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' } |
18 | sub SIG_SIZE () { 1 } |
19 | |
612969fb |
20 | sub precalc_sizes { |
beac1dff |
21 | ## |
22 | # Precalculate index, bucket and bucket list sizes |
23 | ## |
251dfd0e |
24 | my $self = shift; |
1bf65be7 |
25 | |
251dfd0e |
26 | $self->{index_size} = (2**8) * $self->{long_size}; |
27 | $self->{bucket_size} = $self->{hash_size} + $self->{long_size}; |
28 | $self->{bucket_list_size} = $self->{max_buckets} * $self->{bucket_size}; |
1bf65be7 |
29 | |
251dfd0e |
30 | return 1; |
1bf65be7 |
31 | } |
32 | |
33 | sub set_pack { |
beac1dff |
34 | ## |
35 | # Set pack/unpack modes (see file header for more) |
36 | ## |
251dfd0e |
37 | my $self = shift; |
1bf65be7 |
38 | my ($long_s, $long_p, $data_s, $data_p) = @_; |
39 | |
81d16922 |
40 | ## |
75be6413 |
41 | # Set to 4 and 'N' for 32-bit offset tags (default). Theoretical limit of 4 |
42 | # GB per file. |
beac1dff |
43 | # (Perl must be compiled with largefile support for files > 2 GB) |
81d16922 |
44 | # |
45 | # Set to 8 and 'Q' for 64-bit offsets. Theoretical limit of 16 XB per file. |
beac1dff |
46 | # (Perl must be compiled with largefile and 64-bit long support) |
81d16922 |
47 | ## |
251dfd0e |
48 | $self->{long_size} = $long_s ? $long_s : 4; |
49 | $self->{long_pack} = $long_p ? $long_p : 'N'; |
1bf65be7 |
50 | |
81d16922 |
51 | ## |
75be6413 |
52 | # Set to 4 and 'N' for 32-bit data length prefixes. Limit of 4 GB for each |
d5d7c51d |
53 | # key/value. Upgrading this is possible (see above) but probably not |
54 | # necessary. If you need more than 4 GB for a single key or value, this |
55 | # module is really not for you :-) |
81d16922 |
56 | ## |
251dfd0e |
57 | $self->{data_size} = $data_s ? $data_s : 4; |
58 | $self->{data_pack} = $data_p ? $data_p : 'N'; |
1bf65be7 |
59 | |
beac1dff |
60 | return $self->precalc_sizes(); |
1bf65be7 |
61 | } |
62 | |
63 | sub set_digest { |
beac1dff |
64 | ## |
65 | # Set key digest function (default is MD5) |
66 | ## |
251dfd0e |
67 | my $self = shift; |
1bf65be7 |
68 | my ($digest_func, $hash_size) = @_; |
69 | |
d0b74c17 |
70 | $self->{digest} = $digest_func ? $digest_func : \&Digest::MD5::md5; |
251dfd0e |
71 | $self->{hash_size} = $hash_size ? $hash_size : 16; |
612969fb |
72 | |
beac1dff |
73 | return $self->precalc_sizes(); |
612969fb |
74 | } |
75 | |
76 | sub new { |
77 | my $class = shift; |
78 | my ($args) = @_; |
79 | |
80 | my $self = bless { |
81 | long_size => 4, |
82 | long_pack => 'N', |
83 | data_size => 4, |
84 | data_pack => 'N', |
251dfd0e |
85 | |
612969fb |
86 | digest => \&Digest::MD5::md5, |
87 | hash_size => 16, |
251dfd0e |
88 | |
81d16922 |
89 | ## |
d5d7c51d |
90 | # Maximum number of buckets per list before another level of indexing is |
91 | # done. |
92 | # Increase this value for slightly greater speed, but larger database |
93 | # files. DO NOT decrease this value below 16, due to risk of recursive |
94 | # reindex overrun. |
81d16922 |
95 | ## |
612969fb |
96 | max_buckets => 16, |
97 | }, $class; |
98 | |
251dfd0e |
99 | $self->precalc_sizes; |
612969fb |
100 | |
101 | return $self; |
1bf65be7 |
102 | } |
103 | |
70b55428 |
104 | sub setup_fh { |
105 | my $self = shift; |
106 | my ($obj) = @_; |
107 | |
108 | $self->open( $obj ) if !defined $obj->_fh; |
109 | |
673464d9 |
110 | #XXX We have to make sure we don't mess up when autoflush isn't turned on |
70b55428 |
111 | unless ( $obj->_root->{inode} ) { |
112 | my @stats = stat($obj->_fh); |
113 | $obj->_root->{inode} = $stats[1]; |
114 | $obj->_root->{end} = $stats[7]; |
115 | } |
116 | |
117 | return 1; |
118 | } |
119 | |
a20d9a3f |
120 | sub open { |
20f7b20c |
121 | ## |
122 | # Open a fh to the database, create if nonexistent. |
