package DBM::Deep::Engine;
-use 5.6.0;
+use 5.006_000;
use strict;
-use warnings;
+use warnings FATAL => 'all';
-our $VERSION = q(0.99_01);
-
-use Fcntl qw( :DEFAULT :flock );
-use Scalar::Util ();
+use DBM::Deep::Iterator ();
# File-wide notes:
-# * To add to bucket_size, make sure you modify the following:
-# - calculate_sizes()
-# - _get_key_subloc()
-# - add_bucket() - where the buckets are printed
-
-##
-# Setup file and tag signatures. These should never change.
-##
-sub SIG_FILE () { 'DPDB' }
-sub SIG_HEADER () { 'h' }
-sub SIG_INTERNAL () { 'i' }
-sub SIG_HASH () { 'H' }
-sub SIG_ARRAY () { 'A' }
-sub SIG_NULL () { 'N' }
-sub SIG_DATA () { 'D' }
-sub SIG_INDEX () { 'I' }
-sub SIG_BLIST () { 'B' }
-sub SIG_FREE () { 'F' }
-sub SIG_KEYS () { 'K' }
-sub SIG_SIZE () { 1 }
-
-sub new {
- my $class = shift;
- my ($args) = @_;
-
- my $self = bless {
- long_size => 4,
- long_pack => 'N',
- data_size => 4,
- data_pack => 'N',
-
- digest => \&Digest::MD5::md5,
- hash_size => 16,
-
- ##
- # Maximum number of buckets per blist before another level of indexing is
- # done. Increase this value for slightly greater speed, but larger database
- # files. DO NOT decrease this value below 16, due to risk of recursive
- # reindex overrun.
- ##
- max_buckets => 16,
-
- fileobj => undef,
- obj => undef,
- }, $class;
-
- if ( defined $args->{pack_size} ) {
- if ( lc $args->{pack_size} eq 'small' ) {
- $args->{long_size} = 2;
- $args->{long_pack} = 'n';
- }
- elsif ( lc $args->{pack_size} eq 'medium' ) {
- $args->{long_size} = 4;
- $args->{long_pack} = 'N';
- }
- elsif ( lc $args->{pack_size} eq 'large' ) {
- $args->{long_size} = 8;
- $args->{long_pack} = 'Q';
- }
- else {
- die "Unknown pack_size value: '$args->{pack_size}'\n";
- }
- }
+# * Every method in here assumes that the storage has been appropriately
+# safeguarded. This can be anything from flock() to some sort of manual
+# mutex. But, it's the caller's responsability to make sure that this has
+# been done.
- # Grab the parameters we want to use
- foreach my $param ( keys %$self ) {
- next unless exists $args->{$param};
- $self->{$param} = $args->{$param};
- }
- Scalar::Util::weaken( $self->{obj} ) if $self->{obj};
+sub SIG_HASH () { 'H' }
+sub SIG_ARRAY () { 'A' }
- if ( $self->{max_buckets} < 16 ) {
- warn "Floor of max_buckets is 16. Setting it to 16 from '$self->{max_buckets}'\n";
- $self->{max_buckets} = 16;
- }
+=head1 NAME
- return $self;
-}
+DBM::Deep::Engine
-sub _fileobj { return $_[0]{fileobj} }
+=head1 PURPOSE
-sub calculate_sizes {
- my $self = shift;
+This is an internal-use-only object for L<DBM::Deep/>. It mediates the low-level
+mapping between the L<DBM::Deep/> objects and the storage medium.
- # The 2**8 here indicates the number of different characters in the
- # current hashing algorithm
- #XXX Does this need to be updated with different hashing algorithms?
- $self->{hash_chars_used} = (2**8);
- $self->{index_size} = $self->{hash_chars_used} * $self->{long_size};
+The purpose of this documentation is to provide low-level documentation for
+developers. It is B<not> intended to be used by the general public. This
+documentation and what it documents can and will change without notice.
