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} . ": $!");
257 #XXX Can we remove this by using the right sysopen() flags?
258 binmode $fh; # for win32
260 if ($self->root->{autoflush}) {
261 # $self->fh->autoflush();
262 my $old = select( $fh );
269 my $bytes_read = read( $fh, $signature, length(SIG_FILE));
272 # File is empty -- write signature and master index
276 $fh->print(SIG_FILE);
277 $self->root->{end} = length(SIG_FILE);
278 $self->_create_tag($self->base_offset, $self->type, chr(0) x $INDEX_SIZE);
280 my $plain_key = "[base]";
281 $fh->print( pack($DATA_LENGTH_PACK, length($plain_key)) . $plain_key );
282 $self->root->{end} += $DATA_LENGTH_SIZE + length($plain_key);
285 my $old_fh = select $fh;
294 # Check signature was valid
296 unless ($signature eq SIG_FILE) {
298 return $self->_throw_error("Signature not found -- file is not a Deep DB");
301 $self->root->{end} = (stat($fh))[7];
304 # Get our type from master index signature
306 my $tag = $self->_load_tag($self->base_offset);
307 #XXX We probably also want to store the hash algorithm name and not assume anything
309 return $self->_throw_error("Corrupted file, no master index record");
311 if ($self->{type} ne $tag->{signature}) {
312 return $self->_throw_error("File type mismatch");
320 # Close database FileHandle
322 my $self = _get_self($_[0]);
323 undef $self->root->{fh};
328 # Given offset, signature and content, create tag and write to disk
330 my ($self, $offset, $sig, $content) = @_;
331 my $size = length($content);
335 seek($fh, $offset, 0);
336 $fh->print( $sig . pack($DATA_LENGTH_PACK, $size) . $content );
338 if ($offset == $self->root->{end}) {
339 $self->root->{end} += SIG_SIZE + $DATA_LENGTH_SIZE + $size;
345 offset => $offset + SIG_SIZE + $DATA_LENGTH_SIZE,
352 # Given offset, load single tag and return signature, size and data
359 seek($fh, $offset, 0);
360 if (eof $fh) { return undef; }
363 read( $fh, $sig, SIG_SIZE);
366 read( $fh, $size, $DATA_LENGTH_SIZE);
367 $size = unpack($DATA_LENGTH_PACK, $size);
370 read( $fh, $buffer, $size);
375 offset => $offset + SIG_SIZE + $DATA_LENGTH_SIZE,
382 # Given index tag, lookup single entry in index and return .
385 my ($tag, $index) = @_;
387 my $location = unpack($LONG_PACK, substr($tag->{content}, $index * $LONG_SIZE, $LONG_SIZE) );
388 if (!$location) { return; }
390 return $self->_load_tag( $location );
395 # Adds one key/value pair to bucket list, given offset, MD5 digest of key,
396 # plain (undigested) key and value.
399 my ($tag, $md5, $plain_key, $value) = @_;
400 my $keys = $tag->{content};
404 my $is_dbm_deep = eval { $value->isa( 'DBM::Deep' ) };
405 my $internal_ref = $is_dbm_deep && ($value->root eq $self->root);
410 # Iterate through buckets, seeing if this is a new entry or a replace.
412 for (my $i=0; $i<$MAX_BUCKETS; $i++) {
413 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
414 my $subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
417 # Found empty bucket (end of list). Populate and exit loop.
421 $location = $internal_ref
422 ? $value->base_offset
423 : $self->root->{end};
425 seek($fh, $tag->{offset} + ($i * $BUCKET_SIZE), 0);
426 $fh->print( $md5 . pack($LONG_PACK, $location) );
429 elsif ($md5 eq $key) {
431 # Found existing bucket with same key. Replace with new value.
436 $location = $value->base_offset;
437 seek($fh, $tag->{offset} + ($i * $BUCKET_SIZE), 0);
438 $fh->print( $md5 . pack($LONG_PACK, $location) );
441 seek($fh, $subloc + SIG_SIZE, 0);
443 read( $fh, $size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
446 # If value is a hash, array, or raw value with equal or less size, we can
447 # reuse the same content area of the database. Otherwise, we have to create
448 # a new content area at the EOF.
451 my $r = Scalar::Util::reftype( $value ) || '';
452 if ( $r eq 'HASH' || $r eq 'ARRAY' ) { $actual_length = $INDEX_SIZE; }
453 else { $actual_length = length($value); }
455 if ($actual_length <= $size) {
459 $location = $self->root->{end};
460 seek($fh, $tag->{offset} + ($i * $BUCKET_SIZE) + $HASH_SIZE, 0);
461 $fh->print( pack($LONG_PACK, $location) );
469 # If this is an internal reference, return now.
470 # No need to write value or plain key
477 # If bucket didn't fit into list, split into a new index level
480 seek($fh, $tag->{ref_loc}, 0);
481 $fh->print( pack($LONG_PACK, $self->root->{end}) );
483 my $index_tag = $self->_create_tag($self->root->{end}, SIG_INDEX, chr(0) x $INDEX_SIZE);
486 $keys .= $md5 . pack($LONG_PACK, 0);
488 for (my $i=0; $i<=$MAX_BUCKETS; $i++) {
489 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
491 my $old_subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
492 my $num = ord(substr($key, $tag->{ch} + 1, 1));
494 if ($offsets[$num]) {
495 my $offset = $offsets[$num] + SIG_SIZE + $DATA_LENGTH_SIZE;
496 seek($fh, $offset, 0);
498 read( $fh, $subkeys, $BUCKET_LIST_SIZE);
500 for (my $k=0; $k<$MAX_BUCKETS; $k++) {
501 my $subloc = unpack($LONG_PACK, substr($subkeys, ($k * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
503 seek($fh, $offset + ($k * $BUCKET_SIZE), 0);
504 $fh->print( $key . pack($LONG_PACK, $old_subloc || $self->root->{end}) );
510 $offsets[$num] = $self->root->{end};
511 seek($fh, $index_tag->{offset} + ($num * $LONG_SIZE), 0);
512 $fh->print( pack($LONG_PACK, $self->root->{end}) );
514 my $blist_tag = $self->_create_tag($self->root->{end}, SIG_BLIST, chr(0) x $BUCKET_LIST_SIZE);
516 seek($fh, $blist_tag->{offset}, 0);
517 $fh->print( $key . pack($LONG_PACK, $old_subloc || $self->root->{end}) );
522 $location ||= $self->root->{end};
523 } # re-index bucket list
526 # Seek to content area and store signature, value and plaintext key
530 seek($fh, $location, 0);
533 # Write signature based on content type, set content length and write actual value.
535 my $r = Scalar::Util::reftype($value) || '';
537 $fh->print( TYPE_HASH );
538 $fh->print( pack($DATA_LENGTH_PACK, $INDEX_SIZE) . chr(0) x $INDEX_SIZE );
539 $content_length = $INDEX_SIZE;
541 elsif ($r eq 'ARRAY') {
542 $fh->print( TYPE_ARRAY );
543 $fh->print( pack($DATA_LENGTH_PACK, $INDEX_SIZE) . chr(0) x $INDEX_SIZE );
544 $content_length = $INDEX_SIZE;
546 elsif (!defined($value)) {
547 $fh->print( SIG_NULL );
548 $fh->print( pack($DATA_LENGTH_PACK, 0) );
552 $fh->print( SIG_DATA );
553 $fh->print( pack($DATA_LENGTH_PACK, length($value)) . $value );
554 $content_length = length($value);
558 # Plain key is stored AFTER value, as keys are typically fetched less often.
560 $fh->print( pack($DATA_LENGTH_PACK, length($plain_key)) . $plain_key );
563 # If value is blessed, preserve class name
565 if ( $self->root->{autobless} ) {
566 my $value_class = Scalar::Util::blessed($value);
567 if ( defined $value_class && $value_class ne 'DBM::Deep' ) {
569 # Blessed ref -- will restore later
571 $fh->print( chr(1) );
572 $fh->print( pack($DATA_LENGTH_PACK, length($value_class)) . $value_class );
573 $content_length += 1;
574 $content_length += $DATA_LENGTH_SIZE + length($value_class);
577 $fh->print( chr(0) );
578 $content_length += 1;
583 # If this is a new content area, advance EOF counter
585 if ($location == $self->root->{end}) {
586 $self->root->{end} += SIG_SIZE;
587 $self->root->{end} += $DATA_LENGTH_SIZE + $content_length;
588 $self->root->{end} += $DATA_LENGTH_SIZE + length($plain_key);
592 # If content is a hash or array, create new child DeepDB object and
593 # pass each key or element to it.
596 my $branch = DBM::Deep->new(
598 base_offset => $location,
601 foreach my $key (keys %{$value}) {
602 $branch->{$key} = $value->{$key};
605 elsif ($r eq 'ARRAY') {
606 my $branch = DBM::Deep->new(
608 base_offset => $location,
612 foreach my $element (@{$value}) {
613 $branch->[$index] = $element;
621 return $self->_throw_error("Fatal error: indexing failed -- possibly due to corruption in file");
624 sub _get_bucket_value {
626 # Fetch single value given tag and MD5 digested key.
