Commit | Line | Data |
a20d9a3f |
1 | package DBM::Deep::Engine; |
2 | |
3 | use strict; |
4 | |
5 | use Fcntl qw( :DEFAULT :flock :seek ); |
6 | |
8db25060 |
7 | ## |
8 | # Setup file and tag signatures. These should never change. |
9 | ## |
10 | sub SIG_FILE () { 'DPDB' } |
11 | sub SIG_INTERNAL () { 'i' } |
12 | sub SIG_HASH () { 'H' } |
13 | sub SIG_ARRAY () { 'A' } |
8db25060 |
14 | sub SIG_NULL () { 'N' } |
15 | sub SIG_DATA () { 'D' } |
16 | sub SIG_INDEX () { 'I' } |
17 | sub SIG_BLIST () { 'B' } |
7b1e1aa1 |
18 | sub SIG_FREE () { 'F' } |
8db25060 |
19 | sub SIG_SIZE () { 1 } |
20 | |
612969fb |
21 | sub precalc_sizes { |
beac1dff |
22 | ## |
23 | # Precalculate index, bucket and bucket list sizes |
24 | ## |
251dfd0e |
25 | my $self = shift; |
1bf65be7 |
26 | |
251dfd0e |
27 | $self->{index_size} = (2**8) * $self->{long_size}; |
6ed2f3df |
28 | $self->{bucket_size} = $self->{hash_size} + $self->{long_size} * 2; |
251dfd0e |
29 | $self->{bucket_list_size} = $self->{max_buckets} * $self->{bucket_size}; |
1bf65be7 |
30 | |
251dfd0e |
31 | return 1; |
1bf65be7 |
32 | } |
33 | |
34 | sub set_pack { |
beac1dff |
35 | ## |
36 | # Set pack/unpack modes (see file header for more) |
37 | ## |
251dfd0e |
38 | my $self = shift; |
1bf65be7 |
39 | my ($long_s, $long_p, $data_s, $data_p) = @_; |
40 | |
81d16922 |
41 | ## |
75be6413 |
42 | # Set to 4 and 'N' for 32-bit offset tags (default). Theoretical limit of 4 |
43 | # GB per file. |
beac1dff |
44 | # (Perl must be compiled with largefile support for files > 2 GB) |
81d16922 |
45 | # |
46 | # Set to 8 and 'Q' for 64-bit offsets. Theoretical limit of 16 XB per file. |
beac1dff |
47 | # (Perl must be compiled with largefile and 64-bit long support) |
81d16922 |
48 | ## |
251dfd0e |
49 | $self->{long_size} = $long_s ? $long_s : 4; |
50 | $self->{long_pack} = $long_p ? $long_p : 'N'; |
1bf65be7 |
51 | |
81d16922 |
52 | ## |
75be6413 |
53 | # Set to 4 and 'N' for 32-bit data length prefixes. Limit of 4 GB for each |
d5d7c51d |
54 | # key/value. Upgrading this is possible (see above) but probably not |
55 | # necessary. If you need more than 4 GB for a single key or value, this |
56 | # module is really not for you :-) |
81d16922 |
57 | ## |
251dfd0e |
58 | $self->{data_size} = $data_s ? $data_s : 4; |
59 | $self->{data_pack} = $data_p ? $data_p : 'N'; |
1bf65be7 |
60 | |
beac1dff |
61 | return $self->precalc_sizes(); |
1bf65be7 |
62 | } |
63 | |
64 | sub set_digest { |
beac1dff |
65 | ## |
66 | # Set key digest function (default is MD5) |
67 | ## |
251dfd0e |
68 | my $self = shift; |
1bf65be7 |
69 | my ($digest_func, $hash_size) = @_; |
70 | |
d0b74c17 |
71 | $self->{digest} = $digest_func ? $digest_func : \&Digest::MD5::md5; |
251dfd0e |
72 | $self->{hash_size} = $hash_size ? $hash_size : 16; |
612969fb |
73 | |
beac1dff |
74 | return $self->precalc_sizes(); |
612969fb |
75 | } |
76 | |
77 | sub new { |
78 | my $class = shift; |
79 | my ($args) = @_; |
80 | |
81 | my $self = bless { |
82 | long_size => 4, |
83 | long_pack => 'N', |
84 | data_size => 4, |
85 | data_pack => 'N', |
251dfd0e |
86 | |
612969fb |
87 | digest => \&Digest::MD5::md5, |
88 | hash_size => 16, |
251dfd0e |
89 | |
81d16922 |
90 | ## |
d5d7c51d |
91 | # Maximum number of buckets per list before another level of indexing is |
92 | # done. |
93 | # Increase this value for slightly greater speed, but larger database |
94 | # files. DO NOT decrease this value below 16, due to risk of recursive |
95 | # reindex overrun. |
81d16922 |
96 | ## |
612969fb |
97 | max_buckets => 16, |
98 | }, $class; |
99 | |
251dfd0e |
100 | $self->precalc_sizes; |
612969fb |
101 | |
102 | return $self; |
1bf65be7 |
103 | } |
104 | |
0d0f3d5d |
105 | sub write_file_signature { |
106 | my $self = shift; |
107 | my ($obj) = @_; |
108 | |
109 | my $fh = $obj->_fh; |
110 | |
42f79e07 |
111 | my $loc = $self->_request_space( |
112 | $obj, length( SIG_FILE ) + $self->{data_size}, |
113 | ); |
0d0f3d5d |
114 | seek($fh, $loc + $obj->_root->{file_offset}, SEEK_SET); |
42f79e07 |
115 | print( $fh SIG_FILE, pack($self->{data_pack}, 0) ); |
0d0f3d5d |
116 | |
117 | return; |
118 | } |
119 | |
e064ccd1 |
120 | sub read_file_signature { |
121 | my $self = shift; |
122 | my ($obj) = @_; |
123 | |
124 | my $fh = $obj->_fh; |
125 | |
126 | seek($fh, 0 + $obj->_root->{file_offset}, SEEK_SET); |
42f79e07 |
