use Fcntl qw( :DEFAULT :flock :seek );
+##
+# Setup file and tag signatures. These should never change.
+##
+sub SIG_FILE () { 'DPDB' }
+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 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_size} = $self->{hash_size} + $self->{long_size} * 2;
$self->{bucket_list_size} = $self->{max_buckets} * $self->{bucket_size};
return 1;
my ($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.
+ # 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.
$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 :-)
+ # 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';
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,
}, $class;
$self->open( $obj ) if !defined $obj->_fh;
- #XXX Is {end} busted? Can't we just seek( $fh, 0, SEEK_END ) ?
+ my $fh = $obj->_fh;
+ flock $fh, LOCK_EX;
+
+ unless ( $obj->{base_offset} ) {
+ seek($fh, 0 + $obj->_root->{file_offset}, SEEK_SET);
+ my $signature;
+ my $bytes_read = read( $fh, $signature, length(SIG_FILE));
+
+ ##
+ # File is empty -- write signature and master index
+ ##
+ if (!$bytes_read) {
+ my $loc = $self->_request_space( $obj, length( SIG_FILE ) );
+ seek($fh, $loc + $obj->_root->{file_offset}, SEEK_SET);
+ print( $fh SIG_FILE);
+
+ $obj->{base_offset} = $self->_request_space(
+ $obj, $self->tag_size( $self->{index_size} ),
+ );
+
+ $self->write_tag(
+ $obj, $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} = $bytes_read;
+
+ ##
+ # Check signature was valid
+ ##
+ unless ($signature eq SIG_FILE) {
+ $self->close_fh( $obj );
+ $obj->_throw_error("Signature not found -- file is not a Deep DB");
+ }
+
+ ##
+ # Get our type from master index signature
+ ##
+ my $tag = $self->load_tag($obj, $obj->_base_offset)
+ or $obj->_throw_error("Corrupted file, no master index record");
+
+ unless ($obj->{type} eq $tag->{signature}) {
+ $obj->_throw_error("File type mismatch");
+ }
+ }
+ }
+
+ #XXX We have to make sure we don't mess up when autoflush isn't turned on
unless ( $obj->_root->{inode} ) {
my @stats = stat($obj->_fh);
$obj->_root->{inode} = $stats[1];
$obj->_root->{end} = $stats[7];
}
+ flock $fh, LOCK_UN;
+
return 1;
}
my $self = shift;
my ($obj) = @_;
- if (defined($obj->_fh)) { $self->close_fh( $obj ); }
+ # 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;
- 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;
- 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} . ": $!");
- }
-
- my $fh = $obj->_fh;
+ my $fh;
+ my $filename = $obj->_root->{file};
+ sysopen( $fh, $filename, $flags )
+ or $obj->_throw_error("Cannot sysopen file '$filename': $!");
+ $obj->_root->{fh} = $fh;
#XXX Can we remove this by using the right sysopen() flags?
# Maybe ... q.v. above
select $old;
}
- seek($fh, 0 + $obj->_root->{file_offset}, SEEK_SET);
-
- my $signature;
- my $bytes_read = read( $fh, $signature, length(DBM::Deep->SIG_FILE));
-
- ##
- # File is empty -- write signature and master index
- ##
- if (!$bytes_read) {
- seek($fh, 0 + $obj->_root->{file_offset}, SEEK_SET);
- print( $fh DBM::Deep->SIG_FILE);
-
- $self->create_tag($obj, $obj->_base_offset, $obj->_type, chr(0) x $self->{index_size});
-
- my $plain_key = "[base]";
- print( $fh pack($self->{data_pack}, length($plain_key)) . $plain_key );
-
- # Flush the filehandle
- my $old_fh = select $fh;
- my $old_af = $|; $| = 1; $| = $old_af;
- select $old_fh;
-
- return 1;
- }
-
- ##
- # Check signature was valid
- ##
- unless ($signature eq DBM::Deep->SIG_FILE) {
- $self->close_fh( $obj );
- return $obj->_throw_error("Signature not found -- file is not a Deep DB");
- }
-
- ##
- # Get our type from master index signature
- ##
- my $tag = $self->load_tag($obj, $obj->_base_offset);
- if (!$tag) {
- return $obj->_throw_error("Corrupted file, no master index record");
- }
-
- if ($obj->{type} ne $tag->{signature}) {
- return $obj->_throw_error("File type mismatch");
- }
-
-#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
-
return 1;
}
return 1;
}
-sub create_tag {
+sub tag_size {
+ my $self = shift;
+ my ($size) = @_;
+ return SIG_SIZE + $self->{data_size} + $size;
+}
+
+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 $size = length( $content );
my $fh = $obj->_fh;
- seek($fh, $offset + $obj->_root->{file_offset}, SEEK_SET);