123 | # Make sure file signature matches DBM::Deep spec. |
124 | ## |
a20d9a3f |
125 | my $self = shift; |
70b55428 |
126 | my ($obj) = @_; |
a20d9a3f |
127 | |
3d1b8be9 |
128 | if (defined($obj->_fh)) { $self->close_fh( $obj ); } |
20f7b20c |
129 | |
673464d9 |
130 | # Theoretically, adding O_BINARY should remove the need for the binmode |
131 | # Of course, testing it is going to be ... interesting. |
132 | my $flags = O_RDWR | O_CREAT | O_BINARY; |
a20d9a3f |
133 | |
673464d9 |
134 | my $fh; |
d5d7c51d |
135 | my $filename = $obj->_root->{file}; |
136 | sysopen( $fh, $filename, $flags ) |
137 | or $obj->_throw_error("Cannot sysopen file '$filename': $!"); |
673464d9 |
138 | $obj->_root->{fh} = $fh; |
a20d9a3f |
139 | |
140 | #XXX Can we remove this by using the right sysopen() flags? |
141 | # Maybe ... q.v. above |
142 | binmode $fh; # for win32 |
143 | |
cd59cad8 |
144 | if ($obj->_root->{autoflush}) { |
a20d9a3f |
145 | my $old = select $fh; |
146 | $|=1; |
147 | select $old; |
148 | } |
20f7b20c |
149 | |
cd59cad8 |
150 | seek($fh, 0 + $obj->_root->{file_offset}, SEEK_SET); |
a20d9a3f |
151 | my $signature; |
8db25060 |
152 | my $bytes_read = read( $fh, $signature, length(SIG_FILE)); |
20f7b20c |
153 | |
a20d9a3f |
154 | ## |
155 | # File is empty -- write signature and master index |
156 | ## |
157 | if (!$bytes_read) { |
cd59cad8 |
158 | seek($fh, 0 + $obj->_root->{file_offset}, SEEK_SET); |
8db25060 |
159 | print( $fh SIG_FILE); |
d0b74c17 |
160 | |
d5d7c51d |
161 | $self->create_tag( |
162 | $obj, $obj->_base_offset, $obj->_type, |
163 | chr(0) x $self->{index_size}, |
164 | ); |
a20d9a3f |
165 | |
a20d9a3f |
166 | # Flush the filehandle |
167 | my $old_fh = select $fh; |
168 | my $old_af = $|; $| = 1; $| = $old_af; |
169 | select $old_fh; |
170 | |
a20d9a3f |
171 | return 1; |
172 | } |
20f7b20c |
173 | |
a20d9a3f |
174 | ## |
175 | # Check signature was valid |
176 | ## |
8db25060 |
177 | unless ($signature eq SIG_FILE) { |
3d1b8be9 |
178 | $self->close_fh( $obj ); |
e5fc7e69 |
179 | $obj->_throw_error("Signature not found -- file is not a Deep DB"); |
a20d9a3f |
180 | } |
181 | |
a20d9a3f |
182 | ## |
183 | # Get our type from master index signature |
184 | ## |
e5fc7e69 |
185 | my $tag = $self->load_tag($obj, $obj->_base_offset) |
186 | or $obj->_throw_error("Corrupted file, no master index record"); |
d0b74c17 |
187 | |
e5fc7e69 |
188 | unless ($obj->{type} eq $tag->{signature}) { |
189 | $obj->_throw_error("File type mismatch"); |
a20d9a3f |
190 | } |
20f7b20c |
191 | |
d0b74c17 |
192 | #XXX We probably also want to store the hash algorithm name and not assume anything |
193 | #XXX The cool thing would be to allow a different hashing algorithm at every level |
194 | |
a20d9a3f |
195 | return 1; |
196 | } |
197 | |
3d1b8be9 |
198 | sub close_fh { |
cd59cad8 |
199 | my $self = shift; |
a21f2d90 |
200 | my ($obj) = @_; |
cd59cad8 |
201 | |
202 | if ( my $fh = $obj->_root->{fh} ) { |
203 | close $fh; |
204 | } |
205 | $obj->_root->{fh} = undef; |
206 | |
207 | return 1; |
208 | } |
209 | |
d4b1166e |
210 | sub create_tag { |
20f7b20c |
211 | ## |
212 | # Given offset, signature and content, create tag and write to disk |
213 | ## |
d4b1166e |
214 | my $self = shift; |
20f7b20c |
215 | my ($obj, $offset, $sig, $content) = @_; |
216 | my $size = length($content); |
217 | |
d4b1166e |
218 | my $fh = $obj->_fh; |
219 | |
20f7b20c |
220 | seek($fh, $offset + $obj->_root->{file_offset}, SEEK_SET); |
251dfd0e |
221 | print( $fh $sig . pack($self->{data_pack}, $size) . $content ); |
20f7b20c |
222 | |
223 | if ($offset == $obj->_root->{end}) { |
8db25060 |
224 | $obj->_root->{end} += SIG_SIZE + $self->{data_size} + $size; |
20f7b20c |
225 | } |