127 | my $buffer; |
128 | my $bytes_read = read( |
129 | $fh, $buffer, length(SIG_FILE) + $self->{data_size}, |
130 | ); |
e064ccd1 |
131 | |
132 | if ( $bytes_read ) { |
42f79e07 |
133 | my ($signature, $version) = unpack( "A4 $self->{data_pack}", $buffer ); |
e064ccd1 |
134 | unless ($signature eq SIG_FILE) { |
135 | $self->close_fh( $obj ); |
136 | $obj->_throw_error("Signature not found -- file is not a Deep DB"); |
137 | } |
138 | } |
139 | |
140 | return $bytes_read; |
141 | } |
142 | |
70b55428 |
143 | sub setup_fh { |
144 | my $self = shift; |
145 | my ($obj) = @_; |
146 | |
147 | $self->open( $obj ) if !defined $obj->_fh; |
148 | |
6fde4ed2 |
149 | my $fh = $obj->_fh; |
150 | flock $fh, LOCK_EX; |
118ba343 |
151 | |
6fde4ed2 |
152 | unless ( $obj->{base_offset} ) { |
e064ccd1 |
153 | my $bytes_read = $self->read_file_signature( $obj ); |
118ba343 |
154 | |
155 | ## |
156 | # File is empty -- write signature and master index |
157 | ## |
158 | if (!$bytes_read) { |
0d0f3d5d |
159 | $self->write_file_signature( $obj ); |
118ba343 |
160 | |
c9ec091a |
161 | $obj->{base_offset} = $self->_request_space( |
16d1ad9b |
162 | $obj, $self->tag_size( $self->{index_size} ), |
c9ec091a |
163 | ); |
118ba343 |
164 | |
9e4f83a0 |
165 | $self->write_tag( |
c9ec091a |
166 | $obj, $obj->_base_offset, $obj->_type, |
f37c15ab |
167 | chr(0)x$self->{index_size}, |
118ba343 |
168 | ); |
169 | |
170 | # Flush the filehandle |
171 | my $old_fh = select $fh; |
172 | my $old_af = $|; $| = 1; $| = $old_af; |
173 | select $old_fh; |
174 | } |
175 | else { |
176 | $obj->{base_offset} = $bytes_read; |
177 | |
178 | ## |
118ba343 |
179 | # Get our type from master index signature |
180 | ## |
181 | my $tag = $self->load_tag($obj, $obj->_base_offset) |
182 | or $obj->_throw_error("Corrupted file, no master index record"); |
183 | |
184 | unless ($obj->{type} eq $tag->{signature}) { |
185 | $obj->_throw_error("File type mismatch"); |
186 | } |
187 | } |
118ba343 |
188 | } |
e06824f8 |
189 | |
673464d9 |
190 | #XXX We have to make sure we don't mess up when autoflush isn't turned on |
70b55428 |
191 | unless ( $obj->_root->{inode} ) { |
192 | my @stats = stat($obj->_fh); |
193 | $obj->_root->{inode} = $stats[1]; |
194 | $obj->_root->{end} = $stats[7]; |
195 | } |
196 | |
6fde4ed2 |
197 | flock $fh, LOCK_UN; |
198 | |
70b55428 |
199 | return 1; |
200 | } |
201 | |
a20d9a3f |
202 | sub open { |
20f7b20c |
203 | ## |
204 | # Open a fh to the database, create if nonexistent. |
205 | # Make sure file signature matches DBM::Deep spec. |
206 | ## |
a20d9a3f |
207 | my $self = shift; |
70b55428 |
208 | my ($obj) = @_; |
a20d9a3f |
209 | |
673464d9 |
210 | # Theoretically, adding O_BINARY should remove the need for the binmode |
211 | # Of course, testing it is going to be ... interesting. |
212 | my $flags = O_RDWR | O_CREAT | O_BINARY; |
a20d9a3f |
213 | |
673464d9 |
214 | my $fh; |
d5d7c51d |
215 | my $filename = $obj->_root->{file}; |
216 | sysopen( $fh, $filename, $flags ) |
217 | or $obj->_throw_error("Cannot sysopen file '$filename': $!"); |
673464d9 |
218 | $obj->_root->{fh} = $fh; |
a20d9a3f |
219 | |
42f79e07 |
220 | # Even though we use O_BINARY, better be safe than sorry. |
221 | binmode $fh; |
a20d9a3f |
222 | |
cd59cad8 |
223 | if ($obj->_root->{autoflush}) { |
a20d9a3f |
224 | my $old = select $fh; |
225 | $|=1; |
226 | select $old; |
227 | } |
20f7b20c |
228 | |
a20d9a3f |
229 | return 1; |
230 | } |
231 | |
3d1b8be9 |
232 | sub close_fh { |
cd59cad8 |
233 | my $self = shift; |
a21f2d90 |
234 | my ($obj) = @_; |
cd59cad8 |
235 | |
236 | if ( my $fh = $obj->_root->{fh} ) { |
237 | close $fh; |
238 | } |
239 | $obj->_root->{fh} = undef; |
240 | |
241 | return 1; |
242 | } |
243 | |
16d1ad9b |
244 | sub tag_size { |
245 | my $self = shift; |
246 | my ($size) = @_; |
247 | return SIG_SIZE + $self->{data_size} + $size; |
248 | } |
249 | |
9e4f83a0 |
250 | sub write_tag { |
20f7b20c |
251 | ## |
252 | # Given offset, signature and content, create tag and write to disk |
253 | ## |
d4b1166e |
254 | my $self = shift; |
20f7b20c |
255 | my ($obj, $offset, $sig, $content) = @_; |
f37c15ab |
256 | my $size = length( $content ); |
20f7b20c |
257 | |
d4b1166e |
258 | my $fh = $obj->_fh; |
259 | |
f37c15ab |
260 | if ( defined $offset ) { |
261 | seek($fh, $offset + $obj->_root->{file_offset}, SEEK_SET); |
