, C<pop()>, C<shift()>, C<unshift()> and C<splice()> functions.
1828 The object must have first been created using type C<DBM::Deep-E<gt>TYPE_ARRAY>,
1829 or simply be a nested array reference inside a hash. Example:
1831 my $db = DBM::Deep->new(
1832 file => "foo-array.db",
1833 type => DBM::Deep->TYPE_ARRAY
1837 push @$db, "bar", "baz";
1838 unshift @$db, "bah";
1840 my $last_elem = pop @$db; # baz
1841 my $first_elem = shift @$db; # bah
1842 my $second_elem = $db->[1]; # bar
1844 my $num_elements = scalar @$db;
1848 In addition to the I<tie()> interface, you can also use a standard OO interface
1849 to manipulate all aspects of DBM::Deep databases. Each type of object (hash or
1850 array) has its own methods, but both types share the following common methods:
1851 C<put()>, C<get()>, C<exists()>, C<delete()> and C<clear()>.
1857 Stores a new hash key/value pair, or sets an array element value. Takes two
1858 arguments, the hash key or array index, and the new value. The value can be
1859 a scalar, hash ref or array ref. Returns true on success, false on failure.
1861 $db->put("foo", "bar"); # for hashes
1862 $db->put(1, "bar"); # for arrays
1866 Fetches the value of a hash key or array element. Takes one argument: the hash
1867 key or array index. Returns a scalar, hash ref or array ref, depending on the
1870 my $value = $db->get("foo"); # for hashes
1871 my $value = $db->get(1); # for arrays
1875 Checks if a hash key or array index exists. Takes one argument: the hash key
1876 or array index. Returns true if it exists, false if not.
1878 if ($db->exists("foo")) { print "yay!\n"; } # for hashes
1879 if ($db->exists(1)) { print "yay!\n"; } # for arrays
1883 Deletes one hash key/value pair or array element. Takes one argument: the hash
1884 key or array index. Returns true on success, false if not found. For arrays,
1885 the remaining elements located after the deleted element are NOT moved over.
1886 The deleted element is essentially just undefined, which is exactly how Perl's
1887 internal arrays work. Please note that the space occupied by the deleted
1888 key/value or element is B<not> reused again -- see L<UNUSED SPACE RECOVERY>
1889 below for details and workarounds.
1891 $db->delete("foo"); # for hashes
1892 $db->delete(1); # for arrays
1896 Deletes B<all> hash keys or array elements. Takes no arguments. No return
1897 value. Please note that the space occupied by the deleted keys/values or
1898 elements is B<not> reused again -- see L<UNUSED SPACE RECOVERY> below for
1899 details and workarounds.
1901 $db->clear(); # hashes or arrays
1907 For hashes, DBM::Deep supports all the common methods described above, and the
1908 following additional methods: C<first_key()> and C<next_key()>.
1914 Returns the "first" key in the hash. As with built-in Perl hashes, keys are
1915 fetched in an undefined order (which appears random). Takes no arguments,
1916 returns the key as a scalar value.
1918 my $key = $db->first_key();
1922 Returns the "next" key in the hash, given the previous one as the sole argument.
1923 Returns undef if there are no more keys to be fetched.
1925 $key = $db->next_key($key);
1929 Here are some examples of using hashes:
1931 my $db = DBM::Deep->new( "foo.db" );
1933 $db->put("foo", "bar");
1934 print "foo: " . $db->get("foo") . "\n";
1936 $db->put("baz", {}); # new child hash ref
1937 $db->get("baz")->put("buz", "biz");
1938 print "buz: " . $db->get("baz")->get("buz") . "\n";
1940 my $key = $db->first_key();
1942 print "$key: " . $db->get($key) . "\n";
1943 $key = $db->next_key($key);
1946 if ($db->exists("foo")) { $db->delete("foo"); }
1950 For arrays, DBM::Deep supports all the common methods described above, and the
1951 following additional methods: C<length()>, C<push()>, C<pop()>, C<shift()>,
1952 C<unshift()> and C<splice()>.
1958 Returns the number of elements in the array. Takes no arguments.
1960 my $len = $db->length();
1964 Adds one or more elements onto the end of the array. Accepts scalars, hash
1965 refs or array refs. No return value.
