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(); }
244 if (!(-e $self->root->{file}) && $self->root->{mode} eq 'r+') {
245 my $temp = FileHandle->new( $self->root->{file}, 'w' );
248 #XXX Convert to set_fh()
249 $self->root->{fh} = FileHandle->new( $self->root->{file}, $self->root->{mode} );
250 # }; if ($@ ) { $self->_throw_error( "Received error: $@\n" ); }
251 if (! defined($self->fh)) {
252 return $self->_throw_error("Cannot open file: " . $self->root->{file} . ": $!");
256 binmode $fh; # for win32
257 if ($self->root->{autoflush}) {
258 my $old = select( $fh );
261 # $self->fh->autoflush();
266 my $bytes_read = read( $fh, $signature, length(SIG_FILE));
269 # File is empty -- write signature and master index
273 $fh->print(SIG_FILE);
274 $self->root->{end} = length(SIG_FILE);
275 $self->_create_tag($self->base_offset, $self->type, chr(0) x $INDEX_SIZE);
277 my $plain_key = "[base]";
278 $fh->print( pack($DATA_LENGTH_PACK, length($plain_key)) . $plain_key );
279 $self->root->{end} += $DATA_LENGTH_SIZE + length($plain_key);
286 # Check signature was valid
288 unless ($signature eq SIG_FILE) {
290 return $self->_throw_error("Signature not found -- file is not a Deep DB");
293 $self->root->{end} = (stat($fh))[7];
296 # Get our type from master index signature
298 my $tag = $self->_load_tag($self->base_offset);
299 #XXX We probably also want to store the hash algorithm name and not assume anything
301 return $self->_throw_error("Corrupted file, no master index record");
303 if ($self->{type} ne $tag->{signature}) {
304 return $self->_throw_error("File type mismatch");
312 # Close database FileHandle
314 my $self = _get_self($_[0]);
315 undef $self->root->{fh};
320 # Given offset, signature and content, create tag and write to disk
322 my ($self, $offset, $sig, $content) = @_;
323 my $size = length($content);
327 seek($fh, $offset, 0);
328 $fh->print( $sig . pack($DATA_LENGTH_PACK, $size) . $content );
330 if ($offset == $self->root->{end}) {
331 $self->root->{end} += SIG_SIZE + $DATA_LENGTH_SIZE + $size;
337 offset => $offset + SIG_SIZE + $DATA_LENGTH_SIZE,
344 # Given offset, load single tag and return signature, size and data
351 seek($fh, $offset, 0);
352 if (eof $fh) { return undef; }
355 read( $fh, $sig, SIG_SIZE);
358 read( $fh, $size, $DATA_LENGTH_SIZE);
359 $size = unpack($DATA_LENGTH_PACK, $size);
362 read( $fh, $buffer, $size);
367 offset => $offset + SIG_SIZE + $DATA_LENGTH_SIZE,
374 # Given index tag, lookup single entry in index and return .
377 my ($tag, $index) = @_;
379 my $location = unpack($LONG_PACK, substr($tag->{content}, $index * $LONG_SIZE, $LONG_SIZE) );
380 if (!$location) { return; }
382 return $self->_load_tag( $location );
387 # Adds one key/value pair to bucket list, given offset, MD5 digest of key,
388 # plain (undigested) key and value.
391 my ($tag, $md5, $plain_key, $value) = @_;
392 my $keys = $tag->{content};
396 my $is_dbm_deep = 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), 0);
418 $fh->print( $md5 . pack($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), 0);
430 $fh->print( $md5 . pack($LONG_PACK, $location) );
433 seek($fh, $subloc + SIG_SIZE, 0);
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' ) { $actual_length = $INDEX_SIZE; }
445 else { $actual_length = length($value); }
447 if ($actual_length <= $size) {
451 $location = $self->root->{end};
452 seek($fh, $tag->{offset} + ($i * $BUCKET_SIZE) + $HASH_SIZE, 0);
453 $fh->print( pack($LONG_PACK, $location) );
461 # If this is an internal reference, return now.
462 # No need to write value or plain key
469 # If bucket didn't fit into list, split into a new index level
472 seek($fh, $tag->{ref_loc}, 0);
473 $fh->print( pack($LONG_PACK, $self->root->{end}) );
475 my $index_tag = $self->_create_tag($self->root->{end}, SIG_INDEX, chr(0) x $INDEX_SIZE);
478 $keys .= $md5 . pack($LONG_PACK, 0);
480 for (my $i=0; $i<=$MAX_BUCKETS; $i++) {
481 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
483 my $old_subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
484 my $num = ord(substr($key, $tag->{ch} + 1, 1));
486 if ($offsets[$num]) {
487 my $offset = $offsets[$num] + SIG_SIZE + $DATA_LENGTH_SIZE;
488 seek($fh, $offset, 0);
490 read( $fh, $subkeys, $BUCKET_LIST_SIZE);
492 for (my $k=0; $k<$MAX_BUCKETS; $k++) {
493 my $subloc = unpack($LONG_PACK, substr($subkeys, ($k * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
495 seek($fh, $offset + ($k * $BUCKET_SIZE), 0);
496 $fh->print( $key . pack($LONG_PACK, $old_subloc || $self->root->{end}) );
502 $offsets[$num] = $self->root->{end};
503 seek($fh, $index_tag->{offset} + ($num * $LONG_SIZE), 0);
504 $fh->print( pack($LONG_PACK, $self->root->{end}) );
506 my $blist_tag = $self->_create_tag($self->root->{end}, SIG_BLIST, chr(0) x $BUCKET_LIST_SIZE);
508 seek($fh, $blist_tag->{offset}, 0);
509 $fh->print( $key . pack($LONG_PACK, $old_subloc || $self->root->{end}) );
514 $location ||= $self->root->{end};
515 } # re-index bucket list
518 # Seek to content area and store signature, value and plaintext key
522 seek($fh, $location, 0);
525 # Write signature based on content type, set content length and write actual value.
527 my $r = Scalar::Util::reftype($value) || '';
529 $fh->print( TYPE_HASH );
530 $fh->print( pack($DATA_LENGTH_PACK, $INDEX_SIZE) . chr(0) x $INDEX_SIZE );
531 $content_length = $INDEX_SIZE;
533 elsif ($r eq 'ARRAY') {
534 $fh->print( TYPE_ARRAY );
535 $fh->print( pack($DATA_LENGTH_PACK, $INDEX_SIZE) . chr(0) x $INDEX_SIZE );
536 $content_length = $INDEX_SIZE;
538 elsif (!defined($value)) {
539 $fh->print( SIG_NULL );
540 $fh->print( pack($DATA_LENGTH_PACK, 0) );
544 $fh->print( SIG_DATA );
545 $fh->print( pack($DATA_LENGTH_PACK, length($value)) . $value );
546 $content_length = length($value);
550 # Plain key is stored AFTER value, as keys are typically fetched less often.
552 $fh->print( pack($DATA_LENGTH_PACK, length($plain_key)) . $plain_key );
555 # If value is blessed, preserve class name
557 if ( $self->root->{autobless} ) {
558 my $value_class = Scalar::Util::blessed($value);
559 if ( defined $value_class && $value_class ne 'DBM::Deep' ) {
561 # Blessed ref -- will restore later
563 $fh->print( chr(1) );
564 $fh->print( pack($DATA_LENGTH_PACK, length($value_class)) . $value_class );
565 $content_length += 1;
566 $content_length += $DATA_LENGTH_SIZE + length($value_class);
569 $fh->print( chr(0) );
570 $content_length += 1;
575 # If this is a new content area, advance EOF counter
577 if ($location == $self->root->{end}) {
578 $self->root->{end} += SIG_SIZE;
579 $self->root->{end} += $DATA_LENGTH_SIZE + $content_length;
580 $self->root->{end} += $DATA_LENGTH_SIZE + length($plain_key);
584 # If content is a hash or array, create new child DeepDB object and
585 # pass each key or element to it.
588 my $branch = DBM::Deep->new(
590 base_offset => $location,
593 foreach my $key (keys %{$value}) {
594 $branch->{$key} = $value->{$key};
597 elsif ($r eq 'ARRAY') {
598 my $branch = DBM::Deep->new(
600 base_offset => $location,
604 foreach my $element (@{$value}) {
605 $branch->[$index] = $element;
613 return $self->_throw_error("Fatal error: indexing failed -- possibly due to corruption in file");
616 sub _get_bucket_value {
618 # Fetch single value given tag and MD5 digested key.
621 my ($tag, $md5) = @_;
622 my $keys = $tag->{content};
627 # Iterate through buckets, looking for a key match
630 for (my $i=0; $i<$MAX_BUCKETS; $i++) {
631 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
632 my $subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
636 # Hit end of list, no match
641 if ( $md5 ne $key ) {
646 # Found match -- seek to offset and read signature
649 seek($fh, $subloc, 0);
650 read( $fh, $signature, SIG_SIZE);
653 # If value is a hash or array, return new DeepDB object with correct offset
655 if (($signature eq TYPE_HASH) || ($signature eq TYPE_ARRAY)) {
656 my $obj = DBM::Deep->new(
658 base_offset => $subloc,
662 if ($self->root->{autobless}) {
664 # Skip over value and plain key to see if object needs
667 seek($fh, $DATA_LENGTH_SIZE + $INDEX_SIZE, 1);
670 read( $fh, $size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
671 if ($size) { seek($fh, $size, 1); }
674 read( $fh, $bless_bit, 1);
675 if (ord($bless_bit)) {
677 # Yes, object needs to be re-blessed
680 read( $fh, $size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
681 if ($size) { read( $fh, $class_name, $size); }
682 if ($class_name) { $obj = bless( $obj, $class_name ); }
690 # Otherwise return actual value
692 elsif ($signature eq SIG_DATA) {
695 read( $fh, $size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
696 if ($size) { read( $fh, $value, $size); }
701 # Key exists, but content is null
711 # Delete single key/value pair given tag and MD5 digested key.