262 | } |
263 | |
251dfd0e |
264 | print( $fh $sig . pack($self->{data_pack}, $size) . $content ); |
20f7b20c |
265 | |
f37c15ab |
266 | return unless defined $offset; |
267 | |
20f7b20c |
268 | return { |
269 | signature => $sig, |
270 | size => $size, |
8db25060 |
271 | offset => $offset + SIG_SIZE + $self->{data_size}, |
20f7b20c |
272 | content => $content |
273 | }; |
d4b1166e |
274 | } |
275 | |
276 | sub load_tag { |
20f7b20c |
277 | ## |
278 | # Given offset, load single tag and return signature, size and data |
279 | ## |
d4b1166e |
280 | my $self = shift; |
20f7b20c |
281 | my ($obj, $offset) = @_; |
282 | |
e06824f8 |
283 | # print join(':',map{$_||''}caller(1)), $/; |
284 | |
d4b1166e |
285 | my $fh = $obj->_fh; |
286 | |
20f7b20c |
287 | seek($fh, $offset + $obj->_root->{file_offset}, SEEK_SET); |
e5fc7e69 |
288 | |
75be6413 |
289 | #XXX I'm not sure this check will work if autoflush isn't enabled ... |
e5fc7e69 |
290 | return if eof $fh; |
20f7b20c |
291 | |
d4b1166e |
292 | my $b; |
8db25060 |
293 | read( $fh, $b, SIG_SIZE + $self->{data_size} ); |
251dfd0e |
294 | my ($sig, $size) = unpack( "A $self->{data_pack}", $b ); |
20f7b20c |
295 | |
296 | my $buffer; |
297 | read( $fh, $buffer, $size); |
298 | |
299 | return { |
300 | signature => $sig, |
301 | size => $size, |
8db25060 |
302 | offset => $offset + SIG_SIZE + $self->{data_size}, |
20f7b20c |
303 | content => $buffer |
304 | }; |
d4b1166e |
305 | } |
306 | |
56ec4340 |
307 | sub _get_dbm_object { |
308 | my $item = shift; |
309 | |
310 | my $obj = eval { |
311 | local $SIG{__DIE__}; |
312 | if ($item->isa( 'DBM::Deep' )) { |
313 | return $item; |
314 | } |
315 | return; |
316 | }; |
317 | return $obj if $obj; |
318 | |
319 | my $r = Scalar::Util::reftype( $item ) || ''; |
320 | if ( $r eq 'HASH' ) { |
321 | my $obj = eval { |
322 | local $SIG{__DIE__}; |
323 | my $obj = tied(%$item); |
324 | if ($obj->isa( 'DBM::Deep' )) { |
325 | return $obj; |
326 | } |
327 | return; |
328 | }; |
329 | return $obj if $obj; |
330 | } |
331 | elsif ( $r eq 'ARRAY' ) { |
332 | my $obj = eval { |
333 | local $SIG{__DIE__}; |
334 | my $obj = tied(@$item); |
335 | if ($obj->isa( 'DBM::Deep' )) { |
336 | return $obj; |
337 | } |
338 | return; |
339 | }; |
340 | return $obj if $obj; |
341 | } |
342 | |
343 | return; |
344 | } |
345 | |
29b01632 |
346 | sub _length_needed { |
347 | my $self = shift; |
f37c15ab |
348 | my ($obj, $value, $key) = @_; |
29b01632 |
349 | |
350 | my $is_dbm_deep = eval { |
351 | local $SIG{'__DIE__'}; |
352 | $value->isa( 'DBM::Deep' ); |
353 | }; |
354 | |
f37c15ab |
355 | my $len = SIG_SIZE + $self->{data_size} |
356 | + $self->{data_size} + length( $key ); |
29b01632 |
357 | |
f37c15ab |
358 | if ( $is_dbm_deep && $value->_root eq $obj->_root ) { |
359 | return $len + $self->{long_size}; |
29b01632 |
360 | } |
361 | |
362 | my $r = Scalar::Util::reftype( $value ) || ''; |
9a187d8c |
363 | if ( $obj->_root->{autobless} ) { |
364 | # This is for the bit saying whether or not this thing is blessed. |
365 | $len += 1; |
366 | } |
367 | |
29b01632 |
368 | unless ( $r eq 'HASH' || $r eq 'ARRAY' ) { |
f37c15ab |
369 | if ( defined $value ) { |
370 | $len += length( $value ); |
371 | } |
372 | return $len; |
29b01632 |
373 | } |
374 | |
f37c15ab |
375 | $len += $self->{index_size}; |
29b01632 |
376 | |
377 | # if autobless is enabled, must also take into consideration |
f37c15ab |
378 | # the class name as it is stored after the key. |
29b01632 |
379 | if ( $obj->_root->{autobless} ) { |
56ec4340 |
380 | my $c = Scalar::Util::blessed($value); |
381 | if ( defined $c && !$is_dbm_deep ) { |
382 | $len += $self->{data_size} + length($c); |
29b01632 |
383 | } |
384 | } |
385 | |
f37c15ab |
386 | return $len; |
29b01632 |
387 | } |
388 | |
20f7b20c |
389 | sub add_bucket { |
390 | ## |
391 | # Adds one key/value pair to bucket list, given offset, MD5 digest of key, |
392 | # plain (undigested) key and value. |
393 | ## |
d4b1166e |
394 | my $self = shift; |
20f7b20c |
395 | my ($obj, $tag, $md5, $plain_key, $value) = @_; |
75be6413 |
396 | |
eea0d863 |
397 | # This verifies that only supported values will be stored. |
398 | { |
399 | my $r = Scalar::Util::reftype( $value ); |
400 | last if !defined $r; |
401 | |
402 | last if $r eq 'HASH'; |
403 | last if $r eq 'ARRAY'; |
404 | |
405 | $obj->_throw_error( |