1967 $db->push("foo", "bar", {});
1971 Fetches the last element in the array, and deletes it. Takes no arguments.
1972 Returns undef if array is empty. Returns the element value.
1974 my $elem = $db->pop();
1978 Fetches the first element in the array, deletes it, then shifts all the
1979 remaining elements over to take up the space. Returns the element value. This
1980 method is not recommended with large arrays -- see L<LARGE ARRAYS> below for
1983 my $elem = $db->shift();
1987 Inserts one or more elements onto the beginning of the array, shifting all
1988 existing elements over to make room. Accepts scalars, hash refs or array refs.
1989 No return value. This method is not recommended with large arrays -- see
1990 <LARGE ARRAYS> below for details.
1992 $db->unshift("foo", "bar", {});
1996 Performs exactly like Perl's built-in function of the same name. See L<perldoc
1997 -f splice> for usage -- it is too complicated to document here. This method is
1998 not recommended with large arrays -- see L<LARGE ARRAYS> below for details.
2002 Here are some examples of using arrays:
2004 my $db = DBM::Deep->new(
2006 type => DBM::Deep->TYPE_ARRAY
2009 $db->push("bar", "baz");
2010 $db->unshift("foo");
2013 my $len = $db->length();
2014 print "length: $len\n"; # 4
2016 for (my $k=0; $k<$len; $k++) {
2017 print "$k: " . $db->get($k) . "\n";
2020 $db->splice(1, 2, "biz", "baf");
2022 while (my $elem = shift @$db) {
2023 print "shifted: $elem\n";
2028 Enable automatic file locking by passing a true value to the C<locking>
2029 parameter when constructing your DBM::Deep object (see L<SETUP> above).
2031 my $db = DBM::Deep->new(
2036 This causes DBM::Deep to C<flock()> the underlying FileHandle object with exclusive
2037 mode for writes, and shared mode for reads. This is required if you have
2038 multiple processes accessing the same database file, to avoid file corruption.
2039 Please note that C<flock()> does NOT work for files over NFS. See L<DB OVER
2040 NFS> below for more.
2042 =head2 EXPLICIT LOCKING
2044 You can explicitly lock a database, so it remains locked for multiple
2045 transactions. This is done by calling the C<lock()> method, and passing an
2046 optional lock mode argument (defaults to exclusive mode). This is particularly
2047 useful for things like counters, where the current value needs to be fetched,
2048 then incremented, then stored again.
2051 my $counter = $db->get("counter");
2053 $db->put("counter", $counter);
2062 You can pass C<lock()> an optional argument, which specifies which mode to use
2063 (exclusive or shared). Use one of these two constants: C<DBM::Deep-E<gt>LOCK_EX>
2064 or C<DBM::Deep-E<gt>LOCK_SH>. These are passed directly to C<flock()>, and are the
2065 same as the constants defined in Perl's C<Fcntl> module.
2067 $db->lock( DBM::Deep->LOCK_SH );
2071 If you want to implement your own file locking scheme, be sure to create your
2072 DBM::Deep objects setting the C<volatile> option to true. This hints to DBM::Deep
2073 that the DB file may change between transactions. See L<LOW-LEVEL ACCESS>
2076 =head1 IMPORTING/EXPORTING
2078 You can import existing complex structures by calling the C<import()> method,
2079 and export an entire database into an in-memory structure using the C<export()>
2080 method. Both are examined here.
2084 Say you have an existing hash with nested hashes/arrays inside it. Instead of
2085 walking the structure and adding keys/elements to the database as you go,
2086 simply pass a reference to the C<import()> method. This recursively adds
2087 everything to an existing DBM::Deep object for you. Here is an example:
2092 array1 => [ "elem0", "elem1", "elem2" ],
2094 subkey1 => "subvalue1",
2095 subkey2 => "subvalue2"
2099 my $db = DBM::Deep->new( "foo.db" );
2100 $db->import( $struct );
2102 print $db->{key1} . "\n"; # prints "value1"
2104 This recursively imports the entire C<$struct> object into C<$db>, including
2105 all nested hashes and arrays. If the DBM::Deep object contains exsiting data,
2106 keys are merged with the existing ones, replacing if they already exist.
2107 The C<import()> method can be called on any database level (not just the base
2108 level), and works with both hash and array DB types.