- $self->{bucket_size} = $self->{hash_size} + $self->{long_size} * 2;
- $self->{bucket_list_size} = $self->{max_buckets} * $self->{bucket_size};
+=head1 OVERVIEW
- $self->{key_size} = $self->{long_size} * 2;
- $self->{keyloc_size} = $self->{max_buckets} * $self->{key_size};
+The engine exposes an API to the DBM::Deep objects (DBM::Deep, DBM::Deep::Array,
+and DBM::Deep::Hash) for their use to access the actual stored values. This API
+is the following:
- return;
-}
+=over 4
-sub write_file_header {
- my $self = shift;
+=item * new
- my $loc = $self->_fileobj->request_space( length( SIG_FILE ) + 33 );
+=item * read_value
- $self->_fileobj->print_at( $loc,
- SIG_FILE,
- SIG_HEADER,
- pack('N', 1), # header version
- pack('N', 24), # header size
- pack('N4', 0, 0, 0, 0), # currently running transaction IDs
- pack('n', $self->{long_size}),
- pack('A', $self->{long_pack}),
- pack('n', $self->{data_size}),
- pack('A', $self->{data_pack}),
- pack('n', $self->{max_buckets}),
- );
+=item * get_classname
- $self->_fileobj->set_transaction_offset( 13 );
+=item * make_reference
- return;
-}
+=item * key_exists
-sub read_file_header {
- my $self = shift;
+=item * delete_key
- my $buffer = $self->_fileobj->read_at( 0, length(SIG_FILE) + 9 );
- return unless length($buffer);
+=item * write_value
- my ($file_signature, $sig_header, $header_version, $size) = unpack(
- 'A4 A N N', $buffer
- );
+=item * get_next_key
- unless ( $file_signature eq SIG_FILE ) {
- $self->_fileobj->close;
- $self->_throw_error( "Signature not found -- file is not a Deep DB" );
- }
+=item * setup
- unless ( $sig_header eq SIG_HEADER ) {
- $self->_fileobj->close;
- $self->_throw_error( "Old file version found." );
- }
+=item * begin_work
- my $buffer2 = $self->_fileobj->read_at( undef, $size );
- my ($a1, $a2, $a3, $a4, @values) = unpack( 'N4 n A n A n', $buffer2 );
+=item * commit
- $self->_fileobj->set_transaction_offset( 13 );
+=item * rollback
- if ( @values < 5 || grep { !defined } @values ) {
- $self->_fileobj->close;
- $self->_throw_error("Corrupted file - bad header");
- }
+=item * lock_exclusive
- #XXX Add warnings if values weren't set right
- @{$self}{qw(long_size long_pack data_size data_pack max_buckets)} = @values;
+=item * lock_shared
- return length($buffer) + length($buffer2);
-}
+=item * unlock
-sub setup_fh {
- my $self = shift;
- my ($obj) = @_;
+=back
- # Need to remove use of $fh here
- my $fh = $self->_fileobj->{fh};
- flock $fh, LOCK_EX;
-
- #XXX The duplication of calculate_sizes needs to go away
- unless ( $obj->{base_offset} ) {
- my $bytes_read = $self->read_file_header;
-
- $self->calculate_sizes;
-
- ##
- # File is empty -- write header and master index
- ##
- if (!$bytes_read) {
- $self->_fileobj->audit( "# Database created on" );
-
- $self->write_file_header;
-
- $obj->{base_offset} = $self->_fileobj->request_space(
- $self->tag_size( $self->{index_size} ),
- );
-
- $self->write_tag(
- $obj->_base_offset, $obj->_type,
- chr(0)x$self->{index_size},
- );
-
- # Flush the filehandle
- my $old_fh = select $fh;
- my $old_af = $|; $| = 1; $| = $old_af;
- select $old_fh;
- }
- else {
- $obj->{base_offset} = $bytes_read;
-
- ##
- # Get our type from master index header
- ##
- my $tag = $self->load_tag($obj->_base_offset);
- unless ( $tag ) {
- flock $fh, LOCK_UN;
- $self->_throw_error("Corrupted file, no master index record");
- }
-
- unless ($obj->_type eq $tag->{signature}) {
- flock $fh, LOCK_UN;
- $self->_throw_error("File type mismatch");
- }
- }
- }
- else {
- $self->calculate_sizes;
- }
+They are explained in their own sections below. These methods, in turn, may
+provide some bounds-checking, but primarily act to instantiate objects in the
+Engine::Sector::* hierarchy and dispatch to them.
- #XXX We have to make sure we don't mess up when autoflush isn't turned on
- $self->_fileobj->set_inode;
+=head1 TRANSACTIONS
- flock $fh, LOCK_UN;
+Transactions in DBM::Deep are implemented using a variant of MVCC. This attempts
+to keep the amount of actual work done against the file low while stil providing
+Atomicity, Consistency, and Isolation. Durability, unfortunately, cannot be done
+with only one file.
- return 1;
-}
+=head2 STALENESS
-sub tag_size {
- my $self = shift;
- my ($size) = @_;
- return SIG_SIZE + $self->{data_size} + $size;
-}
+If another process uses a transaction slot and writes stuff to it, then
+terminates, the data that process wrote it still within the file. In order to
+address this, there is also a transaction staleness counter associated within
+every write. Each time a transaction is started, that process increments that
+transaction's staleness counter. If, when it reads a value, the staleness
+counters aren't identical, DBM::Deep will consider the value on disk to be stale
+and discard it.
-sub write_tag {
- ##
- # Given offset, signature and content, create tag and write to disk
- ##
- my $self = shift;
- my ($offset, $sig, $content) = @_;
- my $size = length( $content );
-
- $self->_fileobj->print_at(
- $offset,
- $sig, pack($self->{data_pack}, $size), $content,
- );
-
- return unless defined $offset;
-
- return {
- signature => $sig,
- #XXX Is this even used?
- size => $size,
- offset => $offset + SIG_SIZE + $self->{data_size},
- content => $content
- };
-}
+=head2 DURABILITY
-sub load_tag {
- ##
- # Given offset, load single tag and return signature, size and data
- ##
- my $self = shift;
- my ($offset) = @_;
-
- my $fileobj = $self->_fileobj;
-
- my ($sig, $size) = unpack(
- "A $self->{data_pack}",
- $fileobj->read_at( $offset, SIG_SIZE + $self->{data_size} ),
- );
-
- return {
- signature => $sig,
- #XXX Is this even used?