226 | |
227 | return { |
228 | signature => $sig, |
229 | size => $size, |
8db25060 |
230 | offset => $offset + SIG_SIZE + $self->{data_size}, |
20f7b20c |
231 | content => $content |
232 | }; |
d4b1166e |
233 | } |
234 | |
235 | sub load_tag { |
20f7b20c |
236 | ## |
237 | # Given offset, load single tag and return signature, size and data |
238 | ## |
d4b1166e |
239 | my $self = shift; |
20f7b20c |
240 | my ($obj, $offset) = @_; |
241 | |
d4b1166e |
242 | my $fh = $obj->_fh; |
243 | |
20f7b20c |
244 | seek($fh, $offset + $obj->_root->{file_offset}, SEEK_SET); |
e5fc7e69 |
245 | |
75be6413 |
246 | #XXX I'm not sure this check will work if autoflush isn't enabled ... |
e5fc7e69 |
247 | return if eof $fh; |
20f7b20c |
248 | |
d4b1166e |
249 | my $b; |
8db25060 |
250 | read( $fh, $b, SIG_SIZE + $self->{data_size} ); |
251dfd0e |
251 | my ($sig, $size) = unpack( "A $self->{data_pack}", $b ); |
20f7b20c |
252 | |
253 | my $buffer; |
254 | read( $fh, $buffer, $size); |
255 | |
256 | return { |
257 | signature => $sig, |
258 | size => $size, |
8db25060 |
259 | offset => $offset + SIG_SIZE + $self->{data_size}, |
20f7b20c |
260 | content => $buffer |
261 | }; |
d4b1166e |
262 | } |
263 | |
20f7b20c |
264 | sub add_bucket { |
265 | ## |
266 | # Adds one key/value pair to bucket list, given offset, MD5 digest of key, |
267 | # plain (undigested) key and value. |
268 | ## |
d4b1166e |
269 | my $self = shift; |
20f7b20c |
270 | my ($obj, $tag, $md5, $plain_key, $value) = @_; |
75be6413 |
271 | |
eea0d863 |
272 | # This verifies that only supported values will be stored. |
273 | { |
274 | my $r = Scalar::Util::reftype( $value ); |
275 | last if !defined $r; |
276 | |
277 | last if $r eq 'HASH'; |
278 | last if $r eq 'ARRAY'; |
279 | |
280 | $obj->_throw_error( |
281 | "Storage of variables of type '$r' is not supported." |
282 | ); |
283 | } |
284 | |
20f7b20c |
285 | my $location = 0; |
286 | my $result = 2; |
287 | |
288 | my $root = $obj->_root; |
289 | |
d5d7c51d |
290 | my $is_dbm_deep = eval { |
291 | local $SIG{'__DIE__'}; |
292 | $value->isa( 'DBM::Deep' ); |
293 | }; |
294 | |
20f7b20c |
295 | my $internal_ref = $is_dbm_deep && ($value->_root eq $root); |
296 | |
297 | my $fh = $obj->_fh; |
298 | |
386bab6c |
299 | my ($subloc, $offset) = $self->_find_in_buckets( $tag, $md5 ); |
75be6413 |
300 | |
386bab6c |
301 | # Updating a known md5 |
302 | if ( $subloc ) { |
303 | $result = 1; |
20f7b20c |
304 | |
386bab6c |
305 | ## |
d5d7c51d |
306 | # If value is a hash, array, or raw value with equal or less size, we |
307 | # can reuse the same content area of the database. Otherwise, we have |
308 | # to create a new content area at the EOF. |
386bab6c |
309 | ## |
310 | my $actual_length; |
311 | if ( $internal_ref ) { |
312 | $actual_length = $self->{long_size}; |
313 | } |
314 | else { |
315 | my $r = Scalar::Util::reftype( $value ) || ''; |
316 | if ( $r eq 'HASH' || $r eq 'ARRAY' ) { |
317 | $actual_length = $self->{index_size}; |
318 | |
319 | # if autobless is enabled, must also take into consideration |
320 | # the class name, as it is stored along with key/value. |
321 | if ( $root->{autobless} ) { |
322 | my $value_class = Scalar::Util::blessed($value); |
323 | if ( defined $value_class && !$value->isa('DBM::Deep') ) { |
324 | $actual_length += length($value_class); |
8db25060 |
325 | } |
75be6413 |
326 | } |
20f7b20c |
327 | } |
386bab6c |
328 | else { $actual_length = length($value); } |
75be6413 |
329 | } |
20f7b20c |
330 | |
d5d7c51d |
331 | seek($fh, $subloc + SIG_SIZE + $root->{file_offset}, SEEK_SET); |
332 | my $size; |
333 | read( $fh, $size, $self->{data_size}); |
334 | $size = unpack($self->{data_pack}, $size); |
335 | |
386bab6c |