2112 B<Note:> Make sure your existing structure has no circular references in it.
2113 These will cause an infinite loop when importing.
2117 Calling the C<export()> method on an existing DBM::Deep object will return
2118 a reference to a new in-memory copy of the database. The export is done
2119 recursively, so all nested hashes/arrays are all exported to standard Perl
2120 objects. Here is an example:
2122 my $db = DBM::Deep->new( "foo.db" );
2124 $db->{key1} = "value1";
2125 $db->{key2} = "value2";
2127 $db->{hash1}->{subkey1} = "subvalue1";
2128 $db->{hash1}->{subkey2} = "subvalue2";
2130 my $struct = $db->export();
2132 print $struct->{key1} . "\n"; # prints "value1"
2134 This makes a complete copy of the database in memory, and returns a reference
2135 to it. The C<export()> method can be called on any database level (not just
2136 the base level), and works with both hash and array DB types. Be careful of
2137 large databases -- you can store a lot more data in a DBM::Deep object than an
2138 in-memory Perl structure.
2142 B<Note:> Make sure your database has no circular references in it.
2143 These will cause an infinite loop when exporting.
2147 DBM::Deep has a number of hooks where you can specify your own Perl function
2148 to perform filtering on incoming or outgoing data. This is a perfect
2149 way to extend the engine, and implement things like real-time compression or
2150 encryption. Filtering applies to the base DB level, and all child hashes /
2151 arrays. Filter hooks can be specified when your DBM::Deep object is first
2152 constructed, or by calling the C<set_filter()> method at any time. There are
2153 four available filter hooks, described below:
2157 =item * filter_store_key
2159 This filter is called whenever a hash key is stored. It
2160 is passed the incoming key, and expected to return a transformed key.
2162 =item * filter_store_value
2164 This filter is called whenever a hash key or array element is stored. It
2165 is passed the incoming value, and expected to return a transformed value.
2167 =item * filter_fetch_key
2169 This filter is called whenever a hash key is fetched (i.e. via
2170 C<first_key()> or C<next_key()>). It is passed the transformed key,
2171 and expected to return the plain key.
2173 =item * filter_fetch_value
2175 This filter is called whenever a hash key or array element is fetched.
2176 It is passed the transformed value, and expected to return the plain value.
2180 Here are the two ways to setup a filter hook:
2182 my $db = DBM::Deep->new(
2184 filter_store_value => \&my_filter_store,
2185 filter_fetch_value => \&my_filter_fetch
2190 $db->set_filter( "filter_store_value", \&my_filter_store );
2191 $db->set_filter( "filter_fetch_value", \&my_filter_fetch );
2193 Your filter function will be called only when dealing with SCALAR keys or
2194 values. When nested hashes and arrays are being stored/fetched, filtering
2195 is bypassed. Filters are called as static functions, passed a single SCALAR
2196 argument, and expected to return a single SCALAR value. If you want to
2197 remove a filter, set the function reference to C<undef>:
2199 $db->set_filter( "filter_store_value", undef );
2201 =head2 REAL-TIME ENCRYPTION EXAMPLE
2203 Here is a working example that uses the I<Crypt::Blowfish> module to
2204 do real-time encryption / decryption of keys & values with DBM::Deep Filters.
2205 Please visit L<http://search.cpan.org/search?module=Crypt::Blowfish> for more
2206 on I<Crypt::Blowfish>. You'll also need the I<Crypt::CBC> module.
2209 use Crypt::Blowfish;
2212 my $cipher = Crypt::CBC->new({
2213 'key' => 'my secret key',
2214 'cipher' => 'Blowfish',
2216 'regenerate_key' => 0,
2217 'padding' => 'space',
2221 my $db = DBM::Deep->new(
2222 file => "foo-encrypt.db",
2223 filter_store_key => \&my_encrypt,
2224 filter_store_value => \&my_encrypt,
2225 filter_fetch_key => \&my_decrypt,
2226 filter_fetch_value => \&my_decrypt,
2229 $db->{key1} = "value1";
2230 $db->{key2} = "value2";
2231 print "key1: " . $db->{key1} . "\n";
2232 print "key2: " . $db->{key2} . "\n";
2238 return $cipher->encrypt( $_[0] );
2241 return $cipher->decrypt( $_[0] );
2244 =head2 REAL-TIME COMPRESSION EXAMPLE
2246 Here is a working example that uses the I<Compress::Zlib> module to do real-time
2247 compression / decompression of keys & values with DBM::Deep Filters.
2248 Please visit L<http://search.cpan.org/search?module=Compress::Zlib> for
2249 more on I<Compress::Zlib>.