406 | "Storage of variables of type '$r' is not supported." |
407 | ); |
408 | } |
409 | |
20f7b20c |
410 | my $location = 0; |
411 | my $result = 2; |
412 | |
413 | my $root = $obj->_root; |
f37c15ab |
414 | my $fh = $obj->_fh; |
20f7b20c |
415 | |
f37c15ab |
416 | my $actual_length = $self->_length_needed( $obj, $value, $plain_key ); |
20f7b20c |
417 | |
9a187d8c |
418 | my ($subloc, $offset, $size) = $self->_find_in_buckets( $tag, $md5 ); |
75be6413 |
419 | |
f9c33187 |
420 | # $self->_release_space( $obj, $size, $subloc ); |
386bab6c |
421 | # Updating a known md5 |
f9c33187 |
422 | #XXX This needs updating to use _release_space |
386bab6c |
423 | if ( $subloc ) { |
424 | $result = 1; |
20f7b20c |
425 | |
386bab6c |
426 | if ($actual_length <= $size) { |
427 | $location = $subloc; |
20f7b20c |
428 | } |
75be6413 |
429 | else { |
f37c15ab |
430 | $location = $self->_request_space( $obj, $actual_length ); |
386bab6c |
431 | seek( |
432 | $fh, |
9a187d8c |
433 | $tag->{offset} + $offset |
434 | + $self->{hash_size} + $root->{file_offset}, |
386bab6c |
435 | SEEK_SET, |
436 | ); |
9a187d8c |
437 | print( $fh pack($self->{long_pack}, $location ) ); |
438 | print( $fh pack($self->{long_pack}, $actual_length ) ); |
75be6413 |
439 | } |
75be6413 |
440 | } |
386bab6c |
441 | # Adding a new md5 |
442 | elsif ( defined $offset ) { |
f37c15ab |
443 | $location = $self->_request_space( $obj, $actual_length ); |
386bab6c |
444 | |
445 | seek( $fh, $tag->{offset} + $offset + $root->{file_offset}, SEEK_SET ); |
9a187d8c |
446 | print( $fh $md5 . pack($self->{long_pack}, $location ) ); |
447 | print( $fh pack($self->{long_pack}, $actual_length ) ); |
386bab6c |
448 | } |
449 | # If bucket didn't fit into list, split into a new index level |
f9c33187 |
450 | # split_index() will do the _request_space() call |
386bab6c |
451 | else { |
f9c33187 |
452 | $location = $self->split_index( $obj, $md5, $tag ); |
386bab6c |
453 | } |
20f7b20c |
454 | |
d5d7c51d |
455 | $self->write_value( $obj, $location, $plain_key, $value ); |
456 | |
457 | return $result; |
458 | } |
459 | |
460 | sub write_value { |
461 | my $self = shift; |
462 | my ($obj, $location, $key, $value) = @_; |
463 | |
464 | my $fh = $obj->_fh; |
465 | my $root = $obj->_root; |
466 | |
9d4fa373 |
467 | my $dbm_deep_obj = _get_dbm_object( $value ); |
468 | if ( $dbm_deep_obj && $dbm_deep_obj->_root ne $obj->_root ) { |
469 | $obj->_throw_error( "Cannot cross-reference. Use export() instead" ); |
470 | } |
d5d7c51d |
471 | |
472 | seek($fh, $location + $root->{file_offset}, SEEK_SET); |
473 | |
20f7b20c |
474 | ## |
d5d7c51d |
475 | # Write signature based on content type, set content length and write |
476 | # actual value. |
20f7b20c |
477 | ## |
9d4fa373 |
478 | my $r = Scalar::Util::reftype( $value ) || ''; |
479 | if ( $dbm_deep_obj ) { |
480 | $self->write_tag( $obj, undef, SIG_INTERNAL,pack($self->{long_pack}, $dbm_deep_obj->_base_offset) ); |
f37c15ab |
481 | } |
482 | elsif ($r eq 'HASH') { |
9d4fa373 |
483 | if ( !$dbm_deep_obj && tied %{$value} ) { |
019ab3a1 |
484 | $obj->_throw_error( "Cannot store something that is tied" ); |
485 | } |
9e4f83a0 |
486 | $self->write_tag( $obj, undef, SIG_HASH, chr(0)x$self->{index_size} ); |
f37c15ab |
487 | } |
488 | elsif ($r eq 'ARRAY') { |
9d4fa373 |
489 | if ( !$dbm_deep_obj && tied @{$value} ) { |
019ab3a1 |
490 | $obj->_throw_error( "Cannot store something that is tied" ); |
491 | } |
9e4f83a0 |
492 | $self->write_tag( $obj, undef, SIG_ARRAY, chr(0)x$self->{index_size} ); |
f37c15ab |
493 | } |
494 | elsif (!defined($value)) { |
9e4f83a0 |
495 | $self->write_tag( $obj, undef, SIG_NULL, '' ); |
d5d7c51d |
496 | } |
497 | else { |
9e4f83a0 |
498 | $self->write_tag( $obj, undef, SIG_DATA, $value ); |
d5d7c51d |
499 | } |
20f7b20c |
500 | |
d5d7c51d |
501 | ## |
502 | # Plain key is stored AFTER value, as keys are typically fetched less often. |
503 | ## |
504 | print( $fh pack($self->{data_pack}, length($key)) . $key ); |
20f7b20c |
505 | |
9a187d8c |
506 | # Internal references don't care about autobless |
9d4fa373 |
507 | return 1 if $dbm_deep_obj; |
9a187d8c |
508 | |
d5d7c51d |
509 | ## |
510 | # If value is blessed, preserve class name |
511 | ## |
512 | if ( $root->{autobless} ) { |
56ec4340 |
513 | my $c = Scalar::Util::blessed($value); |
514 | if ( defined $c && !$dbm_deep_obj ) { |