+ if ( defined $offset ) {
+ seek($fh, $offset + $obj->_root->{file_offset}, SEEK_SET);
+ }
+
print( $fh $sig . pack($self->{data_pack}, $size) . $content );
- if ($offset == $obj->_root->{end}) {
- $obj->_root->{end} += DBM::Deep->SIG_SIZE + $self->{data_size} + $size;
- }
+ 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
};
}
my $self = shift;
my ($obj, $offset) = @_;
+# print join(':',map{$_||''}caller(1)), $/;
+
my $fh = $obj->_fh;
seek($fh, $offset + $obj->_root->{file_offset}, SEEK_SET);
- if (eof $fh) { return undef; }
+
+ #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;
return {
signature => $sig,
size => $size,
- offset => $offset + DBM::Deep->SIG_SIZE + $self->{data_size},
+ offset => $offset + SIG_SIZE + $self->{data_size},
content => $buffer
};
}
+sub _length_needed {
+ my $self = shift;
+ my ($obj, $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->_root eq $obj->_root ) {
+ return $len + $self->{long_size};
+ }
+
+ my $r = Scalar::Util::reftype( $value ) || '';
+ if ( $obj->_root->{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 ( $obj->_root->{autobless} ) {
+ my $value_class = Scalar::Util::blessed($value);
+ if ( defined $value_class && !$is_dbm_deep ) {
+ $len += $self->{data_size} + length($value_class);
+ }
+ }
+
+ return $len;
+}
+
sub add_bucket {
##
# Adds one key/value pair to bucket list, given offset, MD5 digest of key,
##
my $self = shift;
my ($obj, $tag, $md5, $plain_key, $value) = @_;
- my $keys = $tag->{content};
+
+ # 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';
+
+ $obj->_throw_error(
+ "Storage of variables of type '$r' is not supported."
+ );
+ }
+
my $location = 0;
my $result = 2;
my $root = $obj->_root;
+ my $fh = $obj->_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( $obj, $value, $plain_key );
- my $fh = $obj->_fh;
+ my ($subloc, $offset, $size) = $self->_find_in_buckets( $tag, $md5 );
- ##
- # 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;
+# $self->_release_space( $obj, $size, $subloc );
+ # Updating a known md5
+#XXX This needs updating to use _release_space
+ if ( $subloc ) {
+ $result = 1;
- $location = $internal_ref
- ? $value->_base_offset
- : $root->{end};
-
- seek($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( $obj, $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 ) );
}
+ }
+ # Adding a new md5
+ elsif ( defined $offset ) {
+ $location = $self->_request_space( $obj, $actual_length );
- 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;
+ 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 ) );
+ }
+ # 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( $obj, $md5, $tag );
+ }
- if ($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;
- }
+ $self->write_value( $obj, $location, $plain_key, $value );
- seek($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);
+ return $result;
+}
- ##
- # 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); }
+sub write_value {
+ my $self = shift;
+ my ($obj, $location, $key, $value) = @_;
- if ($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 = $obj->_fh;
+ my $root = $obj->_root;
- last;
- }
- }
+ my $is_dbm_deep = eval {
+ local $SIG{'__DIE__'};
+ $value->isa( 'DBM::Deep' );
+ };
+
+ my $is_internal_ref = $is_dbm_deep && ($value->_root eq $root);
+
+ seek($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 ( $is_internal_ref ) {
+ $self->write_tag( $obj, undef, SIG_INTERNAL,pack($self->{long_pack}, $value->_base_offset) );
+ }
+ elsif ($r eq 'HASH') {
+ $self->write_tag( $obj, undef, SIG_HASH, chr(0)x$self->{index_size} );
+ }
+ elsif ($r eq 'ARRAY') {
+ $self->write_tag( $obj, undef, SIG_ARRAY, chr(0)x$self->{index_size} );
+ }
+ elsif (!defined($value)) {
+ $self->write_tag( $obj, undef, SIG_NULL, '' );
+ }
+ else {
+ $self->write_tag( $obj, 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);
-
- ##
- # 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;
- }
- else {
- print( $fh DBM::Deep->SIG_DATA );
- print( $fh pack($self->{data_pack}, length($value)) . $value );
- $content_length = length($value);
- }
+ print( $fh pack($self->{data_pack}, length($key)) . $key );
- ##
- # Plain key is stored AFTER value, as keys are typically fetched less often.