629 my ($tag, $md5) = @_;
630 my $keys = $tag->{content};
635 # Iterate through buckets, looking for a key match
638 for (my $i=0; $i<$MAX_BUCKETS; $i++) {
639 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
640 my $subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
644 # Hit end of list, no match
649 if ( $md5 ne $key ) {
654 # Found match -- seek to offset and read signature
657 seek($fh, $subloc, 0);
658 read( $fh, $signature, SIG_SIZE);
661 # If value is a hash or array, return new DeepDB object with correct offset
663 if (($signature eq TYPE_HASH) || ($signature eq TYPE_ARRAY)) {
664 my $obj = DBM::Deep->new(
666 base_offset => $subloc,
670 if ($self->root->{autobless}) {
672 # Skip over value and plain key to see if object needs
675 seek($fh, $DATA_LENGTH_SIZE + $INDEX_SIZE, 1);
678 read( $fh, $size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
679 if ($size) { seek($fh, $size, 1); }
682 read( $fh, $bless_bit, 1);
683 if (ord($bless_bit)) {
685 # Yes, object needs to be re-blessed
688 read( $fh, $size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
689 if ($size) { read( $fh, $class_name, $size); }
690 if ($class_name) { $obj = bless( $obj, $class_name ); }
698 # Otherwise return actual value
700 elsif ($signature eq SIG_DATA) {
703 read( $fh, $size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
704 if ($size) { read( $fh, $value, $size); }
709 # Key exists, but content is null
719 # Delete single key/value pair given tag and MD5 digested key.
722 my ($tag, $md5) = @_;
723 my $keys = $tag->{content};
728 # Iterate through buckets, looking for a key match
731 for (my $i=0; $i<$MAX_BUCKETS; $i++) {
732 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
733 my $subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
737 # Hit end of list, no match
742 if ( $md5 ne $key ) {
747 # Matched key -- delete bucket and return
749 seek($fh, $tag->{offset} + ($i * $BUCKET_SIZE), 0);
750 $fh->print( substr($keys, ($i+1) * $BUCKET_SIZE ) );
751 $fh->print( chr(0) x $BUCKET_SIZE );
761 # Check existence of single key given tag and MD5 digested key.
764 my ($tag, $md5) = @_;
765 my $keys = $tag->{content};
768 # Iterate through buckets, looking for a key match
771 for (my $i=0; $i<$MAX_BUCKETS; $i++) {
772 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
773 my $subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
777 # Hit end of list, no match
782 if ( $md5 ne $key ) {
787 # Matched key -- return true
795 sub _find_bucket_list {
797 # Locate offset for bucket list, given digested key
803 # Locate offset for bucket list using digest index system
806 my $tag = $self->_load_tag($self->base_offset);
807 if (!$tag) { return; }
809 while ($tag->{signature} ne SIG_BLIST) {
810 $tag = $self->_index_lookup($tag, ord(substr($md5, $ch, 1)));
811 if (!$tag) { return; }
818 sub _traverse_index {
820 # Scan index and recursively step into deeper levels, looking for next key.
822 my ($self, $offset, $ch, $force_return_next) = @_;
823 $force_return_next = undef unless $force_return_next;
825 my $tag = $self->_load_tag( $offset );
829 if ($tag->{signature} ne SIG_BLIST) {
830 my $content = $tag->{content};
832 if ($self->{return_next}) { $start = 0; }
833 else { $start = ord(substr($self->{prev_md5}, $ch, 1)); }
835 for (my $index = $start; $index < 256; $index++) {
836 my $subloc = unpack($LONG_PACK, substr($content, $index * $LONG_SIZE, $LONG_SIZE) );
838 my $result = $self->_traverse_index( $subloc, $ch + 1, $force_return_next );
839 if (defined($result)) { return $result; }
843 $self->{return_next} = 1;
846 elsif ($tag->{signature} eq SIG_BLIST) {
847 my $keys = $tag->{content};
848 if ($force_return_next) { $self->{return_next} = 1; }
851 # Iterate through buckets, looking for a key match
853 for (my $i=0; $i<$MAX_BUCKETS; $i++) {
854 my $key = substr($keys, $i * $BUCKET_SIZE, $HASH_SIZE);
855 my $subloc = unpack($LONG_PACK, substr($keys, ($i * $BUCKET_SIZE) + $HASH_SIZE, $LONG_SIZE));
859 # End of bucket list -- return to outer loop
861 $self->{return_next} = 1;
864 elsif ($key eq $self->{prev_md5}) {
866 # Located previous key -- return next one found
868 $self->{return_next} = 1;
871 elsif ($self->{return_next}) {
873 # Seek to bucket location and skip over signature
875 seek($fh, $subloc + SIG_SIZE, 0);
878 # Skip over value to get to plain key
881 read( $fh, $size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
882 if ($size) { seek($fh, $size, 1); }
885 # Read in plain key and return as scalar
888 read( $fh, $size, $DATA_LENGTH_SIZE); $size = unpack($DATA_LENGTH_PACK, $size);
889 if ($size) { read( $fh, $plain_key, $size); }
895 $self->{return_next} = 1;
896 } # tag is a bucket list
903 # Locate next key, given digested previous one
905 my $self = _get_self($_[0]);
907 $self->{prev_md5} = $_[1] ? $_[1] : undef;
908 $self->{return_next} = 0;
911 # If the previous key was not specifed, start at the top and
912 # return the first one found.
914 if (!$self->{prev_md5}) {
915 $self->{prev_md5} = chr(0) x $HASH_SIZE;
916 $self->{return_next} = 1;
919 return $self->_traverse_index( $self->base_offset, 0 );
924 # If db locking is set, flock() the db file. If called multiple
925 # times before unlock(), then the same number of unlocks() must
926 # be called before the lock is released.
928 my $self = _get_self($_[0]);
930 $type = LOCK_EX unless defined $type;
932 if ($self->root->{locking}) {
933 if (!$self->root->{locked}) { flock($self->fh, $type); }
934 $self->root->{locked}++;
940 # If db locking is set, unlock the db file. See note in lock()
941 # regarding calling lock() multiple times.
943 my $self = _get_self($_[0]);
945 if ($self->root->{locking} && $self->root->{locked} > 0) {
946 $self->root->{locked}--;
947 if (!$self->root->{locked}) { flock($self->fh, LOCK_UN); }
951 #XXX These uses of ref() need verified
954 # Copy single level of keys or elements to new DB handle.
955 # Recurse for nested structures
957 my $self = _get_self($_[0]);
960 if ($self->type eq TYPE_HASH) {
961 my $key = $self->first_key();
963 my $value = $self->get($key);
964 #XXX This doesn't work with autobless
965 if (!ref($value)) { $db_temp->{$key} = $value; }
967 my $type = $value->type;
968 if ($type eq TYPE_HASH) { $db_temp->{$key} = {}; }
969 else { $db_temp->{$key} = []; }
970 $value->_copy_node( $db_temp->{$key} );
972 $key = $self->next_key($key);
976 my $length = $self->length();
977 for (my $index = 0; $index < $length; $index++) {
978 my $value = $self->get($index);
979 if (!ref($value)) { $db_temp->[$index] = $value; }
980 #XXX NO tests for this code
982 my $type = $value->type;
983 if ($type eq TYPE_HASH) { $db_temp->[$index] = {}; }
984 else { $db_temp->[$index] = []; }
985 $value->_copy_node( $db_temp->[$index] );
993 # Recursively export into standard Perl hashes and arrays.
995 my $self = _get_self($_[0]);
998 if ($self->type eq TYPE_HASH) { $temp = {}; }
999 elsif ($self->type eq TYPE_ARRAY) { $temp = []; }
1002 $self->_copy_node( $temp );
1010 # Recursively import Perl hash/array structure
1012 #XXX This use of ref() seems to be ok
1013 if (!ref($_[0])) { return; } # Perl calls import() on use -- ignore
1015 my $self = _get_self($_[0]);
1018 #XXX This use of ref() seems to be ok
1019 if (!ref($struct)) {
1021 # struct is not a reference, so just import based on our type
1025 if ($self->type eq TYPE_HASH) { $struct = {@_}; }
1026 elsif ($self->type eq TYPE_ARRAY) { $struct = [@_]; }
1029 my $r = Scalar::Util::reftype($struct) || '';
1030 if ($r eq "HASH" && $self->type eq TYPE_HASH) {
1031 foreach my $key (keys %$struct) { $self->put($key, $struct->{$key}); }
1033 elsif ($r eq "ARRAY" && $self->type eq TYPE_ARRAY) {
1034 $self->push( @$struct );
1037 return $self->_throw_error("Cannot import: type mismatch");
1045 # Rebuild entire database into new file, then move
1046 # it back on top of original.
1048 my $self = _get_self($_[0]);
1049 if ($self->root->{links} > 1) {
1050 return $self->_throw_error("Cannot optimize: reference count is greater than 1");
1053 my $db_temp = DBM::Deep->new(
1054 file => $self->root->{file} . '.tmp',
1058 return $self->_throw_error("Cannot optimize: failed to open temp file: $!");
1062 $self->_copy_node( $db_temp );
1066 # Attempt to copy user, group and permissions over to new file
1068 my @stats = stat($self->fh);
1069 my $perms = $stats[2] & 07777;
1070 my $uid = $stats[4];
1071 my $gid = $stats[5];
1072 chown( $uid, $gid, $self->root->{file} . '.tmp' );
1073 chmod( $perms, $self->root->{file} . '.tmp' );
1075 # q.v. perlport for more information on this variable
1076 if ( $^O eq 'MSWin32' ) {
1078 # Potential race condition when optmizing on Win32 with locking.