2254 my $db = DBM::Deep->new(
2255 file => "foo-compress.db",
2256 filter_store_key => \&my_compress,
2257 filter_store_value => \&my_compress,
2258 filter_fetch_key => \&my_decompress,
2259 filter_fetch_value => \&my_decompress,
2262 $db->{key1} = "value1";
2263 $db->{key2} = "value2";
2264 print "key1: " . $db->{key1} . "\n";
2265 print "key2: " . $db->{key2} . "\n";
2271 return Compress::Zlib::memGzip( $_[0] ) ;
2274 return Compress::Zlib::memGunzip( $_[0] ) ;
2277 B<Note:> Filtering of keys only applies to hashes. Array "keys" are
2278 actually numerical index numbers, and are not filtered.
2280 =head1 ERROR HANDLING
2282 Most DBM::Deep methods return a true value for success, and call die() on
2283 failure. You can wrap calls in an eval block to catch the die. Also, the
2284 actual error message is stored in an internal scalar, which can be fetched by
2285 calling the C<error()> method.
2287 my $db = DBM::Deep->new( "foo.db" ); # create hash
2288 eval { $db->push("foo"); }; # ILLEGAL -- push is array-only call
2290 print $db->error(); # prints error message
2292 You can then call C<clear_error()> to clear the current error state.
2296 If you set the C<debug> option to true when creating your DBM::Deep object,
2297 all errors are considered NON-FATAL, and dumped to STDERR. This is only
2298 for debugging purposes.
2300 =head1 LARGEFILE SUPPORT
2302 If you have a 64-bit system, and your Perl is compiled with both LARGEFILE
2303 and 64-bit support, you I<may> be able to create databases larger than 2 GB.
2304 DBM::Deep by default uses 32-bit file offset tags, but these can be changed
2305 by calling the static C<set_pack()> method before you do anything else.
2307 DBM::Deep::set_pack(8, 'Q');
2309 This tells DBM::Deep to pack all file offsets with 8-byte (64-bit) quad words
2310 instead of 32-bit longs. After setting these values your DB files have a
2311 theoretical maximum size of 16 XB (exabytes).
2315 B<Note:> Changing these values will B<NOT> work for existing database files.
2316 Only change this for new files, and make sure it stays set consistently
2317 throughout the file's life. If you do set these values, you can no longer
2318 access 32-bit DB files. You can, however, call C<set_pack(4, 'N')> to change
2319 back to 32-bit mode.
2323 B<Note:> I have not personally tested files > 2 GB -- all my systems have
2324 only a 32-bit Perl. However, I have received user reports that this does
2327 =head1 LOW-LEVEL ACCESS
2329 If you require low-level access to the underlying FileHandle that DBM::Deep uses,
2330 you can call the C<fh()> method, which returns the handle:
2334 This method can be called on the root level of the datbase, or any child
2335 hashes or arrays. All levels share a I<root> structure, which contains things
2336 like the FileHandle, a reference counter, and all your options you specified
2337 when you created the object. You can get access to this root structure by
2338 calling the C<root()> method.
2340 my $root = $db->root();
2342 This is useful for changing options after the object has already been created,
2343 such as enabling/disabling locking, volatile or debug modes. You can also
2344 store your own temporary user data in this structure (be wary of name
2345 collision), which is then accessible from any child hash or array.
2347 =head1 CUSTOM DIGEST ALGORITHM
2349 DBM::Deep by default uses the I<Message Digest 5> (MD5) algorithm for hashing
2350 keys. However you can override this, and use another algorithm (such as SHA-256)
2351 or even write your own. But please note that DBM::Deep currently expects zero
2352 collisions, so your algorithm has to be I<perfect>, so to speak.
2353 Collision detection may be introduced in a later version.
2357 You can specify a custom digest algorithm by calling the static C<set_digest()>
2358 function, passing a reference to a subroutine, and the length of the algorithm's
2359 hashes (in bytes). This is a global static function, which affects ALL DBM::Deep
2360 objects. Here is a working example that uses a 256-bit hash from the
2361 I<Digest::SHA256> module. Please see
2362 L<http://search.cpan.org/search?module=Digest::SHA256> for more.
2367 my $context = Digest::SHA256::new(256);
2369 DBM::Deep::set_digest( \&my_digest, 32 );
2371 my $db = DBM::Deep->new( "foo-sha.db" );
2373 $db->{key1} = "value1";
2374 $db->{key2} = "value2";
2375 print "key1: " . $db->{key1} . "\n";
2376 print "key2: " . $db->{key2} . "\n";
2382 return substr( $context->hash($_[0]), 0, 32 );
2385 B<Note:> Your returned digest strings must be B<EXACTLY> the number
2386 of bytes you specify in the C<set_digest()> function (in this case 32).