d5d7c51d |
515 | print( $fh chr(1) ); |
56ec4340 |
516 | print( $fh pack($self->{data_pack}, length($c)) . $c ); |
20f7b20c |
517 | } |
d5d7c51d |
518 | else { |
519 | print( $fh chr(0) ); |
20f7b20c |
520 | } |
d5d7c51d |
521 | } |
20f7b20c |
522 | |
d5d7c51d |
523 | ## |
56ec4340 |
524 | # Tie the passed in reference so that changes to it are reflected in the |
525 | # datafile. The use of $location as the base_offset will act as the |
526 | # the linkage between parent and child. |
527 | # |
528 | # The overall assignment is a hack around the fact that just tying doesn't |
529 | # store the values. This may not be the wrong thing to do. |
d5d7c51d |
530 | ## |
9d4fa373 |
531 | if ($r eq 'HASH') { |
532 | my %x = %$value; |
533 | tie %$value, 'DBM::Deep', { |
534 | base_offset => $location, |
535 | root => $root, |
536 | }; |
537 | %$value = %x; |
538 | } |
539 | elsif ($r eq 'ARRAY') { |
540 | my @x = @$value; |
541 | tie @$value, 'DBM::Deep', { |
542 | base_offset => $location, |
543 | root => $root, |
544 | }; |
545 | @$value = @x; |
20f7b20c |
546 | } |
d4b1166e |
547 | |
d5d7c51d |
548 | return 1; |
d4b1166e |
549 | } |
550 | |
75be6413 |
551 | sub split_index { |
552 | my $self = shift; |
553 | my ($obj, $md5, $tag) = @_; |
554 | |
555 | my $fh = $obj->_fh; |
556 | my $root = $obj->_root; |
16d1ad9b |
557 | |
558 | my $loc = $self->_request_space( |
559 | $obj, $self->tag_size( $self->{index_size} ), |
560 | ); |
561 | |
7b1e1aa1 |
562 | seek($fh, $tag->{ref_loc} + $root->{file_offset}, SEEK_SET); |
16d1ad9b |
563 | print( $fh pack($self->{long_pack}, $loc) ); |
75be6413 |
564 | |
9e4f83a0 |
565 | my $index_tag = $self->write_tag( |
16d1ad9b |
566 | $obj, $loc, SIG_INDEX, |
f37c15ab |
567 | chr(0)x$self->{index_size}, |
75be6413 |
568 | ); |
569 | |
f9c33187 |
570 | my $newtag_loc = $self->_request_space( |
571 | $obj, $self->tag_size( $self->{bucket_list_size} ), |
572 | ); |
75be6413 |
573 | |
7b1e1aa1 |
574 | my $keys = $tag->{content} |
f9c33187 |
575 | . $md5 . pack($self->{long_pack}, $newtag_loc) |
576 | . pack($self->{long_pack}, 0); |
75be6413 |
577 | |
f9c33187 |
578 | my @newloc = (); |
75be6413 |
579 | BUCKET: |
580 | for (my $i = 0; $i <= $self->{max_buckets}; $i++) { |
9a187d8c |
581 | my ($key, $old_subloc, $size) = $self->_get_key_subloc( $keys, $i ); |
75be6413 |
582 | |
f9c33187 |
583 | die "[INTERNAL ERROR]: No key in split_index()\n" unless $key; |
584 | die "[INTERNAL ERROR]: No subloc in split_index()\n" unless $old_subloc; |
75be6413 |
585 | |
75be6413 |
586 | my $num = ord(substr($key, $tag->{ch} + 1, 1)); |
587 | |
f9c33187 |
588 | if ($newloc[$num]) { |
589 | seek($fh, $newloc[$num] + $root->{file_offset}, SEEK_SET); |
75be6413 |
590 | my $subkeys; |
591 | read( $fh, $subkeys, $self->{bucket_list_size}); |
592 | |
f9c33187 |
593 | # This is looking for the first empty spot |
7b1e1aa1 |
594 | my ($subloc, $offset, $size) = $self->_find_in_buckets( |
f9c33187 |
595 | { content => $subkeys }, '', |
7b1e1aa1 |
596 | ); |
75be6413 |
597 | |
f9c33187 |
598 | seek($fh, $newloc[$num] + $offset + $root->{file_offset}, SEEK_SET); |
599 | print( $fh $key . pack($self->{long_pack}, $old_subloc) ); |
7b1e1aa1 |
600 | |
601 | next; |
75be6413 |
602 | } |
75be6413 |
603 | |
7b1e1aa1 |
604 | seek($fh, $index_tag->{offset} + ($num * $self->{long_size}) + $root->{file_offset}, SEEK_SET); |
2603d86e |
605 | |
7b1e1aa1 |
606 | my $loc = $self->_request_space( |
607 | $obj, $self->tag_size( $self->{bucket_list_size} ), |
608 | ); |
2603d86e |
609 | |
7b1e1aa1 |
610 | print( $fh pack($self->{long_pack}, $loc) ); |
75be6413 |
611 | |
7b1e1aa1 |
612 | my $blist_tag = $self->write_tag( |
613 | $obj, $loc, SIG_BLIST, |
614 | chr(0)x$self->{bucket_list_size}, |
615 | ); |
616 | |
617 | seek($fh, $blist_tag->{offset} + $root->{file_offset}, SEEK_SET); |
f9c33187 |
618 | print( $fh $key . pack($self->{long_pack}, $old_subloc) ); |
7b1e1aa1 |
619 | |
f9c33187 |
620 | $newloc[$num] = $blist_tag->{offset}; |
7b1e1aa1 |
621 | } |
622 | |
623 | $self->_release_space( |
f9c33187 |
624 | $obj, $self->tag_size( $self->{bucket_list_size} ), |
7b1e1aa1 |
625 | $tag->{offset} - SIG_SIZE - $self->{data_size}, |
626 | ); |
75be6413 |
627 | |
f9c33187 |
628 | return $newtag_loc; |
75be6413 |
629 | } |
630 | |
8db25060 |
631 | sub read_from_loc { |
