X-Git-Url: http://git.shadowcat.co.uk/gitweb/gitweb.cgi?a=blobdiff_plain;f=lib%2FDBM%2FDeep%2FEngine.pm;h=309c2748fc8ce48717626ff35dbe57cf1306ae86;hb=4768a58079c3e2e519f4693cf0c78ed6dee5ca90;hp=7c8d13949939c6bbe922b98c0816b99cf9c71e35;hpb=a21f2d90935286a81dbaa6299707e140060e52d3;p=dbsrgits%2FDBM-Deep.git diff --git a/lib/DBM/Deep/Engine.pm b/lib/DBM/Deep/Engine.pm index 7c8d139..309c274 100644 --- a/lib/DBM/Deep/Engine.pm +++ b/lib/DBM/Deep/Engine.pm @@ -1,62 +1,34 @@ package DBM::Deep::Engine; +use 5.6.0; + use strict; +use warnings; use Fcntl qw( :DEFAULT :flock :seek ); - -sub precalc_sizes { - ## - # Precalculate index, bucket and bucket list sizes - ## - my $self = shift; - - $self->{index_size} = (2**8) * $self->{long_size}; - $self->{bucket_size} = $self->{hash_size} + $self->{long_size}; - $self->{bucket_list_size} = $self->{max_buckets} * $self->{bucket_size}; - - return 1; -} - -sub set_pack { - ## - # Set pack/unpack modes (see file header for more) - ## - my $self = shift; - my ($long_s, $long_p, $data_s, $data_p) = @_; - - ## - # Set to 4 and 'N' for 32-bit offset tags (default). Theoretical limit of 4 GB per file. - # (Perl must be compiled with largefile support for files > 2 GB) - # - # Set to 8 and 'Q' for 64-bit offsets. Theoretical limit of 16 XB per file. - # (Perl must be compiled with largefile and 64-bit long support) - ## - $self->{long_size} = $long_s ? $long_s : 4; - $self->{long_pack} = $long_p ? $long_p : 'N'; - - ## - # Set to 4 and 'N' for 32-bit data length prefixes. Limit of 4 GB for each key/value. - # Upgrading this is possible (see above) but probably not necessary. If you need - # more than 4 GB for a single key or value, this module is really not for you :-) - ## - $self->{data_size} = $data_s ? $data_s : 4; - $self->{data_pack} = $data_p ? $data_p : 'N'; - - return $self->precalc_sizes(); -} - -sub set_digest { - ## - # Set key digest function (default is MD5) - ## - my $self = shift; - my ($digest_func, $hash_size) = @_; - - $self->{digest} = $digest_func ? $digest_func : \&Digest::MD5::md5; - $self->{hash_size} = $hash_size ? $hash_size : 16; - - return $self->precalc_sizes(); -} +use Scalar::Util (); + +# File-wide notes: +# * All the local($/,$\); are to protect read() and print() from -l. +# * 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_SIZE () { 1 } sub new { my $class = shift; @@ -72,155 +44,234 @@ sub new { hash_size => 16, ## - # Maximum number of buckets per list 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. + # Maximum number of buckets per list 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; - $self->precalc_sizes; + 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"; + } + } + + # 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}; + + 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; + } return $self; } -sub setup_fh { +sub _fileobj { return $_[0]{fileobj} } +sub _fh { return $_[0]->_fileobj->{fh} } + +sub calculate_sizes { my $self = shift; - my ($obj) = @_; - $self->open( $obj ) if !defined $obj->_fh; + #XXX Does this need to be updated with different hashing algorithms? + $self->{index_size} = (2**8) * $self->{long_size}; + $self->{bucket_size} = $self->{hash_size} + $self->{long_size} * 3; + $self->{bucket_list_size} = $self->{max_buckets} * $self->{bucket_size}; - #XXX Is {end} busted? Can't we just seek( $fh, 0, SEEK_END ) ? - unless ( $obj->_root->{inode} ) { - my @stats = stat($obj->_fh); - $obj->_root->{inode} = $stats[1]; - $obj->_root->{end} = $stats[7]; - } + return; +} - return 1; +sub write_file_header { + my $self = shift; + + local($/,$\); + + my $fh = $self->_fh; + + my $loc = $self->_request_space( length( SIG_FILE ) + 21 ); + seek($fh, $loc + $self->_fileobj->{file_offset}, SEEK_SET); + print( $fh + SIG_FILE, + SIG_HEADER, + pack('N', 1), # header version + pack('N', 12), # header size + pack('N', 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}), + ); + + $self->_fileobj->set_transaction_offset( 13 ); + + return; } -sub open { - ## - # Open a fh to the database, create if nonexistent. - # Make sure file signature matches DBM::Deep spec. - ## +sub read_file_header { my $self = shift; - my ($obj) = @_; - if (defined($obj->_fh)) { $self->close_fh( $obj ); } + local($/,$\); - eval { - local $SIG{'__DIE__'}; - # Theoretically, adding O_BINARY should remove the need for the binmode - # Of course, testing it is going to be ... interesting. - my $flags = O_RDWR | O_CREAT | O_BINARY; + my $fh = $self->_fh; + + seek($fh, 0 + $self->_fileobj->{file_offset}, SEEK_SET); + my $buffer; + my $bytes_read = read( $fh, $buffer, length(SIG_FILE) + 9 ); + + return unless $bytes_read; + + my ($file_signature, $sig_header, $header_version, $size) = unpack( + 'A4 A N N', $buffer + ); + + unless ( $file_signature eq SIG_FILE ) { + $self->_fileobj->close; + $self->_throw_error( "Signature not found -- file is not a Deep DB" ); + } - my $fh; - sysopen( $fh, $obj->_root->{file}, $flags ) - or $fh = undef; - $obj->_root->{fh} = $fh; - }; if ($@ ) { $obj->_throw_error( "Received error: $@\n" ); } - if (! defined($obj->_fh)) { - return $obj->_throw_error("Cannot sysopen file: " . $obj->_root->{file} . ": $!"); + unless ( $sig_header eq SIG_HEADER ) { + $self->_fileobj->close; + $self->_throw_error( "Old file version found." ); } - my $fh = $obj->_fh; + my $buffer2; + $bytes_read += read( $fh, $buffer2, $size ); + my ($running_transactions, @values) = unpack( 'N n A n A n', $buffer2 ); - #XXX Can we remove this by using the right sysopen() flags? - # Maybe ... q.v. above - binmode $fh; # for win32 + $self->_fileobj->set_transaction_offset( 13 ); - if ($obj->_root->{autoflush}) { - my $old = select $fh; - $|=1; - select $old; + if ( @values < 5 || grep { !defined } @values ) { + $self->_fileobj->close; + $self->_throw_error("Corrupted file - bad header"); } - seek($fh, 0 + $obj->_root->{file_offset}, SEEK_SET); + #XXX Add warnings if values weren't set right + @{$self}{qw(long_size long_pack data_size data_pack max_buckets)} = @values; - my $signature; - my $bytes_read = read( $fh, $signature, length(DBM::Deep->SIG_FILE)); + return $bytes_read; +} - ## - # File is empty -- write signature and master index - ## - if (!$bytes_read) { - seek($fh, 0 + $obj->_root->{file_offset}, SEEK_SET); - print( $fh DBM::Deep->SIG_FILE); +sub setup_fh { + my $self = shift; + my ($obj) = @_; - $self->create_tag($obj, $obj->_base_offset, $obj->_type, chr(0) x $self->{index_size}); + local($/,$\); - my $plain_key = "[base]"; - print( $fh pack($self->{data_pack}, length($plain_key)) . $plain_key ); + my $fh = $self->_fh; + flock $fh, LOCK_EX; - # Flush the filehandle - my $old_fh = select $fh; - my $old_af = $|; $| = 1; $| = $old_af; - select $old_fh; + #XXX The duplication of calculate_sizes needs to go away + unless ( $obj->{base_offset} ) { + my $bytes_read = $self->read_file_header; - return 1; - } + $self->calculate_sizes; - ## - # Check signature was valid - ## - unless ($signature eq DBM::Deep->SIG_FILE) { - $self->close_fh( $obj ); - return $obj->_throw_error("Signature not found -- file is not a Deep DB"); - } + ## + # File is empty -- write header and master index + ## + if (!$bytes_read) { + $self->_fileobj->audit( "# Database created on" ); - ## - # Get our type from master index signature - ## - my $tag = $self->load_tag($obj, $obj->_base_offset); - if (!$tag) { - return $obj->_throw_error("Corrupted file, no master index record"); + $self->write_file_header; + + $obj->{base_offset} = $self->_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 { + $self->calculate_sizes; } - if ($obj->{type} ne $tag->{signature}) { - return $obj->_throw_error("File type mismatch"); + #XXX We have to make sure we don't mess up when autoflush isn't turned on + unless ( $self->_fileobj->{inode} ) { + my @stats = stat($fh); + $self->_fileobj->{inode} = $stats[1]; + $self->_fileobj->{end} = $stats[7]; } -#XXX We probably also want to store the hash algorithm name and not assume anything -#XXX The cool thing would be to allow a different hashing algorithm at every level + flock $fh, LOCK_UN; return 1; } -sub close_fh { +sub tag_size { my $self = shift; - my ($obj) = @_; - - if ( my $fh = $obj->_root->{fh} ) { - close $fh; - } - $obj->_root->{fh} = undef; - - return 1; + my ($size) = @_; + return SIG_SIZE + $self->{data_size} + $size; } -sub create_tag { +sub write_tag { ## # Given offset, signature and content, create tag and write to disk ## my $self = shift; - my ($obj, $offset, $sig, $content) = @_; - my $size = length($content); + my ($offset, $sig, $content) = @_; + my $size = length( $content ); - my $fh = $obj->_fh; + local($/,$\); - seek($fh, $offset + $obj->_root->{file_offset}, SEEK_SET); - print( $fh $sig . pack($self->{data_pack}, $size) . $content ); + my $fh = $self->_fh; - if ($offset == $obj->_root->{end}) { - $obj->_root->{end} += DBM::Deep->SIG_SIZE + $self->{data_size} + $size; + if ( defined $offset ) { + seek($fh, $offset + $self->_fileobj->{file_offset}, SEEK_SET); } + print( $fh $sig . pack($self->{data_pack}, $size) . $content ); + + return unless defined $offset; + return { signature => $sig, size => $size, - offset => $offset + DBM::Deep->SIG_SIZE + $self->{data_size}, + offset => $offset + SIG_SIZE + $self->{data_size}, content => $content }; } @@ -230,15 +281,21 @@ sub load_tag { # Given offset, load single tag and return signature, size and data ## my $self = shift; - my ($obj, $offset) = @_; + my ($offset) = @_; + + local($/,$\); - my $fh = $obj->_fh; +# print join(':',map{$_||''}caller(1)), $/; - seek($fh, $offset + $obj->_root->{file_offset}, SEEK_SET); - if (eof $fh) { return undef; } + my $fh = $self->_fh; + + seek($fh, $offset + $self->_fileobj->{file_offset}, SEEK_SET); + + #XXX I'm not sure this check will work if autoflush isn't enabled ... + return if eof $fh; my $b; - read( $fh, $b, DBM::Deep->SIG_SIZE + $self->{data_size} ); + read( $fh, $b, SIG_SIZE + $self->{data_size} ); my ($sig, $size) = unpack( "A $self->{data_pack}", $b ); my $buffer; @@ -247,391 +304,487 @@ sub load_tag { return { signature => $sig, size => $size, - offset => $offset + DBM::Deep->SIG_SIZE + $self->{data_size}, + offset => $offset + SIG_SIZE + $self->{data_size}, content => $buffer }; } +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; + } + + return; +} + +sub _length_needed { + my $self = shift; + my ($value, $key) = @_; + + my $is_dbm_deep = eval { + local $SIG{'__DIE__'}; + $value->isa( 'DBM::Deep' ); + }; + + my $len = SIG_SIZE + $self->{data_size} + + $self->{data_size} + length( $key ); + + if ( $is_dbm_deep && $value->_fileobj eq $self->_fileobj ) { + return $len + $self->{long_size}; + } + + my $r = Scalar::Util::reftype( $value ) || ''; + if ( $self->_fileobj->{autobless} ) { + # This is for the bit saying whether or not this thing is blessed. + $len += 1; + } + + unless ( $r eq 'HASH' || $r eq 'ARRAY' ) { + if ( defined $value ) { + $len += length( $value ); + } + return $len; + } + + $len += $self->{index_size}; + + # 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); + } + } + + return $len; +} + sub add_bucket { ## # Adds one key/value pair to bucket list, given offset, MD5 digest of key, # plain (undigested) key and value. ## my $self = shift; - my ($obj, $tag, $md5, $plain_key, $value) = @_; - my $keys = $tag->{content}; + my ($tag, $md5, $plain_key, $value, $deleted, $orig_key) = @_; + $deleted ||= 0; + + local($/,$\); + + # This verifies that only supported values will be stored. + { + my $r = Scalar::Util::reftype( $value ); + last if !defined $r; + + last if $r eq 'HASH'; + last if $r eq 'ARRAY'; + + $self->_throw_error( + "Storage of variables of type '$r' is not supported." + ); + } + my $location = 0; my $result = 2; - my $root = $obj->_root; + my $root = $self->_fileobj; + my $fh = $self->_fh; - my $is_dbm_deep = eval { local $SIG{'__DIE__'}; $value->isa( 'DBM::Deep' ) }; - my $internal_ref = $is_dbm_deep && ($value->_root eq $root); + my $actual_length = $self->_length_needed( $value, $plain_key ); - my $fh = $obj->_fh; + #ACID - This is a mutation. Must only find the exact transaction + my ($subloc, $offset, $size,$is_deleted) = $self->_find_in_buckets( $tag, $md5, 1 ); - ## - # Iterate through buckets, seeing if this is a new entry or a replace. - ## - for (my $i = 0; $i < $self->{max_buckets}; $i++) { - my $subloc = unpack($self->{long_pack}, substr($keys, ($i * $self->{bucket_size}) + $self->{hash_size}, $self->{long_size})); - if (!$subloc) { - ## - # Found empty bucket (end of list). Populate and exit loop. - ## - $result = 2; + my @transactions; + if ( $self->_fileobj->transaction_id == 0 ) { + @transactions = $self->_fileobj->current_transactions; + } - $location = $internal_ref - ? $value->_base_offset - : $root->{end}; +# $self->_release_space( $size, $subloc ); + # Updating a known md5 +#XXX This needs updating to use _release_space + if ( $subloc ) { + $result = 1; - seek($fh, $tag->{offset} + ($i * $self->{bucket_size}) + $root->{file_offset}, SEEK_SET); - print( $fh $md5 . pack($self->{long_pack}, $location) ); - last; + if ($actual_length <= $size) { + $location = $subloc; } + else { + $location = $self->_request_space( $actual_length ); + seek( + $fh, + $tag->{offset} + $offset + + $self->{hash_size} + $root->{file_offset}, + SEEK_SET, + ); + print( $fh pack($self->{long_pack}, $location ) ); + print( $fh pack($self->{long_pack}, $actual_length ) ); + print( $fh pack('n n', $root->transaction_id, $deleted ) ); + } + } + # Adding a new md5 + elsif ( defined $offset ) { + $location = $self->_request_space( $actual_length ); + + seek( $fh, $tag->{offset} + $offset + $root->{file_offset}, SEEK_SET ); + print( $fh $md5 . pack($self->{long_pack}, $location ) ); + print( $fh pack($self->{long_pack}, $actual_length ) ); + print( $fh pack('n n', $root->transaction_id, $deleted ) ); + + for ( @transactions ) { + my $tag2 = $self->load_tag( $tag->{offset} - SIG_SIZE - $self->{data_size} ); + $self->_fileobj->{transaction_id} = $_; + $self->add_bucket( $tag2, $md5, '', '', 1, $orig_key ); + $self->_fileobj->{transaction_id} = 0; + } + } + # If bucket didn't fit into list, split into a new index level + # split_index() will do the _request_space() call + else { + $location = $self->split_index( $md5, $tag ); + } - my $key = substr($keys, $i * $self->{bucket_size}, $self->{hash_size}); - if ($md5 eq $key) { - ## - # Found existing bucket with same key. Replace with new value. - ## - $result = 1; + $self->write_value( $location, $plain_key, $value, $orig_key ); - if ($internal_ref) { - $location = $value->_base_offset; - seek($fh, $tag->{offset} + ($i * $self->{bucket_size}) + $root->{file_offset}, SEEK_SET); - print( $fh $md5 . pack($self->{long_pack}, $location) ); - return $result; - } + return $result; +} - seek($fh, $subloc + DBM::Deep->SIG_SIZE + $root->{file_offset}, SEEK_SET); - my $size; - read( $fh, $size, $self->{data_size}); $size = unpack($self->{data_pack}, $size); +sub write_value { + my $self = shift; + my ($location, $key, $value, $orig_key) = @_; - ## - # If value is a hash, array, or raw value with equal or less size, we can - # reuse the same content area of the database. Otherwise, we have to create - # a new content area at the EOF. - ## - my $actual_length; - my $r = Scalar::Util::reftype( $value ) || ''; - if ( $r eq 'HASH' || $r eq 'ARRAY' ) { - $actual_length = $self->{index_size}; - - # if autobless is enabled, must also take into consideration - # the class name, as it is stored along with key/value. - if ( $root->{autobless} ) { - my $value_class = Scalar::Util::blessed($value); - if ( defined $value_class && !$value->isa('DBM::Deep') ) { - $actual_length += length($value_class); - } - } - } - else { $actual_length = length($value); } + local($/,$\); - if ($actual_length <= $size) { - $location = $subloc; - } - else { - $location = $root->{end}; - seek($fh, $tag->{offset} + ($i * $self->{bucket_size}) + $self->{hash_size} + $root->{file_offset}, SEEK_SET); - print( $fh pack($self->{long_pack}, $location) ); - } + my $fh = $self->_fh; + my $root = $self->_fileobj; - last; - } + my $dbm_deep_obj = _get_dbm_object( $value ); + if ( $dbm_deep_obj && $dbm_deep_obj->_fileobj ne $self->_fileobj ) { + $self->_throw_error( "Cannot cross-reference. Use export() instead" ); } + seek($fh, $location + $root->{file_offset}, SEEK_SET); + ## - # If this is an internal reference, return now. - # No need to write value or plain key + # Write signature based on content type, set content length and write + # actual value. ## - if ($internal_ref) { - return $result; + my $r = Scalar::Util::reftype( $value ) || ''; + if ( $dbm_deep_obj ) { + $self->write_tag( undef, SIG_INTERNAL,pack($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" ); + } + $self->write_tag( undef, SIG_HASH, chr(0)x$self->{index_size} ); + } + elsif ($r eq 'ARRAY') { + if ( !$dbm_deep_obj && tied @{$value} ) { + $self->_throw_error( "Cannot store something that is tied" ); + } + $self->write_tag( undef, SIG_ARRAY, chr(0)x$self->{index_size} ); + } + elsif (!defined($value)) { + $self->write_tag( undef, SIG_NULL, '' ); + } + else { + $self->write_tag( undef, SIG_DATA, $value ); } ## - # If bucket didn't fit into list, split into a new index level + # Plain key is stored AFTER value, as keys are typically fetched less often. ## - if (!$location) { - seek($fh, $tag->{ref_loc} + $root->{file_offset}, SEEK_SET); - print( $fh pack($self->{long_pack}, $root->{end}) ); - - my $index_tag = $self->create_tag($obj, $root->{end}, DBM::Deep->SIG_INDEX, chr(0) x $self->{index_size}); - my @offsets = (); - - $keys .= $md5 . pack($self->{long_pack}, 0); - - for (my $i=0; $i<=$self->{max_buckets}; $i++) { - my $key = substr($keys, $i * $self->{bucket_size}, $self->{hash_size}); - if ($key) { - my $old_subloc = unpack($self->{long_pack}, substr($keys, ($i * $self->{bucket_size}) + - $self->{hash_size}, $self->{long_size})); - my $num = ord(substr($key, $tag->{ch} + 1, 1)); - - if ($offsets[$num]) { - my $offset = $offsets[$num] + DBM::Deep->SIG_SIZE + $self->{data_size}; - seek($fh, $offset + $root->{file_offset}, SEEK_SET); - my $subkeys; - read( $fh, $subkeys, $self->{bucket_list_size}); - - for (my $k=0; $k<$self->{max_buckets}; $k++) { - my $subloc = unpack($self->{long_pack}, substr($subkeys, ($k * $self->{bucket_size}) + - $self->{hash_size}, $self->{long_size})); - if (!$subloc) { - seek($fh, $offset + ($k * $self->{bucket_size}) + $root->{file_offset}, SEEK_SET); - print( $fh $key . pack($self->{long_pack}, $old_subloc || $root->{end}) ); - last; - } - } # k loop - } - else { - $offsets[$num] = $root->{end}; - seek($fh, $index_tag->{offset} + ($num * $self->{long_size}) + $root->{file_offset}, SEEK_SET); - print( $fh pack($self->{long_pack}, $root->{end}) ); - - my $blist_tag = $self->create_tag($obj, $root->{end}, DBM::Deep->SIG_BLIST, chr(0) x $self->{bucket_list_size}); - - seek($fh, $blist_tag->{offset} + $root->{file_offset}, SEEK_SET); - print( $fh $key . pack($self->{long_pack}, $old_subloc || $root->{end}) ); - } - } # key is real - } # i loop - - $location ||= $root->{end}; - } # re-index bucket list - - ## - # Seek to content area and store signature, value and plaintext key - ## - if ($location) { - my $content_length; - seek($fh, $location + $root->{file_offset}, SEEK_SET); + print( $fh pack($self->{data_pack}, length($key)) . $key ); - ## - # Write signature based on content type, set content length and write actual value. - ## - my $r = Scalar::Util::reftype($value) || ''; - if ($r eq 'HASH') { - print( $fh DBM::Deep->TYPE_HASH ); - print( $fh pack($self->{data_pack}, $self->{index_size}) . chr(0) x $self->{index_size} ); - $content_length = $self->{index_size}; - } - elsif ($r eq 'ARRAY') { - print( $fh DBM::Deep->TYPE_ARRAY ); - print( $fh pack($self->{data_pack}, $self->{index_size}) . chr(0) x $self->{index_size} ); - $content_length = $self->{index_size}; - } - elsif (!defined($value)) { - print( $fh DBM::Deep->SIG_NULL ); - print( $fh pack($self->{data_pack}, 0) ); - $content_length = 0; + # Internal references don't care about autobless + return 1 if $dbm_deep_obj; + + ## + # If value is blessed, preserve class name + ## + if ( $root->{autobless} ) { + my $c = Scalar::Util::blessed($value); + if ( defined $c && !