Commit | Line | Data |
596596d5 |
1 | ;# $Id: Storable.pm,v 1.0.1.13 2001/12/01 13:34:49 ram Exp $ |
7a6a85bf |
2 | ;# |
3 | ;# Copyright (c) 1995-2000, Raphael Manfredi |
4 | ;# |
9e21b3d0 |
5 | ;# You may redistribute only under the same terms as Perl 5, as specified |
6 | ;# in the README file that comes with the distribution. |
7a6a85bf |
7 | ;# |
8 | ;# $Log: Storable.pm,v $ |
596596d5 |
9 | ;# Revision 1.0.1.13 2001/12/01 13:34:49 ram |
10 | ;# patch14: avoid requiring Fcntl upfront, useful to embedded runtimes |
11 | ;# patch14: store_fd() will now correctly autoflush file if needed |
12 | ;# |
6e0ac6f5 |
13 | ;# Revision 1.0.1.12 2001/08/28 21:51:51 ram |
14 | ;# patch13: fixed truncation race with lock_retrieve() in lock_store() |
15 | ;# |
e993d95c |
16 | ;# Revision 1.0.1.11 2001/07/01 11:22:14 ram |
17 | ;# patch12: systematically use "=over 4" for POD linters |
18 | ;# patch12: updated version number |
19 | ;# |
8be2b38b |
20 | ;# Revision 1.0.1.10 2001/03/15 00:20:25 ram |
21 | ;# patch11: updated version number |
22 | ;# |
23 | ;# Revision 1.0.1.9 2001/02/17 12:37:32 ram |
24 | ;# patch10: forgot to increase version number at previous patch |
25 | ;# |
b12202d0 |
26 | ;# Revision 1.0.1.8 2001/02/17 12:24:37 ram |
27 | ;# patch8: fixed incorrect error message |
28 | ;# |
862382c7 |
29 | ;# Revision 1.0.1.7 2001/01/03 09:39:02 ram |
30 | ;# patch7: added CAN_FLOCK to determine whether we can flock() or not |
31 | ;# |
90826881 |
32 | ;# Revision 1.0.1.6 2000/11/05 17:20:25 ram |
33 | ;# patch6: increased version number |
34 | ;# |
212e9bde |
35 | ;# Revision 1.0.1.5 2000/10/26 17:10:18 ram |
36 | ;# patch5: documented that store() and retrieve() can return undef |
37 | ;# patch5: added paragraph explaining the auto require for thaw hooks |
38 | ;# |
39 | ;# Revision 1.0.1.4 2000/10/23 18:02:57 ram |
40 | ;# patch4: protected calls to flock() for dos platform |
41 | ;# patch4: added logcarp emulation if they don't have Log::Agent |
42 | ;# |
8be2b38b |
43 | ;# Revision 1.0.1.3 2000/09/29 19:49:01 ram |
44 | ;# patch3: updated version number |
45 | ;# |
46 | ;# Revision 1.0.1.2 2000/09/28 21:42:51 ram |
47 | ;# patch2: added lock_store lock_nstore lock_retrieve |
48 | ;# |
49 | ;# Revision 1.0.1.1 2000/09/17 16:46:21 ram |
50 | ;# patch1: documented that doubles are stringified by nstore() |
51 | ;# patch1: added Salvador Ortiz Garcia in CREDITS section |
52 | ;# |
9e21b3d0 |
53 | ;# Revision 1.0 2000/09/01 19:40:41 ram |
54 | ;# Baseline for first official release. |
7a6a85bf |
55 | ;# |
56 | |
57 | require DynaLoader; |
58 | require Exporter; |
59 | package Storable; @ISA = qw(Exporter DynaLoader); |
60 | |
61 | @EXPORT = qw(store retrieve); |
62 | @EXPORT_OK = qw( |
9e21b3d0 |
63 | nstore store_fd nstore_fd fd_retrieve |
7a6a85bf |
64 | freeze nfreeze thaw |
65 | dclone |
9e21b3d0 |
66 | retrieve_fd |
dd19458b |
67 | lock_store lock_nstore lock_retrieve |
7a6a85bf |
68 | ); |
69 | |
70 | use AutoLoader; |
71 | use vars qw($forgive_me $VERSION); |
72 | |
0a0da639 |
73 | $VERSION = '1.015'; |
7a6a85bf |
74 | *AUTOLOAD = \&AutoLoader::AUTOLOAD; # Grrr... |
75 | |
76 | # |
77 | # Use of Log::Agent is optional |
78 | # |
79 | |
80 | eval "use Log::Agent"; |
81 | |
530b72ba |
82 | require Carp; |
7a6a85bf |
83 | |
dd19458b |
84 | # |
85 | # They might miss :flock in Fcntl |
86 | # |
87 | |
88 | BEGIN { |