336 | if ($actual_length <= $size) { |
337 | $location = $subloc; |
20f7b20c |
338 | } |
75be6413 |
339 | else { |
340 | $location = $root->{end}; |
386bab6c |
341 | seek( |
342 | $fh, |
343 | $tag->{offset} + $offset + $self->{hash_size} + $root->{file_offset}, |
344 | SEEK_SET, |
345 | ); |
346 | print( $fh pack($self->{long_pack}, $location) ); |
75be6413 |
347 | } |
75be6413 |
348 | } |
386bab6c |
349 | # Adding a new md5 |
350 | elsif ( defined $offset ) { |
386bab6c |
351 | $location = $root->{end}; |
352 | |
353 | seek( $fh, $tag->{offset} + $offset + $root->{file_offset}, SEEK_SET ); |
354 | print( $fh $md5 . pack($self->{long_pack}, $location) ); |
355 | } |
356 | # If bucket didn't fit into list, split into a new index level |
357 | else { |
358 | $self->split_index( $obj, $md5, $tag ); |
359 | |
360 | $location = $root->{end}; |
361 | } |
20f7b20c |
362 | |
d5d7c51d |
363 | $self->write_value( $obj, $location, $plain_key, $value ); |
364 | |
365 | return $result; |
366 | } |
367 | |
368 | sub write_value { |
369 | my $self = shift; |
370 | my ($obj, $location, $key, $value) = @_; |
371 | |
372 | my $fh = $obj->_fh; |
373 | my $root = $obj->_root; |
374 | |
375 | my $is_dbm_deep = eval { |
376 | local $SIG{'__DIE__'}; |
377 | $value->isa( 'DBM::Deep' ); |
378 | }; |
379 | |
380 | my $internal_ref = $is_dbm_deep && ($value->_root eq $root); |
381 | |
382 | seek($fh, $location + $root->{file_offset}, SEEK_SET); |
383 | |
20f7b20c |
384 | ## |
d5d7c51d |
385 | # Write signature based on content type, set content length and write |
386 | # actual value. |
20f7b20c |
387 | ## |
d5d7c51d |
388 | my $r = Scalar::Util::reftype($value) || ''; |
389 | my $content_length; |
390 | if ( $internal_ref ) { |
391 | print( $fh SIG_INTERNAL ); |
392 | print( $fh pack($self->{data_pack}, $self->{long_size}) ); |
393 | print( $fh pack($self->{long_pack}, $value->_base_offset) ); |
394 | $content_length = $self->{long_size}; |
395 | } |
396 | else { |
397 | if ($r eq 'HASH') { |
398 | print( $fh SIG_HASH ); |
399 | print( $fh pack($self->{data_pack}, $self->{index_size}) . chr(0) x $self->{index_size} ); |
400 | $content_length = $self->{index_size}; |
401 | } |
402 | elsif ($r eq 'ARRAY') { |
403 | print( $fh SIG_ARRAY ); |
404 | print( $fh pack($self->{data_pack}, $self->{index_size}) . chr(0) x $self->{index_size} ); |
405 | $content_length = $self->{index_size}; |
406 | } |
407 | elsif (!defined($value)) { |
408 | print( $fh SIG_NULL ); |
409 | print( $fh pack($self->{data_pack}, 0) ); |
410 | $content_length = 0; |
20f7b20c |
411 | } |
412 | else { |
d5d7c51d |
413 | print( $fh SIG_DATA ); |
414 | print( $fh pack($self->{data_pack}, length($value)) . $value ); |
415 | $content_length = length($value); |
20f7b20c |
416 | } |
d5d7c51d |
417 | } |
20f7b20c |
418 | |
d5d7c51d |
419 | ## |
420 | # Plain key is stored AFTER value, as keys are typically fetched less often. |
421 | ## |
422 | print( $fh pack($self->{data_pack}, length($key)) . $key ); |
20f7b20c |
423 | |
d5d7c51d |
424 | ## |
425 | # If value is blessed, preserve class name |
426 | ## |
427 | if ( $root->{autobless} ) { |
428 | my $value_class = Scalar::Util::blessed($value); |
429 | if ( defined $value_class && !$value->isa( 'DBM::Deep' ) ) { |
430 | ## |
431 | # Blessed ref -- will restore later |
432 | ## |
433 | print( $fh chr(1) ); |
434 | print( $fh pack($self->{data_pack}, length($value_class)) . $value_class ); |
435 | $content_length += 1; |
436 | $content_length += $self->{data_size} + length($value_class); |
20f7b20c |
437 | } |
d5d7c51d |
438 | else { |
439 | print( $fh chr(0) ); |
440 | $content_length += 1; |
20f7b20c |
441 | } |
d5d7c51d |
442 | } |
20f7b20c |
443 | |
d5d7c51d |
444 | ## |