$dbm_deep_obj ) { + print( $fh chr(1) ); + print( $fh pack($self->{data_pack}, length($c)) . $c ); } else { - print( $fh DBM::Deep->SIG_DATA ); - print( $fh pack($self->{data_pack}, length($value)) . $value ); - $content_length = length($value); + print( $fh chr(0) ); } + } - ## - # Plain key is stored AFTER value, as keys are typically fetched less often. - ## - print( $fh pack($self->{data_pack}, length($plain_key)) . $plain_key ); + ## + # 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 => $root, + parent => $self->{obj}, + parent_key => $orig_key, + }; + %$value = %x; + } + elsif ($r eq 'ARRAY') { + my @x = @$value; + tie @$value, 'DBM::Deep', { + base_offset => $location, + fileobj => $root, + parent => $self->{obj}, + parent_key => $orig_key, + }; + @$value = @x; + } - ## - # If value is blessed, preserve class name - ## - if ( $root->{autobless} ) { - my $value_class = Scalar::Util::blessed($value); - if ( defined $value_class && $value_class ne 'DBM::Deep' ) { - ## - # Blessed ref -- will restore later - ## - print( $fh chr(1) ); - print( $fh pack($self->{data_pack}, length($value_class)) . $value_class ); - $content_length += 1; - $content_length += $self->{data_size} + length($value_class); - } - else { - print( $fh chr(0) ); - $content_length += 1; - } - } + return 1; +} - ## - # If this is a new content area, advance EOF counter - ## - if ($location == $root->{end}) { - $root->{end} += DBM::Deep->SIG_SIZE; - $root->{end} += $self->{data_size} + $content_length; - $root->{end} += $self->{data_size} + length($plain_key); - } +sub split_index { + my $self = shift; + my ($md5, $tag) = @_; - ## - # If content is a hash or array, create new child DBM::Deep object and - # pass each key or element to it. - ## - if ($r eq 'HASH') { - my $branch = DBM::Deep->new( - type => DBM::Deep->TYPE_HASH, - base_offset => $location, - root => $root, - ); - foreach my $key (keys %{$value}) { - $branch->STORE( $key, $value->{$key} ); - } - } - elsif ($r eq 'ARRAY') { - my $branch = DBM::Deep->new( - type => DBM::Deep->TYPE_ARRAY, - base_offset => $location, - root => $root, + local($/,$\); + + my $fh = $self->_fh; + my $root = $self->_fileobj; + + my $loc = $self->_request_space( + $self->tag_size( $self->{index_size} ), + ); + + seek($fh, $tag->{ref_loc} + $root->{file_offset}, SEEK_SET); + print( $fh pack($self->{long_pack}, $loc) ); + + my $index_tag = $self->write_tag( + $loc, SIG_INDEX, + chr(0)x$self->{index_size}, + ); + + my $newtag_loc = $self->_request_space( + $self->tag_size( $self->{bucket_list_size} ), + ); + + my $keys = $tag->{content} + . $md5 . pack($self->{long_pack}, $newtag_loc) + . pack($self->{long_pack}, 0) # size + . pack($self->{long_pack}, 0); # transaction ID + + my @newloc = (); + BUCKET: + for (my $i = 0; $i <= $self->{max_buckets}; $i++) { + my ($key, $old_subloc, $size) = $self->_get_key_subloc( $keys, $i ); + + die "[INTERNAL ERROR]: No key in split_index()\n" unless $key; + die "[INTERNAL ERROR]: No subloc in split_index()\n" unless $old_subloc; + + my $num = ord(substr($key, $tag->{ch} + 1, 1)); + + if ($newloc[$num]) { + seek($fh, $newloc[$num] + $root->{file_offset}, SEEK_SET); + my $subkeys; + read( $fh, $subkeys, $self->{bucket_list_size}); + + # This is looking for the first empty spot + my ($subloc, $offset, $size) = $self->_find_in_buckets( + { content => $subkeys }, '', ); - my $index = 0; - foreach my $element (@{$value}) { - $branch->STORE( $index, $element ); - $index++; - } + + seek($fh, $newloc[$num] + $offset + $root->{file_offset}, SEEK_SET); + print( $fh $key . pack($self->{long_pack}, $old_subloc) ); + + next; } - return $result; + seek($fh, $index_tag->{offset} + ($num * $self->{long_size}) + $root->{file_offset}, SEEK_SET); + + my $loc = $self->_request_space( + $self->tag_size( $self->{bucket_list_size} ), + ); + + print( $fh pack($self->{long_pack}, $loc) ); + + my $blist_tag = $self->write_tag( + $loc, SIG_BLIST, + chr(0)x$self->{bucket_list_size}, + ); + + seek($fh, $blist_tag->{offset} + $root->{file_offset}, SEEK_SET); + print( $fh $key . pack($self->{long_pack}, $old_subloc) ); + + $newloc[$num] = $blist_tag->{offset}; } - return $obj->_throw_error("Fatal error: indexing failed -- possibly due to corruption in file"); + $self->_release_space( + $self->tag_size( $self->{bucket_list_size} ), + $tag->{offset} - SIG_SIZE - $self->{data_size}, + ); + + return $newtag_loc; } -sub get_bucket_value { - ## - # Fetch single value given tag and MD5 digested key. - ## +sub read_from_loc { my $self = shift; - my ($obj, $tag, $md5) = @_; - my $keys = $tag->{content}; + my ($subloc, $orig_key) = @_; + + local($/,$\); - my $fh = $obj->_fh; + my $fh = $self->_fh; ## - # Iterate through buckets, looking for a key match + # Found match -- seek to offset and read signature ## - BUCKET: - for (my $i=0; $i<$self->{max_buckets}; $i++) { - my $key = substr($keys, $i * $self->{bucket_size}, $self->{hash_size}); - my $subloc = unpack($self->{long_pack}, substr($keys, ($i * $self->{bucket_size}) + $self->{hash_size}, $self->{long_size})); + my $signature; + seek($fh, $subloc + $self->_fileobj->{file_offset}, SEEK_SET); + read( $fh, $signature, SIG_SIZE); - if (!