- size => $size,
- offset => $offset + SIG_SIZE + $self->{data_size},
- content => $fileobj->read_at( undef, $size ),
- };
-}
+The fourth leg of ACID is Durability, the guarantee that when a commit returns,
+the data will be there the next time you read from it. This should be regardless
+of any crashes or powerdowns in between the commit and subsequent read.
+DBM::Deep does provide that guarantee; once the commit returns, all of the data
+has been transferred from the transaction shadow to the HEAD. The issue arises
+with partial commits - a commit that is interrupted in some fashion. In keeping
+with DBM::Deep's "tradition" of very light error-checking and non-existent
+error-handling, there is no way to recover from a partial commit. (This is
+probably a failure in Consistency as well as Durability.)
-sub find_keyloc {
- my $self = shift;
- my ($tag, $transaction_id) = @_;
- $transaction_id = $self->_fileobj->transaction_id
- unless defined $transaction_id;
+Other DBMSes use transaction logs (a separate file, generally) to achieve
+Durability. As DBM::Deep is a single-file, we would have to do something
+similar to what SQLite and BDB do in terms of committing using synchonized
+writes. To do this, we would have to use a much higher RAM footprint and some
+serious programming that make my head hurts just to think about it.
- for ( my $i = 0; $i < $self->{max_buckets}; $i++ ) {
- my ($loc, $trans_id, $is_deleted) = unpack(
- "$self->{long_pack} C C",
- substr( $tag->{content}, $i * $self->{key_size}, $self->{key_size} ),
- );
+=cut
- if ( $loc == 0 ) {
- return ( $loc, $is_deleted, $i * $self->{key_size} );
- }
+=head2 read_value( $obj, $key )
- next if $transaction_id != $trans_id;
+This takes an object that provides _base_offset() and a string. It returns the
+value stored in the corresponding Sector::Value's data section.
- return ( $loc, $is_deleted, $i * $self->{key_size} );
- }
+=cut
- return;
-}
+sub read_value { die "read_value must be implemented in a child class" }
-sub add_bucket {
- ##
- # Adds one key/value pair to bucket list, given offset, MD5 digest of key,
- # plain (undigested) key and value.
- ##
- my $self = shift;
- my ($tag, $md5, $plain_key, $value, $deleted, $orig_key) = @_;
+=head2 get_classname( $obj )
- # This verifies that only supported values will be stored.
- {
- my $r = Scalar::Util::reftype( $value );
+This takes an object that provides _base_offset() and returns the classname (if
+any) associated with it.
- last if !defined $r;
- last if $r eq 'HASH';
- last if $r eq 'ARRAY';
+It delegates to Sector::Reference::get_classname() for the heavy lifting.
- $self->_throw_error(
- "Storage of references of type '$r' is not supported."
- );
- }
+It performs a staleness check.
- my $fileobj = $self->_fileobj;
+=cut
- #ACID - This is a mutation. Must only find the exact transaction
- my ($keyloc, $offset) = $self->_find_in_buckets( $tag, $md5, 1 );
+sub get_classname { die "get_classname must be implemented in a child class" }
- my @transactions;
- if ( $fileobj->transaction_id == 0 ) {
- @transactions = $fileobj->current_transactions;
- }
+=head2 make_reference( $obj, $old_key, $new_key )
-# $self->_release_space( $size, $subloc );
-#XXX This needs updating to use _release_space
-
- my $location;
- my $size = $self->_length_needed( $value, $plain_key );
-
- # Updating a known md5
- if ( $keyloc ) {
- my $keytag = $self->load_tag( $keyloc );
- my ($subloc, $is_deleted, $offset) = $self->find_keyloc( $keytag );
-
- if ( $subloc && !$is_deleted && @transactions ) {
- my $old_value = $self->read_from_loc( $subloc, $orig_key );
- my $old_size = $self->_length_needed( $old_value, $plain_key );
-
- for my $trans_id ( @transactions ) {
- my ($loc, $is_deleted, $offset2) = $self->find_keyloc( $keytag, $trans_id );
- unless ($loc) {
- my $location2 = $fileobj->request_space( $old_size );
- $fileobj->print_at( $keytag->{offset} + $offset2,
- pack($self->{long_pack}, $location2 ),
- pack( 'C C', $trans_id, 0 ),
- );
- $self->write_value( $location2, $plain_key, $old_value, $orig_key );
- }
- }
- }
-
- $location = $self->_fileobj->request_space( $size );
- #XXX This needs to be transactionally-aware in terms of which keytag->{offset} to use
- $fileobj->print_at( $keytag->{offset} + $offset,
- pack($self->{long_pack}, $location ),
- pack( 'C C', $fileobj->transaction_id, 0 ),
- );
- }
- # Adding a new md5
- else {
- my $keyloc = $fileobj->request_space( $self->tag_size( $self->{keyloc_size} ) );
-
- # The bucket fit into list
- if ( defined $offset ) {
- $fileobj->print_at( $tag->{offset} + $offset,
- $md5, pack( $self->{long_pack}, $keyloc ),
- );
- }
- # If bucket didn't fit into list, split into a new index level
- else {
- $self->split_index( $tag, $md5, $keyloc );
- }
-
- my $keytag = $self->write_tag(
- $keyloc, SIG_KEYS, chr(0)x$self->{keyloc_size},
- );
-
- $location = $self->_fileobj->request_space( $size );
- $fileobj->print_at( $keytag->{offset},
- pack( $self->{long_pack}, $location ),
- pack( 'C C', $fileobj->transaction_id, 0 ),
- );
-
- my $offset = 1;
- for my $trans_id ( @transactions ) {
- $fileobj->print_at( $keytag->{offset} + $self->{key_size} * $offset++,
- pack( $self->{long_pack}, 0 ),
- pack( 'C C', $trans_id, 1 ),
- );
- }
- }
+This takes an object that provides _base_offset() and two strings. The
+strings correspond to the old key and new key, respectively. This operation
+is equivalent to (given C<< $db->{foo} = []; >>) C<< $db->{bar} = $db->{foo} >>.