714 my ($tag, $md5) = @_;
715 my $keys = $tag->{content};
720 # Iterate through buckets, looking for a key match
723 for (my $i=0; $i<$MAX_BUCKETS; $i++) {
724 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
725 my $subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
729 # Hit end of list, no match
734 if ( $md5 ne $key ) {
739 # Matched key -- delete bucket and return
741 seek($fh, $tag->{offset} + ($i * $BUCKET_SIZE), 0);
742 $fh->print( substr($keys, ($i+1) * $BUCKET_SIZE ) );
743 $fh->print( chr(0) x $BUCKET_SIZE );
753 # Check existence of single key given tag and MD5 digested key.
756 my ($tag, $md5) = @_;
757 my $keys = $tag->{content};
760 # Iterate through buckets, looking for a key match
763 for (my $i=0; $i<$MAX_BUCKETS; $i++) {
764 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
765 my $subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
769 # Hit end of list, no match
774 if ( $md5 ne $key ) {
779 # Matched key -- return true
787 sub _find_bucket_list {
789 # Locate offset for bucket list, given digested key
795 # Locate offset for bucket list using digest index system
798 my $tag = $self->_load_tag($self->base_offset);
799 if (!$tag) { return; }
801 while ($tag->{signature} ne SIG_BLIST) {
802 $tag = $self->_index_lookup($tag, ord(substr($md5, $ch, 1)));
803 if (!$tag) { return; }
810 sub _traverse_index {
812 # Scan index and recursively step into deeper levels, looking for next key.
814 my ($self, $offset, $ch, $force_return_next) = @_;
815 $force_return_next = undef unless $force_return_next;
817 my $tag = $self->_load_tag( $offset );
821 if ($tag->{signature} ne SIG_BLIST) {
822 my $content = $tag->{content};
824 if ($self->{return_next}) { $start = 0; }
825 else { $start = ord(substr($self->{prev_md5}, $ch, 1)); }
827 for (my $index = $start; $index < 256; $index++) {
828 my $subloc = unpack($LONG_PACK, substr($content, $index * $LONG_SIZE, $LONG_SIZE) );
830 my $result = $self->_traverse_index( $subloc, $ch + 1, $force_return_next );
831 if (defined($result)) { return $result; }
835 $self->{return_next} = 1;
838 elsif ($tag->{signature} eq SIG_BLIST) {
839 my $keys = $tag->{content};
840 if ($force_return_next) { $self->{return_next} = 1; }
843 # Iterate through buckets, looking for a key match
845 for (my $i=0; $i<$MAX_BUCKETS; $i++) {
846 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
847 my $subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
851 # End of bucket list -- return to outer loop
853 $self->{return_next} = 1;
856 elsif ($key eq $self->{prev_md5}) {
858 # Located previous key -- return next one found
860 $self->{return_next} = 1;
863 elsif ($self->{return_next}) {
865 # Seek to bucket location and skip over signature
867 seek($fh, $subloc + SIG_SIZE, 0);
870 # Skip over value to get to plain key
873 read( $fh, $size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
874 if ($size) { seek($fh, $size, 1); }
877 # Read in plain key and return as scalar
880 read( $fh, $size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
881 if ($size) { read( $fh, $plain_key, $size); }
887 $self->{return_next} = 1;
888 } # tag is a bucket list
895 # Locate next key, given digested previous one
897 my $self = _get_self($_[0]);
899 $self->{prev_md5} = $_[1] ? $_[1] : undef;
900 $self->{return_next} = 0;
903 # If the previous key was not specifed, start at the top and
904 # return the first one found.
906 if (!$self->{prev_md5}) {
907 $self->{prev_md5} = chr(0) x $HASH_SIZE;
908 $self->{return_next} = 1;
911 return $self->_traverse_index( $self->base_offset, 0 );
916 # If db locking is set, flock() the db file. If called multiple
917 # times before unlock(), then the same number of unlocks() must
918 # be called before the lock is released.
920 my $self = _get_self($_[0]);
922 $type = LOCK_EX unless defined $type;
924 if ($self->root->{locking}) {
925 if (!$self->root->{locked}) { flock($self->fh, $type); }
926 $self->root->{locked}++;
932 # If db locking is set, unlock the db file. See note in lock()
933 # regarding calling lock() multiple times.
935 my $self = _get_self($_[0]);
937 if ($self->root->{locking} && $self->root->{locked} > 0) {
938 $self->root->{locked}--;
939 if (!$self->root->{locked}) { flock($self->fh, LOCK_UN); }
943 #XXX These uses of ref() need verified
946 # Copy single level of keys or elements to new DB handle.
947 # Recurse for nested structures
949 my $self = _get_self($_[0]);
952 if ($self->type eq TYPE_HASH) {
953 my $key = $self->first_key();
955 my $value = $self->get($key);
956 #XXX This doesn't work with autobless
957 if (!ref($value)) { $db_temp->{$key} = $value; }
959 my $type = $value->type;
960 if ($type eq TYPE_HASH) { $db_temp->{$key} = {}; }
961 else { $db_temp->{$key} = []; }
962 $value->_copy_node( $db_temp->{$key} );
964 $key = $self->next_key($key);
968 my $length = $self->length();
969 for (my $index = 0; $index < $length; $index++) {
970 my $value = $self->get($index);
971 if (!ref($value)) { $db_temp->[$index] = $value; }
972 #XXX NO tests for this code
974 my $type = $value->type;
975 if ($type eq TYPE_HASH) { $db_temp->[$index] = {}; }
976 else { $db_temp->[$index] = []; }
977 $value->_copy_node( $db_temp->[$index] );
985 # Recursively export into standard Perl hashes and arrays.
987 my $self = _get_self($_[0]);
990 if ($self->type eq TYPE_HASH) { $temp = {}; }
991 elsif ($self->type eq TYPE_ARRAY) { $temp = []; }
994 $self->_copy_node( $temp );
1002 # Recursively import Perl hash/array structure
1004 #XXX This use of ref() seems to be ok
1005 if (!ref($_[0])) { return; } # Perl calls import() on use -- ignore
1007 my $self = _get_self($_[0]);
1010 #XXX This use of ref() seems to be ok
1011 if (!ref($struct)) {
1013 # struct is not a reference, so just import based on our type
1017 if ($self->type eq TYPE_HASH) { $struct = {@_}; }
1018 elsif ($self->type eq TYPE_ARRAY) { $struct = [@_]; }
1021 my $r = Scalar::Util::reftype($struct) || '';
1022 if ($r eq "HASH" && $self->type eq TYPE_HASH) {
1023 foreach my $key (keys %$struct) { $self->put($key, $struct->{$key}); }
1025 elsif ($r eq "ARRAY" && $self->type eq TYPE_ARRAY) {
1026 $self->push( @$struct );
1029 return $self->_throw_error("Cannot import: type mismatch");
1037 # Rebuild entire database into new file, then move
1038 # it back on top of original.
1040 my $self = _get_self($_[0]);
1041 if ($self->root->{links} > 1) {
1042 return $self->_throw_error("Cannot optimize: reference count is greater than 1");
1045 my $db_temp = DBM::Deep->new(
1046 file => $self->root->{file} . '.tmp',
1050 return $self->_throw_error("Cannot optimize: failed to open temp file: $!");
1054 $self->_copy_node( $db_temp );
1058 # Attempt to copy user, group and permissions over to new file
1060 my @stats = stat($self->fh);
1061 my $perms = $stats[2] & 07777;
1062 my $uid = $stats[4];
1063 my $gid = $stats[5];
1064 chown( $uid, $gid, $self->root->{file} . '.tmp' );
1065 chmod( $perms, $self->root->{file} . '.tmp' );
1067 # q.v. perlport for more information on this variable
1068 if ( $^O eq 'MSWin32' ) {
1070 # Potential race condition when optmizing on Win32 with locking.
1071 # The Windows filesystem requires that the filehandle be closed
1072 # before it is overwritten with rename(). This could be redone
1079 if (!rename $self->root->{file} . '.tmp', $self->root->{file}) {
1080 unlink $self->root->{file} . '.tmp';
1082 return $self->_throw_error("Optimize failed: Cannot copy temp file over original: $!");
1094 # Make copy of object and return
1096 my $self = _get_self($_[0]);
1098 return DBM::Deep->new(
1099 type => $self->type,
1100 base_offset => $self->base_offset,
1106 my %is_legal_filter = map {
1109 store_key store_value
1110 fetch_key fetch_value
1115 # Setup filter function for storing or fetching the key or value
1117 my $self = _get_self($_[0]);
1118 my $type = lc $_[1];
1119 my $func = $_[2] ? $_[2] : undef;
1121 if ( $is_legal_filter{$type} ) {
1122 $self->root->{"filter_$type"} = $func;
1136 # Get access to the root structure
1138 my $self = _get_self($_[0]);
1139 return $self->{root};
1144 # Get access to the raw FileHandle
1146 #XXX It will be useful, though, when we split out HASH and ARRAY
1147 my $self = _get_self($_[0]);
1148 return $self->root->{fh};
1153 # Get type of current node (TYPE_HASH or TYPE_ARRAY)
1155 my $self = _get_self($_[0]);
1156 return $self->{type};
1161 # Get base_offset of current node (TYPE_HASH or TYPE_ARRAY)
1163 my $self = _get_self($_[0]);
1164 return $self->{base_offset};
1169 # Get last error string, or undef if no error
1172 ? ( _get_self($_[0])->{root}->{error} or undef )
1182 # Store error string in self
1184 my $self = _get_self($_[0]);
1185 my $error_text = $_[1];
1187 $self->root->{error} = $error_text;
1189 unless ($self->root->{debug}) {
1190 die "DBM::Deep: $error_text\n";
1193 warn "DBM::Deep: $error_text\n";
1201 my $self = _get_self($_[0]);
1203 undef $self->root->{error};
1208 # Precalculate index, bucket and bucket list sizes
1211 #XXX I don't like this ...