$subloc) { + ## + # 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 => $orig_key, + }); + + if ($new_obj->_fileobj->{autobless}) { ## - # Hit end of list, no match + # Skip over value and plain key to see if object needs + # to be re-blessed ## - return; - } + seek($fh, $self->{data_size} + $self->{index_size}, SEEK_CUR); - if ( $md5 ne $key ) { - next BUCKET; - } - - ## - # Found match -- seek to offset and read signature - ## - my $signature; - seek($fh, $subloc + $obj->_root->{file_offset}, SEEK_SET); - read( $fh, $signature, DBM::Deep->SIG_SIZE); - - ## - # If value is a hash or array, return new DBM::Deep object with correct offset - ## - if (($signature eq DBM::Deep->TYPE_HASH) || ($signature eq DBM::Deep->TYPE_ARRAY)) { - my $obj = DBM::Deep->new( - type => $signature, - base_offset => $subloc, - root => $obj->_root, - ); + my $size; + read( $fh, $size, $self->{data_size}); + $size = unpack($self->{data_pack}, $size); + if ($size) { seek($fh, $size, SEEK_CUR); } - if ($obj->_root->{autobless}) { + my $bless_bit; + read( $fh, $bless_bit, 1); + if (ord($bless_bit)) { ## - # Skip over value and plain key to see if object needs - # to be re-blessed + # Yes, object needs to be re-blessed ## - seek($fh, $self->{data_size} + $self->{index_size}, SEEK_CUR); + my $class_name; + read( $fh, $size, $self->{data_size}); + $size = unpack($self->{data_pack}, $size); + if ($size) { read( $fh, $class_name, $size); } + if ($class_name) { $new_obj = bless( $new_obj, $class_name ); } + } + } - my $size; - read( $fh, $size, $self->{data_size}); $size = unpack($self->{data_pack}, $size); - if ($size) { seek($fh, $size, SEEK_CUR); } + return $new_obj; + } + elsif ( $signature eq SIG_INTERNAL ) { + my $size; + read( $fh, $size, $self->{data_size}); + $size = unpack($self->{data_pack}, $size); - my $bless_bit; - read( $fh, $bless_bit, 1); - if (ord($bless_bit)) { - ## - # Yes, object needs to be re-blessed - ## - my $class_name; - read( $fh, $size, $self->{data_size}); $size = unpack($self->{data_pack}, $size); - if ($size) { read( $fh, $class_name, $size); } - if ($class_name) { $obj = bless( $obj, $class_name ); } - } - } + if ( $size ) { + my $new_loc; + read( $fh, $new_loc, $size ); + $new_loc = unpack( $self->{long_pack}, $new_loc ); - return $obj; + return $self->read_from_loc( $new_loc, $orig_key ); } - - ## - # Otherwise return actual value - ## - elsif ($signature eq DBM::Deep->SIG_DATA) { - my $size; - my $value = ''; - read( $fh, $size, $self->{data_size}); $size = unpack($self->{data_pack}, $size); - if ($size) { read( $fh, $value, $size); } - return $value; + else { + return; } + } + ## + # Otherwise return actual value + ## + elsif ( $signature eq SIG_DATA ) { + my $size; + read( $fh, $size, $self->{data_size}); + $size = unpack($self->{data_pack}, $size); + + my $value = ''; + if ($size) { read( $fh, $value, $size); } + return $value; + } - ## - # Key exists, but content is null - ## - else { return; } - } # i loop - + ## + # Key exists, but content is null + ## return; } -sub delete_bucket { +sub get_bucket_value { ## - # Delete single key/value pair given tag and MD5 digested key. + # Fetch single value given tag and MD5 digested key. ## my $self = shift; - my ($obj, $tag, $md5) = @_; - my $keys = $tag->{content}; + my ($tag, $md5, $orig_key) = @_; - my $fh = $obj->_fh; + #ACID - This is a read. Can find exact or HEAD + my ($subloc, $offset, $size,$is_deleted) = $self->_find_in_buckets( $tag, $md5 ); + if ( $subloc && !$is_deleted ) { + return $self->read_from_loc( $subloc, $orig_key ); + } + return; +} +sub delete_bucket { ## - # Iterate through buckets, looking for a key match + # Delete single key/value pair given tag and MD5 digested key. ## - BUCKET: - for (my $i=0; $i<$self->{max_buckets}; $i++) { - my $key = substr($keys, $i * $self->{bucket_size}, $self->{hash_size}); - my $subloc = unpack($self->{long_pack}, substr($keys, ($i * $self->{bucket_size}) + $self->{hash_size}, $self->{long_size})); - - if (!