- ##
- print( $fh pack($self->{data_pack}, length($plain_key)) . $plain_key );
+ # Internal references don't care about autobless
+ return 1 if $is_internal_ref;
- ##
- # 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;
- }
+ ##
+ # If value is blessed, preserve class name
+ ##
+ if ( $root->{autobless} ) {
+ my $value_class = Scalar::Util::blessed($value);
+ if ( defined $value_class && !$is_dbm_deep ) {
+ print( $fh chr(1) );
+ print( $fh pack($self->{data_pack}, length($value_class)) . $value_class );
}
-
- ##
- # If this is a new content area, advance EOF counter
- ##
- if ($location == $root->{end}) {
- $root->{end} += DBM::Deep->SIG_SIZE;
- $root->{end} += $self->{data_size} + $content_length;
- $root->{end} += $self->{data_size} + length($plain_key);
+ else {
+ print( $fh chr(0) );
}
+ }
- ##
- # If content is a hash or array, create new child DBM::Deep object and
- # pass each key or element to it.
- ##
+ ##
+ # If content is a hash or array, create new child DBM::Deep object and
+ # pass each key or element to it.
+ ##
+ if ( !$is_internal_ref ) {
if ($r eq 'HASH') {
my $branch = DBM::Deep->new(
type => DBM::Deep->TYPE_HASH,
$index++;
}
}
-
- return $result;
}
- return $obj->_throw_error("Fatal error: indexing failed -- possibly due to corruption in file");
+ return 1;
}
-sub get_bucket_value {
- ##
- # Fetch single value given tag and MD5 digested key.
- ##
+sub split_index {
my $self = shift;
- my ($obj, $tag, $md5) = @_;
- my $keys = $tag->{content};
+ my ($obj, $md5, $tag) = @_;
my $fh = $obj->_fh;
+ my $root = $obj->_root;
- ##
- # Iterate through buckets, looking for a key match
- ##
+ my $loc = $self->_request_space(
+ $obj, $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(
+ $obj, $loc, SIG_INDEX,
+ chr(0)x$self->{index_size},
+ );
+
+ my $newtag_loc = $self->_request_space(
+ $obj, $self->tag_size( $self->{bucket_list_size} ),
+ );
+
+ my $keys = $tag->{content}
+ . $md5 . pack($self->{long_pack}, $newtag_loc)
+ . pack($self->{long_pack}, 0);
+
+ my @newloc = ();
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}));
+ for (my $i = 0; $i <= $self->{max_buckets}; $i++) {
+ my ($key, $old_subloc, $size) = $self->_get_key_subloc( $keys, $i );
- if (!$subloc) {
- ##
- # Hit end of list, no match
- ##
- return;
- }
+ die "[INTERNAL ERROR]: No key in split_index()\n" unless $key;
+ die "[INTERNAL ERROR]: No subloc in split_index()\n" unless $old_subloc;
- if ( $md5 ne $key ) {
- next BUCKET;
- }
+ my $num = ord(substr($key, $tag->{ch} + 1, 1));
- ##
- # Found match -- seek to offset and read signature
- ##
- my $signature;
- seek($fh, $subloc + $obj->_root->{file_offset}, SEEK_SET);
- read( $fh, $signature, DBM::Deep->SIG_SIZE);
+ if ($newloc[$num]) {
+ seek($fh, $newloc[$num] + $root->{file_offset}, SEEK_SET);
+ my $subkeys;
+ read( $fh, $subkeys, $self->{bucket_list_size});
- ##
- # If value is a hash or array, return new DBM::Deep object with correct offset
- ##
- if (($signature eq DBM::Deep->TYPE_HASH) || ($signature eq DBM::Deep->TYPE_ARRAY)) {
- my $obj = DBM::Deep->new(
- type => $signature,
- base_offset => $subloc,
- root => $obj->_root,
+ # This is looking for the first empty spot
+ my ($subloc, $offset, $size) = $self->_find_in_buckets(