1212 set_pack() unless defined $LONG_SIZE;
1214 $INDEX_SIZE = 256 * $LONG_SIZE;
1215 $BUCKET_SIZE = $HASH_SIZE + $LONG_SIZE;
1216 $BUCKET_LIST_SIZE = $MAX_BUCKETS * $BUCKET_SIZE;
1221 # Set pack/unpack modes (see file header for more)
1223 my ($long_s, $long_p, $data_s, $data_p) = @_;
1225 $LONG_SIZE = $long_s ? $long_s : 4;
1226 $LONG_PACK = $long_p ? $long_p : 'N';
1228 $DATA_LENGTH_SIZE = $data_s ? $data_s : 4;
1229 $DATA_LENGTH_PACK = $data_p ? $data_p : 'N';
1236 # Set key digest function (default is MD5)
1238 my ($digest_func, $hash_size) = @_;
1240 $DIGEST_FUNC = $digest_func ? $digest_func : \&Digest::MD5::md5;
1241 $HASH_SIZE = $hash_size ? $hash_size : 16;
1247 # tie() methods (hashes and arrays)
1252 # Store single hash key/value or array element in database.
1254 my $self = _get_self($_[0]);
1255 my $key = ($self->root->{filter_store_key} && $self->type eq TYPE_HASH) ? $self->root->{filter_store_key}->($_[1]) : $_[1];
1256 #XXX What is ref() checking here?
1257 #YYY User may be storing a hash, in which case we do not want it run
1258 #YYY through the filtering system
1259 my $value = ($self->root->{filter_store_value} && !ref($_[2])) ? $self->root->{filter_store_value}->($_[2]) : $_[2];
1261 my $unpacked_key = $key;
1262 if (($self->type eq TYPE_ARRAY) && ($key =~ /^\d+$/)) { $key = pack($LONG_PACK, $key); }
1263 my $md5 = $DIGEST_FUNC->($key);
1266 # Make sure file is open
1268 if (!defined($self->fh) && !$self->_open()) {
1275 # Request exclusive lock for writing
1277 $self->lock( LOCK_EX );
1280 # If locking is enabled, set 'end' parameter again, in case another
1281 # DB instance appended to our file while we were unlocked.
1283 if ($self->root->{locking} || $self->root->{volatile}) {
1284 $self->root->{end} = (stat($fh))[7];
1288 # Locate offset for bucket list using digest index system
1290 my $tag = $self->_load_tag($self->base_offset);
1292 $tag = $self->_create_tag($self->base_offset, SIG_INDEX, chr(0) x $INDEX_SIZE);
1296 while ($tag->{signature} ne SIG_BLIST) {
1297 my $num = ord(substr($md5, $ch, 1));
1298 my $new_tag = $self->_index_lookup($tag, $num);
1300 my $ref_loc = $tag->{offset} + ($num * $LONG_SIZE);
1301 seek($fh, $ref_loc, 0);
1302 $fh->print( pack($LONG_PACK, $self->root->{end}) );
1304 $tag = $self->_create_tag($self->root->{end}, SIG_BLIST, chr(0) x $BUCKET_LIST_SIZE);
1305 $tag->{ref_loc} = $ref_loc;
1310 my $ref_loc = $tag->{offset} + ($num * $LONG_SIZE);
1312 $tag->{ref_loc} = $ref_loc;
1319 # Add key/value to bucket list
1321 my $result = $self->_add_bucket( $tag, $md5, $key, $value );
1324 # If this object is an array, and bucket was not a replace, and key is numerical,
1325 # and index is equal or greater than current length, advance length variable.
1327 if (($result == 2) && ($self->type eq TYPE_ARRAY) && ($unpacked_key =~ /^\d+$/) && ($unpacked_key >= $self->FETCHSIZE())) {
1328 $self->STORESIZE( $unpacked_key + 1 );
1338 # Fetch single value or element given plain key or array index
1340 my $self = _get_self($_[0]);
1343 if ( $self->type eq TYPE_HASH ) {
1344 if ( my $filter = $self->root->{filter_store_key} ) {
1345 $key = $filter->( $key );
1348 elsif ( $self->type eq TYPE_ARRAY ) {
1349 if ( $key =~ /^\d+$/ ) {
1350 $key = pack($LONG_PACK, $key);
1354 my $md5 = $DIGEST_FUNC->($key);
1357 # Make sure file is open
1359 if (!defined($self->fh)) { $self->_open(); }
1362 # Request shared lock for reading
1364 $self->lock( LOCK_SH );
1366 my $tag = $self->_find_bucket_list( $md5 );
1373 # Get value from bucket list
1375 my $result = $self->_get_bucket_value( $tag, $md5 );
1379 #XXX What is ref() checking here?
1380 return ($result && !ref($result) && $self->root->{filter_fetch_value}) ? $self->root->{filter_fetch_value}->($result) : $result;
1385 # Delete single key/value pair or element given plain key or array index
1387 my $self = _get_self($_[0]);
1388 my $key = ($self->root->{filter_store_key} && $self->type eq TYPE_HASH) ? $self->root->{filter_store_key}->($_[1]) : $_[1];
1390 my $unpacked_key = $key;
1391 if (($self->type eq TYPE_ARRAY) && ($key =~ /^\d+$/)) { $key = pack($LONG_PACK, $key); }
1392 my $md5 = $DIGEST_FUNC->($key);
1395 # Make sure file is open
1397 if (!defined($self->fh)) { $self->_open(); }
1400 # Request exclusive lock for writing
1402 $self->lock( LOCK_EX );
1404 my $tag = $self->_find_bucket_list( $md5 );
1413 my $result = $self->_delete_bucket( $tag, $md5 );
1416 # If this object is an array and the key deleted was on the end of the stack,
1417 # decrement the length variable.
1419 if ($result && ($self->type eq TYPE_ARRAY) && ($unpacked_key == $self->FETCHSIZE() - 1)) {
1420 $self->STORESIZE( $unpacked_key );
1430 # Check if a single key or element exists given plain key or array index
1432 my $self = _get_self($_[0]);
1433 my $key = ($self->root->{filter_store_key} && $self->type eq TYPE_HASH) ? $self->root->{filter_store_key}->($_[1]) : $_[1];
1435 if (($self->type eq TYPE_ARRAY) && ($key =~ /^\d+$/)) { $key = pack($LONG_PACK, $key); }
1436 my $md5 = $DIGEST_FUNC->($key);
1439 # Make sure file is open
1441 if (!defined($self->fh)) { $self->_open(); }
1444 # Request shared lock for reading
1446 $self->lock( LOCK_SH );
1448 my $tag = $self->_find_bucket_list( $md5 );
1451 # For some reason, the built-in exists() function returns '' for false
1459 # Check if bucket exists and return 1 or ''
1461 my $result = $self->_bucket_exists( $tag, $md5 ) || '';
1470 # Clear all keys from hash, or all elements from array.
1472 my $self = _get_self($_[0]);
1475 # Make sure file is open
1477 if (!defined($self->fh)) { $self->_open(); }
1480 # Request exclusive lock for writing
1482 $self->lock( LOCK_EX );
1486 seek($fh, $self->base_offset, 0);
1492 $self->_create_tag($self->base_offset, $self->type, chr(0) x $INDEX_SIZE);
1501 # Locate and return first key (in no particular order)
1503 my $self = _get_self($_[0]);
1504 if ($self->type ne TYPE_HASH) {
1505 return $self->_throw_error("FIRSTKEY method only supported for hashes");
1509 # Make sure file is open
1511 if (!defined($self->fh)) { $self->_open(); }
1514 # Request shared lock for reading
1516 $self->lock( LOCK_SH );
1518 my $result = $self->_get_next_key();
1522 return ($result && $self->root->{filter_fetch_key}) ? $self->root->{filter_fetch_key}->($result) : $result;
1527 # Return next key (in no particular order), given previous one
1529 my $self = _get_self($_[0]);
1530 if ($self->type ne TYPE_HASH) {
1531 return $self->_throw_error("NEXTKEY method only supported for hashes");
1533 my $prev_key = ($self->root->{filter_store_key} && $self->type eq TYPE_HASH) ? $self->root->{filter_store_key}->($_[1]) : $_[1];
1534 my $prev_md5 = $DIGEST_FUNC->($prev_key);
1537 # Make sure file is open
1539 if (!defined($self->fh)) { $self->_open(); }
1542 # Request shared lock for reading
1544 $self->lock( LOCK_SH );
1546 my $result = $self->_get_next_key( $prev_md5 );
1550 return ($result && $self->root->{filter_fetch_key}) ? $self->root->{filter_fetch_key}->($result) : $result;
1554 # The following methods are for arrays only
1559 # Return the length of the array
1561 my $self = _get_self($_[0]);
1562 if ($self->type ne TYPE_ARRAY) {
1563 return $self->_throw_error("FETCHSIZE method only supported for arrays");
1566 my $SAVE_FILTER = $self->root->{filter_fetch_value};
1567 $self->root->{filter_fetch_value} = undef;
1569 my $packed_size = $self->FETCH('length');
1571 $self->root->{filter_fetch_value} = $SAVE_FILTER;
1573 if ($packed_size) { return int(unpack($LONG_PACK, $packed_size)); }
1579 # Set the length of the array
1581 my $self = _get_self($_[0]);
1582 if ($self->type ne TYPE_ARRAY) {
1583 return $self->_throw_error("STORESIZE method only supported for arrays");
1585 my $new_length = $_[1];
1587 my $SAVE_FILTER = $self->root->{filter_store_value};
1588 $self->root->{filter_store_value} = undef;
1590 my $result = $self->STORE('length', pack($LONG_PACK, $new_length));
1592 $self->root->{filter_store_value} = $SAVE_FILTER;
1599 # Remove and return the last element on the array
1601 my $self = _get_self($_[0]);
1602 if ($self->type ne TYPE_ARRAY) {
1603 return $self->_throw_error("POP method only supported for arrays");
1605 my $length = $self->FETCHSIZE();
1608 my $content = $self->FETCH( $length - 1 );
1609 $self->DELETE( $length - 1 );
1619 # Add new element(s) to the end of the array
1621 my $self = _get_self(shift);
1622 if ($self->type ne TYPE_ARRAY) {
1623 return $self->_throw_error("PUSH method only supported for arrays");
1625 my $length = $self->FETCHSIZE();
1627 while (my $content = shift @_) {
1628 $self->STORE( $length, $content );
1635 # Remove and return first element on the array.