- $self->write_value( $location, $plain_key, $value, $orig_key );
+This returns nothing.
- return 1;
-}
+=cut
-sub write_value {
- my $self = shift;
- my ($location, $key, $value, $orig_key) = @_;
+sub make_reference { die "make_reference must be implemented in a child class" }
- my $fileobj = $self->_fileobj;
+=head2 key_exists( $obj, $key )
- my $dbm_deep_obj = _get_dbm_object( $value );
- if ( $dbm_deep_obj && $dbm_deep_obj->_fileobj ne $fileobj ) {
- $self->_throw_error( "Cannot cross-reference. Use export() instead" );
- }
+This takes an object that provides _base_offset() and a string for
+the key to be checked. This returns 1 for true and "" for false.
- ##
- # Write signature based on content type, set content length and write
- # actual value.
- ##
- my $r = Scalar::Util::reftype( $value ) || '';
- if ( $dbm_deep_obj ) {
- $self->write_tag( $location, SIG_INTERNAL,pack($self->{long_pack}, $dbm_deep_obj->_base_offset) );
- }
- elsif ($r eq 'HASH') {
- if ( !$dbm_deep_obj && tied %{$value} ) {
- $self->_throw_error( "Cannot store something that is tied" );
- }
- $self->write_tag( $location, SIG_HASH, chr(0)x$self->{index_size} );
- }
- elsif ($r eq 'ARRAY') {
- if ( !$dbm_deep_obj && tied @{$value} ) {
- $self->_throw_error( "Cannot store something that is tied" );
- }
- $self->write_tag( $location, SIG_ARRAY, chr(0)x$self->{index_size} );
- }
- elsif (!defined($value)) {
- $self->write_tag( $location, SIG_NULL, '' );
- }
- else {
- $self->write_tag( $location, SIG_DATA, $value );
- }
+=cut
- ##
- # Plain key is stored AFTER value, as keys are typically fetched less often.
- ##
- $fileobj->print_at( undef, pack($self->{data_pack}, length($key)) . $key );
-
- # Internal references don't care about autobless
- return 1 if $dbm_deep_obj;
-
- ##
- # If value is blessed, preserve class name
- ##
- if ( $fileobj->{autobless} ) {
- if ( defined( my $c = Scalar::Util::blessed($value) ) ) {
- $fileobj->print_at( undef, chr(1), pack($self->{data_pack}, length($c)) . $c );
- }
- else {
- $fileobj->print_at( undef, chr(0) );
- }
- }
+sub key_exists { die "key_exists must be implemented in a child class" }
- ##
- # Tie the passed in reference so that changes to it are reflected in the
- # datafile. The use of $location as the base_offset will act as the
- # the linkage between parent and child.
- #
- # The overall assignment is a hack around the fact that just tying doesn't
- # store the values. This may not be the wrong thing to do.
- ##
- if ($r eq 'HASH') {
- my %x = %$value;
- tie %$value, 'DBM::Deep', {
- base_offset => $location,
- fileobj => $fileobj,
- parent => $self->{obj},
- parent_key => $orig_key,
- };
- %$value = %x;
- }
- elsif ($r eq 'ARRAY') {
- my @x = @$value;
- tie @$value, 'DBM::Deep', {
- base_offset => $location,
- fileobj => $fileobj,
- parent => $self->{obj},
- parent_key => $orig_key,
- };
- @$value = @x;
- }
+=head2 delete_key( $obj, $key )
- return 1;
-}
+This takes an object that provides _base_offset() and a string for
+the key to be deleted. This returns the result of the Sector::Reference
+delete_key() method.
-sub split_index {
- my $self = shift;
- my ($tag, $md5, $keyloc) = @_;
+=cut
- my $fileobj = $self->_fileobj;
+sub delete_key { die "delete_key must be implemented in a child class" }
- my $loc = $fileobj->request_space(
- $self->tag_size( $self->{index_size} ),
- );
+=head2 write_value( $obj, $key, $value )
- $fileobj->print_at( $tag->{ref_loc}, pack($self->{long_pack}, $loc) );
+This takes an object that provides _base_offset(), a string for the
+key, and a value. This value can be anything storable within L<DBM::Deep/>.