632 | my $self = shift; |
633 | my ($obj, $subloc) = @_; |
634 | |
635 | my $fh = $obj->_fh; |
636 | |
637 | ## |
638 | # Found match -- seek to offset and read signature |
639 | ## |
640 | my $signature; |
641 | seek($fh, $subloc + $obj->_root->{file_offset}, SEEK_SET); |
642 | read( $fh, $signature, SIG_SIZE); |
643 | |
644 | ## |
645 | # If value is a hash or array, return new DBM::Deep object with correct offset |
646 | ## |
647 | if (($signature eq SIG_HASH) || ($signature eq SIG_ARRAY)) { |
685e40f1 |
648 | my $new_obj = DBM::Deep->new({ |
8db25060 |
649 | type => $signature, |
650 | base_offset => $subloc, |
651 | root => $obj->_root, |
685e40f1 |
652 | }); |
8db25060 |
653 | |
685e40f1 |
654 | if ($new_obj->_root->{autobless}) { |
8db25060 |
655 | ## |
656 | # Skip over value and plain key to see if object needs |
657 | # to be re-blessed |
658 | ## |
659 | seek($fh, $self->{data_size} + $self->{index_size}, SEEK_CUR); |
660 | |
661 | my $size; |
c6ea6b6c |
662 | read( $fh, $size, $self->{data_size}); |
663 | $size = unpack($self->{data_pack}, $size); |
8db25060 |
664 | if ($size) { seek($fh, $size, SEEK_CUR); } |
665 | |
666 | my $bless_bit; |
667 | read( $fh, $bless_bit, 1); |
668 | if (ord($bless_bit)) { |
669 | ## |
670 | # Yes, object needs to be re-blessed |
671 | ## |
672 | my $class_name; |
c6ea6b6c |
673 | read( $fh, $size, $self->{data_size}); |
674 | $size = unpack($self->{data_pack}, $size); |
8db25060 |
675 | if ($size) { read( $fh, $class_name, $size); } |
685e40f1 |
676 | if ($class_name) { $new_obj = bless( $new_obj, $class_name ); } |
8db25060 |
677 | } |
678 | } |
679 | |
685e40f1 |
680 | return $new_obj; |
8db25060 |
681 | } |
682 | elsif ( $signature eq SIG_INTERNAL ) { |
683 | my $size; |
684 | read( $fh, $size, $self->{data_size}); |
685 | $size = unpack($self->{data_pack}, $size); |
686 | |
687 | if ( $size ) { |
688 | my $new_loc; |
689 | read( $fh, $new_loc, $size ); |
690 | $new_loc = unpack( $self->{long_pack}, $new_loc ); |
691 | |
692 | return $self->read_from_loc( $obj, $new_loc ); |
693 | } |
694 | else { |
695 | return; |
696 | } |
697 | } |
698 | ## |
699 | # Otherwise return actual value |
700 | ## |
701 | elsif ($signature eq SIG_DATA) { |
702 | my $size; |
703 | read( $fh, $size, $self->{data_size}); |
704 | $size = unpack($self->{data_pack}, $size); |
705 | |
706 | my $value = ''; |
707 | if ($size) { read( $fh, $value, $size); } |
708 | return $value; |
709 | } |
710 | |
711 | ## |
712 | # Key exists, but content is null |
713 | ## |
714 | return; |
715 | } |
716 | |
9020ee8c |
717 | sub get_bucket_value { |
beac1dff |
718 | ## |
719 | # Fetch single value given tag and MD5 digested key. |
720 | ## |
721 | my $self = shift; |
722 | my ($obj, $tag, $md5) = @_; |
9020ee8c |
723 | |
9a187d8c |
724 | my ($subloc, $offset, $size) = $self->_find_in_buckets( $tag, $md5 ); |
386bab6c |
725 | if ( $subloc ) { |
8db25060 |
726 | return $self->read_from_loc( $obj, $subloc ); |
386bab6c |
727 | } |
beac1dff |
728 | return; |
9020ee8c |
729 | } |
ab0e4957 |
730 | |
731 | sub delete_bucket { |
beac1dff |
732 | ## |
733 | # Delete single key/value pair given tag and MD5 digested key. |
734 | ## |
735 | my $self = shift; |
736 | my ($obj, $tag, $md5) = @_; |
ab0e4957 |
737 | |
9a187d8c |
738 | my ($subloc, $offset, $size) = $self->_find_in_buckets( $tag, $md5 ); |
f9c33187 |
739 | #XXX This needs _release_space() |
386bab6c |
740 | if ( $subloc ) { |
741 | my $fh = $obj->_fh; |
742 | seek($fh, $tag->{offset} + $offset + $obj->_root->{file_offset}, SEEK_SET); |
743 | print( $fh substr($tag->{content}, $offset + $self->{bucket_size} ) ); |
251dfd0e |
744 | print( $fh chr(0) x $self->{bucket_size} ); |
d0b74c17 |
745 | |
ab0e4957 |
746 | return 1; |
386bab6c |
747 | } |
beac1dff |
748 | return; |
ab0e4957 |
749 | } |
750 | |
912d50b1 |
751 | sub bucket_exists { |
beac1dff |
752 | ## |
753 | # Check existence of single key given tag and MD5 digested key. |
754 | ## |
755 | my $self = shift; |
756 | my ($obj, $tag, $md5) = @_; |
912d50b1 |
757 | |
9a187d8c |
758 | my ($subloc, $offset, $size) = $self->_find_in_buckets( $tag, $md5 ); |
d5d7c51d |
759 | return $subloc && 1; |
912d50b1 |
760 | } |
761 | |
6736c116 |
762 | sub find_bucket_list { |
beac1dff |
763 | ## |
764 | # Locate offset for bucket list, given digested key |
765 | ## |
766 | my $self = shift; |