1079 # The Windows filesystem requires that the filehandle be closed
1080 # before it is overwritten with rename(). This could be redone
1087 if (!rename $self->root->{file} . '.tmp', $self->root->{file}) {
1088 unlink $self->root->{file} . '.tmp';
1090 return $self->_throw_error("Optimize failed: Cannot copy temp file over original: $!");
1102 # Make copy of object and return
1104 my $self = _get_self($_[0]);
1106 return DBM::Deep->new(
1107 type => $self->type,
1108 base_offset => $self->base_offset,
1114 my %is_legal_filter = map {
1117 store_key store_value
1118 fetch_key fetch_value
1123 # Setup filter function for storing or fetching the key or value
1125 my $self = _get_self($_[0]);
1126 my $type = lc $_[1];
1127 my $func = $_[2] ? $_[2] : undef;
1129 if ( $is_legal_filter{$type} ) {
1130 $self->root->{"filter_$type"} = $func;
1144 # Get access to the root structure
1146 my $self = _get_self($_[0]);
1147 return $self->{root};
1152 # Get access to the raw FileHandle
1154 #XXX It will be useful, though, when we split out HASH and ARRAY
1155 my $self = _get_self($_[0]);
1156 return $self->root->{fh};
1161 # Get type of current node (TYPE_HASH or TYPE_ARRAY)
1163 my $self = _get_self($_[0]);
1164 return $self->{type};
1169 # Get base_offset of current node (TYPE_HASH or TYPE_ARRAY)
1171 my $self = _get_self($_[0]);
1172 return $self->{base_offset};
1177 # Get last error string, or undef if no error
1180 ? ( _get_self($_[0])->{root}->{error} or undef )
1190 # Store error string in self
1192 my $self = _get_self($_[0]);
1193 my $error_text = $_[1];
1195 $self->root->{error} = $error_text;
1197 unless ($self->root->{debug}) {
1198 die "DBM::Deep: $error_text\n";
1201 warn "DBM::Deep: $error_text\n";
1209 my $self = _get_self($_[0]);
1211 undef $self->root->{error};
1216 # Precalculate index, bucket and bucket list sizes
1219 #XXX I don't like this ...
1220 set_pack() unless defined $LONG_SIZE;
1222 $INDEX_SIZE = 256 * $LONG_SIZE;
1223 $BUCKET_SIZE = $HASH_SIZE + $LONG_SIZE;
1224 $BUCKET_LIST_SIZE = $MAX_BUCKETS * $BUCKET_SIZE;
1229 # Set pack/unpack modes (see file header for more)
1231 my ($long_s, $long_p, $data_s, $data_p) = @_;
1233 $LONG_SIZE = $long_s ? $long_s : 4;
1234 $LONG_PACK = $long_p ? $long_p : 'N';
1236 $DATA_LENGTH_SIZE = $data_s ? $data_s : 4;
1237 $DATA_LENGTH_PACK = $data_p ? $data_p : 'N';
1244 # Set key digest function (default is MD5)
1246 my ($digest_func, $hash_size) = @_;
1248 $DIGEST_FUNC = $digest_func ? $digest_func : \&Digest::MD5::md5;
1249 $HASH_SIZE = $hash_size ? $hash_size : 16;
1255 # tie() methods (hashes and arrays)
1260 # Store single hash key/value or array element in database.
1262 my $self = _get_self($_[0]);
1263 my $key = ($self->root->{filter_store_key} && $self->type eq TYPE_HASH) ? $self->root->{filter_store_key}->($_[1]) : $_[1];
1264 #XXX What is ref() checking here?
1265 #YYY User may be storing a hash, in which case we do not want it run
1266 #YYY through the filtering system
1267 my $value = ($self->root->{filter_store_value} && !ref($_[2])) ? $self->root->{filter_store_value}->($_[2]) : $_[2];
1269 my $unpacked_key = $key;
1270 if (($self->type eq TYPE_ARRAY) && ($key =~ /^\d+$/)) { $key = pack($LONG_PACK, $key); }
1271 my $md5 = $DIGEST_FUNC->($key);
1274 # Make sure file is open
1276 if (!defined($self->fh) && !$self->_open()) {
1283 # Request exclusive lock for writing
1285 $self->lock( LOCK_EX );
1288 # If locking is enabled, set 'end' parameter again, in case another
1289 # DB instance appended to our file while we were unlocked.
1291 if ($self->root->{locking} || $self->root->{volatile}) {
1292 $self->root->{end} = (stat($fh))[7];
1296 # Locate offset for bucket list using digest index system
1298 my $tag = $self->_load_tag($self->base_offset);
1300 $tag = $self->_create_tag($self->base_offset, SIG_INDEX, chr(0) x $INDEX_SIZE);
1304 while ($tag->{signature} ne SIG_BLIST) {
1305 my $num = ord(substr($md5, $ch, 1));
1306 my $new_tag = $self->_index_lookup($tag, $num);
1308 my $ref_loc = $tag->{offset} + ($num * $LONG_SIZE);
1309 seek($fh, $ref_loc, 0);
1310 $fh->print( pack($LONG_PACK, $self->root->{end}) );
1312 $tag = $self->_create_tag($self->root->{end}, SIG_BLIST, chr(0) x $BUCKET_LIST_SIZE);
1313 $tag->{ref_loc} = $ref_loc;
1318 my $ref_loc = $tag->{offset} + ($num * $LONG_SIZE);
1320 $tag->{ref_loc} = $ref_loc;
1327 # Add key/value to bucket list
1329 my $result = $self->_add_bucket( $tag, $md5, $key, $value );
1332 # If this object is an array, and bucket was not a replace, and key is numerical,
1333 # and index is equal or greater than current length, advance length variable.
1335 if (($result == 2) && ($self->type eq TYPE_ARRAY) && ($unpacked_key =~ /^\d+$/) && ($unpacked_key >= $self->FETCHSIZE())) {
1336 $self->STORESIZE( $unpacked_key + 1 );
1346 # Fetch single value or element given plain key or array index
1348 my $self = _get_self($_[0]);
1351 if ( $self->type eq TYPE_HASH ) {
1352 if ( my $filter = $self->root->{filter_store_key} ) {
1353 $key = $filter->( $key );
1356 elsif ( $self->type eq TYPE_ARRAY ) {
1357 if ( $key =~ /^\d+$/ ) {
1358 $key = pack($LONG_PACK, $key);
1362 my $md5 = $DIGEST_FUNC->($key);
1365 # Make sure file is open
1367 if (!defined($self->fh)) { $self->_open(); }
1370 # Request shared lock for reading
1372 $self->lock( LOCK_SH );
1374 my $tag = $self->_find_bucket_list( $md5 );
1381 # Get value from bucket list
1383 my $result = $self->_get_bucket_value( $tag, $md5 );
1387 #XXX What is ref() checking here?
1388 return ($result && !ref($result) && $self->root->{filter_fetch_value}) ? $self->root->{filter_fetch_value}->($result) : $result;
1393 # Delete single key/value pair or element given plain key or array index
1395 my $self = _get_self($_[0]);
1396 my $key = ($self->root->{filter_store_key} && $self->type eq TYPE_HASH) ? $self->root->{filter_store_key}->($_[1]) : $_[1];
1398 my $unpacked_key = $key;
1399 if (($self->type eq TYPE_ARRAY) && ($key =~ /^\d+$/)) { $key = pack($LONG_PACK, $key); }
1400 my $md5 = $DIGEST_FUNC->($key);
1403 # Make sure file is open
1405 if (!defined($self->fh)) { $self->_open(); }
1408 # Request exclusive lock for writing
1410 $self->lock( LOCK_EX );
1412 my $tag = $self->_find_bucket_list( $md5 );
1421 my $result = $self->_delete_bucket( $tag, $md5 );
1424 # If this object is an array and the key deleted was on the end of the stack,
1425 # decrement the length variable.
1427 if ($result && ($self->type eq TYPE_ARRAY) && ($unpacked_key == $self->FETCHSIZE() - 1)) {
1428 $self->STORESIZE( $unpacked_key );
1438 # Check if a single key or element exists given plain key or array index
1440 my $self = _get_self($_[0]);
1441 my $key = ($self->root->{filter_store_key} && $self->type eq TYPE_HASH) ? $self->root->{filter_store_key}->($_[1]) : $_[1];
1443 if (($self->type eq TYPE_ARRAY) && ($key =~ /^\d+$/)) { $key = pack($LONG_PACK, $key); }
1444 my $md5 = $DIGEST_FUNC->($key);
1447 # Make sure file is open
1449 if (!defined($self->fh)) { $self->_open(); }
1452 # Request shared lock for reading
1454 $self->lock( LOCK_SH );
1456 my $tag = $self->_find_bucket_list( $md5 );
1459 # For some reason, the built-in exists() function returns '' for false
1467 # Check if bucket exists and return 1 or ''
1469 my $result = $self->_bucket_exists( $tag, $md5 ) || '';
1478 # Clear all keys from hash, or all elements from array.