1636 # Shift over remaining elements to take up space.
1638 my $self = _get_self($_[0]);
1639 if ($self->type ne TYPE_ARRAY) {
1640 return $self->_throw_error("SHIFT method only supported for arrays");
1642 my $length = $self->FETCHSIZE();
1645 my $content = $self->FETCH( 0 );
1648 # Shift elements over and remove last one.
1650 for (my $i = 0; $i < $length - 1; $i++) {
1651 $self->STORE( $i, $self->FETCH($i + 1) );
1653 $self->DELETE( $length - 1 );
1664 # Insert new element(s) at beginning of array.
1665 # Shift over other elements to make space.
1667 my $self = _get_self($_[0]);shift @_;
1668 if ($self->type ne TYPE_ARRAY) {
1669 return $self->_throw_error("UNSHIFT method only supported for arrays");
1671 my @new_elements = @_;
1672 my $length = $self->FETCHSIZE();
1673 my $new_size = scalar @new_elements;
1676 for (my $i = $length - 1; $i >= 0; $i--) {
1677 $self->STORE( $i + $new_size, $self->FETCH($i) );
1681 for (my $i = 0; $i < $new_size; $i++) {
1682 $self->STORE( $i, $new_elements[$i] );
1688 # Splices section of array with optional new section.
1689 # Returns deleted section, or last element deleted in scalar context.
1691 my $self = _get_self($_[0]);shift @_;
1692 if ($self->type ne TYPE_ARRAY) {
1693 return $self->_throw_error("SPLICE method only supported for arrays");
1695 my $length = $self->FETCHSIZE();
1698 # Calculate offset and length of splice
1700 my $offset = shift || 0;
1701 if ($offset < 0) { $offset += $length; }
1704 if (scalar @_) { $splice_length = shift; }
1705 else { $splice_length = $length - $offset; }
1706 if ($splice_length < 0) { $splice_length += ($length - $offset); }
1709 # Setup array with new elements, and copy out old elements for return
1711 my @new_elements = @_;
1712 my $new_size = scalar @new_elements;
1714 my @old_elements = ();
1715 for (my $i = $offset; $i < $offset + $splice_length; $i++) {
1716 push @old_elements, $self->FETCH( $i );
1720 # Adjust array length, and shift elements to accomodate new section.
1722 if ( $new_size != $splice_length ) {
1723 if ($new_size > $splice_length) {
1724 for (my $i = $length - 1; $i >= $offset + $splice_length; $i--) {
1725 $self->STORE( $i + ($new_size - $splice_length), $self->FETCH($i) );
1729 for (my $i = $offset + $splice_length; $i < $length; $i++) {
1730 $self->STORE( $i + ($new_size - $splice_length), $self->FETCH($i) );
1732 for (my $i = 0; $i < $splice_length - $new_size; $i++) {
1733 $self->DELETE( $length - 1 );
1740 # Insert new elements into array
1742 for (my $i = $offset; $i < $offset + $new_size; $i++) {
1743 $self->STORE( $i, shift @new_elements );
1747 # Return deleted section, or last element in scalar context.
1749 return wantarray ? @old_elements : $old_elements[-1];
1752 #XXX We don't need to define it.
1753 #XXX It will be useful, though, when we split out HASH and ARRAY
1756 # Perl will call EXTEND() when the array is likely to grow.
1757 # We don't care, but include it for compatibility.
1762 # Public method aliases
1764 *put = *store = *STORE;
1765 *get = *fetch = *FETCH;
1769 *first_key = *FIRSTKEY;
1770 *next_key = *NEXTKEY;
1771 *length = *FETCHSIZE;
1775 *unshift = *UNSHIFT;
1784 DBM::Deep - A pure perl multi-level hash/array DBM
1789 my $db = DBM::Deep->new( "foo.db" );
1791 $db->{key} = 'value'; # tie() style
1794 $db->put('key', 'value'); # OO style
1795 print $db->get('key');
1797 # true multi-level support
1798 $db->{my_complex} = [
1799 'hello', { perl => 'rules' },
1804 A unique flat-file database module, written in pure perl. True
1805 multi-level hash/array support (unlike MLDBM, which is faked), hybrid
1806 OO / tie() interface, cross-platform FTPable files, and quite fast. Can
1807 handle millions of keys and unlimited hash levels without significant
1808 slow-down. Written from the ground-up in pure perl -- this is NOT a
1809 wrapper around a C-based DBM. Out-of-the-box compatibility with Unix,
1810 Mac OS X and Windows.
1814 Hopefully you are using CPAN's excellent Perl module, which will download
1815 and install the module for you. If not, get the tarball, and run these
1827 Construction can be done OO-style (which is the recommended way), or using
1828 Perl's tie() function. Both are examined here.
1830 =head2 OO CONSTRUCTION
1832 The recommended way to construct a DBM::Deep object is to use the new()
1833 method, which gets you a blessed, tied hash or array reference.
1835 my $db = DBM::Deep->new( "foo.db" );
1837 This opens a new database handle, mapped to the file "foo.db". If this
1838 file does not exist, it will automatically be created. DB files are
1839 opened in "r+" (read/write) mode, and the type of object returned is a
1840 hash, unless otherwise specified (see L<OPTIONS> below).
1844 You can pass a number of options to the constructor to specify things like
1845 locking, autoflush, etc. This is done by passing an inline hash:
1847 my $db = DBM::Deep->new(
1853 Notice that the filename is now specified I<inside> the hash with
1854 the "file" parameter, as opposed to being the sole argument to the
1855 constructor. This is required if any options are specified.
1856 See L<OPTIONS> below for the complete list.
1860 You can also start with an array instead of a hash. For this, you must
1861 specify the C<type> parameter:
1863 my $db = DBM::Deep->new(
1865 type => DBM::Deep->TYPE_ARRAY
1868 B<Note:> Specifing the C<type> parameter only takes effect when beginning
1869 a new DB file. If you create a DBM::Deep object with an existing file, the
1870 C<type> will be loaded from the file header, and ignored if it is passed
1873 =head2 TIE CONSTRUCTION
1875 Alternatively, you can create a DBM::Deep handle by using Perl's built-in
1876 tie() function. This is not ideal, because you get only a basic, tied hash
1877 (or array) which is not blessed, so you can't call any functions on it.
1880 tie %hash, "DBM::Deep", "foo.db";
1883 tie @array, "DBM::Deep", "bar.db";
1885 As with the OO constructor, you can replace the DB filename parameter with
1886 a hash containing one or more options (see L<OPTIONS> just below for the
1889 tie %hash, "DBM::Deep", {
1897 There are a number of options that can be passed in when constructing your
1898 DBM::Deep objects. These apply to both the OO- and tie- based approaches.
1904 Filename of the DB file to link the handle to. You can pass a full absolute
1905 filesystem path, partial path, or a plain filename if the file is in the
1906 current working directory. This is a required parameter.
1910 File open mode (read-only, read-write, etc.) string passed to Perl's FileHandle
1911 module. This is an optional parameter, and defaults to "r+" (read/write).
1912 B<Note:> If the default (r+) mode is selected, the file will also be auto-
1913 created if it doesn't exist.
1917 This parameter specifies what type of object to create, a hash or array. Use
1918 one of these two constants: C<DBM::Deep-E<gt>TYPE_HASH> or C<DBM::Deep-E<gt>TYPE_ARRAY>.
1919 This only takes effect when beginning a new file. This is an optional
1920 parameter, and defaults to C<DBM::Deep-E<gt>TYPE_HASH>.
1924 Specifies whether locking is to be enabled. DBM::Deep uses Perl's Fnctl flock()
1925 function to lock the database in exclusive mode for writes, and shared mode for
1926 reads. Pass any true value to enable. This affects the base DB handle I<and
1927 any child hashes or arrays> that use the same DB file. This is an optional
1928 parameter, and defaults to 0 (disabled). See L<LOCKING> below for more.