- my $index_tag = $self->write_tag(
- $loc, SIG_INDEX,
- chr(0)x$self->{index_size},
- );
+This returns 1 upon success.
- my $keys = $tag->{content}
- . $md5 . pack($self->{long_pack}, $keyloc);
+=cut
- my @newloc = ();
- BUCKET:
- # The <= here is deliberate - we have max_buckets+1 keys to iterate
- # through, unlike every other loop that uses max_buckets as a stop.
- for (my $i = 0; $i <= $self->{max_buckets}; $i++) {
- my ($key, $old_subloc) = $self->_get_key_subloc( $keys, $i );
+sub write_value { die "write_value must be implemented in a child class" }
- die "[INTERNAL ERROR]: No key in split_index()\n" unless $key;
- die "[INTERNAL ERROR]: No subloc in split_index()\n" unless $old_subloc;
+=head2 setup( $obj )
- my $num = ord(substr($key, $tag->{ch} + 1, 1));
+This takes an object that provides _base_offset(). It will do everything needed
+in order to properly initialize all values for necessary functioning. If this is
+called upon an already initialized object, this will also reset the inode.
- if ($newloc[$num]) {
- my $subkeys = $fileobj->read_at( $newloc[$num], $self->{bucket_list_size} );
+This returns 1.
- # This is looking for the first empty spot
- my ($subloc, $offset) = $self->_find_in_buckets(
- { content => $subkeys }, '',
- );
+=cut
- $fileobj->print_at(
- $newloc[$num] + $offset,
- $key, pack($self->{long_pack}, $old_subloc),
- );
+sub setup { die "setup must be implemented in a child class" }
- next;
- }
+=head2 begin_work( $obj )
- my $loc = $fileobj->request_space(
- $self->tag_size( $self->{bucket_list_size} ),
- );
+This takes an object that provides _base_offset(). It will set up all necessary
+bookkeeping in order to run all work within a transaction.
- $fileobj->print_at(
- $index_tag->{offset} + ($num * $self->{long_size}),
- pack($self->{long_pack}, $loc),
- );
+If $obj is already within a transaction, an error wiill be thrown. If there are
+no more available transactions, an error will be thrown.
- my $blist_tag = $self->write_tag(
- $loc, SIG_BLIST,
- chr(0)x$self->{bucket_list_size},
- );
+This returns undef.
- $fileobj->print_at( $blist_tag->{offset}, $key . pack($self->{long_pack}, $old_subloc) );
+=cut
- $newloc[$num] = $blist_tag->{offset};
- }
+sub begin_work { die "begin_work must be implemented in a child class" }
- $self->_release_space(
- $self->tag_size( $self->{bucket_list_size} ),
- $tag->{offset} - SIG_SIZE - $self->{data_size},
- );
+=head2 rollback( $obj )
- return 1;
-}
+This takes an object that provides _base_offset(). It will revert all
+actions taken within the running transaction.
-sub read_from_loc {
- my $self = shift;
- my ($subloc, $orig_key) = @_;
-
- my $fileobj = $self->_fileobj;
-
- my $signature = $fileobj->read_at( $subloc, SIG_SIZE );
-
- ##
- # If value is a hash or array, return new DBM::Deep object with correct offset
- ##
- if (($signature eq SIG_HASH) || ($signature eq SIG_ARRAY)) {
- my $new_obj = DBM::Deep->new({
- type => $signature,
- base_offset => $subloc,
- fileobj => $self->_fileobj,
- parent => $self->{obj},
- parent_key => $orig_key,
- });
+If $obj is not within a transaction, an error will be thrown.
- if ($new_obj->_fileobj->{autobless}) {
- ##
- # Skip over value and plain key to see if object needs
- # to be re-blessed
- ##
- $fileobj->increment_pointer( $self->{data_size} + $self->{index_size} );
-
- my $size = $fileobj->read_at( undef, $self->{data_size} );
- $size = unpack($self->{data_pack}, $size);
- if ($size) { $fileobj->increment_pointer( $size ); }
-
- my $bless_bit = $fileobj->read_at( undef, 1 );
- if ( ord($bless_bit) ) {
- my $size = unpack(
- $self->{data_pack},
- $fileobj->read_at( undef, $self->{data_size} ),
- );
-
- if ( $size ) {
- $new_obj = bless $new_obj, $fileobj->read_at( undef, $size );
- }
- }
- }
-
- return $new_obj;
- }
- elsif ( $signature eq SIG_INTERNAL ) {
- my $size = $fileobj->read_at( undef, $self->{data_size} );
- $size = unpack($self->{data_pack}, $size);
-
- if ( $size ) {
- my $new_loc = $fileobj->read_at( undef, $size );
- $new_loc = unpack( $self->{long_pack}, $new_loc );
- return $self->read_from_loc( $new_loc, $orig_key );
- }
- else {
- return;
- }
- }
- ##
- # Otherwise return actual value
- ##
- elsif ( $signature eq SIG_DATA ) {
- my $size = $fileobj->read_at( undef, $self->{data_size} );
- $size = unpack($self->{data_pack}, $size);
-
- my $value = $size ? $fileobj->read_at( undef, $size ) : '';
- return $value;
- }
+This returns 1.