1480 my $self = _get_self($_[0]);
1483 # Make sure file is open
1485 if (!defined($self->fh)) { $self->_open(); }
1488 # Request exclusive lock for writing
1490 $self->lock( LOCK_EX );
1494 seek($fh, $self->base_offset, 0);
1500 $self->_create_tag($self->base_offset, $self->type, chr(0) x $INDEX_SIZE);
1509 # Locate and return first key (in no particular order)
1511 my $self = _get_self($_[0]);
1512 if ($self->type ne TYPE_HASH) {
1513 return $self->_throw_error("FIRSTKEY method only supported for hashes");
1517 # Make sure file is open
1519 if (!defined($self->fh)) { $self->_open(); }
1522 # Request shared lock for reading
1524 $self->lock( LOCK_SH );
1526 my $result = $self->_get_next_key();
1530 return ($result && $self->root->{filter_fetch_key}) ? $self->root->{filter_fetch_key}->($result) : $result;
1535 # Return next key (in no particular order), given previous one
1537 my $self = _get_self($_[0]);
1538 if ($self->type ne TYPE_HASH) {
1539 return $self->_throw_error("NEXTKEY method only supported for hashes");
1541 my $prev_key = ($self->root->{filter_store_key} && $self->type eq TYPE_HASH) ? $self->root->{filter_store_key}->($_[1]) : $_[1];
1542 my $prev_md5 = $DIGEST_FUNC->($prev_key);
1545 # Make sure file is open
1547 if (!defined($self->fh)) { $self->_open(); }
1550 # Request shared lock for reading
1552 $self->lock( LOCK_SH );
1554 my $result = $self->_get_next_key( $prev_md5 );
1558 return ($result && $self->root->{filter_fetch_key}) ? $self->root->{filter_fetch_key}->($result) : $result;
1562 # The following methods are for arrays only
1567 # Return the length of the array
1569 my $self = _get_self($_[0]);
1570 if ($self->type ne TYPE_ARRAY) {
1571 return $self->_throw_error("FETCHSIZE method only supported for arrays");
1574 my $SAVE_FILTER = $self->root->{filter_fetch_value};
1575 $self->root->{filter_fetch_value} = undef;
1577 my $packed_size = $self->FETCH('length');
1579 $self->root->{filter_fetch_value} = $SAVE_FILTER;
1581 if ($packed_size) { return int(unpack($LONG_PACK, $packed_size)); }
1587 # Set the length of the array
1589 my $self = _get_self($_[0]);
1590 if ($self->type ne TYPE_ARRAY) {
1591 return $self->_throw_error("STORESIZE method only supported for arrays");
1593 my $new_length = $_[1];
1595 my $SAVE_FILTER = $self->root->{filter_store_value};
1596 $self->root->{filter_store_value} = undef;
1598 my $result = $self->STORE('length', pack($LONG_PACK, $new_length));
1600 $self->root->{filter_store_value} = $SAVE_FILTER;
1607 # Remove and return the last element on the array
1609 my $self = _get_self($_[0]);
1610 if ($self->type ne TYPE_ARRAY) {
1611 return $self->_throw_error("POP method only supported for arrays");
1613 my $length = $self->FETCHSIZE();
1616 my $content = $self->FETCH( $length - 1 );
1617 $self->DELETE( $length - 1 );
1627 # Add new element(s) to the end of the array
1629 my $self = _get_self(shift);
1630 if ($self->type ne TYPE_ARRAY) {
1631 return $self->_throw_error("PUSH method only supported for arrays");
1633 my $length = $self->FETCHSIZE();
1635 while (my $content = shift @_) {
1636 $self->STORE( $length, $content );
1643 # Remove and return first element on the array.
1644 # Shift over remaining elements to take up space.
1646 my $self = _get_self($_[0]);
1647 if ($self->type ne TYPE_ARRAY) {
1648 return $self->_throw_error("SHIFT method only supported for arrays");
1650 my $length = $self->FETCHSIZE();
1653 my $content = $self->FETCH( 0 );
1656 # Shift elements over and remove last one.
1658 for (my $i = 0; $i < $length - 1; $i++) {
1659 $self->STORE( $i, $self->FETCH($i + 1) );
1661 $self->DELETE( $length - 1 );
1672 # Insert new element(s) at beginning of array.
1673 # Shift over other elements to make space.
1675 my $self = _get_self($_[0]);shift @_;
1676 if ($self->type ne TYPE_ARRAY) {
1677 return $self->_throw_error("UNSHIFT method only supported for arrays");
1679 my @new_elements = @_;
1680 my $length = $self->FETCHSIZE();
1681 my $new_size = scalar @new_elements;
1684 for (my $i = $length - 1; $i >= 0; $i--) {
1685 $self->STORE( $i + $new_size, $self->FETCH($i) );
1689 for (my $i = 0; $i < $new_size; $i++) {
1690 $self->STORE( $i, $new_elements[$i] );
1696 # Splices section of array with optional new section.
1697 # Returns deleted section, or last element deleted in scalar context.
1699 my $self = _get_self($_[0]);shift @_;
1700 if ($self->type ne TYPE_ARRAY) {
1701 return $self->_throw_error("SPLICE method only supported for arrays");
1703 my $length = $self->FETCHSIZE();
1706 # Calculate offset and length of splice
1708 my $offset = shift || 0;
1709 if ($offset < 0) { $offset += $length; }
1712 if (scalar @_) { $splice_length = shift; }
1713 else { $splice_length = $length - $offset; }
1714 if ($splice_length < 0) { $splice_length += ($length - $offset); }
1717 # Setup array with new elements, and copy out old elements for return
1719 my @new_elements = @_;
1720 my $new_size = scalar @new_elements;
1722 my @old_elements = ();
1723 for (my $i = $offset; $i < $offset + $splice_length; $i++) {
1724 push @old_elements, $self->FETCH( $i );
1728 # Adjust array length, and shift elements to accomodate new section.
1730 if ( $new_size != $splice_length ) {
1731 if ($new_size > $splice_length) {
1732 for (my $i = $length - 1; $i >= $offset + $splice_length; $i--) {
1733 $self->STORE( $i + ($new_size - $splice_length), $self->FETCH($i) );
1737 for (my $i = $offset + $splice_length; $i < $length; $i++) {
1738 $self->STORE( $i + ($new_size - $splice_length), $self->FETCH($i) );
1740 for (my $i = 0; $i < $splice_length - $new_size; $i++) {
1741 $self->DELETE( $length - 1 );
1748 # Insert new elements into array
1750 for (my $i = $offset; $i < $offset + $new_size; $i++) {
1751 $self->STORE( $i, shift @new_elements );
1755 # Return deleted section, or last element in scalar context.
1757 return wantarray ? @old_elements : $old_elements[-1];
1760 #XXX We don't need to define it.
1761 #XXX It will be useful, though, when we split out HASH and ARRAY
1764 # Perl will call EXTEND() when the array is likely to grow.
1765 # We don't care, but include it for compatibility.
1770 # Public method aliases
1772 *put = *store = *STORE;
1773 *get = *fetch = *FETCH;
1777 *first_key = *FIRSTKEY;
1778 *next_key = *NEXTKEY;
1779 *length = *FETCHSIZE;
1783 *unshift = *UNSHIFT;
1792 DBM::Deep - A pure perl multi-level hash/array DBM
1797 my $db = DBM::Deep->new( "foo.db" );
1799 $db->{key} = 'value'; # tie() style
1802 $db->put('key', 'value'); # OO style
1803 print $db->get('key');
1805 # true multi-level support
1806 $db->{my_complex} = [
1807 'hello', { perl => 'rules' },
1812 A unique flat-file database module, written in pure perl. True
1813 multi-level hash/array support (unlike MLDBM, which is faked), hybrid
1814 OO / tie() interface, cross-platform FTPable files, and quite fast. Can
1815 handle millions of keys and unlimited hash levels without significant
1816 slow-down. Written from the ground-up in pure perl -- this is NOT a
1817 wrapper around a C-based DBM. Out-of-the-box compatibility with Unix,
1818 Mac OS X and Windows.
1822 Hopefully you are using CPAN's excellent Perl module, which will download
1823 and install the module for you. If not, get the tarball, and run these
1835 Construction can be done OO-style (which is the recommended way), or using
1836 Perl's tie() function. Both are examined here.
1838 =head2 OO CONSTRUCTION
1840 The recommended way to construct a DBM::Deep object is to use the new()
1841 method, which gets you a blessed, tied hash or array reference.
1843 my $db = DBM::Deep->new( "foo.db" );
1845 This opens a new database handle, mapped to the file "foo.db". If this
1846 file does not exist, it will automatically be created. DB files are
1847 opened in "r+" (read/write) mode, and the type of object returned is a
1848 hash, unless otherwise specified (see L<OPTIONS> below).
1852 You can pass a number of options to the constructor to specify things like
1853 locking, autoflush, etc. This is done by passing an inline hash:
1855 my $db = DBM::Deep->new(
1861 Notice that the filename is now specified I<inside> the hash with
1862 the "file" parameter, as opposed to being the sole argument to the
1863 constructor. This is required if any options are specified.
1864 See L<OPTIONS> below for the complete list.
1868 You can also start with an array instead of a hash. For this, you must
1869 specify the C<type> parameter:
1871 my $db = DBM::Deep->new(
1873 type => DBM::Deep->TYPE_ARRAY
1876 B<Note:> Specifing the C<type> parameter only takes effect when beginning
1877 a new DB file. If you create a DBM::Deep object with an existing file, the
1878 C<type> will be loaded from the file header, and ignored if it is passed
1881 =head2 TIE CONSTRUCTION
1883 Alternatively, you can create a DBM::Deep handle by using Perl's built-in
1884 tie() function. This is not ideal, because you get only a basic, tied hash
1885 (or array) which is not blessed, so you can't call any functions on it.
1888 tie %hash, "DBM::Deep", "foo.db";
1891 tie @array, "DBM::Deep", "bar.db";
1893 As with the OO constructor, you can replace the DB filename parameter with
1894 a hash containing one or more options (see L<OPTIONS> just below for the
1897 tie %hash, "DBM::Deep", {
1905 There are a number of options that can be passed in when constructing your
1906 DBM::Deep objects. These apply to both the OO- and tie- based approaches.
1912 Filename of the DB file to link the handle to. You can pass a full absolute
1913 filesystem path, partial path, or a plain filename if the file is in the
1914 current working directory. This is a required parameter.
1918 File open mode (read-only, read-write, etc.) string passed to Perl's FileHandle
1919 module. This is an optional parameter, and defaults to "r+" (read/write).
1920 B<Note:> If the default (r+) mode is selected, the file will also be auto-
1921 created if it doesn't exist.
1925 This parameter specifies what type of object to create, a hash or array. Use
1926 one of these two constants: C<DBM::Deep-E<gt>TYPE_HASH> or C<DBM::Deep-E<gt>TYPE_ARRAY>.