d0b74c17 |
767 | my ($obj, $md5, $args) = @_; |
768 | $args = {} unless $args; |
769 | |
beac1dff |
770 | ## |
771 | # Locate offset for bucket list using digest index system |
772 | ## |
e5fc7e69 |
773 | my $tag = $self->load_tag($obj, $obj->_base_offset) |
d5d7c51d |
774 | or $obj->_throw_error( "INTERNAL ERROR - Cannot find tag" ); |
d0b74c17 |
775 | |
e5fc7e69 |
776 | my $ch = 0; |
8db25060 |
777 | while ($tag->{signature} ne SIG_BLIST) { |
d0b74c17 |
778 | my $num = ord substr($md5, $ch, 1); |
779 | |
780 | my $ref_loc = $tag->{offset} + ($num * $self->{long_size}); |
781 | $tag = $self->index_lookup( $obj, $tag, $num ); |
782 | |
783 | if (!$tag) { |
29b01632 |
784 | return if !$args->{create}; |
d0b74c17 |
785 | |
16d1ad9b |
786 | my $loc = $self->_request_space( |
787 | $obj, $self->tag_size( $self->{bucket_list_size} ), |
788 | ); |
789 | |
7b1e1aa1 |
790 | my $fh = $obj->_fh; |
791 | seek($fh, $ref_loc + $obj->_root->{file_offset}, SEEK_SET); |
16d1ad9b |
792 | print( $fh pack($self->{long_pack}, $loc) ); |
d0b74c17 |
793 | |
9e4f83a0 |
794 | $tag = $self->write_tag( |
16d1ad9b |
795 | $obj, $loc, SIG_BLIST, |
f37c15ab |
796 | chr(0)x$self->{bucket_list_size}, |
d5d7c51d |
797 | ); |
798 | |
799 | $tag->{ref_loc} = $ref_loc; |
800 | $tag->{ch} = $ch; |
801 | |
802 | last; |
d0b74c17 |
803 | } |
804 | |
16d1ad9b |
805 | $tag->{ch} = $ch++; |
d0b74c17 |
806 | $tag->{ref_loc} = $ref_loc; |
beac1dff |
807 | } |
d0b74c17 |
808 | |
beac1dff |
809 | return $tag; |
6736c116 |
810 | } |
811 | |
d0b74c17 |
812 | sub index_lookup { |
813 | ## |
814 | # Given index tag, lookup single entry in index and return . |
815 | ## |
816 | my $self = shift; |
817 | my ($obj, $tag, $index) = @_; |
818 | |
819 | my $location = unpack( |
820 | $self->{long_pack}, |
821 | substr( |
822 | $tag->{content}, |
823 | $index * $self->{long_size}, |
824 | $self->{long_size}, |
825 | ), |
826 | ); |
827 | |
828 | if (!$location) { return; } |
829 | |
830 | return $self->load_tag( $obj, $location ); |
831 | } |
832 | |
6736c116 |
833 | sub traverse_index { |
beac1dff |
834 | ## |
835 | # Scan index and recursively step into deeper levels, looking for next key. |
836 | ## |
6736c116 |
837 | my $self = shift; |
838 | my ($obj, $offset, $ch, $force_return_next) = @_; |
d0b74c17 |
839 | |
beac1dff |
840 | my $tag = $self->load_tag($obj, $offset ); |
6736c116 |
841 | |
842 | my $fh = $obj->_fh; |
d0b74c17 |
843 | |
8db25060 |
844 | if ($tag->{signature} ne SIG_BLIST) { |
beac1dff |
845 | my $content = $tag->{content}; |
e5fc7e69 |
846 | my $start = $obj->{return_next} ? 0 : ord(substr($obj->{prev_md5}, $ch, 1)); |
d0b74c17 |
847 | |
d5d7c51d |
848 | for (my $idx = $start; $idx < (2**8); $idx++) { |
e5fc7e69 |
849 | my $subloc = unpack( |
850 | $self->{long_pack}, |
e06824f8 |
851 | substr( |
852 | $content, |
853 | $idx * $self->{long_size}, |
854 | $self->{long_size}, |
855 | ), |
e5fc7e69 |
856 | ); |
857 | |
beac1dff |
858 | if ($subloc) { |
e5fc7e69 |
859 | my $result = $self->traverse_index( |
860 | $obj, $subloc, $ch + 1, $force_return_next, |
861 | ); |
862 | |
beac1dff |
863 | if (defined($result)) { return $result; } |
864 | } |
865 | } # index loop |
d0b74c17 |
866 | |
beac1dff |
867 | $obj->{return_next} = 1; |
868 | } # tag is an index |
d0b74c17 |
869 | |
e5fc7e69 |
870 | else { |
beac1dff |
871 | my $keys = $tag->{content}; |
872 | if ($force_return_next) { $obj->{return_next} = 1; } |
d0b74c17 |
873 | |
beac1dff |
874 | ## |
875 | # Iterate through buckets, looking for a key match |
876 | ## |
8db25060 |
877 | for (my $i = 0; $i < $self->{max_buckets}; $i++) { |
9cec1360 |
878 | my ($key, $subloc) = $self->_get_key_subloc( $keys, $i ); |
d0b74c17 |
879 | |
8db25060 |
880 | # End of bucket list -- return to outer loop |
beac1dff |
881 | if (!$subloc) { |
beac1dff |
882 | $obj->{return_next} = 1; |
883 | last; |
884 | } |
8db25060 |
885 | # Located previous key -- return next one found |
beac1dff |
886 | elsif ($key eq $obj->{prev_md5}) { |
beac1dff |
887 | $obj->{return_next} = 1; |
888 | next; |
889 | } |
8db25060 |
890 | # Seek to bucket location and skip over signature |
beac1dff |
891 | elsif ($obj->{return_next}) { |
8db25060 |
892 | seek($fh, $subloc + $obj->_root->{file_offset}, SEEK_SET); |
d0b74c17 |
893 | |
beac1dff |
894 | # Skip over value to get to plain key |
8db25060 |