1932 Specifies whether autoflush is to be enabled on the underlying FileHandle.
1933 This obviously slows down write operations, but is required if you may have
1934 multiple processes accessing the same DB file (also consider enable I<locking>
1935 or at least I<volatile>). Pass any true value to enable. This is an optional
1936 parameter, and defaults to 0 (disabled).
1940 If I<volatile> mode is enabled, DBM::Deep will stat() the DB file before each
1941 STORE() operation. This is required if an outside force may change the size of
1942 the file between transactions. Locking also implicitly enables volatile. This
1943 is useful if you want to use a different locking system or write your own. Pass
1944 any true value to enable. This is an optional parameter, and defaults to 0
1949 If I<autobless> mode is enabled, DBM::Deep will preserve blessed hashes, and
1950 restore them when fetched. This is an B<experimental> feature, and does have
1951 side-effects. Basically, when hashes are re-blessed into their original
1952 classes, they are no longer blessed into the DBM::Deep class! So you won't be
1953 able to call any DBM::Deep methods on them. You have been warned.
1954 This is an optional parameter, and defaults to 0 (disabled).
1958 See L<FILTERS> below.
1962 Setting I<debug> mode will make all errors non-fatal, dump them out to
1963 STDERR, and continue on. This is for debugging purposes only, and probably
1964 not what you want. This is an optional parameter, and defaults to 0 (disabled).
1968 Instead of passing a file path, you can instead pass a handle to an pre-opened
1969 filehandle. Note: Beware of using the magick *DATA handle, as this actually
1970 contains your entire Perl script, as well as the data following the __DATA__
1971 marker. This will not work, because DBM::Deep uses absolute seek()s into the
1972 file. Instead, consider reading *DATA into an IO::Scalar handle, then passing
1977 =head1 TIE INTERFACE
1979 With DBM::Deep you can access your databases using Perl's standard hash/array
1980 syntax. Because all DBM::Deep objects are I<tied> to hashes or arrays, you can treat
1981 them as such. DBM::Deep will intercept all reads/writes and direct them to the right
1982 place -- the DB file. This has nothing to do with the L<TIE CONSTRUCTION>
1983 section above. This simply tells you how to use DBM::Deep using regular hashes
1984 and arrays, rather than calling functions like C<get()> and C<put()> (although those
1985 work too). It is entirely up to you how to want to access your databases.
1989 You can treat any DBM::Deep object like a normal Perl hash reference. Add keys,
1990 or even nested hashes (or arrays) using standard Perl syntax:
1992 my $db = DBM::Deep->new( "foo.db" );
1994 $db->{mykey} = "myvalue";
1996 $db->{myhash}->{subkey} = "subvalue";
1998 print $db->{myhash}->{subkey} . "\n";
2000 You can even step through hash keys using the normal Perl C<keys()> function:
2002 foreach my $key (keys %$db) {
2003 print "$key: " . $db->{$key} . "\n";
2006 Remember that Perl's C<keys()> function extracts I<every> key from the hash and
2007 pushes them onto an array, all before the loop even begins. If you have an
2008 extra large hash, this may exhaust Perl's memory. Instead, consider using
2009 Perl's C<each()> function, which pulls keys/values one at a time, using very
2012 while (my ($key, $value) = each %$db) {
2013 print "$key: $value\n";
2016 Please note that when using C<each()>, you should always pass a direct
2017 hash reference, not a lookup. Meaning, you should B<never> do this:
2020 while (my ($key, $value) = each %{$db->{foo}}) { # BAD
2022 This causes an infinite loop, because for each iteration, Perl is calling
2023 FETCH() on the $db handle, resulting in a "new" hash for foo every time, so
2024 it effectively keeps returning the first key over and over again. Instead,
2025 assign a temporary variable to C<$db->{foo}>, then pass that to each().
2029 As with hashes, you can treat any DBM::Deep object like a normal Perl array
2030 reference. This includes inserting, removing and manipulating elements,
2031 and the C<push()>, C<pop()>, C<shift()>, C<unshift()> and C<splice()> functions.
2032 The object must have first been created using type C<DBM::Deep-E<gt>TYPE_ARRAY>,
2033 or simply be a nested array reference inside a hash. Example:
2035 my $db = DBM::Deep->new(
2036 file => "foo-array.db",
2037 type => DBM::Deep->TYPE_ARRAY
2041 push @$db, "bar", "baz";
2042 unshift @$db, "bah";
2044 my $last_elem = pop @$db; # baz
2045 my $first_elem = shift @$db; # bah
2046 my $second_elem = $db->[1]; # bar
2048 my $num_elements = scalar @$db;
2052 In addition to the I<tie()> interface, you can also use a standard OO interface
2053 to manipulate all aspects of DBM::Deep databases. Each type of object (hash or
2054 array) has its own methods, but both types share the following common methods:
2055 C<put()>, C<get()>, C<exists()>, C<delete()> and C<clear()>.
2061 Stores a new hash key/value pair, or sets an array element value. Takes two
2062 arguments, the hash key or array index, and the new value. The value can be
2063 a scalar, hash ref or array ref. Returns true on success, false on failure.
2065 $db->put("foo", "bar"); # for hashes
2066 $db->put(1, "bar"); # for arrays
2070 Fetches the value of a hash key or array element. Takes one argument: the hash
2071 key or array index. Returns a scalar, hash ref or array ref, depending on the
2074 my $value = $db->get("foo"); # for hashes
2075 my $value = $db->get(1); # for arrays
2079 Checks if a hash key or array index exists. Takes one argument: the hash key
2080 or array index. Returns true if it exists, false if not.
2082 if ($db->exists("foo")) { print "yay!\n"; } # for hashes
2083 if ($db->exists(1)) { print "yay!\n"; } # for arrays
2087 Deletes one hash key/value pair or array element. Takes one argument: the hash
2088 key or array index. Returns true on success, false if not found. For arrays,
2089 the remaining elements located after the deleted element are NOT moved over.
2090 The deleted element is essentially just undefined, which is exactly how Perl's
2091 internal arrays work. Please note that the space occupied by the deleted
2092 key/value or element is B<not> reused again -- see L<UNUSED SPACE RECOVERY>
2093 below for details and workarounds.
2095 $db->delete("foo"); # for hashes
2096 $db->delete(1); # for arrays
2100 Deletes B<all> hash keys or array elements. Takes no arguments. No return
2101 value. Please note that the space occupied by the deleted keys/values or
2102 elements is B<not> reused again -- see L<UNUSED SPACE RECOVERY> below for
2103 details and workarounds.
2105 $db->clear(); # hashes or arrays
2111 For hashes, DBM::Deep supports all the common methods described above, and the
2112 following additional methods: C<first_key()> and C<next_key()>.
2118 Returns the "first" key in the hash. As with built-in Perl hashes, keys are
2119 fetched in an undefined order (which appears random). Takes no arguments,
2120 returns the key as a scalar value.
2122 my $key = $db->first_key();
2126 Returns the "next" key in the hash, given the previous one as the sole argument.
2127 Returns undef if there are no more keys to be fetched.
2129 $key = $db->next_key($key);
2133 Here are some examples of using hashes:
2135 my $db = DBM::Deep->new( "foo.db" );
2137 $db->put("foo", "bar");
2138 print "foo: " . $db->get("foo") . "\n";
2140 $db->put("baz", {}); # new child hash ref
2141 $db->get("baz")->put("buz", "biz");
2142 print "buz: " . $db->get("baz")->get("buz") . "\n";
2144 my $key = $db->first_key();
2146 print "$key: " . $db->get($key) . "\n";
2147 $key = $db->next_key($key);
2150 if ($db->exists("foo")) { $db->delete("foo"); }
2154 For arrays, DBM::Deep supports all the common methods described above, and the
2155 following additional methods: C<length()>, C<push()>, C<pop()>, C<shift()>,
2156 C<unshift()> and C<splice()>.
2162 Returns the number of elements in the array. Takes no arguments.
2164 my $len = $db->length();
2168 Adds one or more elements onto the end of the array. Accepts scalars, hash
2169 refs or array refs. No return value.
2171 $db->push("foo", "bar", {});
2175 Fetches the last element in the array, and deletes it. Takes no arguments.
2176 Returns undef if array is empty. Returns the element value.
2178 my $elem = $db->pop();
2182 Fetches the first element in the array, deletes it, then shifts all the
2183 remaining elements over to take up the space. Returns the element value. This
2184 method is not recommended with large arrays -- see L<LARGE ARRAYS> below for
2187 my $elem = $db->shift();
2191 Inserts one or more elements onto the beginning of the array, shifting all
2192 existing elements over to make room. Accepts scalars, hash refs or array refs.
2193 No return value. This method is not recommended with large arrays -- see
2194 <LARGE ARRAYS> below for details.
2196 $db->unshift("foo", "bar", {});
2200 Performs exactly like Perl's built-in function of the same name. See L<perldoc
2201 -f splice> for usage -- it is too complicated to document here. This method is
2202 not recommended with large arrays -- see L<LARGE ARRAYS> below for details.
2206 Here are some examples of using arrays:
2208 my $db = DBM::Deep->new(
2210 type => DBM::Deep->TYPE_ARRAY
2213 $db->push("bar", "baz");
2214 $db->unshift("foo");
2217 my $len = $db->length();
2218 print "length: $len\n"; # 4
2220 for (my $k=0; $k<$len; $k++) {
2221 print "$k: " . $db->get($k) . "\n";
2224 $db->splice(1, 2, "biz", "baf");
2226 while (my $elem = shift @$db) {
2227 print "shifted: $elem\n";
2232 Enable automatic file locking by passing a true value to the C<locking>
2233 parameter when constructing your DBM::Deep object (see L<SETUP> above).