1927 This only takes effect when beginning a new file. This is an optional
1928 parameter, and defaults to C<DBM::Deep-E<gt>TYPE_HASH>.
1932 Specifies whether locking is to be enabled. DBM::Deep uses Perl's Fnctl flock()
1933 function to lock the database in exclusive mode for writes, and shared mode for
1934 reads. Pass any true value to enable. This affects the base DB handle I<and
1935 any child hashes or arrays> that use the same DB file. This is an optional
1936 parameter, and defaults to 0 (disabled). See L<LOCKING> below for more.
1940 Specifies whether autoflush is to be enabled on the underlying FileHandle.
1941 This obviously slows down write operations, but is required if you may have
1942 multiple processes accessing the same DB file (also consider enable I<locking>
1943 or at least I<volatile>). Pass any true value to enable. This is an optional
1944 parameter, and defaults to 0 (disabled).
1948 If I<volatile> mode is enabled, DBM::Deep will stat() the DB file before each
1949 STORE() operation. This is required if an outside force may change the size of
1950 the file between transactions. Locking also implicitly enables volatile. This
1951 is useful if you want to use a different locking system or write your own. Pass
1952 any true value to enable. This is an optional parameter, and defaults to 0
1957 If I<autobless> mode is enabled, DBM::Deep will preserve blessed hashes, and
1958 restore them when fetched. This is an B<experimental> feature, and does have
1959 side-effects. Basically, when hashes are re-blessed into their original
1960 classes, they are no longer blessed into the DBM::Deep class! So you won't be
1961 able to call any DBM::Deep methods on them. You have been warned.
1962 This is an optional parameter, and defaults to 0 (disabled).
1966 See L<FILTERS> below.
1970 Setting I<debug> mode will make all errors non-fatal, dump them out to
1971 STDERR, and continue on. This is for debugging purposes only, and probably
1972 not what you want. This is an optional parameter, and defaults to 0 (disabled).
1976 Instead of passing a file path, you can instead pass a handle to an pre-opened
1977 filehandle. Note: Beware of using the magick *DATA handle, as this actually
1978 contains your entire Perl script, as well as the data following the __DATA__
1979 marker. This will not work, because DBM::Deep uses absolute seek()s into the
1980 file. Instead, consider reading *DATA into an IO::Scalar handle, then passing
1985 =head1 TIE INTERFACE
1987 With DBM::Deep you can access your databases using Perl's standard hash/array
1988 syntax. Because all DBM::Deep objects are I<tied> to hashes or arrays, you can treat
1989 them as such. DBM::Deep will intercept all reads/writes and direct them to the right
1990 place -- the DB file. This has nothing to do with the L<TIE CONSTRUCTION>
1991 section above. This simply tells you how to use DBM::Deep using regular hashes
1992 and arrays, rather than calling functions like C<get()> and C<put()> (although those
1993 work too). It is entirely up to you how to want to access your databases.
1997 You can treat any DBM::Deep object like a normal Perl hash reference. Add keys,
1998 or even nested hashes (or arrays) using standard Perl syntax:
2000 my $db = DBM::Deep->new( "foo.db" );
2002 $db->{mykey} = "myvalue";
2004 $db->{myhash}->{subkey} = "subvalue";
2006 print $db->{myhash}->{subkey} . "\n";
2008 You can even step through hash keys using the normal Perl C<keys()> function:
2010 foreach my $key (keys %$db) {
2011 print "$key: " . $db->{$key} . "\n";
2014 Remember that Perl's C<keys()> function extracts I<every> key from the hash and
2015 pushes them onto an array, all before the loop even begins. If you have an
2016 extra large hash, this may exhaust Perl's memory. Instead, consider using
2017 Perl's C<each()> function, which pulls keys/values one at a time, using very
2020 while (my ($key, $value) = each %$db) {
2021 print "$key: $value\n";
2024 Please note that when using C<each()>, you should always pass a direct
2025 hash reference, not a lookup. Meaning, you should B<never> do this:
2028 while (my ($key, $value) = each %{$db->{foo}}) { # BAD
2030 This causes an infinite loop, because for each iteration, Perl is calling
2031 FETCH() on the $db handle, resulting in a "new" hash for foo every time, so
2032 it effectively keeps returning the first key over and over again. Instead,
2033 assign a temporary variable to C<$db->{foo}>, then pass that to each().
2037 As with hashes, you can treat any DBM::Deep object like a normal Perl array
2038 reference. This includes inserting, removing and manipulating elements,
2039 and the C<push()>, C<pop()>, C<shift()>, C<unshift()> and C<splice()> functions.
2040 The object must have first been created using type C<DBM::Deep-E<gt>TYPE_ARRAY>,
2041 or simply be a nested array reference inside a hash. Example:
2043 my $db = DBM::Deep->new(
2044 file => "foo-array.db",
2045 type => DBM::Deep->TYPE_ARRAY
2049 push @$db, "bar", "baz";
2050 unshift @$db, "bah";
2052 my $last_elem = pop @$db; # baz
2053 my $first_elem = shift @$db; # bah
2054 my $second_elem = $db->[1]; # bar
2056 my $num_elements = scalar @$db;
2060 In addition to the I<tie()> interface, you can also use a standard OO interface
2061 to manipulate all aspects of DBM::Deep databases. Each type of object (hash or
2062 array) has its own methods, but both types share the following common methods:
2063 C<put()>, C<get()>, C<exists()>, C<delete()> and C<clear()>.
2069 Stores a new hash key/value pair, or sets an array element value. Takes two
2070 arguments, the hash key or array index, and the new value. The value can be
2071 a scalar, hash ref or array ref. Returns true on success, false on failure.
2073 $db->put("foo", "bar"); # for hashes
2074 $db->put(1, "bar"); # for arrays
2078 Fetches the value of a hash key or array element. Takes one argument: the hash
2079 key or array index. Returns a scalar, hash ref or array ref, depending on the
2082 my $value = $db->get("foo"); # for hashes
2083 my $value = $db->get(1); # for arrays
2087 Checks if a hash key or array index exists. Takes one argument: the hash key
2088 or array index. Returns true if it exists, false if not.
2090 if ($db->exists("foo")) { print "yay!\n"; } # for hashes
2091 if ($db->exists(1)) { print "yay!\n"; } # for arrays
2095 Deletes one hash key/value pair or array element. Takes one argument: the hash
2096 key or array index. Returns true on success, false if not found. For arrays,
2097 the remaining elements located after the deleted element are NOT moved over.
2098 The deleted element is essentially just undefined, which is exactly how Perl's
2099 internal arrays work. Please note that the space occupied by the deleted
2100 key/value or element is B<not> reused again -- see L<UNUSED SPACE RECOVERY>
2101 below for details and workarounds.
2103 $db->delete("foo"); # for hashes
2104 $db->delete(1); # for arrays
2108 Deletes B<all> hash keys or array elements. Takes no arguments. No return
2109 value. Please note that the space occupied by the deleted keys/values or
2110 elements is B<not> reused again -- see L<UNUSED SPACE RECOVERY> below for
2111 details and workarounds.
2113 $db->clear(); # hashes or arrays
2119 For hashes, DBM::Deep supports all the common methods described above, and the
2120 following additional methods: C<first_key()> and C<next_key()>.
2126 Returns the "first" key in the hash. As with built-in Perl hashes, keys are
2127 fetched in an undefined order (which appears random). Takes no arguments,
2128 returns the key as a scalar value.
2130 my $key = $db->first_key();
2134 Returns the "next" key in the hash, given the previous one as the sole argument.
2135 Returns undef if there are no more keys to be fetched.
2137 $key = $db->next_key($key);
2141 Here are some examples of using hashes:
2143 my $db = DBM::Deep->new( "foo.db" );
2145 $db->put("foo", "bar");
2146 print "foo: " . $db->get("foo") . "\n";
2148 $db->put("baz", {}); # new child hash ref
2149 $db->get("baz")->put("buz", "biz");
2150 print "buz: " . $db->get("baz")->get("buz") . "\n";
2152 my $key = $db->first_key();
2154 print "$key: " . $db->get($key) . "\n";
2155 $key = $db->next_key($key);
2158 if ($db->exists("foo")) { $db->delete("foo"); }
2162 For arrays, DBM::Deep supports all the common methods described above, and the
2163 following additional methods: C<length()>, C<push()>, C<pop()>, C<shift()>,
2164 C<unshift()> and C<splice()>.
2170 Returns the number of elements in the array. Takes no arguments.
2172 my $len = $db->length();
2176 Adds one or more elements onto the end of the array. Accepts scalars, hash
2177 refs or array refs. No return value.
2179 $db->push("foo", "bar", {});
2183 Fetches the last element in the array, and deletes it. Takes no arguments.
2184 Returns undef if array is empty. Returns the element value.
2186 my $elem = $db->pop();
2190 Fetches the first element in the array, deletes it, then shifts all the
2191 remaining elements over to take up the space. Returns the element value. This
2192 method is not recommended with large arrays -- see L<LARGE ARRAYS> below for
2195 my $elem = $db->shift();
2199 Inserts one or more elements onto the beginning of the array, shifting all
2200 existing elements over to make room. Accepts scalars, hash refs or array refs.
2201 No return value. This method is not recommended with large arrays -- see
2202 <LARGE ARRAYS> below for details.
2204 $db->unshift("foo", "bar", {});
2208 Performs exactly like Perl's built-in function of the same name. See L<perldoc
2209 -f splice> for usage -- it is too complicated to document here. This method is
2210 not recommended with large arrays -- see L<LARGE ARRAYS> below for details.