445 | # If this is a new content area, advance EOF counter |
446 | ## |
447 | if ($location == $root->{end}) { |
448 | $root->{end} += SIG_SIZE; |
449 | $root->{end} += $self->{data_size} + $content_length; |
450 | $root->{end} += $self->{data_size} + length($key); |
451 | } |
452 | |
453 | ## |
454 | # If content is a hash or array, create new child DBM::Deep object and |
455 | # pass each key or element to it. |
456 | ## |
457 | if ( ! $internal_ref ) { |
458 | if ($r eq 'HASH') { |
459 | my $branch = DBM::Deep->new( |
460 | type => DBM::Deep->TYPE_HASH, |
461 | base_offset => $location, |
462 | root => $root, |
463 | ); |
464 | foreach my $key (keys %{$value}) { |
465 | $branch->STORE( $key, $value->{$key} ); |
20f7b20c |
466 | } |
d5d7c51d |
467 | } |
468 | elsif ($r eq 'ARRAY') { |
469 | my $branch = DBM::Deep->new( |
470 | type => DBM::Deep->TYPE_ARRAY, |
471 | base_offset => $location, |
472 | root => $root, |
473 | ); |
474 | my $index = 0; |
475 | foreach my $element (@{$value}) { |
476 | $branch->STORE( $index, $element ); |
477 | $index++; |
20f7b20c |
478 | } |
479 | } |
20f7b20c |
480 | } |
d4b1166e |
481 | |
d5d7c51d |
482 | return 1; |
d4b1166e |
483 | } |
484 | |
75be6413 |
485 | sub split_index { |
486 | my $self = shift; |
487 | my ($obj, $md5, $tag) = @_; |
488 | |
489 | my $fh = $obj->_fh; |
490 | my $root = $obj->_root; |
491 | my $keys = $tag->{content}; |
492 | |
493 | seek($fh, $tag->{ref_loc} + $root->{file_offset}, SEEK_SET); |
494 | print( $fh pack($self->{long_pack}, $root->{end}) ); |
495 | |
496 | my $index_tag = $self->create_tag( |
497 | $obj, |
498 | $root->{end}, |
8db25060 |
499 | SIG_INDEX, |
75be6413 |
500 | chr(0) x $self->{index_size}, |
501 | ); |
502 | |
503 | my @offsets = (); |
504 | |
505 | $keys .= $md5 . pack($self->{long_pack}, 0); |
506 | |
507 | BUCKET: |
508 | for (my $i = 0; $i <= $self->{max_buckets}; $i++) { |
9cec1360 |
509 | my ($key, $old_subloc) = $self->_get_key_subloc( $keys, $i ); |
75be6413 |
510 | |
511 | next BUCKET unless $key; |
512 | |
75be6413 |
513 | my $num = ord(substr($key, $tag->{ch} + 1, 1)); |
514 | |
515 | if ($offsets[$num]) { |
8db25060 |
516 | my $offset = $offsets[$num] + SIG_SIZE + $self->{data_size}; |
75be6413 |
517 | seek($fh, $offset + $root->{file_offset}, SEEK_SET); |
518 | my $subkeys; |
519 | read( $fh, $subkeys, $self->{bucket_list_size}); |
520 | |
521 | for (my $k=0; $k<$self->{max_buckets}; $k++) { |
9cec1360 |
522 | my ($temp, $subloc) = $self->_get_key_subloc( $subkeys, $k ); |
75be6413 |
523 | |
524 | if (!$subloc) { |
525 | seek($fh, $offset + ($k * $self->{bucket_size}) + $root->{file_offset}, SEEK_SET); |
526 | print( $fh $key . pack($self->{long_pack}, $old_subloc || $root->{end}) ); |
527 | last; |
528 | } |
529 | } # k loop |
530 | } |
531 | else { |
532 | $offsets[$num] = $root->{end}; |
533 | seek($fh, $index_tag->{offset} + ($num * $self->{long_size}) + $root->{file_offset}, SEEK_SET); |
534 | print( $fh pack($self->{long_pack}, $root->{end}) ); |
535 | |
8db25060 |
536 | my $blist_tag = $self->create_tag($obj, $root->{end}, SIG_BLIST, chr(0) x $self->{bucket_list_size}); |
75be6413 |
537 | |
538 | seek($fh, $blist_tag->{offset} + $root->{file_offset}, SEEK_SET); |
539 | print( $fh $key . pack($self->{long_pack}, $old_subloc || $root->{end}) ); |
540 | } |
541 | } # i loop |
542 | |
543 | return; |
544 | } |
545 | |
8db25060 |
546 | sub read_from_loc { |
547 | my $self = shift; |
548 | my ($obj, $subloc) = @_; |
549 | |
550 | my $fh = $obj->_fh; |
551 | |
552 | ## |
553 | # Found match -- seek to offset and read signature |
554 | ## |
555 | my $signature; |
556 | seek($fh, $subloc + $obj->_root->{file_offset}, SEEK_SET); |
557 | read( $fh, $signature, SIG_SIZE); |
558 | |