+ { content => $subkeys }, '',
);
- if ($obj->_root->{autobless}) {
- ##
- # Skip over value and plain key to see if object needs
- # to be re-blessed
- ##
- seek($fh, $self->{data_size} + $self->{index_size}, SEEK_CUR);
+ seek($fh, $newloc[$num] + $offset + $root->{file_offset}, SEEK_SET);
+ print( $fh $key . pack($self->{long_pack}, $old_subloc) );
- my $size;
- read( $fh, $size, $self->{data_size}); $size = unpack($self->{data_pack}, $size);
- if ($size) { seek($fh, $size, SEEK_CUR); }
+ next;
+ }
- 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, $class_name ); }
- }
- }
+ seek($fh, $index_tag->{offset} + ($num * $self->{long_size}) + $root->{file_offset}, SEEK_SET);
- return $obj;
- }
+ my $loc = $self->_request_space(
+ $obj, $self->tag_size( $self->{bucket_list_size} ),
+ );
- ##
- # Otherwise return actual value
- ##
- elsif ($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;
- }
+ print( $fh pack($self->{long_pack}, $loc) );
- ##
- # Key exists, but content is null
- ##
- else { return; }
- } # i loop
+ my $blist_tag = $self->write_tag(
+ $obj, $loc, SIG_BLIST,
+ chr(0)x$self->{bucket_list_size},
+ );
- return;
+ seek($fh, $blist_tag->{offset} + $root->{file_offset}, SEEK_SET);
+ print( $fh $key . pack($self->{long_pack}, $old_subloc) );
+
+ $newloc[$num] = $blist_tag->{offset};
+ }
+
+ $self->_release_space(
+ $obj, $self->tag_size( $self->{bucket_list_size} ),
+ $tag->{offset} - SIG_SIZE - $self->{data_size},
+ );
+
+ return $newtag_loc;
}
-sub delete_bucket {
- ##
- # Delete single key/value pair given tag and MD5 digested key.
- ##
+sub read_from_loc {
my $self = shift;
- my ($obj, $tag, $md5) = @_;
- my $keys = $tag->{content};
+ my ($obj, $subloc) = @_;
my $fh = $obj->_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 + $obj->_root->{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 $obj = DBM::Deep->new(
+ type => $signature,
+ base_offset => $subloc,
+ root => $obj->_root,
+ );
+
+ if ($obj->_root->{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;
+ my $size;
+ read( $fh, $size, $self->{data_size}); $size = unpack($self->{data_pack}, $size);
+ if ($size) { seek($fh, $size, SEEK_CUR); }
+
+ 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, $class_name ); }
+ }
}
- ##
- # Matched key -- delete bucket and return
- ##
- seek($fh, $tag->{offset} + ($i * $self->{bucket_size}) + $obj->_root->{file_offset}, SEEK_SET);
- print( $fh substr($keys, ($i+1) * $self->{bucket_size} ) );
- print( $fh chr(0) x $self->{bucket_size} );
+ return $obj;
+ }
+ elsif ( $signature eq SIG_INTERNAL ) {
+ my $size;
+ read( $fh, $size, $self->{data_size});
+ $size = unpack($self->{data_pack}, $size);
- return 1;
- } # i loop
+ if ( $size ) {
+ my $new_loc;
+ read( $fh, $new_loc, $size );
+ $new_loc = unpack( $self->{long_pack}, $new_loc );
+
+ return $self->read_from_loc( $obj, $new_loc );
+ }
+ else {
+ return;
+ }
+ }
+ ##
+ # Otherwise return actual value
+ ##
+ elsif ($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
+ ##
return;
}
-sub bucket_exists {
+sub get_bucket_value {
##
- # Check existence of single key given tag and MD5 digested key.
+ # Fetch single value given tag and MD5 digested key.