- ##
- # Key exists, but content is null
- ##
- return;
-}
+=cut
-sub get_bucket_value {
- ##
- # Fetch single value given tag and MD5 digested key.
- ##
- my $self = shift;
- my ($tag, $md5, $orig_key) = @_;
+sub rollback { die "rollback must be implemented in a child class" }
- #ACID - This is a read. Can find exact or HEAD
- my ($keyloc, $offset) = $self->_find_in_buckets( $tag, $md5 );
+=head2 commit( $obj )
- if ( !$keyloc ) {
- #XXX Need to use real key
-# $self->add_bucket( $tag, $md5, $orig_key, undef, $orig_key );
-# return;
- }
-# elsif ( !$is_deleted ) {
- else {
- my $keytag = $self->load_tag( $keyloc );
- my ($subloc, $is_deleted) = $self->find_keyloc( $keytag );
- if (!$subloc && !$is_deleted) {
- ($subloc, $is_deleted) = $self->find_keyloc( $keytag, 0 );
- }
- if ( $subloc && !$is_deleted ) {
- return $self->read_from_loc( $subloc, $orig_key );
- }
- }
+This takes an object that provides _base_offset(). It will apply all
+actions taken within the transaction to the HEAD.
- return;
-}
+If $obj is not within a transaction, an error will be thrown.
-sub delete_bucket {
- ##
- # Delete single key/value pair given tag and MD5 digested key.
- ##
- my $self = shift;
- my ($tag, $md5, $orig_key) = @_;
+This returns 1.
- #ACID - Although this is a mutation, we must find any transaction.
- # This is because we need to mark something as deleted that is in the HEAD.
- my ($keyloc, $offset) = $self->_find_in_buckets( $tag, $md5 );
+=cut
- return if !$keyloc;
+sub commit { die "commit must be implemented in a child class" }
- my $fileobj = $self->_fileobj;
+=head2 get_next_key( $obj, $prev_key )
- my @transactions;
- if ( $fileobj->transaction_id == 0 ) {
- @transactions = $fileobj->current_transactions;
- }
+This takes an object that provides _base_offset() and an optional string
+representing the prior key returned via a prior invocation of this method.
- if ( $fileobj->transaction_id == 0 ) {
- my $keytag = $self->load_tag( $keyloc );
-
- my ($subloc, $is_deleted, $offset) = $self->find_keyloc( $keytag );
- return if !$subloc || $is_deleted;
-
- my $value = $self->read_from_loc( $subloc, $orig_key );
-
- my $size = $self->_length_needed( $value, $orig_key );
-
- for my $trans_id ( @transactions ) {
- my ($loc, $is_deleted, $offset2) = $self->find_keyloc( $keytag, $trans_id );
- unless ($loc) {
- my $location2 = $fileobj->request_space( $size );
- $fileobj->print_at( $keytag->{offset} + $offset2,
- pack($self->{long_pack}, $location2 ),
- pack( 'C C', $trans_id, 0 ),
- );
- $self->write_value( $location2, $orig_key, $value, $orig_key );
- }
- }
-
- $keytag = $self->load_tag( $keyloc );
- ($subloc, $is_deleted, $offset) = $self->find_keyloc( $keytag );
- $fileobj->print_at( $keytag->{offset} + $offset,
- substr( $keytag->{content}, $offset + $self->{key_size} ),
- chr(0) x $self->{key_size},
- );
- }
- else {
- my $keytag = $self->load_tag( $keyloc );
+This method delegates to C<< DBM::Deep::Iterator->get_next_key() >>.
- my ($subloc, $is_deleted, $offset) = $self->find_keyloc( $keytag );
+=cut
- $fileobj->print_at( $keytag->{offset} + $offset,
- pack($self->{long_pack}, 0 ),
- pack( 'C C', $fileobj->transaction_id, 1 ),
- );
- }
-
- return 1;
-}
-
-sub bucket_exists {
- ##
- # Check existence of single key given tag and MD5 digested key.