596596d5 |
89 | if (eval { require Fcntl; 1 } && exists $Fcntl::EXPORT_TAGS{'flock'}) { |
dd19458b |
90 | Fcntl->import(':flock'); |
91 | } else { |
92 | eval q{ |
93 | sub LOCK_SH () {1} |
94 | sub LOCK_EX () {2} |
95 | }; |
96 | } |
97 | } |
98 | |
b8778c7c |
99 | # Can't Autoload cleanly as this clashes 8.3 with &retrieve |
9e21b3d0 |
100 | sub retrieve_fd { &fd_retrieve } # Backward compatibility |
cb3d9de5 |
101 | |
530b72ba |
102 | # By default restricted hashes are downgraded on earlier perls. |
103 | |
104 | $Storable::downgrade_restricted = 1; |
e8189732 |
105 | $Storable::accept_future_minor = 1; |
b8778c7c |
106 | bootstrap Storable; |
107 | 1; |
108 | __END__ |
530b72ba |
109 | # |
110 | # Use of Log::Agent is optional. If it hasn't imported these subs then |
111 | # Autoloader will kindly supply our fallback implementation. |
112 | # |
113 | |
114 | sub logcroak { |
115 | Carp::croak(@_); |
116 | } |
117 | |
118 | sub logcarp { |
119 | Carp::carp(@_); |
120 | } |
b8778c7c |
121 | |
862382c7 |
122 | # |
123 | # Determine whether locking is possible, but only when needed. |
124 | # |
125 | |
530b72ba |
126 | sub CAN_FLOCK; my $CAN_FLOCK; sub CAN_FLOCK { |
862382c7 |
127 | return $CAN_FLOCK if defined $CAN_FLOCK; |
128 | require Config; import Config; |
129 | return $CAN_FLOCK = |
130 | $Config{'d_flock'} || |
131 | $Config{'d_fcntl_can_lock'} || |
132 | $Config{'d_lockf'}; |
133 | } |
134 | |
0a0da639 |
135 | sub show_file_magic { |
136 | print <<EOM; |
137 | # |
138 | # To recognize the data files of the Perl module Storable, |
139 | # the following lines need to be added to the local magic(5) file, |
140 | # usually either /usr/share/misc/magic or /etc/magic. |
0a0da639 |
141 | # |
142 | 0 string perl-store perl Storable(v0.6) data |
8b793558 |
143 | >4 byte >0 (net-order %d) |
144 | >>4 byte &01 (network-ordered) |
145 | >>4 byte =3 (major 1) |
146 | >>4 byte =2 (major 1) |
147 | |
0a0da639 |
148 | 0 string pst0 perl Storable(v0.7) data |
8b793558 |
149 | >4 byte >0 |
150 | >>4 byte &01 (network-ordered) |
151 | >>4 byte =5 (major 2) |
152 | >>4 byte =4 (major 2) |
153 | >>5 byte >0 (minor %d) |
0a0da639 |
154 | EOM |
155 | } |
156 | |
b8778c7c |
157 | sub read_magic { |
158 | my $header = shift; |
159 | return unless defined $header and length $header > 11; |
160 | my $result; |
161 | if ($header =~ s/^perl-store//) { |
162 | die "Can't deal with version 0 headers"; |
163 | } elsif ($header =~ s/^pst0//) { |
164 | $result->{file} = 1; |
165 | } |
166 | # Assume it's a string. |
167 | my ($major, $minor, $bytelen) = unpack "C3", $header; |
168 | |
169 | my $net_order = $major & 1; |
170 | $major >>= 1; |
171 | @$result{qw(major minor netorder)} = ($major, $minor, $net_order); |
172 | |
173 | return $result if $net_order; |
174 | |
175 | # I assume that it is rare to find v1 files, so this is an intentionally |
176 | # inefficient way of doing it, to make the rest of the code constant. |
177 | if ($major < 2) { |
178 | delete $result->{minor}; |
179 | $header = '.' . $header; |
180 | $bytelen = $minor; |
181 | } |
182 | |
183 | @$result{qw(byteorder intsize longsize ptrsize)} = |
184 | unpack "x3 A$bytelen C3", $header; |
185 | |
186 | if ($major >= 2 and $minor >= 2) { |
187 | $result->{nvsize} = unpack "x6 x$bytelen C", $header; |
188 | } |
189 | $result; |
190 | } |
7a6a85bf |
191 | |
192 | # |
193 | # store |