559 | ## |
560 | # If value is a hash or array, return new DBM::Deep object with correct offset |
561 | ## |
562 | if (($signature eq SIG_HASH) || ($signature eq SIG_ARRAY)) { |
563 | my $obj = DBM::Deep->new( |
564 | type => $signature, |
565 | base_offset => $subloc, |
566 | root => $obj->_root, |
567 | ); |
568 | |
569 | if ($obj->_root->{autobless}) { |
570 | ## |
571 | # Skip over value and plain key to see if object needs |
572 | # to be re-blessed |
573 | ## |
574 | seek($fh, $self->{data_size} + $self->{index_size}, SEEK_CUR); |
575 | |
576 | my $size; |
577 | read( $fh, $size, $self->{data_size}); $size = unpack($self->{data_pack}, $size); |
578 | if ($size) { seek($fh, $size, SEEK_CUR); } |
579 | |
580 | my $bless_bit; |
581 | read( $fh, $bless_bit, 1); |
582 | if (ord($bless_bit)) { |
583 | ## |
584 | # Yes, object needs to be re-blessed |
585 | ## |
586 | my $class_name; |
587 | read( $fh, $size, $self->{data_size}); $size = unpack($self->{data_pack}, $size); |
588 | if ($size) { read( $fh, $class_name, $size); } |
589 | if ($class_name) { $obj = bless( $obj, $class_name ); } |
590 | } |
591 | } |
592 | |
593 | return $obj; |
594 | } |
595 | elsif ( $signature eq SIG_INTERNAL ) { |
596 | my $size; |
597 | read( $fh, $size, $self->{data_size}); |
598 | $size = unpack($self->{data_pack}, $size); |
599 | |
600 | if ( $size ) { |
601 | my $new_loc; |
602 | read( $fh, $new_loc, $size ); |
603 | $new_loc = unpack( $self->{long_pack}, $new_loc ); |
604 | |
605 | return $self->read_from_loc( $obj, $new_loc ); |
606 | } |
607 | else { |
608 | return; |
609 | } |
610 | } |
611 | ## |
612 | # Otherwise return actual value |
613 | ## |
614 | elsif ($signature eq SIG_DATA) { |
615 | my $size; |
616 | read( $fh, $size, $self->{data_size}); |
617 | $size = unpack($self->{data_pack}, $size); |
618 | |
619 | my $value = ''; |
620 | if ($size) { read( $fh, $value, $size); } |
621 | return $value; |
622 | } |
623 | |
624 | ## |
625 | # Key exists, but content is null |
626 | ## |
627 | return; |
628 | } |
629 | |
9020ee8c |
630 | sub get_bucket_value { |
beac1dff |
631 | ## |
632 | # Fetch single value given tag and MD5 digested key. |
633 | ## |
634 | my $self = shift; |
635 | my ($obj, $tag, $md5) = @_; |
9020ee8c |
636 | |
386bab6c |
637 | my ($subloc, $offset) = $self->_find_in_buckets( $tag, $md5 ); |
638 | if ( $subloc ) { |
8db25060 |
639 | return $self->read_from_loc( $obj, $subloc ); |
386bab6c |
640 | } |
beac1dff |
641 | return; |
9020ee8c |
642 | } |
ab0e4957 |
643 | |
644 | sub delete_bucket { |
beac1dff |
645 | ## |
646 | # Delete single key/value pair given tag and MD5 digested key. |
647 | ## |
648 | my $self = shift; |
649 | my ($obj, $tag, $md5) = @_; |
ab0e4957 |
650 | |
386bab6c |
651 | my ($subloc, $offset) = $self->_find_in_buckets( $tag, $md5 ); |
652 | if ( $subloc ) { |
653 | my $fh = $obj->_fh; |
654 | seek($fh, $tag->{offset} + $offset + $obj->_root->{file_offset}, SEEK_SET); |
655 | print( $fh substr($tag->{content}, $offset + $self->{bucket_size} ) ); |
251dfd0e |
656 | print( $fh chr(0) x $self->{bucket_size} ); |
d0b74c17 |
657 | |
ab0e4957 |
658 | return 1; |
386bab6c |
659 | } |
beac1dff |
660 | return; |
ab0e4957 |
661 | } |
662 | |
912d50b1 |
663 | sub bucket_exists { |
beac1dff |
664 | ## |
665 | # Check existence of single key given tag and MD5 digested key. |
666 | ## |
667 | my $self = shift; |
668 | my ($obj, $tag, $md5) = @_; |
912d50b1 |
669 | |
386bab6c |
670 | my ($subloc, $offset) = $self->_find_in_buckets( $tag, $md5 ); |
d5d7c51d |
671 | return $subloc && 1; |
912d50b1 |
672 | } |
673 | |
6736c116 |
674 | sub find_bucket_list { |
beac1dff |
675 | ## |
676 | # Locate offset for bucket list, given digested key |