2214 Here are some examples of using arrays:
2216 my $db = DBM::Deep->new(
2218 type => DBM::Deep->TYPE_ARRAY
2221 $db->push("bar", "baz");
2222 $db->unshift("foo");
2225 my $len = $db->length();
2226 print "length: $len\n"; # 4
2228 for (my $k=0; $k<$len; $k++) {
2229 print "$k: " . $db->get($k) . "\n";
2232 $db->splice(1, 2, "biz", "baf");
2234 while (my $elem = shift @$db) {
2235 print "shifted: $elem\n";
2240 Enable automatic file locking by passing a true value to the C<locking>
2241 parameter when constructing your DBM::Deep object (see L<SETUP> above).
2243 my $db = DBM::Deep->new(
2248 This causes DBM::Deep to C<flock()> the underlying FileHandle object with exclusive
2249 mode for writes, and shared mode for reads. This is required if you have
2250 multiple processes accessing the same database file, to avoid file corruption.
2251 Please note that C<flock()> does NOT work for files over NFS. See L<DB OVER
2252 NFS> below for more.
2254 =head2 EXPLICIT LOCKING
2256 You can explicitly lock a database, so it remains locked for multiple
2257 transactions. This is done by calling the C<lock()> method, and passing an
2258 optional lock mode argument (defaults to exclusive mode). This is particularly
2259 useful for things like counters, where the current value needs to be fetched,
2260 then incremented, then stored again.
2263 my $counter = $db->get("counter");
2265 $db->put("counter", $counter);
2274 You can pass C<lock()> an optional argument, which specifies which mode to use
2275 (exclusive or shared). Use one of these two constants: C<DBM::Deep-E<gt>LOCK_EX>
2276 or C<DBM::Deep-E<gt>LOCK_SH>. These are passed directly to C<flock()>, and are the
2277 same as the constants defined in Perl's C<Fcntl> module.
2279 $db->lock( DBM::Deep->LOCK_SH );
2283 If you want to implement your own file locking scheme, be sure to create your
2284 DBM::Deep objects setting the C<volatile> option to true. This hints to DBM::Deep
2285 that the DB file may change between transactions. See L<LOW-LEVEL ACCESS>
2288 =head1 IMPORTING/EXPORTING
2290 You can import existing complex structures by calling the C<import()> method,
2291 and export an entire database into an in-memory structure using the C<export()>
2292 method. Both are examined here.
2296 Say you have an existing hash with nested hashes/arrays inside it. Instead of
2297 walking the structure and adding keys/elements to the database as you go,
2298 simply pass a reference to the C<import()> method. This recursively adds
2299 everything to an existing DBM::Deep object for you. Here is an example:
2304 array1 => [ "elem0", "elem1", "elem2" ],
2306 subkey1 => "subvalue1",
2307 subkey2 => "subvalue2"
2311 my $db = DBM::Deep->new( "foo.db" );
2312 $db->import( $struct );
2314 print $db->{key1} . "\n"; # prints "value1"
2316 This recursively imports the entire C<$struct> object into C<$db>, including
2317 all nested hashes and arrays. If the DBM::Deep object contains exsiting data,
2318 keys are merged with the existing ones, replacing if they already exist.
2319 The C<import()> method can be called on any database level (not just the base
2320 level), and works with both hash and array DB types.
2324 B<Note:> Make sure your existing structure has no circular references in it.
2325 These will cause an infinite loop when importing.
2329 Calling the C<export()> method on an existing DBM::Deep object will return
2330 a reference to a new in-memory copy of the database. The export is done
2331 recursively, so all nested hashes/arrays are all exported to standard Perl
2332 objects. Here is an example:
2334 my $db = DBM::Deep->new( "foo.db" );
2336 $db->{key1} = "value1";
2337 $db->{key2} = "value2";
2339 $db->{hash1}->{subkey1} = "subvalue1";
2340 $db->{hash1}->{subkey2} = "subvalue2";
2342 my $struct = $db->export();
2344 print $struct->{key1} . "\n"; # prints "value1"
2346 This makes a complete copy of the database in memory, and returns a reference
2347 to it. The C<export()> method can be called on any database level (not just
2348 the base level), and works with both hash and array DB types. Be careful of
2349 large databases -- you can store a lot more data in a DBM::Deep object than an
2350 in-memory Perl structure.
2354 B<Note:> Make sure your database has no circular references in it.
2355 These will cause an infinite loop when exporting.
2359 DBM::Deep has a number of hooks where you can specify your own Perl function
2360 to perform filtering on incoming or outgoing data. This is a perfect
2361 way to extend the engine, and implement things like real-time compression or
2362 encryption. Filtering applies to the base DB level, and all child hashes /
2363 arrays. Filter hooks can be specified when your DBM::Deep object is first
2364 constructed, or by calling the C<set_filter()> method at any time. There are
2365 four available filter hooks, described below:
2369 =item * filter_store_key
2371 This filter is called whenever a hash key is stored. It
2372 is passed the incoming key, and expected to return a transformed key.
2374 =item * filter_store_value
2376 This filter is called whenever a hash key or array element is stored. It
2377 is passed the incoming value, and expected to return a transformed value.
2379 =item * filter_fetch_key
2381 This filter is called whenever a hash key is fetched (i.e. via
2382 C<first_key()> or C<next_key()>). It is passed the transformed key,
2383 and expected to return the plain key.
2385 =item * filter_fetch_value
2387 This filter is called whenever a hash key or array element is fetched.
2388 It is passed the transformed value, and expected to return the plain value.
2392 Here are the two ways to setup a filter hook:
2394 my $db = DBM::Deep->new(
2396 filter_store_value => \&my_filter_store,
2397 filter_fetch_value => \&my_filter_fetch
2402 $db->set_filter( "filter_store_value", \&my_filter_store );
2403 $db->set_filter( "filter_fetch_value", \&my_filter_fetch );
2405 Your filter function will be called only when dealing with SCALAR keys or
2406 values. When nested hashes and arrays are being stored/fetched, filtering
2407 is bypassed. Filters are called as static functions, passed a single SCALAR
2408 argument, and expected to return a single SCALAR value. If you want to
2409 remove a filter, set the function reference to C<undef>:
2411 $db->set_filter( "filter_store_value", undef );
2413 =head2 REAL-TIME ENCRYPTION EXAMPLE
2415 Here is a working example that uses the I<Crypt::Blowfish> module to
2416 do real-time encryption / decryption of keys & values with DBM::Deep Filters.
2417 Please visit L<http://search.cpan.org/search?module=Crypt::Blowfish> for more
2418 on I<Crypt::Blowfish>. You'll also need the I<Crypt::CBC> module.
2421 use Crypt::Blowfish;
2424 my $cipher = Crypt::CBC->new({
2425 'key' => 'my secret key',
2426 'cipher' => 'Blowfish',
2428 'regenerate_key' => 0,
2429 'padding' => 'space',
2433 my $db = DBM::Deep->new(
2434 file => "foo-encrypt.db",
2435 filter_store_key => \&my_encrypt,
2436 filter_store_value => \&my_encrypt,
2437 filter_fetch_key => \&my_decrypt,
2438 filter_fetch_value => \&my_decrypt,
2441 $db->{key1} = "value1";
2442 $db->{key2} = "value2";
2443 print "key1: " . $db->{key1} . "\n";
2444 print "key2: " . $db->{key2} . "\n";
2450 return $cipher->encrypt( $_[0] );
2453 return $cipher->decrypt( $_[0] );
2456 =head2 REAL-TIME COMPRESSION EXAMPLE
2458 Here is a working example that uses the I<Compress::Zlib> module to do real-time
2459 compression / decompression of keys & values with DBM::Deep Filters.
2460 Please visit L<http://search.cpan.org/search?module=Compress::Zlib> for
2461 more on I<Compress::Zlib>.
2466 my $db = DBM::Deep->new(
2467 file => "foo-compress.db",
2468 filter_store_key => \&my_compress,
2469 filter_store_value => \&my_compress,
2470 filter_fetch_key => \&my_decompress,
2471 filter_fetch_value => \&my_decompress,
2474 $db->{key1} = "value1";
2475 $db->{key2} = "value2";
2476 print "key1: " . $db->{key1} . "\n";
2477 print "key2: " . $db->{key2} . "\n";
2483 return Compress::Zlib::memGzip( $_[0] ) ;
2486 return Compress::Zlib::memGunzip( $_[0] ) ;
2489 B<Note:> Filtering of keys only applies to hashes. Array "keys" are
2490 actually numerical index numbers, and are not filtered.
2492 =head1 ERROR HANDLING
2494 Most DBM::Deep methods return a true value for success, and call die() on
2495 failure. You can wrap calls in an eval block to catch the die. Also, the
2496 actual error message is stored in an internal scalar, which can be fetched by
2497 calling the C<error()> method.
2499 my $db = DBM::Deep->new( "foo.db" ); # create hash
2500 eval { $db->push("foo"); }; # ILLEGAL -- push is array-only call
2502 print $db->error(); # prints error message
2504 You can then call C<clear_error()> to clear the current error state.
2508 If you set the C<debug> option to true when creating your DBM::Deep object,
2509 all errors are considered NON-FATAL, and dumped to STDERR. This is only
2510 for debugging purposes.
2512 =head1 LARGEFILE SUPPORT
2514 If you have a 64-bit system, and your Perl is compiled with both LARGEFILE
2515 and 64-bit support, you I<may> be able to create databases larger than 2 GB.
2516 DBM::Deep by default uses 32-bit file offset tags, but these can be changed
2517 by calling the static C<set_pack()> method before you do anything else.