$subloc) { - ## - # Hit end of list, no match - ## - return; - } + my $self = shift; + my ($tag, $md5, $orig_key) = @_; - if ( $md5 ne $key ) { - next BUCKET; - } + local($/,$\); - ## - # Matched key -- delete bucket and return - ## - seek($fh, $tag->{offset} + ($i * $self->{bucket_size}) + $obj->_root->{file_offset}, SEEK_SET); - print( $fh substr($keys, ($i+1) * $self->{bucket_size} ) ); + #ACID - This is a mutation. Must only find the exact transaction + my ($subloc, $offset, $size) = $self->_find_in_buckets( $tag, $md5, 1 ); +#XXX This needs _release_space() + if ( $subloc ) { + my $fh = $self->_fh; + seek($fh, $tag->{offset} + $offset + $self->_fileobj->{file_offset}, SEEK_SET); + print( $fh substr($tag->{content}, $offset + $self->{bucket_size} ) ); print( $fh chr(0) x $self->{bucket_size} ); return 1; - } # i loop - + } return; } @@ -640,35 +793,11 @@ sub bucket_exists { # Check existence of single key given tag and MD5 digested key. ## my $self = shift; - my ($obj, $tag, $md5) = @_; - my $keys = $tag->{content}; + my ($tag, $md5) = @_; - ## - # Iterate through buckets, looking for a key match - ## - BUCKET: - for (my $i=0; $i<$self->{max_buckets}; $i++) { - my $key = substr($keys, $i * $self->{bucket_size}, $self->{hash_size}); - my $subloc = unpack($self->{long_pack}, substr($keys, ($i * $self->{bucket_size}) + $self->{hash_size}, $self->{long_size})); - - if (!$subloc) { - ## - # Hit end of list, no match - ## - return; - } - - if ( $md5 ne $key ) { - next BUCKET; - } - - ## - # Matched key -- return true - ## - return 1; - } # i loop - - return; + #ACID - This is a read. Can find exact or HEAD + my ($subloc, $offset, $size, $is_deleted) = $self->_find_in_buckets( $tag, $md5 ); + return ($subloc && !$is_deleted) && 1; } sub find_bucket_list { @@ -676,50 +805,48 @@ sub find_bucket_list { # Locate offset for bucket list, given digested key ## my $self = shift; - my ($obj, $md5, $args) = @_; + my ($offset, $md5, $args) = @_; $args = {} unless $args; + local($/,$\); + ## # Locate offset for bucket list using digest index system ## - my $ch = 0; - my $tag = $self->load_tag($obj, $obj->_base_offset); - if (!$tag) { - return $self->_throw_error( "INTERNAL ERROR - Cannot find tag" ); - } + my $tag = $self->load_tag( $offset ) + or $self->_throw_error( "INTERNAL ERROR - Cannot find tag" ); - while ($tag->{signature} ne DBM::Deep->SIG_BLIST) { + 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( $obj, $tag, $num ); + $tag = $self->index_lookup( $tag, $num ); if (!$tag) { - if ( $args->{create} ) { - my $fh = $obj->_fh; - seek($fh, $ref_loc + $obj->_root->{file_offset}, SEEK_SET); - print( $fh pack($self->{long_pack}, $obj->_root->{end}) ); - - $tag = $self->create_tag( - $obj, $obj->_root->{end}, - DBM::Deep->SIG_BLIST, - chr(0) x $self->{bucket_list_size}, - ); + return if !$args->{create}; - $tag->{ref_loc} = $ref_loc; - $tag->{ch} = $ch; + my $loc = $self->_request_space( + $self->tag_size( $self->{bucket_list_size} ), + ); - last; - } - else { - return; - } + my $fh = $self->_fh; + seek($fh, $ref_loc + $self->_fileobj->{file_offset}, SEEK_SET); + print( $fh pack($self->{long_pack}, $loc) ); + + $tag = $self->write_tag( + $loc, SIG_BLIST, + chr(0)x$self->{bucket_list_size}, + ); + + $tag->{ref_loc} = $ref_loc; + $tag->{ch} = $ch; + + last; } - $tag->{ch} = $ch; + $tag->{ch} = $ch++; $tag->{ref_loc} = $ref_loc; - - $ch++; } return $tag; @@ -730,7 +857,7 @@ sub index_lookup { # Given index tag, lookup single entry in index and return . ## my $self = shift; - my ($obj, $tag, $index) = @_; + my ($tag, $index) = @_; my $location = unpack( $self->{long_pack}, @@ -743,7 +870,7 @@ sub index_lookup { if (!$location) { return; } - return $self->load_tag( $obj, $location ); + return $self->load_tag( $location ); } sub traverse_index { @@ -752,22 +879,32 @@ sub traverse_index { ## my $self = shift; my ($obj, $offset, $ch, $force_return_next) = @_; - $force_return_next = undef unless $force_return_next; - my $tag = $self->load_tag($obj, $offset ); + local($/,$\); - my $fh = $obj->_fh; + my $tag = $self->load_tag( $offset ); - if ($tag->{signature} ne DBM::Deep->SIG_BLIST) { + my $fh = $self->_fh; + + if ($tag->{signature} ne SIG_BLIST) { my $content = $tag->{content}; - my $start; - if ($obj->{return_next}) { $start = 0; } - else { $start = ord(substr($obj->{prev_md5}, $ch, 1)); } + my $start = $obj->{return_next} ? 0 : ord(substr($obj->{prev_md5}, $ch, 1)); + + for (my $idx = $start; $idx < (2**8); $idx++) { + my $subloc = unpack( + $self->{long_pack}, + substr( + $content, + $idx * $self->{long_size}, + $self->{long_size}, + ), + ); - for (my $index = $start; $index < 256; $index++) { - my $subloc = unpack($self->{long_pack}, substr($content, $index * $self->{long_size}, $self->{long_size}) ); if ($subloc) { - my $result = $self->traverse_index( $obj, $subloc, $ch + 1, $force_return_next ); + my $result = $self->traverse_index( + $obj, $subloc, $ch + 1, $force_return_next, + ); + if (defined($result)) { return $result; } } } # index loop @@ -775,54 +912,48 @@ sub traverse_index { $obj->{return_next} = 1; } # tag is an index - elsif ($tag->{signature} eq DBM::Deep->SIG_BLIST) { + else { my $keys = $tag->{content}; if ($force_return_next) { $obj->{return_next} = 1; } ## # Iterate through buckets, looking for a key match ## - for (my $i=0; $i<$self->{max_buckets}; $i++) { - my $key = substr($keys, $i * $self->{bucket_size}, $self->{hash_size}); - my $subloc = unpack($self->{long_pack}, substr($keys, ($i * $self->{bucket_size}) + $self->{hash_size}, $self->{long_size})); + for (my $i = 0; $i < $self->{max_buckets}; $i++) { + my ($key, $subloc) = $self->_get_key_subloc( $keys, $i ); + # End of bucket list -- return to outer loop if (!