2235 my $db = DBM::Deep->new(
2240 This causes DBM::Deep to C<flock()> the underlying FileHandle object with exclusive
2241 mode for writes, and shared mode for reads. This is required if you have
2242 multiple processes accessing the same database file, to avoid file corruption.
2243 Please note that C<flock()> does NOT work for files over NFS. See L<DB OVER
2244 NFS> below for more.
2246 =head2 EXPLICIT LOCKING
2248 You can explicitly lock a database, so it remains locked for multiple
2249 transactions. This is done by calling the C<lock()> method, and passing an
2250 optional lock mode argument (defaults to exclusive mode). This is particularly
2251 useful for things like counters, where the current value needs to be fetched,
2252 then incremented, then stored again.
2255 my $counter = $db->get("counter");
2257 $db->put("counter", $counter);
2266 You can pass C<lock()> an optional argument, which specifies which mode to use
2267 (exclusive or shared). Use one of these two constants: C<DBM::Deep-E<gt>LOCK_EX>
2268 or C<DBM::Deep-E<gt>LOCK_SH>. These are passed directly to C<flock()>, and are the
2269 same as the constants defined in Perl's C<Fcntl> module.
2271 $db->lock( DBM::Deep->LOCK_SH );
2275 If you want to implement your own file locking scheme, be sure to create your
2276 DBM::Deep objects setting the C<volatile> option to true. This hints to DBM::Deep
2277 that the DB file may change between transactions. See L<LOW-LEVEL ACCESS>
2280 =head1 IMPORTING/EXPORTING
2282 You can import existing complex structures by calling the C<import()> method,
2283 and export an entire database into an in-memory structure using the C<export()>
2284 method. Both are examined here.
2288 Say you have an existing hash with nested hashes/arrays inside it. Instead of
2289 walking the structure and adding keys/elements to the database as you go,
2290 simply pass a reference to the C<import()> method. This recursively adds
2291 everything to an existing DBM::Deep object for you. Here is an example:
2296 array1 => [ "elem0", "elem1", "elem2" ],
2298 subkey1 => "subvalue1",
2299 subkey2 => "subvalue2"
2303 my $db = DBM::Deep->new( "foo.db" );
2304 $db->import( $struct );
2306 print $db->{key1} . "\n"; # prints "value1"
2308 This recursively imports the entire C<$struct> object into C<$db>, including
2309 all nested hashes and arrays. If the DBM::Deep object contains exsiting data,
2310 keys are merged with the existing ones, replacing if they already exist.
2311 The C<import()> method can be called on any database level (not just the base
2312 level), and works with both hash and array DB types.
2316 B<Note:> Make sure your existing structure has no circular references in it.
2317 These will cause an infinite loop when importing.
2321 Calling the C<export()> method on an existing DBM::Deep object will return
2322 a reference to a new in-memory copy of the database. The export is done
2323 recursively, so all nested hashes/arrays are all exported to standard Perl
2324 objects. Here is an example:
2326 my $db = DBM::Deep->new( "foo.db" );
2328 $db->{key1} = "value1";
2329 $db->{key2} = "value2";
2331 $db->{hash1}->{subkey1} = "subvalue1";
2332 $db->{hash1}->{subkey2} = "subvalue2";
2334 my $struct = $db->export();
2336 print $struct->{key1} . "\n"; # prints "value1"
2338 This makes a complete copy of the database in memory, and returns a reference
2339 to it. The C<export()> method can be called on any database level (not just
2340 the base level), and works with both hash and array DB types. Be careful of
2341 large databases -- you can store a lot more data in a DBM::Deep object than an
2342 in-memory Perl structure.
2346 B<Note:> Make sure your database has no circular references in it.
2347 These will cause an infinite loop when exporting.
2351 DBM::Deep has a number of hooks where you can specify your own Perl function
2352 to perform filtering on incoming or outgoing data. This is a perfect
2353 way to extend the engine, and implement things like real-time compression or
2354 encryption. Filtering applies to the base DB level, and all child hashes /
2355 arrays. Filter hooks can be specified when your DBM::Deep object is first
2356 constructed, or by calling the C<set_filter()> method at any time. There are
2357 four available filter hooks, described below:
2361 =item * filter_store_key
2363 This filter is called whenever a hash key is stored. It
2364 is passed the incoming key, and expected to return a transformed key.
2366 =item * filter_store_value
2368 This filter is called whenever a hash key or array element is stored. It
2369 is passed the incoming value, and expected to return a transformed value.
2371 =item * filter_fetch_key
2373 This filter is called whenever a hash key is fetched (i.e. via
2374 C<first_key()> or C<next_key()>). It is passed the transformed key,
2375 and expected to return the plain key.
2377 =item * filter_fetch_value
2379 This filter is called whenever a hash key or array element is fetched.
2380 It is passed the transformed value, and expected to return the plain value.
2384 Here are the two ways to setup a filter hook:
2386 my $db = DBM::Deep->new(
2388 filter_store_value => \&my_filter_store,
2389 filter_fetch_value => \&my_filter_fetch
2394 $db->set_filter( "filter_store_value", \&my_filter_store );
2395 $db->set_filter( "filter_fetch_value", \&my_filter_fetch );
2397 Your filter function will be called only when dealing with SCALAR keys or
2398 values. When nested hashes and arrays are being stored/fetched, filtering
2399 is bypassed. Filters are called as static functions, passed a single SCALAR
2400 argument, and expected to return a single SCALAR value. If you want to
2401 remove a filter, set the function reference to C<undef>:
2403 $db->set_filter( "filter_store_value", undef );
2405 =head2 REAL-TIME ENCRYPTION EXAMPLE
2407 Here is a working example that uses the I<Crypt::Blowfish> module to
2408 do real-time encryption / decryption of keys & values with DBM::Deep Filters.
2409 Please visit L<http://search.cpan.org/search?module=Crypt::Blowfish> for more
2410 on I<Crypt::Blowfish>. You'll also need the I<Crypt::CBC> module.
2413 use Crypt::Blowfish;
2416 my $cipher = Crypt::CBC->new({
2417 'key' => 'my secret key',
2418 'cipher' => 'Blowfish',
2420 'regenerate_key' => 0,
2421 'padding' => 'space',
2425 my $db = DBM::Deep->new(
2426 file => "foo-encrypt.db",
2427 filter_store_key => \&my_encrypt,
2428 filter_store_value => \&my_encrypt,
2429 filter_fetch_key => \&my_decrypt,
2430 filter_fetch_value => \&my_decrypt,
2433 $db->{key1} = "value1";
2434 $db->{key2} = "value2";
2435 print "key1: " . $db->{key1} . "\n";
2436 print "key2: " . $db->{key2} . "\n";
2442 return $cipher->encrypt( $_[0] );
2445 return $cipher->decrypt( $_[0] );
2448 =head2 REAL-TIME COMPRESSION EXAMPLE
2450 Here is a working example that uses the I<Compress::Zlib> module to do real-time
2451 compression / decompression of keys & values with DBM::Deep Filters.
2452 Please visit L<http://search.cpan.org/search?module=Compress::Zlib> for
2453 more on I<Compress::Zlib>.
2458 my $db = DBM::Deep->new(
2459 file => "foo-compress.db",
2460 filter_store_key => \&my_compress,
2461 filter_store_value => \&my_compress,
2462 filter_fetch_key => \&my_decompress,
2463 filter_fetch_value => \&my_decompress,
2466 $db->{key1} = "value1";
2467 $db->{key2} = "value2";
2468 print "key1: " . $db->{key1} . "\n";
2469 print "key2: " . $db->{key2} . "\n";
2475 return Compress::Zlib::memGzip( $_[0] ) ;
2478 return Compress::Zlib::memGunzip( $_[0] ) ;
2481 B<Note:> Filtering of keys only applies to hashes. Array "keys" are
2482 actually numerical index numbers, and are not filtered.
2484 =head1 ERROR HANDLING
2486 Most DBM::Deep methods return a true value for success, and call die() on
2487 failure. You can wrap calls in an eval block to catch the die. Also, the
2488 actual error message is stored in an internal scalar, which can be fetched by
2489 calling the C<error()> method.
2491 my $db = DBM::Deep->new( "foo.db" ); # create hash
2492 eval { $db->push("foo"); }; # ILLEGAL -- push is array-only call
2494 print $db->error(); # prints error message
2496 You can then call C<clear_error()> to clear the current error state.
2500 If you set the C<debug> option to true when creating your DBM::Deep object,
2501 all errors are considered NON-FATAL, and dumped to STDERR. This is only
2502 for debugging purposes.
2504 =head1 LARGEFILE SUPPORT
2506 If you have a 64-bit system, and your Perl is compiled with both LARGEFILE
2507 and 64-bit support, you I<may> be able to create databases larger than 2 GB.
2508 DBM::Deep by default uses 32-bit file offset tags, but these can be changed
2509 by calling the static C<set_pack()> method before you do anything else.
2511 DBM::Deep::set_pack(8, 'Q');
2513 This tells DBM::Deep to pack all file offsets with 8-byte (64-bit) quad words
2514 instead of 32-bit longs. After setting these values your DB files have a
2515 theoretical maximum size of 16 XB (exabytes).