677 | ## |
678 | my $self = shift; |
d0b74c17 |
679 | my ($obj, $md5, $args) = @_; |
680 | $args = {} unless $args; |
681 | |
beac1dff |
682 | ## |
683 | # Locate offset for bucket list using digest index system |
684 | ## |
e5fc7e69 |
685 | my $tag = $self->load_tag($obj, $obj->_base_offset) |
d5d7c51d |
686 | or $obj->_throw_error( "INTERNAL ERROR - Cannot find tag" ); |
d0b74c17 |
687 | |
e5fc7e69 |
688 | my $ch = 0; |
8db25060 |
689 | while ($tag->{signature} ne SIG_BLIST) { |
d0b74c17 |
690 | my $num = ord substr($md5, $ch, 1); |
691 | |
692 | my $ref_loc = $tag->{offset} + ($num * $self->{long_size}); |
693 | $tag = $self->index_lookup( $obj, $tag, $num ); |
694 | |
695 | if (!$tag) { |
d5d7c51d |
696 | return if ! $args->{create}; |
d0b74c17 |
697 | |
d5d7c51d |
698 | my $fh = $obj->_fh; |
699 | seek($fh, $ref_loc + $obj->_root->{file_offset}, SEEK_SET); |
700 | print( $fh pack($self->{long_pack}, $obj->_root->{end}) ); |
d0b74c17 |
701 | |
d5d7c51d |
702 | $tag = $self->create_tag( |
703 | $obj, $obj->_root->{end}, |
704 | SIG_BLIST, |
705 | chr(0) x $self->{bucket_list_size}, |
706 | ); |
707 | |
708 | $tag->{ref_loc} = $ref_loc; |
709 | $tag->{ch} = $ch; |
710 | |
711 | last; |
d0b74c17 |
712 | } |
713 | |
714 | $tag->{ch} = $ch; |
715 | $tag->{ref_loc} = $ref_loc; |
716 | |
beac1dff |
717 | $ch++; |
718 | } |
d0b74c17 |
719 | |
beac1dff |
720 | return $tag; |
6736c116 |
721 | } |
722 | |
d0b74c17 |
723 | sub index_lookup { |
724 | ## |
725 | # Given index tag, lookup single entry in index and return . |
726 | ## |
727 | my $self = shift; |
728 | my ($obj, $tag, $index) = @_; |
729 | |
730 | my $location = unpack( |
731 | $self->{long_pack}, |
732 | substr( |
733 | $tag->{content}, |
734 | $index * $self->{long_size}, |
735 | $self->{long_size}, |
736 | ), |
737 | ); |
738 | |
739 | if (!$location) { return; } |
740 | |
741 | return $self->load_tag( $obj, $location ); |
742 | } |
743 | |
6736c116 |
744 | sub traverse_index { |
beac1dff |
745 | ## |
746 | # Scan index and recursively step into deeper levels, looking for next key. |
747 | ## |
6736c116 |
748 | my $self = shift; |
749 | my ($obj, $offset, $ch, $force_return_next) = @_; |
d0b74c17 |
750 | |
beac1dff |
751 | my $tag = $self->load_tag($obj, $offset ); |
6736c116 |
752 | |
753 | my $fh = $obj->_fh; |
d0b74c17 |
754 | |
8db25060 |
755 | if ($tag->{signature} ne SIG_BLIST) { |
beac1dff |
756 | my $content = $tag->{content}; |
e5fc7e69 |
757 | my $start = $obj->{return_next} ? 0 : ord(substr($obj->{prev_md5}, $ch, 1)); |
d0b74c17 |
758 | |
d5d7c51d |
759 | for (my $idx = $start; $idx < (2**8); $idx++) { |
e5fc7e69 |
760 | my $subloc = unpack( |
761 | $self->{long_pack}, |
d5d7c51d |
762 | substr($content, $idx * $self->{long_size}, $self->{long_size}), |
e5fc7e69 |
763 | ); |
764 | |
beac1dff |
765 | if ($subloc) { |
e5fc7e69 |
766 | my $result = $self->traverse_index( |
767 | $obj, $subloc, $ch + 1, $force_return_next, |
768 | ); |
769 | |
beac1dff |
770 | if (defined($result)) { return $result; } |
771 | } |
772 | } # index loop |
d0b74c17 |
773 | |
beac1dff |
774 | $obj->{return_next} = 1; |
775 | } # tag is an index |
d0b74c17 |
776 | |
e5fc7e69 |
777 | else { |
beac1dff |
778 | my $keys = $tag->{content}; |
779 | if ($force_return_next) { $obj->{return_next} = 1; } |
d0b74c17 |
780 | |
beac1dff |
781 | ## |
782 | # Iterate through buckets, looking for a key match |
783 | ## |
8db25060 |
784 | for (my $i = 0; $i < $self->{max_buckets}; $i++) { |
9cec1360 |
785 | my ($key, $subloc) = $self->_get_key_subloc( $keys, $i ); |
d0b74c17 |
786 | |
8db25060 |
787 | # End of bucket list -- return to outer loop |
beac1dff |
788 | if (!$subloc) { |
beac1dff |
789 | $obj->{return_next} = 1; |
790 | last; |