2388 =head1 CIRCULAR REFERENCES
2390 DBM::Deep has B<experimental> support for circular references. Meaning you
2391 can have a nested hash key or array element that points to a parent object.
2392 This relationship is stored in the DB file, and is preserved between sessions.
2395 my $db = DBM::Deep->new( "foo.db" );
2398 $db->{circle} = $db; # ref to self
2400 print $db->{foo} . "\n"; # prints "foo"
2401 print $db->{circle}->{foo} . "\n"; # prints "foo" again
2403 One catch is, passing the object to a function that recursively walks the
2404 object tree (such as I<Data::Dumper> or even the built-in C<optimize()> or
2405 C<export()> methods) will result in an infinite loop. The other catch is,
2406 if you fetch the I<key> of a circular reference (i.e. using the C<first_key()>
2407 or C<next_key()> methods), you will get the I<target object's key>, not the
2408 ref's key. This gets even more interesting with the above example, where
2409 the I<circle> key points to the base DB object, which technically doesn't
2410 have a key. So I made DBM::Deep return "[base]" as the key name in that
2413 =head1 CAVEATS / ISSUES / BUGS
2415 This section describes all the known issues with DBM::Deep. It you have found
2416 something that is not listed here, please send e-mail to L<jhuckaby@cpan.org>.
2418 =head2 UNUSED SPACE RECOVERY
2420 One major caveat with DBM::Deep is that space occupied by existing keys and
2421 values is not recovered when they are deleted. Meaning if you keep deleting
2422 and adding new keys, your file will continuously grow. I am working on this,
2423 but in the meantime you can call the built-in C<optimize()> method from time to
2424 time (perhaps in a crontab or something) to recover all your unused space.
2426 $db->optimize(); # returns true on success
2428 This rebuilds the ENTIRE database into a new file, then moves it on top of
2429 the original. The new file will have no unused space, thus it will take up as
2430 little disk space as possible. Please note that this operation can take
2431 a long time for large files, and you need enough disk space to temporarily hold
2432 2 copies of your DB file. The temporary file is created in the same directory
2433 as the original, named with a ".tmp" extension, and is deleted when the
2434 operation completes. Oh, and if locking is enabled, the DB is automatically
2435 locked for the entire duration of the copy.
2439 B<WARNING:> Only call optimize() on the top-level node of the database, and
2440 make sure there are no child references lying around. DBM::Deep keeps a reference
2441 counter, and if it is greater than 1, optimize() will abort and return undef.
2443 =head2 AUTOVIVIFICATION
2445 Unfortunately, autovivification doesn't work with tied hashes. This appears to
2446 be a bug in Perl's tie() system, as I<Jakob Schmidt> encountered the very same
2447 issue with his I<DWH_FIle> module (see L<http://search.cpan.org/search?module=DWH_File>),
2448 and it is also mentioned in the BUGS section for the I<MLDBM> module <see
2449 L<http://search.cpan.org/search?module=MLDBM>). Basically, on a new db file,
2452 $db->{foo}->{bar} = "hello";
2454 Since "foo" doesn't exist, you cannot add "bar" to it. You end up with "foo"
2455 being an empty hash. Try this instead, which works fine:
2457 $db->{foo} = { bar => "hello" };
2459 As of Perl 5.8.7, this bug still exists. I have walked very carefully through
2460 the execution path, and Perl indeed passes an empty hash to the STORE() method.
2461 Probably a bug in Perl.
2463 =head2 FILE CORRUPTION
2465 The current level of error handling in DBM::Deep is minimal. Files I<are> checked
2466 for a 32-bit signature when opened, but other corruption in files can cause
2467 segmentation faults. DBM::Deep may try to seek() past the end of a file, or get
2468 stuck in an infinite loop depending on the level of corruption. File write
2469 operations are not checked for failure (for speed), so if you happen to run
2470 out of disk space, DBM::Deep will probably fail in a bad way. These things will
2471 be addressed in a later version of DBM::Deep.