194 | # |
195 | # Store target object hierarchy, identified by a reference to its root. |
196 | # The stored object tree may later be retrieved to memory via retrieve. |
197 | # Returns undef if an I/O error occurred, in which case the file is |
198 | # removed. |
199 | # |
200 | sub store { |
dd19458b |
201 | return _store(\&pstore, @_, 0); |
7a6a85bf |
202 | } |
203 | |
204 | # |
205 | # nstore |
206 | # |
207 | # Same as store, but in network order. |
208 | # |
209 | sub nstore { |
dd19458b |
210 | return _store(\&net_pstore, @_, 0); |
211 | } |
212 | |
213 | # |
214 | # lock_store |
215 | # |
216 | # Same as store, but flock the file first (advisory locking). |
217 | # |
218 | sub lock_store { |
219 | return _store(\&pstore, @_, 1); |
220 | } |
221 | |
222 | # |
223 | # lock_nstore |
224 | # |
225 | # Same as nstore, but flock the file first (advisory locking). |
226 | # |
227 | sub lock_nstore { |
228 | return _store(\&net_pstore, @_, 1); |
7a6a85bf |
229 | } |
230 | |
231 | # Internal store to file routine |
232 | sub _store { |
233 | my $xsptr = shift; |
234 | my $self = shift; |
dd19458b |
235 | my ($file, $use_locking) = @_; |
7a6a85bf |
236 | logcroak "not a reference" unless ref($self); |
b12202d0 |
237 | logcroak "wrong argument number" unless @_ == 2; # No @foo in arglist |
7a6a85bf |
238 | local *FILE; |
dd19458b |
239 | if ($use_locking) { |
6e0ac6f5 |
240 | open(FILE, ">>$file") || logcroak "can't write into $file: $!"; |
862382c7 |
241 | unless (&CAN_FLOCK) { |
b29b780f |
242 | logcarp "Storable::lock_store: fcntl/flock emulation broken on $^O"; |
243 | return undef; |
f567092b |
244 | } |
dd19458b |
245 | flock(FILE, LOCK_EX) || |
246 | logcroak "can't get exclusive lock on $file: $!"; |
247 | truncate FILE, 0; |
248 | # Unlocking will happen when FILE is closed |
6e0ac6f5 |
249 | } else { |
250 | open(FILE, ">$file") || logcroak "can't create $file: $!"; |
dd19458b |
251 | } |
6e0ac6f5 |
252 | binmode FILE; # Archaic systems... |
7a6a85bf |
253 | my $da = $@; # Don't mess if called from exception handler |
254 | my $ret; |
255 | # Call C routine nstore or pstore, depending on network order |
256 | eval { $ret = &$xsptr(*FILE, $self) }; |
257 | close(FILE) or $ret = undef; |
258 | unlink($file) or warn "Can't unlink $file: $!\n" if $@ || !defined $ret; |
259 | logcroak $@ if $@ =~ s/\.?\n$/,/; |
260 | $@ = $da; |
261 | return $ret ? $ret : undef; |
262 | } |
263 | |
264 | # |
265 | # store_fd |
266 | # |
267 | # Same as store, but perform on an already opened file descriptor instead. |
268 | # Returns undef if an I/O error occurred. |
269 | # |
270 | sub store_fd { |
271 | return _store_fd(\&pstore, @_); |
272 | } |
273 | |
274 | # |
275 | # nstore_fd |
276 | # |
277 | # Same as store_fd, but in network order. |
278 | # |
279 | sub nstore_fd { |
280 | my ($self, $file) = @_; |
281 | return _store_fd(\&net_pstore, @_); |
282 | } |
283 | |
284 | # Internal store routine on opened file descriptor |
285 | sub _store_fd { |
286 | my $xsptr = shift; |
287 | my $self = shift; |
288 | my ($file) = @_; |
289 | logcroak "not a reference" unless ref($self); |
290 | logcroak "too many arguments" unless @_ == 1; # No @foo in arglist |
291 | my $fd = fileno($file); |
292 | logcroak "not a valid file descriptor" unless defined $fd; |
293 | my $da = $@; # Don't mess if called from exception handler |