791 | } |
8db25060 |
792 | # Located previous key -- return next one found |
beac1dff |
793 | elsif ($key eq $obj->{prev_md5}) { |
beac1dff |
794 | $obj->{return_next} = 1; |
795 | next; |
796 | } |
8db25060 |
797 | # Seek to bucket location and skip over signature |
beac1dff |
798 | elsif ($obj->{return_next}) { |
8db25060 |
799 | seek($fh, $subloc + $obj->_root->{file_offset}, SEEK_SET); |
d0b74c17 |
800 | |
beac1dff |
801 | # Skip over value to get to plain key |
8db25060 |
802 | my $sig; |
803 | read( $fh, $sig, SIG_SIZE ); |
804 | |
beac1dff |
805 | my $size; |
e5fc7e69 |
806 | read( $fh, $size, $self->{data_size}); |
807 | $size = unpack($self->{data_pack}, $size); |
beac1dff |
808 | if ($size) { seek($fh, $size, SEEK_CUR); } |
d0b74c17 |
809 | |
beac1dff |
810 | # Read in plain key and return as scalar |
beac1dff |
811 | my $plain_key; |
e5fc7e69 |
812 | read( $fh, $size, $self->{data_size}); |
813 | $size = unpack($self->{data_pack}, $size); |
beac1dff |
814 | if ($size) { read( $fh, $plain_key, $size); } |
d0b74c17 |
815 | |
beac1dff |
816 | return $plain_key; |
817 | } |
8db25060 |
818 | } |
d0b74c17 |
819 | |
beac1dff |
820 | $obj->{return_next} = 1; |
821 | } # tag is a bucket list |
d0b74c17 |
822 | |
beac1dff |
823 | return; |
6736c116 |
824 | } |
825 | |
826 | sub get_next_key { |
beac1dff |
827 | ## |
828 | # Locate next key, given digested previous one |
829 | ## |
6736c116 |
830 | my $self = shift; |
831 | my ($obj) = @_; |
d0b74c17 |
832 | |
beac1dff |
833 | $obj->{prev_md5} = $_[1] ? $_[1] : undef; |
834 | $obj->{return_next} = 0; |
d0b74c17 |
835 | |
beac1dff |
836 | ## |
837 | # If the previous key was not specifed, start at the top and |
838 | # return the first one found. |
839 | ## |
840 | if (!$obj->{prev_md5}) { |
841 | $obj->{prev_md5} = chr(0) x $self->{hash_size}; |
842 | $obj->{return_next} = 1; |
843 | } |
d0b74c17 |
844 | |
beac1dff |
845 | return $self->traverse_index( $obj, $obj->_base_offset, 0 ); |
6736c116 |
846 | } |
847 | |
75be6413 |
848 | # Utilities |
849 | |
9cec1360 |
850 | sub _get_key_subloc { |
75be6413 |
851 | my $self = shift; |
852 | my ($keys, $idx) = @_; |
853 | |
9cec1360 |
854 | my ($key, $subloc) = unpack( |
855 | "a$self->{hash_size} $self->{long_pack}", |
75be6413 |
856 | substr( |
857 | $keys, |
9cec1360 |
858 | ($idx * $self->{bucket_size}), |
859 | $self->{bucket_size}, |
75be6413 |
860 | ), |
861 | ); |
862 | |
9cec1360 |
863 | return ($key, $subloc); |
75be6413 |
864 | } |
865 | |
d608b06e |
866 | sub _find_in_buckets { |
867 | my $self = shift; |
868 | my ($tag, $md5) = @_; |
869 | |
870 | BUCKET: |
871 | for ( my $i = 0; $i < $self->{max_buckets}; $i++ ) { |
872 | my ($key, $subloc) = $self->_get_key_subloc( $tag->{content}, $i ); |
873 | |
874 | return ($subloc, $i * $self->{bucket_size}) unless $subloc; |
875 | |
876 | next BUCKET if $key ne $md5; |
877 | |
878 | return ($subloc, $i * $self->{bucket_size}); |
879 | } |
880 | |
881 | return; |
882 | } |
883 | |
a20d9a3f |
884 | 1; |
885 | __END__ |
d5d7c51d |
886 | |
887 | # This will be added in later, after more refactoring is done. This is an early |
888 | # attempt at refactoring on the physical level instead of the virtual level. |
889 | sub _read_at { |
890 | my $self = shift; |
891 | my ($obj, $spot, $amount, $unpack) = @_; |
892 | |
893 | my $fh = $obj->_fh; |
894 | seek( $fh, $spot + $obj->_root->{file_offset}, SEEK_SET ); |
895 | |
896 | my $buffer; |
897 | my $bytes_read = read( $fh, $buffer, $amount ); |
898 | |
899 | if ( $unpack ) { |
900 | $buffer = unpack( $unpack, $buffer ); |
901 | } |
902 | |
903 | if ( wantarray ) { |
904 | return ($buffer, $bytes_read); |
905 | } |
906 | else { |
907 | return $buffer; |
908 | } |
909 | } |