2519 B<Note:> Changing these values will B<NOT> work for existing database files.
2520 Only change this for new files, and make sure it stays set consistently
2521 throughout the file's life. If you do set these values, you can no longer
2522 access 32-bit DB files. You can, however, call C<set_pack(4, 'N')> to change
2523 back to 32-bit mode.
2527 B<Note:> I have not personally tested files > 2 GB -- all my systems have
2528 only a 32-bit Perl. However, I have received user reports that this does
2531 =head1 LOW-LEVEL ACCESS
2533 If you require low-level access to the underlying FileHandle that DBM::Deep uses,
2534 you can call the C<fh()> method, which returns the handle:
2538 This method can be called on the root level of the datbase, or any child
2539 hashes or arrays. All levels share a I<root> structure, which contains things
2540 like the FileHandle, a reference counter, and all your options you specified
2541 when you created the object. You can get access to this root structure by
2542 calling the C<root()> method.
2544 my $root = $db->root();
2546 This is useful for changing options after the object has already been created,
2547 such as enabling/disabling locking, volatile or debug modes. You can also
2548 store your own temporary user data in this structure (be wary of name
2549 collision), which is then accessible from any child hash or array.
2551 =head1 CUSTOM DIGEST ALGORITHM
2553 DBM::Deep by default uses the I<Message Digest 5> (MD5) algorithm for hashing
2554 keys. However you can override this, and use another algorithm (such as SHA-256)
2555 or even write your own. But please note that DBM::Deep currently expects zero
2556 collisions, so your algorithm has to be I<perfect>, so to speak.
2557 Collision detection may be introduced in a later version.
2561 You can specify a custom digest algorithm by calling the static C<set_digest()>
2562 function, passing a reference to a subroutine, and the length of the algorithm's
2563 hashes (in bytes). This is a global static function, which affects ALL DBM::Deep
2564 objects. Here is a working example that uses a 256-bit hash from the
2565 I<Digest::SHA256> module. Please see
2566 L<http://search.cpan.org/search?module=Digest::SHA256> for more.
2571 my $context = Digest::SHA256::new(256);
2573 DBM::Deep::set_digest( \&my_digest, 32 );
2575 my $db = DBM::Deep->new( "foo-sha.db" );
2577 $db->{key1} = "value1";
2578 $db->{key2} = "value2";
2579 print "key1: " . $db->{key1} . "\n";
2580 print "key2: " . $db->{key2} . "\n";
2586 return substr( $context->hash($_[0]), 0, 32 );
2589 B<Note:> Your returned digest strings must be B<EXACTLY> the number
2590 of bytes you specify in the C<set_digest()> function (in this case 32).
2592 =head1 CIRCULAR REFERENCES
2594 DBM::Deep has B<experimental> support for circular references. Meaning you
2595 can have a nested hash key or array element that points to a parent object.
2596 This relationship is stored in the DB file, and is preserved between sessions.
2599 my $db = DBM::Deep->new( "foo.db" );
2602 $db->{circle} = $db; # ref to self
2604 print $db->{foo} . "\n"; # prints "foo"
2605 print $db->{circle}->{foo} . "\n"; # prints "foo" again
2607 One catch is, passing the object to a function that recursively walks the
2608 object tree (such as I<Data::Dumper> or even the built-in C<optimize()> or
2609 C<export()> methods) will result in an infinite loop. The other catch is,
2610 if you fetch the I<key> of a circular reference (i.e. using the C<first_key()>
2611 or C<next_key()> methods), you will get the I<target object's key>, not the
2612 ref's key. This gets even more interesting with the above example, where
2613 the I<circle> key points to the base DB object, which technically doesn't
2614 have a key. So I made DBM::Deep return "[base]" as the key name in that
2617 =head1 CAVEATS / ISSUES / BUGS
2619 This section describes all the known issues with DBM::Deep. It you have found
2620 something that is not listed here, please send e-mail to L<jhuckaby@cpan.org>.
2622 =head2 UNUSED SPACE RECOVERY
2624 One major caveat with DBM::Deep is that space occupied by existing keys and
2625 values is not recovered when they are deleted. Meaning if you keep deleting
2626 and adding new keys, your file will continuously grow. I am working on this,
2627 but in the meantime you can call the built-in C<optimize()> method from time to
2628 time (perhaps in a crontab or something) to recover all your unused space.
2630 $db->optimize(); # returns true on success
2632 This rebuilds the ENTIRE database into a new file, then moves it on top of
2633 the original. The new file will have no unused space, thus it will take up as
2634 little disk space as possible. Please note that this operation can take
2635 a long time for large files, and you need enough disk space to temporarily hold
2636 2 copies of your DB file. The temporary file is created in the same directory
2637 as the original, named with a ".tmp" extension, and is deleted when the
2638 operation completes. Oh, and if locking is enabled, the DB is automatically
2639 locked for the entire duration of the copy.
2643 B<WARNING:> Only call optimize() on the top-level node of the database, and
2644 make sure there are no child references lying around. DBM::Deep keeps a reference
2645 counter, and if it is greater than 1, optimize() will abort and return undef.
2647 =head2 AUTOVIVIFICATION
2649 Unfortunately, autovivification doesn't work with tied hashes. This appears to
2650 be a bug in Perl's tie() system, as I<Jakob Schmidt> encountered the very same
2651 issue with his I<DWH_FIle> module (see L<http://search.cpan.org/search?module=DWH_File>),
2652 and it is also mentioned in the BUGS section for the I<MLDBM> module <see
2653 L<http://search.cpan.org/search?module=MLDBM>). Basically, on a new db file,
2656 $db->{foo}->{bar} = "hello";
2658 Since "foo" doesn't exist, you cannot add "bar" to it. You end up with "foo"
2659 being an empty hash. Try this instead, which works fine:
2661 $db->{foo} = { bar => "hello" };
2663 As of Perl 5.8.7, this bug still exists. I have walked very carefully through
2664 the execution path, and Perl indeed passes an empty hash to the STORE() method.
2665 Probably a bug in Perl.
2667 =head2 FILE CORRUPTION
2669 The current level of error handling in DBM::Deep is minimal. Files I<are> checked
2670 for a 32-bit signature when opened, but other corruption in files can cause
2671 segmentation faults. DBM::Deep may try to seek() past the end of a file, or get
2672 stuck in an infinite loop depending on the level of corruption. File write
2673 operations are not checked for failure (for speed), so if you happen to run
2674 out of disk space, DBM::Deep will probably fail in a bad way. These things will
2675 be addressed in a later version of DBM::Deep.
2679 Beware of using DB files over NFS. DBM::Deep uses flock(), which works well on local
2680 filesystems, but will NOT protect you from file corruption over NFS. I've heard
2681 about setting up your NFS server with a locking daemon, then using lockf() to
2682 lock your files, but your milage may vary there as well. From what I
2683 understand, there is no real way to do it. However, if you need access to the
2684 underlying FileHandle in DBM::Deep for using some other kind of locking scheme like
2685 lockf(), see the L<LOW-LEVEL ACCESS> section above.
2687 =head2 COPYING OBJECTS
2689 Beware of copying tied objects in Perl. Very strange things can happen.
2690 Instead, use DBM::Deep's C<clone()> method which safely copies the object and
2691 returns a new, blessed, tied hash or array to the same level in the DB.
2693 my $copy = $db->clone();
2697 Beware of using C<shift()>, C<unshift()> or C<splice()> with large arrays.
2698 These functions cause every element in the array to move, which can be murder
2699 on DBM::Deep, as every element has to be fetched from disk, then stored again in
2700 a different location. This may be addressed in a later version.
2704 This section discusses DBM::Deep's speed and memory usage.
2708 Obviously, DBM::Deep isn't going to be as fast as some C-based DBMs, such as
2709 the almighty I<BerkeleyDB>. But it makes up for it in features like true
2710 multi-level hash/array support, and cross-platform FTPable files. Even so,
2711 DBM::Deep is still pretty fast, and the speed stays fairly consistent, even
2712 with huge databases. Here is some test data:
2714 Adding 1,000,000 keys to new DB file...
2716 At 100 keys, avg. speed is 2,703 keys/sec
2717 At 200 keys, avg. speed is 2,642 keys/sec
2718 At 300 keys, avg. speed is 2,598 keys/sec
2719 At 400 keys, avg. speed is 2,578 keys/sec
2720 At 500 keys, avg. speed is 2,722 keys/sec
2721 At 600 keys, avg. speed is 2,628 keys/sec
2722 At 700 keys, avg. speed is 2,700 keys/sec
2723 At 800 keys, avg. speed is 2,607 keys/sec
2724 At 900 keys, avg. speed is 2,190 keys/sec
2725 At 1,000 keys, avg. speed is 2,570 keys/sec
2726 At 2,000 keys, avg. speed is 2,417 keys/sec
2727 At 3,000 keys, avg. speed is 1,982 keys/sec
2728 At 4,000 keys, avg. speed is 1,568 keys/sec
2729 At 5,000 keys, avg. speed is 1,533 keys/sec
2730 At 6,000 keys, avg. speed is 1,787 keys/sec
2731 At 7,000 keys, avg. speed is 1,977 keys/sec
2732 At 8,000 keys, avg. speed is 2,028 keys/sec
2733 At 9,000 keys, avg. speed is 2,077 keys/sec
2734 At 10,000 keys, avg. speed is 2,031 keys/sec
2735 At 20,000 keys, avg. speed is 1,970 keys/sec
2736 At 30,000 keys, avg. speed is 2,050 keys/sec
2737 At 40,000 keys, avg. speed is 2,073 keys/sec
2738 At 50,000 keys, avg. speed is 1,973 keys/sec
2739 At 60,000 keys, avg. speed is 1,914 keys/sec
2740 At 70,000 keys, avg. speed is 2,091 keys/sec
2741 At 80,000 keys, avg. speed is 2,103 keys/sec
2742 At 90,000 keys, avg. speed is 1,886 keys/sec
2743 At 100,000 keys, avg. speed is 1,970 keys/sec
2744 At 200,000 keys, avg. speed is 2,053 keys/sec
2745 At 300,000 keys, avg. speed is 1,697 keys/sec
2746 At 400,000 keys, avg. speed is 1,838 keys/sec
2747 At 500,000 keys, avg. speed is 1,941 keys/sec
2748 At 600,000 keys, avg. speed is 1,930 keys/sec
2749 At 700,000 keys, avg. speed is 1,735 keys/sec
2750 At 800,000 keys, avg. speed is 1,795 keys/sec
2751 At 900,000 keys, avg. speed is 1,221 keys/sec
2752 At 1,000,000 keys, avg. speed is 1,077 keys/sec
2754 This test was performed on a PowerMac G4 1gHz running Mac OS X 10.3.2 & Perl
2755 5.8.1, with an 80GB Ultra ATA/100 HD spinning at 7200RPM. The hash keys and
2756 values were between 6 - 12 chars in length. The DB file ended up at 210MB.