2519 DBM::Deep::set_pack(8, 'Q');
2521 This tells DBM::Deep to pack all file offsets with 8-byte (64-bit) quad words
2522 instead of 32-bit longs. After setting these values your DB files have a
2523 theoretical maximum size of 16 XB (exabytes).
2527 B<Note:> Changing these values will B<NOT> work for existing database files.
2528 Only change this for new files, and make sure it stays set consistently
2529 throughout the file's life. If you do set these values, you can no longer
2530 access 32-bit DB files. You can, however, call C<set_pack(4, 'N')> to change
2531 back to 32-bit mode.
2535 B<Note:> I have not personally tested files > 2 GB -- all my systems have
2536 only a 32-bit Perl. However, I have received user reports that this does
2539 =head1 LOW-LEVEL ACCESS
2541 If you require low-level access to the underlying FileHandle that DBM::Deep uses,
2542 you can call the C<fh()> method, which returns the handle:
2546 This method can be called on the root level of the datbase, or any child
2547 hashes or arrays. All levels share a I<root> structure, which contains things
2548 like the FileHandle, a reference counter, and all your options you specified
2549 when you created the object. You can get access to this root structure by
2550 calling the C<root()> method.
2552 my $root = $db->root();
2554 This is useful for changing options after the object has already been created,
2555 such as enabling/disabling locking, volatile or debug modes. You can also
2556 store your own temporary user data in this structure (be wary of name
2557 collision), which is then accessible from any child hash or array.
2559 =head1 CUSTOM DIGEST ALGORITHM
2561 DBM::Deep by default uses the I<Message Digest 5> (MD5) algorithm for hashing
2562 keys. However you can override this, and use another algorithm (such as SHA-256)
2563 or even write your own. But please note that DBM::Deep currently expects zero
2564 collisions, so your algorithm has to be I<perfect>, so to speak.
2565 Collision detection may be introduced in a later version.
2569 You can specify a custom digest algorithm by calling the static C<set_digest()>
2570 function, passing a reference to a subroutine, and the length of the algorithm's
2571 hashes (in bytes). This is a global static function, which affects ALL DBM::Deep
2572 objects. Here is a working example that uses a 256-bit hash from the
2573 I<Digest::SHA256> module. Please see
2574 L<http://search.cpan.org/search?module=Digest::SHA256> for more.
2579 my $context = Digest::SHA256::new(256);
2581 DBM::Deep::set_digest( \&my_digest, 32 );
2583 my $db = DBM::Deep->new( "foo-sha.db" );
2585 $db->{key1} = "value1";
2586 $db->{key2} = "value2";
2587 print "key1: " . $db->{key1} . "\n";
2588 print "key2: " . $db->{key2} . "\n";
2594 return substr( $context->hash($_[0]), 0, 32 );
2597 B<Note:> Your returned digest strings must be B<EXACTLY> the number
2598 of bytes you specify in the C<set_digest()> function (in this case 32).
2600 =head1 CIRCULAR REFERENCES
2602 DBM::Deep has B<experimental> support for circular references. Meaning you
2603 can have a nested hash key or array element that points to a parent object.
2604 This relationship is stored in the DB file, and is preserved between sessions.
2607 my $db = DBM::Deep->new( "foo.db" );
2610 $db->{circle} = $db; # ref to self
2612 print $db->{foo} . "\n"; # prints "foo"
2613 print $db->{circle}->{foo} . "\n"; # prints "foo" again
2615 One catch is, passing the object to a function that recursively walks the
2616 object tree (such as I<Data::Dumper> or even the built-in C<optimize()> or
2617 C<export()> methods) will result in an infinite loop. The other catch is,
2618 if you fetch the I<key> of a circular reference (i.e. using the C<first_key()>
2619 or C<next_key()> methods), you will get the I<target object's key>, not the
2620 ref's key. This gets even more interesting with the above example, where
2621 the I<circle> key points to the base DB object, which technically doesn't
2622 have a key. So I made DBM::Deep return "[base]" as the key name in that
2625 =head1 CAVEATS / ISSUES / BUGS
2627 This section describes all the known issues with DBM::Deep. It you have found
2628 something that is not listed here, please send e-mail to L<jhuckaby@cpan.org>.
2630 =head2 UNUSED SPACE RECOVERY
2632 One major caveat with DBM::Deep is that space occupied by existing keys and
2633 values is not recovered when they are deleted. Meaning if you keep deleting
2634 and adding new keys, your file will continuously grow. I am working on this,
2635 but in the meantime you can call the built-in C<optimize()> method from time to
2636 time (perhaps in a crontab or something) to recover all your unused space.
2638 $db->optimize(); # returns true on success
2640 This rebuilds the ENTIRE database into a new file, then moves it on top of
2641 the original. The new file will have no unused space, thus it will take up as
2642 little disk space as possible. Please note that this operation can take
2643 a long time for large files, and you need enough disk space to temporarily hold
2644 2 copies of your DB file. The temporary file is created in the same directory
2645 as the original, named with a ".tmp" extension, and is deleted when the
2646 operation completes. Oh, and if locking is enabled, the DB is automatically
2647 locked for the entire duration of the copy.
2651 B<WARNING:> Only call optimize() on the top-level node of the database, and
2652 make sure there are no child references lying around. DBM::Deep keeps a reference
2653 counter, and if it is greater than 1, optimize() will abort and return undef.
2655 =head2 AUTOVIVIFICATION
2657 Unfortunately, autovivification doesn't work with tied hashes. This appears to
2658 be a bug in Perl's tie() system, as I<Jakob Schmidt> encountered the very same
2659 issue with his I<DWH_FIle> module (see L<http://search.cpan.org/search?module=DWH_File>),
2660 and it is also mentioned in the BUGS section for the I<MLDBM> module <see
2661 L<http://search.cpan.org/search?module=MLDBM>). Basically, on a new db file,
2664 $db->{foo}->{bar} = "hello";
2666 Since "foo" doesn't exist, you cannot add "bar" to it. You end up with "foo"
2667 being an empty hash. Try this instead, which works fine:
2669 $db->{foo} = { bar => "hello" };
2671 As of Perl 5.8.7, this bug still exists. I have walked very carefully through
2672 the execution path, and Perl indeed passes an empty hash to the STORE() method.
2673 Probably a bug in Perl.
2675 =head2 FILE CORRUPTION
2677 The current level of error handling in DBM::Deep is minimal. Files I<are> checked
2678 for a 32-bit signature when opened, but other corruption in files can cause
2679 segmentation faults. DBM::Deep may try to seek() past the end of a file, or get
2680 stuck in an infinite loop depending on the level of corruption. File write
2681 operations are not checked for failure (for speed), so if you happen to run
2682 out of disk space, DBM::Deep will probably fail in a bad way. These things will
2683 be addressed in a later version of DBM::Deep.
2687 Beware of using DB files over NFS. DBM::Deep uses flock(), which works well on local
2688 filesystems, but will NOT protect you from file corruption over NFS. I've heard
2689 about setting up your NFS server with a locking daemon, then using lockf() to
2690 lock your files, but your milage may vary there as well. From what I
2691 understand, there is no real way to do it. However, if you need access to the
2692 underlying FileHandle in DBM::Deep for using some other kind of locking scheme like
2693 lockf(), see the L<LOW-LEVEL ACCESS> section above.
2695 =head2 COPYING OBJECTS
2697 Beware of copying tied objects in Perl. Very strange things can happen.
2698 Instead, use DBM::Deep's C<clone()> method which safely copies the object and
2699 returns a new, blessed, tied hash or array to the same level in the DB.
2701 my $copy = $db->clone();
2705 Beware of using C<shift()>, C<unshift()> or C<splice()> with large arrays.
2706 These functions cause every element in the array to move, which can be murder
2707 on DBM::Deep, as every element has to be fetched from disk, then stored again in
2708 a different location. This may be addressed in a later version.
2712 This section discusses DBM::Deep's speed and memory usage.
2716 Obviously, DBM::Deep isn't going to be as fast as some C-based DBMs, such as
2717 the almighty I<BerkeleyDB>. But it makes up for it in features like true
2718 multi-level hash/array support, and cross-platform FTPable files. Even so,
2719 DBM::Deep is still pretty fast, and the speed stays fairly consistent, even
2720 with huge databases. Here is some test data:
2722 Adding 1,000,000 keys to new DB file...
2724 At 100 keys, avg. speed is 2,703 keys/sec
2725 At 200 keys, avg. speed is 2,642 keys/sec
2726 At 300 keys, avg. speed is 2,598 keys/sec
2727 At 400 keys, avg. speed is 2,578 keys/sec
2728 At 500 keys, avg. speed is 2,722 keys/sec
2729 At 600 keys, avg. speed is 2,628 keys/sec
2730 At 700 keys, avg. speed is 2,700 keys/sec
2731 At 800 keys, avg. speed is 2,607 keys/sec
2732 At 900 keys, avg. speed is 2,190 keys/sec
2733 At 1,000 keys, avg. speed is 2,570 keys/sec
2734 At 2,000 keys, avg. speed is 2,417 keys/sec
2735 At 3,000 keys, avg. speed is 1,982 keys/sec
2736 At 4,000 keys, avg. speed is 1,568 keys/sec
2737 At 5,000 keys, avg. speed is 1,533 keys/sec
2738 At 6,000 keys, avg. speed is 1,787 keys/sec
2739 At 7,000 keys, avg. speed is 1,977 keys/sec
2740 At 8,000 keys, avg. speed is 2,028 keys/sec
2741 At 9,000 keys, avg. speed is 2,077 keys/sec
2742 At 10,000 keys, avg. speed is 2,031 keys/sec
2743 At 20,000 keys, avg. speed is 1,970 keys/sec
2744 At 30,000 keys, avg. speed is 2,050 keys/sec
2745 At 40,000 keys, avg. speed is 2,073 keys/sec
2746 At 50,000 keys, avg. speed is 1,973 keys/sec
2747 At 60,000 keys, avg. speed is 1,914 keys/sec
2748 At 70,000 keys, avg. speed is 2,091 keys/sec
2749 At 80,000 keys, avg. speed is 2,103 keys/sec
2750 At 90,000 keys, avg. speed is 1,886 keys/sec
2751 At 100,000 keys, avg. speed is 1,970 keys/sec
2752 At 200,000 keys, avg. speed is 2,053 keys/sec
2753 At 300,000 keys, avg. speed is 1,697 keys/sec
2754 At 400,000 keys, avg. speed is 1,838 keys/sec
2755 At 500,000 keys, avg. speed is 1,941 keys/sec
2756 At 600,000 keys, avg. speed is 1,930 keys/sec
2757 At 700,000 keys, avg. speed is 1,735 keys/sec
2758 At 800,000 keys, avg. speed is 1,795 keys/sec
2759 At 900,000 keys, avg. speed is 1,221 keys/sec
2760 At 1,000,000 keys, avg. speed is 1,077 keys/sec
2762 This test was performed on a PowerMac G4 1gHz running Mac OS X 10.3.2 & Perl
2763 5.8.1, with an 80GB Ultra ATA/100 HD spinning at 7200RPM. The hash keys and
2764 values were between 6 - 12 chars in length. The DB file ended up at 210MB.
2765 Run time was 12 min 3 sec.
2769 One of the great things about DBM::Deep is that it uses very little memory.
2770 Even with huge databases (1,000,000+ keys) you will not see much increased
2771 memory on your process. DBM::Deep relies solely on the filesystem for storing
2772 and fetching data. Here is output from I</usr/bin/top> before even opening a
2775 PID USER PRI NI SIZE RSS SHARE STAT %CPU %MEM TIME COMMAND
2776 22831 root 11 0 2716 2716 1296 R 0.0 0.2 0:07 perl
2778 Basically the process is taking 2,716K of memory. And here is the same
2779 process after storing and fetching 1,000,000 keys:
2781 PID USER PRI NI SIZE RSS SHARE STAT %CPU %MEM TIME COMMAND
2782 22831 root 14 0 2772 2772 1328 R 0.0 0.2 13:32 perl
2784 Notice the memory usage increased by only 56K. Test was performed on a 700mHz
2785 x86 box running Linux RedHat 7.2 & Perl 5.6.1.
2787 =head1 DB FILE FORMAT
2789 In case you were interested in the underlying DB file format, it is documented
2790 here in this section. You don't need to know this to use the module, it's just
2791 included for reference.
2795 DBM::Deep files always start with a 32-bit signature to identify the file type.
2796 This is at offset 0. The signature is "DPDB" in network byte order. This is
2797 checked when the file is opened.
2801 The DBM::Deep file is in a I<tagged format>, meaning each section of the file
2802 has a standard header containing the type of data, the length of data, and then
2803 the data itself. The type is a single character (1 byte), the length is a
2804 32-bit unsigned long in network byte order, and the data is, well, the data.
2805 Here is how it unfolds:
2809 Immediately after the 32-bit file signature is the I<Master Index> record.
2810 This is a standard tag header followed by 1024 bytes (in 32-bit mode) or 2048
2811 bytes (in 64-bit mode) of data. The type is I<H> for hash or I<A> for array,
2812 depending on how the DBM::Deep object was constructed.
2816 The index works by looking at a I<MD5 Hash> of the hash key (or array index
2817 number). The first 8-bit char of the MD5 signature is the offset into the
2818 index, multipled by 4 in 32-bit mode, or 8 in 64-bit mode. The value of the
2819 index element is a file offset of the next tag for the key/element in question,
2820 which is usually a I<Bucket List> tag (see below).
2824 The next tag I<could> be another index, depending on how many keys/elements
2825 exist. See L<RE-INDEXING> below for details.
2829 A I<Bucket List> is a collection of 16 MD5 hashes for keys/elements, plus
2830 file offsets to where the actual data is stored. It starts with a standard
2831 tag header, with type I<B>, and a data size of 320 bytes in 32-bit mode, or
2832 384 bytes in 64-bit mode. Each MD5 hash is stored in full (16 bytes), plus
2833 the 32-bit or 64-bit file offset for the I<Bucket> containing the actual data.
2834 When the list fills up, a I<Re-Index> operation is performed (See
2835 L<RE-INDEXING> below).
2839 A I<Bucket> is a tag containing a key/value pair (in hash mode), or a
2840 index/value pair (in array mode). It starts with a standard tag header with
2841 type I<D> for scalar data (string, binary, etc.), or it could be a nested
2842 hash (type I<H>) or array (type I<A>). The value comes just after the tag
2843 header. The size reported in the tag header is only for the value, but then,
2844 just after the value is another size (32-bit unsigned long) and then the plain
2845 key itself. Since the value is likely to be fetched more often than the plain
2846 key, I figured it would be I<slightly> faster to store the value first.
2850 If the type is I<H> (hash) or I<A> (array), the value is another I<Master Index>
2851 record for the nested structure, where the process begins all over again.
2855 After a I<Bucket List> grows to 16 records, its allocated space in the file is
2856 exhausted. Then, when another key/element comes in, the list is converted to a
2857 new index record. However, this index will look at the next char in the MD5
2858 hash, and arrange new Bucket List pointers accordingly. This process is called
2859 I<Re-Indexing>. Basically, a new index tag is created at the file EOF, and all
2860 17 (16 + new one) keys/elements are removed from the old Bucket List and
2861 inserted into the new index. Several new Bucket Lists are created in the
2862 process, as a new MD5 char from the key is being examined (it is unlikely that
2863 the keys will all share the same next char of their MD5s).
2867 Because of the way the I<MD5> algorithm works, it is impossible to tell exactly
2868 when the Bucket Lists will turn into indexes, but the first round tends to
2869 happen right around 4,000 keys. You will see a I<slight> decrease in
2870 performance here, but it picks back up pretty quick (see L<SPEED> above). Then
2871 it takes B<a lot> more keys to exhaust the next level of Bucket Lists. It's
2872 right around 900,000 keys. This process can continue nearly indefinitely --
2873 right up until the point the I<MD5> signatures start colliding with each other,
2874 and this is B<EXTREMELY> rare -- like winning the lottery 5 times in a row AND
2875 getting struck by lightning while you are walking to cash in your tickets.
2876 Theoretically, since I<MD5> hashes are 128-bit values, you I<could> have up to
2877 340,282,366,921,000,000,000,000,000,000,000,000,000 keys/elements (I believe
2878 this is 340 unodecillion, but don't quote me).
2882 When a new key/element is stored, the key (or index number) is first ran through
2883 I<Digest::MD5> to get a 128-bit signature (example, in hex:
2884 b05783b0773d894396d475ced9d2f4f6). Then, the I<Master Index> record is checked
2885 for the first char of the signature (in this case I<b>). If it does not exist,
2886 a new I<Bucket List> is created for our key (and the next 15 future keys that
2887 happen to also have I<b> as their first MD5 char). The entire MD5 is written
2888 to the I<Bucket List> along with the offset of the new I<Bucket> record (EOF at
2889 this point, unless we are replacing an existing I<Bucket>), where the actual
2890 data will be stored.
2894 Fetching an existing key/element involves getting a I<Digest::MD5> of the key
2895 (or index number), then walking along the indexes. If there are enough
2896 keys/elements in this DB level, there might be nested indexes, each linked to
2897 a particular char of the MD5. Finally, a I<Bucket List> is pointed to, which
2898 contains up to 16 full MD5 hashes. Each is checked for equality to the key in
2899 question. If we found a match, the I<Bucket> tag is loaded, where the value and
2900 plain key are stored.
2904 Fetching the plain key occurs when calling the I<first_key()> and I<next_key()>
2905 methods. In this process the indexes are walked systematically, and each key
2906 fetched in increasing MD5 order (which is why it appears random). Once the
2907 I<Bucket> is found, the value is skipped the plain key returned instead.
2908 B<Note:> Do not count on keys being fetched as if the MD5 hashes were
2909 alphabetically sorted. This only happens on an index-level -- as soon as the
2910 I<Bucket Lists> are hit, the keys will come out in the order they went in --
2911 so it's pretty much undefined how the keys will come out -- just like Perl's
2914 =head1 CODE COVERAGE
2916 I use B<Devel::Cover> to test the code coverage of my tests, below is the B<Devel::Cover> report on this
2917 module's test suite.
2919 ---------------------------- ------ ------ ------ ------ ------ ------ ------
2920 File stmt bran cond sub pod time total
2921 ---------------------------- ------ ------ ------ ------ ------ ------ ------
2922 blib/lib/DBM/Deep.pm 94.9 84.5 77.8 100.0 11.1 100.0 89.7
2923 Total 94.9 84.5 77.8 100.0 11.1 100.0 89.7
2924 ---------------------------- ------ ------ ------ ------ ------ ------ ------
2928 Joseph Huckaby, L<jhuckaby@cpan.org>
2930 Special thanks to Adam Sah and Rich Gaushell! You know why :-)
2934 perltie(1), Tie::Hash(3), Digest::MD5(3), Fcntl(3), flock(2), lockf(3), nfs(5),
2935 Digest::SHA256(3), Crypt::Blowfish(3), Compress::Zlib(3)
2939 Copyright (c) 2002-2005 Joseph Huckaby. All Rights Reserved.
2940 This is free software, you may use it and distribute it under the
2941 same terms as Perl itself.