- ##
+# XXX Add staleness here
+sub get_next_key {
my $self = shift;
- my ($tag, $md5) = @_;
-
- #ACID - This is a read. Can find exact or HEAD
- my ($keyloc) = $self->_find_in_buckets( $tag, $md5 );
- my $keytag = $self->load_tag( $keyloc );
- my ($subloc, $is_deleted, $offset) = $self->find_keyloc( $keytag );
- if ( !$subloc && !$is_deleted ) {
- ($subloc, $is_deleted, $offset) = $self->find_keyloc( $keytag, 0 );
+ my ($obj, $prev_key) = @_;
+
+ # XXX Need to add logic about resetting the iterator if any key in the
+ # reference has changed
+ unless ( $prev_key ) {
+ $obj->{iterator} = $self->iterator_class->new({
+ base_offset => $obj->_base_offset,
+ engine => $self,
+ });
}
- return ($subloc && !$is_deleted) && 1;
-}
-sub find_blist {
- ##
- # Locate offset for bucket list, given digested key
- ##
- my $self = shift;
- my ($offset, $md5, $args) = @_;
- $args = {} unless $args;
-
- ##
- # Locate offset for bucket list using digest index system
- ##
- my $tag = $self->load_tag( $offset )
- or $self->_throw_error( "INTERNAL ERROR - Cannot find tag" );
-
- my $ch = 0;
- while ($tag->{signature} ne SIG_BLIST) {
- my $num = ord substr($md5, $ch, 1);
-
- my $ref_loc = $tag->{offset} + ($num * $self->{long_size});
- $tag = $self->index_lookup( $tag, $num );
-
- if (!$tag) {
- return if !$args->{create};
-
- my $loc = $self->_fileobj->request_space(
- $self->tag_size( $self->{bucket_list_size} ),
- );
-
- $self->_fileobj->print_at( $ref_loc, pack($self->{long_pack}, $loc) );
+ return $obj->{iterator}->get_next_key( $obj );
+}
- $tag = $self->write_tag(
- $loc, SIG_BLIST,
- chr(0)x$self->{bucket_list_size},
- );
+=head2 lock_exclusive()
- $tag->{ref_loc} = $ref_loc;
- $tag->{ch} = $ch;
+This takes an object that provides _base_offset(). It will guarantee that
+the storage has taken precautions to be safe for a write.
- last;
- }
+This returns nothing.
- $tag->{ch} = $ch++;
- $tag->{ref_loc} = $ref_loc;
- }
+=cut
- return $tag;
+sub lock_exclusive {
+ my $self = shift;
+ my ($obj) = @_;
+ return $self->storage->lock_exclusive( $obj );
}
-sub index_lookup {
- ##
- # Given index tag, lookup single entry in index and return .
- ##
- my $self = shift;
- my ($tag, $index) = @_;
+=head2 lock_shared()
- my $location = unpack(
- $self->{long_pack},
- substr(
- $tag->{content},
- $index * $self->{long_size},
- $self->{long_size},
- ),
- );
+This takes an object that provides _base_offset(). It will guarantee that
+the storage has taken precautions to be safe for a read.
- if (!$location) { return; }
+This returns nothing.
- return $self->load_tag( $location );
-}
+=cut
-sub traverse_index {
- ##
- # Scan index and recursively step into deeper levels, looking for next key.
- ##
+sub lock_shared {
my $self = shift;
- my ($xxxx, $offset, $ch, $force_return_next) = @_;
-
- my $tag = $self->load_tag( $offset );
-
- if ($tag->{signature} ne SIG_BLIST) {
- my $start = $xxxx->{return_next} ? 0 : ord(substr($xxxx->{prev_md5}, $ch, 1));
-
- for (my $idx = $start; $idx < $self->{hash_chars_used}; $idx++) {
- my $subloc = unpack(
- $self->{long_pack},
- substr(
- $tag->{content},
- $idx * $self->{long_size},
- $self->{long_size},
- ),
- );
+ my ($obj) = @_;
+ return $self->storage->lock_shared( $obj );
+}
- if ($subloc) {
- my $result = $self->traverse_index(
- $xxxx, $subloc, $ch + 1, $force_return_next,
- );
+=head2 unlock()
- if (defined $result) { return $result; }
- }
- } # index loop
+This takes an object that provides _base_offset(). It will guarantee that
+the storage has released the most recently-taken lock.
- $xxxx->{return_next} = 1;
- }
- # This is the bucket list
- else {
- my $keys = $tag->{content};
- if ($force_return_next) { $xxxx->{return_next} = 1; }
-
- ##
- # Iterate through buckets, looking for a key match
- ##
- my $transaction_id = $self->_fileobj->transaction_id;
- for (my $i = 0; $i < $self->{max_buckets}; $i++) {
- my ($key, $keyloc) = $self->_get_key_subloc( $keys, $i );
-
- # End of bucket list -- return to outer loop
- if (!$keyloc) {
- $xxxx->{return_next} = 1;
- last;
- }
- # Located previous key -- return next one found
- elsif ($key eq $xxxx->{prev_md5}) {
- $xxxx->{return_next} = 1;
- next;
- }
- # Seek to bucket location and skip over signature
- elsif ($xxxx->{return_next}) {
- my $fileobj = $self->_fileobj;
-
- my $keytag = $self->load_tag( $keyloc );
- my ($subloc, $is_deleted) = $self->find_keyloc( $keytag );
- if ( $subloc == 0 && !$is_deleted ) {
- ($subloc, $is_deleted) = $self->find_keyloc( $keytag, 0 );
- }
- next if $is_deleted;
-
- # Skip over value to get to plain key
- my $sig = $fileobj->read_at( $subloc, SIG_SIZE );
-
- my $size = $fileobj->read_at( undef, $self->{data_size} );
- $size = unpack($self->{data_pack}, $size);
- if ($size) { $fileobj->increment_pointer( $size ); }
-
- # Read in plain key and return as scalar
- $size = $fileobj->read_at( undef, $self->{data_size} );
- $size = unpack($self->{data_pack}, $size);
-
- my $plain_key;
- if ($size) { $plain_key = $fileobj->read_at( undef, $size); }
- return $plain_key;
- }
- }
-
- $xxxx->{return_next} = 1;
- }
+This returns nothing.