895 | my $sig; |
896 | read( $fh, $sig, SIG_SIZE ); |
897 | |
beac1dff |
898 | my $size; |
e5fc7e69 |
899 | read( $fh, $size, $self->{data_size}); |
900 | $size = unpack($self->{data_pack}, $size); |
beac1dff |
901 | if ($size) { seek($fh, $size, SEEK_CUR); } |
d0b74c17 |
902 | |
beac1dff |
903 | # Read in plain key and return as scalar |
beac1dff |
904 | my $plain_key; |
e5fc7e69 |
905 | read( $fh, $size, $self->{data_size}); |
906 | $size = unpack($self->{data_pack}, $size); |
beac1dff |
907 | if ($size) { read( $fh, $plain_key, $size); } |
d0b74c17 |
908 | |
beac1dff |
909 | return $plain_key; |
910 | } |
8db25060 |
911 | } |
d0b74c17 |
912 | |
beac1dff |
913 | $obj->{return_next} = 1; |
914 | } # tag is a bucket list |
d0b74c17 |
915 | |
beac1dff |
916 | return; |
6736c116 |
917 | } |
918 | |
919 | sub get_next_key { |
beac1dff |
920 | ## |
921 | # Locate next key, given digested previous one |
922 | ## |
6736c116 |
923 | my $self = shift; |
924 | my ($obj) = @_; |
d0b74c17 |
925 | |
beac1dff |
926 | $obj->{prev_md5} = $_[1] ? $_[1] : undef; |
927 | $obj->{return_next} = 0; |
d0b74c17 |
928 | |
beac1dff |
929 | ## |
930 | # If the previous key was not specifed, start at the top and |
931 | # return the first one found. |
932 | ## |
933 | if (!$obj->{prev_md5}) { |
934 | $obj->{prev_md5} = chr(0) x $self->{hash_size}; |
935 | $obj->{return_next} = 1; |
936 | } |
d0b74c17 |
937 | |
beac1dff |
938 | return $self->traverse_index( $obj, $obj->_base_offset, 0 ); |
6736c116 |
939 | } |
940 | |
75be6413 |
941 | # Utilities |
942 | |
9cec1360 |
943 | sub _get_key_subloc { |
75be6413 |
944 | my $self = shift; |
945 | my ($keys, $idx) = @_; |
946 | |
6ed2f3df |
947 | my ($key, $subloc, $size) = unpack( |
948 | "a$self->{hash_size} $self->{long_pack} $self->{long_pack}", |
75be6413 |
949 | substr( |
950 | $keys, |
9cec1360 |
951 | ($idx * $self->{bucket_size}), |
952 | $self->{bucket_size}, |
75be6413 |
953 | ), |
954 | ); |
955 | |
6ed2f3df |
956 | return ($key, $subloc, $size); |
75be6413 |
957 | } |
958 | |
d608b06e |
959 | sub _find_in_buckets { |
960 | my $self = shift; |
961 | my ($tag, $md5) = @_; |
962 | |
963 | BUCKET: |
964 | for ( my $i = 0; $i < $self->{max_buckets}; $i++ ) { |
9a187d8c |
965 | my ($key, $subloc, $size) = $self->_get_key_subloc( |
966 | $tag->{content}, $i, |
967 | ); |
d608b06e |
968 | |
9a187d8c |
969 | return ($subloc, $i * $self->{bucket_size}, $size) unless $subloc; |
d608b06e |
970 | |
971 | next BUCKET if $key ne $md5; |
972 | |
9a187d8c |
973 | return ($subloc, $i * $self->{bucket_size}, $size); |
d608b06e |
974 | } |
975 | |
976 | return; |
977 | } |
978 | |
7b1e1aa1 |
979 | #sub _print_at { |
980 | # my $self = shift; |
981 | # my ($obj, $spot, $data) = @_; |
982 | # |
983 | # my $fh = $obj->_fh; |
984 | # seek( $fh, $spot, SEEK_SET ); |
985 | # print( $fh $data ); |
986 | # |
987 | # return; |
988 | #} |
989 | |
994ccd8e |
990 | sub _request_space { |
991 | my $self = shift; |
992 | my ($obj, $size) = @_; |
993 | |
994 | my $loc = $obj->_root->{end}; |
c9ec091a |
995 | $obj->_root->{end} += $size; |
994ccd8e |
996 | |
997 | return $loc; |
998 | } |
999 | |
1000 | sub _release_space { |
1001 | my $self = shift; |
1002 | my ($obj, $size, $loc) = @_; |
1003 | |
7b1e1aa1 |
1004 | my $next_loc = 0; |
1005 | |
1006 | my $fh = $obj->_fh; |
1007 | seek( $fh, $loc + $obj->_root->{file_offset}, SEEK_SET ); |
1008 | print( $fh SIG_FREE |
1009 | . pack($self->{long_pack}, $size ) |
1010 | . pack($self->{long_pack}, $next_loc ) |
1011 | ); |
1012 | |
994ccd8e |
1013 | return; |
1014 | } |
1015 | |
a20d9a3f |
1016 | 1; |
1017 | __END__ |
d5d7c51d |
1018 | |
1019 | # This will be added in later, after more refactoring is done. This is an early |
1020 | # attempt at refactoring on the physical level instead of the virtual level. |
1021 | sub _read_at { |
1022 | my $self = shift; |
1023 | my ($obj, $spot, $amount, $unpack) = @_; |
1024 | |
1025 | my $fh = $obj->_fh; |
1026 | seek( $fh, $spot + $obj->_root->{file_offset}, SEEK_SET ); |
1027 | |
1028 | my $buffer; |
1029 | my $bytes_read = read( $fh, $buffer, $amount ); |
1030 | |
1031 | if ( $unpack ) { |
1032 | $buffer = unpack( $unpack, $buffer ); |
1033 | } |
1034 | |
1035 | if ( wantarray ) { |
1036 | return ($buffer, $bytes_read); |
1037 | } |
1038 | else { |
1039 | return $buffer; |
1040 | } |
1041 | } |