294 | my $ret; |
295 | # Call C routine nstore or pstore, depending on network order |
296 | eval { $ret = &$xsptr($file, $self) }; |
297 | logcroak $@ if $@ =~ s/\.?\n$/,/; |
596596d5 |
298 | local $\; print $file ''; # Autoflush the file if wanted |
7a6a85bf |
299 | $@ = $da; |
300 | return $ret ? $ret : undef; |
301 | } |
302 | |
303 | # |
304 | # freeze |
305 | # |
306 | # Store oject and its hierarchy in memory and return a scalar |
307 | # containing the result. |
308 | # |
309 | sub freeze { |
310 | _freeze(\&mstore, @_); |
311 | } |
312 | |
313 | # |
314 | # nfreeze |
315 | # |
316 | # Same as freeze but in network order. |
317 | # |
318 | sub nfreeze { |
319 | _freeze(\&net_mstore, @_); |
320 | } |
321 | |
322 | # Internal freeze routine |
323 | sub _freeze { |
324 | my $xsptr = shift; |
325 | my $self = shift; |
326 | logcroak "not a reference" unless ref($self); |
327 | logcroak "too many arguments" unless @_ == 0; # No @foo in arglist |
328 | my $da = $@; # Don't mess if called from exception handler |
329 | my $ret; |
330 | # Call C routine mstore or net_mstore, depending on network order |
331 | eval { $ret = &$xsptr($self) }; |
332 | logcroak $@ if $@ =~ s/\.?\n$/,/; |
333 | $@ = $da; |
334 | return $ret ? $ret : undef; |
335 | } |
336 | |
337 | # |
338 | # retrieve |
339 | # |
340 | # Retrieve object hierarchy from disk, returning a reference to the root |
341 | # object of that tree. |
342 | # |
343 | sub retrieve { |
dd19458b |
344 | _retrieve($_[0], 0); |
345 | } |
346 | |
347 | # |
348 | # lock_retrieve |
349 | # |
350 | # Same as retrieve, but with advisory locking. |
351 | # |
352 | sub lock_retrieve { |
353 | _retrieve($_[0], 1); |
354 | } |
355 | |
356 | # Internal retrieve routine |
357 | sub _retrieve { |
358 | my ($file, $use_locking) = @_; |
7a6a85bf |
359 | local *FILE; |
dd19458b |
360 | open(FILE, $file) || logcroak "can't open $file: $!"; |
7a6a85bf |
361 | binmode FILE; # Archaic systems... |
362 | my $self; |
363 | my $da = $@; # Could be from exception handler |
dd19458b |
364 | if ($use_locking) { |
862382c7 |
365 | unless (&CAN_FLOCK) { |
8be2b38b |
366 | logcarp "Storable::lock_store: fcntl/flock emulation broken on $^O"; |
b29b780f |
367 | return undef; |
368 | } |
8be2b38b |
369 | flock(FILE, LOCK_SH) || logcroak "can't get shared lock on $file: $!"; |
dd19458b |
370 | # Unlocking will happen when FILE is closed |
371 | } |
7a6a85bf |
372 | eval { $self = pretrieve(*FILE) }; # Call C routine |
373 | close(FILE); |
374 | logcroak $@ if $@ =~ s/\.?\n$/,/; |
375 | $@ = $da; |
376 | return $self; |
377 | } |
378 | |
379 | # |
9e21b3d0 |
380 | # fd_retrieve |
7a6a85bf |
381 | # |
382 | # Same as retrieve, but perform from an already opened file descriptor instead. |
383 | # |
9e21b3d0 |
384 | sub fd_retrieve { |
7a6a85bf |
385 | my ($file) = @_; |
386 | my $fd = fileno($file); |
387 | logcroak "not a valid file descriptor" unless defined $fd; |
388 | my $self; |
389 | my $da = $@; # Could be from exception handler |
390 | eval { $self = pretrieve($file) }; # Call C routine |
391 | logcroak $@ if $@ =~ s/\.?\n$/,/; |
392 | $@ = $da; |
393 | return $self; |
394 | } |
395 | |
396 | # |
397 | # thaw |
398 | # |
399 | # Recreate objects in memory from an existing frozen image created |
400 | # by freeze. If the frozen image passed is undef, return undef. |
401 | # |
402 | sub thaw { |