$subloc) { - ## - # End of bucket list -- return to outer loop - ## $obj->{return_next} = 1; last; } + # Located previous key -- return next one found elsif ($key eq $obj->{prev_md5}) { - ## - # Located previous key -- return next one found - ## $obj->{return_next} = 1; next; } + # Seek to bucket location and skip over signature elsif ($obj->{return_next}) { - ## - # Seek to bucket location and skip over signature - ## - seek($fh, $subloc + DBM::Deep->SIG_SIZE + $obj->_root->{file_offset}, SEEK_SET); + seek($fh, $subloc + $self->_fileobj->{file_offset}, SEEK_SET); - ## # Skip over value to get to plain key - ## + my $sig; + read( $fh, $sig, SIG_SIZE ); + my $size; - read( $fh, $size, $self->{data_size}); $size = unpack($self->{data_pack}, $size); + read( $fh, $size, $self->{data_size}); + $size = unpack($self->{data_pack}, $size); if ($size) { seek($fh, $size, SEEK_CUR); } - ## # Read in plain key and return as scalar - ## my $plain_key; - read( $fh, $size, $self->{data_size}); $size = unpack($self->{data_pack}, $size); + read( $fh, $size, $self->{data_size}); + $size = unpack($self->{data_pack}, $size); if ($size) { read( $fh, $plain_key, $size); } return $plain_key; } - } # bucket loop + } $obj->{return_next} = 1; } # tag is a bucket list @@ -852,5 +983,164 @@ sub get_next_key { return $self->traverse_index( $obj, $obj->_base_offset, 0 ); } +# Utilities + +sub _get_key_subloc { + my $self = shift; + my ($keys, $idx) = @_; + + my ($key, $subloc, $size, $transaction_id, $is_deleted) = unpack( + # This is 'a', not 'A'. Please read the pack() documentation for the + # difference between the two and why it's important. + "a$self->{hash_size} $self->{long_pack}2 n2", + substr( + $keys, + ($idx * $self->{bucket_size}), + $self->{bucket_size}, + ), + ); + + return ($key, $subloc, $size, $transaction_id, $is_deleted); +} + +sub _find_in_buckets { + my $self = shift; + my ($tag, $md5, $exact) = @_; + + my $trans_id = $self->_fileobj->transaction_id; + + my @zero; + + BUCKET: + for ( my $i = 0; $i < $self->{max_buckets}; $i++ ) { + my ($key, $subloc, $size, $transaction_id, $is_deleted) = $self->_get_key_subloc( + $tag->{content}, $i, + ); + + my @rv = ($subloc, $i * $self->{bucket_size}, $size, $is_deleted); + + unless ( $subloc ) { + if ( !$exact && @zero and $trans_id ) { + @rv = ($zero[2], $zero[0] * $self->{bucket_size},$zero[3],$is_deleted); + } + return @rv; + } + + next BUCKET if $key ne $md5; + + # Save off the HEAD in case we need it. + @zero = ($i,$key,$subloc,$size,$transaction_id,$is_deleted) if $transaction_id == 0; + + next BUCKET if $transaction_id != $trans_id; + + return @rv; + } + + return; +} + +sub _request_space { + my $self = shift; + my ($size) = @_; + + my $loc = $self->_fileobj->{end}; + $self->_fileobj->{end} += $size; + + return $loc; +} + +sub _release_space { + my $self = shift; + my ($size, $loc) = @_; + + local($/,$\); + + my $next_loc = 0; + + my $fh = $self->_fh; + seek( $fh, $loc + $self->_fileobj->{file_offset}, SEEK_SET ); + print( $fh SIG_FREE + . pack($self->{long_pack}, $size ) + . pack($self->{long_pack}, $next_loc ) + ); + + return; +} + +sub _throw_error { + die "DBM::Deep: $_[1]\n"; +} + 1; __END__ + +# This will be added in later, after more refactoring is done. This is an early +# attempt at refactoring on the physical level instead of the virtual level. +sub _read_at { + my $self = shift; + my ($spot, $amount, $unpack) = @_; + + local($/,$\); + + my $fh = $self->_fh; + seek( $fh, $spot + $self->_fileobj->{file_offset}, SEEK_SET ); + + my $buffer; + my $bytes_read = read( $fh, $buffer, $amount ); + + if ( $unpack ) { + $buffer = unpack( $unpack, $buffer ); + } + + if ( wantarray ) { + return ($buffer, $bytes_read); + } + else { + return $buffer; + } +} + +sub _print_at { + my $self = shift; + my ($spot, $data) = @_; + + local($/,$\); + + my $fh = $self->_fh; + seek( $fh, $spot, SEEK_SET ); + print( $fh $data ); + + return; +} + +sub get_file_version { + my $self = shift; + + local($/,$\); + + my $fh = $self->_fh; + + seek( $fh, 13 + $self->_fileobj->{file_offset}, SEEK_SET ); + my $buffer; + my $bytes_read = read( $fh, $buffer, 4 ); + unless ( $bytes_read == 4 ) { + $self->_throw_error( "Cannot read file version" ); + } + + return unpack( 'N', $buffer ); +} + +sub write_file_version { + my $self = shift; + my ($new_version) = @_; + + local($/,$\); + + my $fh = $self->_fh; + + seek( $fh, 13 + $self->_fileobj->{file_offset}, SEEK_SET ); + print( $fh pack( 'N', $new_version ) ); + + return; +} +