##
my $self = shift;
my ($obj, $tag, $md5) = @_;
- my $keys = $tag->{content};
+ my ($subloc, $offset, $size) = $self->_find_in_buckets( $tag, $md5 );
+ if ( $subloc ) {
+ return $self->read_from_loc( $obj, $subloc );
+ }
+ 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 $self = shift;
+ my ($obj, $tag, $md5) = @_;
- if ( $md5 ne $key ) {
- next BUCKET;
- }
+ my ($subloc, $offset, $size) = $self->_find_in_buckets( $tag, $md5 );
+#XXX This needs _release_space()
+ if ( $subloc ) {
+ my $fh = $obj->_fh;
+ seek($fh, $tag->{offset} + $offset + $obj->_root->{file_offset}, SEEK_SET);
+ print( $fh substr($tag->{content}, $offset + $self->{bucket_size} ) );
+ print( $fh chr(0) x $self->{bucket_size} );
- ##
- # Matched key -- return true
- ##
return 1;
- } # i loop
-
+ }
return;
}
+sub bucket_exists {
+ ##
+ # Check existence of single key given tag and MD5 digested key.
+ ##
+ my $self = shift;
+ my ($obj, $tag, $md5) = @_;
+
+ my ($subloc, $offset, $size) = $self->_find_in_buckets( $tag, $md5 );
+ return $subloc && 1;
+}
+
sub find_bucket_list {
##
# Locate offset for bucket list, given digested key
##
# 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 = $self->load_tag($obj, $obj->_base_offset)
+ or $obj->_throw_error( "INTERNAL ERROR - Cannot find tag" );
- while ($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, $num );
if (!$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(
+ $obj, $self->tag_size( $self->{bucket_list_size} ),
+ );
- last;
- }
- else {
- return;
- }
+ my $fh = $obj->_fh;
+ seek($fh, $ref_loc + $obj->_root->{file_offset}, SEEK_SET);
+ print( $fh pack($self->{long_pack}, $loc) );
+
+ $tag = $self->write_tag(
+ $obj, $loc, SIG_BLIST,
+ chr(0)x$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;
##
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 );
my $fh = $obj->_fh;
- if ($tag->{signature} ne DBM::Deep->SIG_BLIST) {
+ 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
$obj->{return_next} = 1;
} # tag is an index
- elsif ($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 + $obj->_root->{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
return $self->traverse_index( $obj, $obj->_base_offset, 0 );
}
+# Utilities
+
+sub _get_key_subloc {
+ my $self = shift;
+ my ($keys, $idx) = @_;
+
+ my ($key, $subloc, $size) = unpack(
+ "a$self->{hash_size} $self->{long_pack} $self->{long_pack}",
+ substr(
+ $keys,
+ ($idx * $self->{bucket_size}),
+ $self->{bucket_size},
+ ),
+ );
+
+ return ($key, $subloc, $size);
+}
+
+sub _find_in_buckets {
+ my $self = shift;
+ my ($tag, $md5) = @_;
+
+ BUCKET:
+ for ( my $i = 0; $i < $self->{max_buckets}; $i++ ) {
+ my ($key, $subloc, $size) = $self->_get_key_subloc(
+ $tag->{content}, $i,
+ );
+
+ return ($subloc, $i * $self->{bucket_size}, $size) unless $subloc;
+
+ next BUCKET if $key ne $md5;
+
+ return ($subloc, $i * $self->{bucket_size}, $size);
+ }
+
+ return;
+}
+
+#sub _print_at {
+# my $self = shift;
+# my ($obj, $spot, $data) = @_;
+#
+# my $fh = $obj->_fh;
+# seek( $fh, $spot, SEEK_SET );
+# print( $fh $data );
+#
+# return;
+#}
+
+sub _request_space {
+ my $self = shift;
+ my ($obj, $size) = @_;
+
+ my $loc = $obj->_root->{end};
+ $obj->_root->{end} += $size;
+
+ return $loc;
+}
+
+sub _release_space {
+ my $self = shift;
+ my ($obj, $size, $loc) = @_;
+
+ my $next_loc = 0;
+
+ my $fh = $obj->_fh;
+ seek( $fh, $loc + $obj->_root->{file_offset}, SEEK_SET );
+ print( $fh SIG_FREE
+ . pack($self->{long_pack}, $size )
+ . pack($self->{long_pack}, $next_loc )
+ );
+
+ return;
+}
+
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 $self = shift;
+ my ($obj, $spot, $amount, $unpack) = @_;
+
+ my $fh = $obj->_fh;
+ seek( $fh, $spot + $obj->_root->{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;
+ }
+}
projects / dbsrgits/DBM-Deep.git / blobdiff