2475 Beware of using DB files over NFS. DBM::Deep uses flock(), which works well on local
2476 filesystems, but will NOT protect you from file corruption over NFS. I've heard
2477 about setting up your NFS server with a locking daemon, then using lockf() to
2478 lock your files, but your milage may vary there as well. From what I
2479 understand, there is no real way to do it. However, if you need access to the
2480 underlying FileHandle in DBM::Deep for using some other kind of locking scheme like
2481 lockf(), see the L<LOW-LEVEL ACCESS> section above.
2483 =head2 COPYING OBJECTS
2485 Beware of copying tied objects in Perl. Very strange things can happen.
2486 Instead, use DBM::Deep's C<clone()> method which safely copies the object and
2487 returns a new, blessed, tied hash or array to the same level in the DB.
2489 my $copy = $db->clone();
2493 Beware of using C<shift()>, C<unshift()> or C<splice()> with large arrays.
2494 These functions cause every element in the array to move, which can be murder
2495 on DBM::Deep, as every element has to be fetched from disk, then stored again in
2496 a different location. This may be addressed in a later version.
2500 This section discusses DBM::Deep's speed and memory usage.
2504 Obviously, DBM::Deep isn't going to be as fast as some C-based DBMs, such as
2505 the almighty I<BerkeleyDB>. But it makes up for it in features like true
2506 multi-level hash/array support, and cross-platform FTPable files. Even so,
2507 DBM::Deep is still pretty fast, and the speed stays fairly consistent, even
2508 with huge databases. Here is some test data:
2510 Adding 1,000,000 keys to new DB file...
2512 At 100 keys, avg. speed is 2,703 keys/sec
2513 At 200 keys, avg. speed is 2,642 keys/sec
2514 At 300 keys, avg. speed is 2,598 keys/sec
2515 At 400 keys, avg. speed is 2,578 keys/sec
2516 At 500 keys, avg. speed is 2,722 keys/sec
2517 At 600 keys, avg. speed is 2,628 keys/sec
2518 At 700 keys, avg. speed is 2,700 keys/sec
2519 At 800 keys, avg. speed is 2,607 keys/sec
2520 At 900 keys, avg. speed is 2,190 keys/sec
2521 At 1,000 keys, avg. speed is 2,570 keys/sec
2522 At 2,000 keys, avg. speed is 2,417 keys/sec
2523 At 3,000 keys, avg. speed is 1,982 keys/sec
2524 At 4,000 keys, avg. speed is 1,568 keys/sec
2525 At 5,000 keys, avg. speed is 1,533 keys/sec
2526 At 6,000 keys, avg. speed is 1,787 keys/sec
2527 At 7,000 keys, avg. speed is 1,977 keys/sec
2528 At 8,000 keys, avg. speed is 2,028 keys/sec
2529 At 9,000 keys, avg. speed is 2,077 keys/sec
2530 At 10,000 keys, avg. speed is 2,031 keys/sec
2531 At 20,000 keys, avg. speed is 1,970 keys/sec
2532 At 30,000 keys, avg. speed is 2,050 keys/sec
2533 At 40,000 keys, avg. speed is 2,073 keys/sec
2534 At 50,000 keys, avg. speed is 1,973 keys/sec
2535 At 60,000 keys, avg. speed is 1,914 keys/sec
2536 At 70,000 keys, avg. speed is 2,091 keys/sec
2537 At 80,000 keys, avg. speed is 2,103 keys/sec
2538 At 90,000 keys, avg. speed is 1,886 keys/sec
2539 At 100,000 keys, avg. speed is 1,970 keys/sec
2540 At 200,000 keys, avg. speed is 2,053 keys/sec
2541 At 300,000 keys, avg. speed is 1,697 keys/sec
2542 At 400,000 keys, avg. speed is 1,838 keys/sec
2543 At 500,000 keys, avg. speed is 1,941 keys/sec
2544 At 600,000 keys, avg. speed is 1,930 keys/sec
2545 At 700,000 keys, avg. speed is 1,735 keys/sec
2546 At 800,000 keys, avg. speed is 1,795 keys/sec
2547 At 900,000 keys, avg. speed is 1,221 keys/sec
2548 At 1,000,000 keys, avg. speed is 1,077 keys/sec
2550 This test was performed on a PowerMac G4 1gHz running Mac OS X 10.3.2 & Perl
2551 5.8.1, with an 80GB Ultra ATA/100 HD spinning at 7200RPM. The hash keys and
2552 values were between 6 - 12 chars in length. The DB file ended up at 210MB.