2757 Run time was 12 min 3 sec.
2761 One of the great things about DBM::Deep is that it uses very little memory.
2762 Even with huge databases (1,000,000+ keys) you will not see much increased
2763 memory on your process. DBM::Deep relies solely on the filesystem for storing
2764 and fetching data. Here is output from I</usr/bin/top> before even opening a
2767 PID USER PRI NI SIZE RSS SHARE STAT %CPU %MEM TIME COMMAND
2768 22831 root 11 0 2716 2716 1296 R 0.0 0.2 0:07 perl
2770 Basically the process is taking 2,716K of memory. And here is the same
2771 process after storing and fetching 1,000,000 keys:
2773 PID USER PRI NI SIZE RSS SHARE STAT %CPU %MEM TIME COMMAND
2774 22831 root 14 0 2772 2772 1328 R 0.0 0.2 13:32 perl
2776 Notice the memory usage increased by only 56K. Test was performed on a 700mHz
2777 x86 box running Linux RedHat 7.2 & Perl 5.6.1.
2779 =head1 DB FILE FORMAT
2781 In case you were interested in the underlying DB file format, it is documented
2782 here in this section. You don't need to know this to use the module, it's just
2783 included for reference.
2787 DBM::Deep files always start with a 32-bit signature to identify the file type.
2788 This is at offset 0. The signature is "DPDB" in network byte order. This is
2789 checked when the file is opened.
2793 The DBM::Deep file is in a I<tagged format>, meaning each section of the file
2794 has a standard header containing the type of data, the length of data, and then
2795 the data itself. The type is a single character (1 byte), the length is a
2796 32-bit unsigned long in network byte order, and the data is, well, the data.
2797 Here is how it unfolds:
2801 Immediately after the 32-bit file signature is the I<Master Index> record.
2802 This is a standard tag header followed by 1024 bytes (in 32-bit mode) or 2048
2803 bytes (in 64-bit mode) of data. The type is I<H> for hash or I<A> for array,
2804 depending on how the DBM::Deep object was constructed.
2808 The index works by looking at a I<MD5 Hash> of the hash key (or array index
2809 number). The first 8-bit char of the MD5 signature is the offset into the
2810 index, multipled by 4 in 32-bit mode, or 8 in 64-bit mode. The value of the
2811 index element is a file offset of the next tag for the key/element in question,
2812 which is usually a I<Bucket List> tag (see below).
2816 The next tag I<could> be another index, depending on how many keys/elements
2817 exist. See L<RE-INDEXING> below for details.
2821 A I<Bucket List> is a collection of 16 MD5 hashes for keys/elements, plus
2822 file offsets to where the actual data is stored. It starts with a standard
2823 tag header, with type I<B>, and a data size of 320 bytes in 32-bit mode, or
2824 384 bytes in 64-bit mode. Each MD5 hash is stored in full (16 bytes), plus
2825 the 32-bit or 64-bit file offset for the I<Bucket> containing the actual data.
2826 When the list fills up, a I<Re-Index> operation is performed (See
2827 L<RE-INDEXING> below).
2831 A I<Bucket> is a tag containing a key/value pair (in hash mode), or a
2832 index/value pair (in array mode). It starts with a standard tag header with
2833 type I<D> for scalar data (string, binary, etc.), or it could be a nested
2834 hash (type I<H>) or array (type I<A>). The value comes just after the tag
2835 header. The size reported in the tag header is only for the value, but then,
2836 just after the value is another size (32-bit unsigned long) and then the plain
2837 key itself. Since the value is likely to be fetched more often than the plain
2838 key, I figured it would be I<slightly> faster to store the value first.
2842 If the type is I<H> (hash) or I<A> (array), the value is another I<Master Index>
2843 record for the nested structure, where the process begins all over again.
2847 After a I<Bucket List> grows to 16 records, its allocated space in the file is
2848 exhausted. Then, when another key/element comes in, the list is converted to a
2849 new index record. However, this index will look at the next char in the MD5
2850 hash, and arrange new Bucket List pointers accordingly. This process is called
2851 I<Re-Indexing>. Basically, a new index tag is created at the file EOF, and all
2852 17 (16 + new one) keys/elements are removed from the old Bucket List and
2853 inserted into the new index. Several new Bucket Lists are created in the
2854 process, as a new MD5 char from the key is being examined (it is unlikely that
2855 the keys will all share the same next char of their MD5s).
2859 Because of the way the I<MD5> algorithm works, it is impossible to tell exactly
2860 when the Bucket Lists will turn into indexes, but the first round tends to
2861 happen right around 4,000 keys. You will see a I<slight> decrease in
2862 performance here, but it picks back up pretty quick (see L<SPEED> above). Then
2863 it takes B<a lot> more keys to exhaust the next level of Bucket Lists. It's
2864 right around 900,000 keys. This process can continue nearly indefinitely --
2865 right up until the point the I<MD5> signatures start colliding with each other,
2866 and this is B<EXTREMELY> rare -- like winning the lottery 5 times in a row AND
2867 getting struck by lightning while you are walking to cash in your tickets.
2868 Theoretically, since I<MD5> hashes are 128-bit values, you I<could> have up to
2869 340,282,366,921,000,000,000,000,000,000,000,000,000 keys/elements (I believe
2870 this is 340 unodecillion, but don't quote me).
2874 When a new key/element is stored, the key (or index number) is first ran through
2875 I<Digest::MD5> to get a 128-bit signature (example, in hex:
2876 b05783b0773d894396d475ced9d2f4f6). Then, the I<Master Index> record is checked
2877 for the first char of the signature (in this case I<b>). If it does not exist,
2878 a new I<Bucket List> is created for our key (and the next 15 future keys that
2879 happen to also have I<b> as their first MD5 char). The entire MD5 is written
2880 to the I<Bucket List> along with the offset of the new I<Bucket> record (EOF at
2881 this point, unless we are replacing an existing I<Bucket>), where the actual
2882 data will be stored.
2886 Fetching an existing key/element involves getting a I<Digest::MD5> of the key
2887 (or index number), then walking along the indexes. If there are enough
2888 keys/elements in this DB level, there might be nested indexes, each linked to
2889 a particular char of the MD5. Finally, a I<Bucket List> is pointed to, which
2890 contains up to 16 full MD5 hashes. Each is checked for equality to the key in
2891 question. If we found a match, the I<Bucket> tag is loaded, where the value and
2892 plain key are stored.
2896 Fetching the plain key occurs when calling the I<first_key()> and I<next_key()>
2897 methods. In this process the indexes are walked systematically, and each key
2898 fetched in increasing MD5 order (which is why it appears random). Once the
2899 I<Bucket> is found, the value is skipped the plain key returned instead.
2900 B<Note:> Do not count on keys being fetched as if the MD5 hashes were
2901 alphabetically sorted. This only happens on an index-level -- as soon as the
2902 I<Bucket Lists> are hit, the keys will come out in the order they went in --
2903 so it's pretty much undefined how the keys will come out -- just like Perl's
2906 =head1 CODE COVERAGE
2908 I use B<Devel::Cover> to test the code coverage of my tests, below is the B<Devel::Cover> report on this
2909 module's test suite.
2911 ---------------------------- ------ ------ ------ ------ ------ ------ ------
2912 File stmt bran cond sub pod time total
2913 ---------------------------- ------ ------ ------ ------ ------ ------ ------
2914 blib/lib/DBM/Deep.pm 94.9 84.5 77.8 100.0 11.1 100.0 89.7
2915 Total 94.9 84.5 77.8 100.0 11.1 100.0 89.7
2916 ---------------------------- ------ ------ ------ ------ ------ ------ ------
2920 Joseph Huckaby, L<jhuckaby@cpan.org>
2922 Special thanks to Adam Sah and Rich Gaushell! You know why :-)
2926 perltie(1), Tie::Hash(3), Digest::MD5(3), Fcntl(3), flock(2), lockf(3), nfs(5),
2927 Digest::SHA256(3), Crypt::Blowfish(3), Compress::Zlib(3)
2931 Copyright (c) 2002-2005 Joseph Huckaby. All Rights Reserved.
2932 This is free software, you may use it and distribute it under the
2933 same terms as Perl itself.