- return;
-}
+=cut
-sub get_next_key {
- ##
- # Locate next key, given digested previous one
- ##
+sub unlock {
my $self = shift;
my ($obj) = @_;
- ##
- # If the previous key was not specifed, start at the top and
- # return the first one found.
- ##
- my $temp;
- if ( @_ > 1 ) {
- $temp = {
- prev_md5 => $_[1],
- return_next => 0,
- };
- }
- else {
- $temp = {
- prev_md5 => chr(0) x $self->{hash_size},
- return_next => 1,
- };
- }
-
- return $self->traverse_index( $temp, $obj->_base_offset, 0 );
-}
+ my $rv = $self->storage->unlock( $obj );
-# Utilities
+ $self->flush if $rv;
-sub _get_key_subloc {
- my $self = shift;
- my ($keys, $idx) = @_;
-
- return unpack(
- # This is 'a', not 'A'. Please read the pack() documentation for the
- # difference between the two and why it's important.
- "a$self->{hash_size} $self->{long_pack}",
- substr(
- $keys,
- ($idx * $self->{bucket_size}),
- $self->{bucket_size},
- ),
- );
+ return $rv;
}
-sub _find_in_buckets {
- my $self = shift;
- my ($tag, $md5) = @_;
+=head1 INTERNAL METHODS
- BUCKET:
- for ( my $i = 0; $i < $self->{max_buckets}; $i++ ) {
- my ($key, $subloc) = $self->_get_key_subloc(
- $tag->{content}, $i,
- );
+The following methods are internal-use-only to DBM::Deep::Engine and its
+child classes.
- my @rv = ($subloc, $i * $self->{bucket_size});
+=cut
- unless ( $subloc ) {
- return @rv;
- }
+=head2 flush()
- next BUCKET if $key ne $md5;
+This takes no arguments. It will do everything necessary to flush all things to
+disk. This is usually called during unlock() and setup().
- return @rv;
- }
+This returns nothing.
- return;
-}
+=cut
-sub _release_space {
+sub flush {
my $self = shift;
- my ($size, $loc) = @_;
- my $next_loc = 0;
-
- $self->_fileobj->print_at( $loc,
- SIG_FREE,
- pack($self->{long_pack}, $size ),
- pack($self->{long_pack}, $next_loc ),
- );
+ # Why do we need to have the storage flush? Shouldn't autoflush take care of
+ # things? -RobK, 2008-06-26
+ $self->storage->flush;
return;
}
-sub _throw_error {
- die "DBM::Deep: $_[1]\n";
-}
+=head2 load_sector( $loc )
-sub _get_dbm_object {
- my $item = shift;
-
- my $obj = eval {
- local $SIG{__DIE__};
- if ($item->isa( 'DBM::Deep' )) {
- return $item;
- }
- return;
- };
- return $obj if $obj;
-
- my $r = Scalar::Util::reftype( $item ) || '';
- if ( $r eq 'HASH' ) {
- my $obj = eval {
- local $SIG{__DIE__};
- my $obj = tied(%$item);
- if ($obj->isa( 'DBM::Deep' )) {
- return $obj;
- }
- return;
- };
- return $obj if $obj;
- }
- elsif ( $r eq 'ARRAY' ) {
- my $obj = eval {
- local $SIG{__DIE__};
- my $obj = tied(@$item);
- if ($obj->isa( 'DBM::Deep' )) {
- return $obj;
- }
- return;
- };
- return $obj if $obj;
- }
+This takes an id/location/offset and loads the sector based on the engine's
+defined sector type.
- return;
-}
+=cut
-sub _length_needed {
- my $self = shift;
- my ($value, $key) = @_;
+sub load_sector { $_[0]->sector_type->load( @_ ) }
- my $is_dbm_deep = eval {
- local $SIG{'__DIE__'};
- $value->isa( 'DBM::Deep' );
- };
+=head2 ACCESSORS
- my $len = SIG_SIZE
- + $self->{data_size} # size for value
- + $self->{data_size} # size for key
- + length( $key ); # length of key
+The following are readonly attributes.
- if ( $is_dbm_deep && $value->_fileobj eq $self->_fileobj ) {
- # long_size is for the internal reference
- return $len + $self->{long_size};
- }
+=over 4
- if ( $self->_fileobj->{autobless} ) {
- # This is for the bit saying whether or not this thing is blessed.
- $len += 1;
- }
+=item * storage
- my $r = Scalar::Util::reftype( $value ) || '';
- unless ( $r eq 'HASH' || $r eq 'ARRAY' ) {
- if ( defined $value ) {
- $len += length( $value );
- }
- return $len;
- }
+=back
- $len += $self->{index_size};
+=cut
- # if autobless is enabled, must also take into consideration
- # the class name as it is stored after the key.
- if ( $self->_fileobj->{autobless} ) {
- my $c = Scalar::Util::blessed($value);
- if ( defined $c && !$is_dbm_deep ) {
- $len += $self->{data_size} + length($c);
- }
- }
+sub storage { $_[0]{storage} }
- return $len;
-}
+sub sector_type { die "sector_type must be implemented in a child class" }
1;
__END__