2553 Run time was 12 min 3 sec.
2557 One of the great things about DBM::Deep is that it uses very little memory.
2558 Even with huge databases (1,000,000+ keys) you will not see much increased
2559 memory on your process. DBM::Deep relies solely on the filesystem for storing
2560 and fetching data. Here is output from I</usr/bin/top> before even opening a
2563 PID USER PRI NI SIZE RSS SHARE STAT %CPU %MEM TIME COMMAND
2564 22831 root 11 0 2716 2716 1296 R 0.0 0.2 0:07 perl
2566 Basically the process is taking 2,716K of memory. And here is the same
2567 process after storing and fetching 1,000,000 keys:
2569 PID USER PRI NI SIZE RSS SHARE STAT %CPU %MEM TIME COMMAND
2570 22831 root 14 0 2772 2772 1328 R 0.0 0.2 13:32 perl
2572 Notice the memory usage increased by only 56K. Test was performed on a 700mHz
2573 x86 box running Linux RedHat 7.2 & Perl 5.6.1.
2575 =head1 DB FILE FORMAT
2577 In case you were interested in the underlying DB file format, it is documented
2578 here in this section. You don't need to know this to use the module, it's just
2579 included for reference.
2583 DBM::Deep files always start with a 32-bit signature to identify the file type.
2584 This is at offset 0. The signature is "DPDB" in network byte order. This is
2585 checked when the file is opened.
2589 The DBM::Deep file is in a I<tagged format>, meaning each section of the file
2590 has a standard header containing the type of data, the length of data, and then
2591 the data itself. The type is a single character (1 byte), the length is a
2592 32-bit unsigned long in network byte order, and the data is, well, the data.
2593 Here is how it unfolds:
2597 Immediately after the 32-bit file signature is the I<Master Index> record.
2598 This is a standard tag header followed by 1024 bytes (in 32-bit mode) or 2048
2599 bytes (in 64-bit mode) of data. The type is I<H> for hash or I<A> for array,
2600 depending on how the DBM::Deep object was constructed.
2604 The index works by looking at a I<MD5 Hash> of the hash key (or array index
2605 number). The first 8-bit char of the MD5 signature is the offset into the
2606 index, multipled by 4 in 32-bit mode, or 8 in 64-bit mode. The value of the
2607 index element is a file offset of the next tag for the key/element in question,
2608 which is usually a I<Bucket List> tag (see below).
2612 The next tag I<could> be another index, depending on how many keys/elements
2613 exist. See L<RE-INDEXING> below for details.
2617 A I<Bucket List> is a collection of 16 MD5 hashes for keys/elements, plus
2618 file offsets to where the actual data is stored. It starts with a standard
2619 tag header, with type I<B>, and a data size of 320 bytes in 32-bit mode, or
2620 384 bytes in 64-bit mode. Each MD5 hash is stored in full (16 bytes), plus
2621 the 32-bit or 64-bit file offset for the I<Bucket> containing the actual data.
2622 When the list fills up, a I<Re-Index> operation is performed (See
2623 L<RE-INDEXING> below).
2627 A I<Bucket> is a tag containing a key/value pair (in hash mode), or a
2628 index/value pair (in array mode). It starts with a standard tag header with
2629 type I<D> for scalar data (string, binary, etc.), or it could be a nested
2630 hash (type I<H>) or array (type I<A>). The value comes just after the tag
2631 header. The size reported in the tag header is only for the value, but then,
2632 just after the value is another size (32-bit unsigned long) and then the plain
2633 key itself. Since the value is likely to be fetched more often than the plain
2634 key, I figured it would be I<slightly> faster to store the value first.
2638 If the type is I<H> (hash) or I<A> (array), the value is another I<Master Index>
2639 record for the nested structure, where the process begins all over again.
2643 After a I<Bucket List> grows to 16 records, its allocated space in the file is
2644 exhausted. Then, when another key/element comes in, the list is converted to a
2645 new index record. However, this index will look at the next char in the MD5
2646 hash, and arrange new Bucket List pointers accordingly. This process is called
2647 I<Re-Indexing>. Basically, a new index tag is created at the file EOF, and all
2648 17 (16 + new one) keys/elements are removed from the old Bucket List and
2649 inserted into the new index. Several new Bucket Lists are created in the
2650 process, as a new MD5 char from the key is being examined (it is unlikely that
2651 the keys will all share the same next char of their MD5s).
2655 Because of the way the I<MD5> algorithm works, it is impossible to tell exactly
2656 when the Bucket Lists will turn into indexes, but the first round tends to
2657 happen right around 4,000 keys. You will see a I<slight> decrease in
2658 performance here, but it picks back up pretty quick (see L<SPEED> above). Then
2659 it takes B<a lot> more keys to exhaust the next level of Bucket Lists. It's
2660 right around 900,000 keys. This process can continue nearly indefinitely --
2661 right up until the point the I<MD5> signatures start colliding with each other,
2662 and this is B<EXTREMELY> rare -- like winning the lottery 5 times in a row AND
2663 getting struck by lightning while you are walking to cash in your tickets.
2664 Theoretically, since I<MD5> hashes are 128-bit values, you I<could> have up to
2665 340,282,366,921,000,000,000,000,000,000,000,000,000 keys/elements (I believe
2666 this is 340 unodecillion, but don't quote me).
2670 When a new key/element is stored, the key (or index number) is first ran through
2671 I<Digest::MD5> to get a 128-bit signature (example, in hex:
2672 b05783b0773d894396d475ced9d2f4f6). Then, the I<Master Index> record is checked
2673 for the first char of the signature (in this case I<b>). If it does not exist,
2674 a new I<Bucket List> is created for our key (and the next 15 future keys that
2675 happen to also have I<b> as their first MD5 char). The entire MD5 is written
2676 to the I<Bucket List> along with the offset of the new I<Bucket> record (EOF at
2677 this point, unless we are replacing an existing I<Bucket>), where the actual
2678 data will be stored.
2682 Fetching an existing key/element involves getting a I<Digest::MD5> of the key
2683 (or index number), then walking along the indexes. If there are enough
2684 keys/elements in this DB level, there might be nested indexes, each linked to
2685 a particular char of the MD5. Finally, a I<Bucket List> is pointed to, which
2686 contains up to 16 full MD5 hashes. Each is checked for equality to the key in
2687 question. If we found a match, the I<Bucket> tag is loaded, where the value and
2688 plain key are stored.
2692 Fetching the plain key occurs when calling the I<first_key()> and I<next_key()>
2693 methods. In this process the indexes are walked systematically, and each key
2694 fetched in increasing MD5 order (which is why it appears random). Once the
2695 I<Bucket> is found, the value is skipped the plain key returned instead.
2696 B<Note:> Do not count on keys being fetched as if the MD5 hashes were
2697 alphabetically sorted. This only happens on an index-level -- as soon as the
2698 I<Bucket Lists> are hit, the keys will come out in the order they went in --
2699 so it's pretty much undefined how the keys will come out -- just like Perl's
2702 =head1 CODE COVERAGE
2704 I use B<Devel::Cover> to test the code coverage of my tests, below is the B<Devel::Cover>
2705 report on this module's test suite.
2707 ---------------------------- ------ ------ ------ ------ ------ ------ ------
2708 File stmt bran cond sub pod time total
2709 ---------------------------- ------ ------ ------ ------ ------ ------ ------
2710 blib/lib/DBM/Deep.pm 93.7 82.5 71.9 96.5 25.9 82.8 87.9
2711 blib/lib/DBM/Deep/Array.pm 98.8 88.0 90.9 100.0 n/a 12.8 96.3
2712 blib/lib/DBM/Deep/Hash.pm 95.2 80.0 100.0 100.0 n/a 4.4 92.3
2713 Total 94.8 83.2 76.5 97.6 25.9 100.0 89.7
2714 ---------------------------- ------ ------ ------ ------ ------ ------ ------
2718 Joseph Huckaby, L<jhuckaby@cpan.org>
2719 Rob Kinyon, L<rkinyon@cpan.org>
2721 Special thanks to Adam Sah and Rich Gaushell! You know why :-)
2725 perltie(1), Tie::Hash(3), Digest::MD5(3), Fcntl(3), flock(2), lockf(3), nfs(5),
2726 Digest::SHA256(3), Crypt::Blowfish(3), Compress::Zlib(3)
2730 Copyright (c) 2002-2006 Joseph Huckaby. All Rights Reserved.
2731 This is free software, you may use it and distribute it under the
2732 same terms as Perl itself.