403 | my ($frozen) = @_; |
404 | return undef unless defined $frozen; |
405 | my $self; |
406 | my $da = $@; # Could be from exception handler |
407 | eval { $self = mretrieve($frozen) }; # Call C routine |
408 | logcroak $@ if $@ =~ s/\.?\n$/,/; |
409 | $@ = $da; |
410 | return $self; |
411 | } |
412 | |
413 | =head1 NAME |
414 | |
415 | Storable - persistency for perl data structures |
416 | |
417 | =head1 SYNOPSIS |
418 | |
419 | use Storable; |
420 | store \%table, 'file'; |
421 | $hashref = retrieve('file'); |
422 | |
423 | use Storable qw(nstore store_fd nstore_fd freeze thaw dclone); |
424 | |
425 | # Network order |
426 | nstore \%table, 'file'; |
427 | $hashref = retrieve('file'); # There is NO nretrieve() |
428 | |
429 | # Storing to and retrieving from an already opened file |
430 | store_fd \@array, \*STDOUT; |
431 | nstore_fd \%table, \*STDOUT; |
9e21b3d0 |
432 | $aryref = fd_retrieve(\*SOCKET); |
433 | $hashref = fd_retrieve(\*SOCKET); |
7a6a85bf |
434 | |
435 | # Serializing to memory |
436 | $serialized = freeze \%table; |
437 | %table_clone = %{ thaw($serialized) }; |
438 | |
439 | # Deep (recursive) cloning |
440 | $cloneref = dclone($ref); |
441 | |
dd19458b |
442 | # Advisory locking |
443 | use Storable qw(lock_store lock_nstore lock_retrieve) |
444 | lock_store \%table, 'file'; |
445 | lock_nstore \%table, 'file'; |
446 | $hashref = lock_retrieve('file'); |
447 | |
7a6a85bf |
448 | =head1 DESCRIPTION |
449 | |
c261f00e |
450 | The Storable package brings persistence to your perl data structures |
7a6a85bf |
451 | containing SCALAR, ARRAY, HASH or REF objects, i.e. anything that can be |
c261f00e |
452 | conveniently stored to disk and retrieved at a later time. |
7a6a85bf |
453 | |
454 | It can be used in the regular procedural way by calling C<store> with |
455 | a reference to the object to be stored, along with the file name where |
456 | the image should be written. |
457 | The routine returns C<undef> for I/O problems or other internal error, |
458 | a true value otherwise. Serious errors are propagated as a C<die> exception. |
459 | |
460 | To retrieve data stored to disk, use C<retrieve> with a file name, |
461 | and the objects stored into that file are recreated into memory for you, |
462 | a I<reference> to the root object being returned. In case an I/O error |
463 | occurs while reading, C<undef> is returned instead. Other serious |
464 | errors are propagated via C<die>. |
465 | |
466 | Since storage is performed recursively, you might want to stuff references |
467 | to objects that share a lot of common data into a single array or hash |
468 | table, and then store that object. That way, when you retrieve back the |
469 | whole thing, the objects will continue to share what they originally shared. |
470 | |
471 | At the cost of a slight header overhead, you may store to an already |
472 | opened file descriptor using the C<store_fd> routine, and retrieve |
9e21b3d0 |
473 | from a file via C<fd_retrieve>. Those names aren't imported by default, |
c261f00e |
474 | so you will have to do that explicitly if you need those routines. |
7a6a85bf |
475 | The file descriptor you supply must be already opened, for read |
476 | if you're going to retrieve and for write if you wish to store. |
477 | |
478 | store_fd(\%table, *STDOUT) || die "can't store to stdout\n"; |
9e21b3d0 |
479 | $hashref = fd_retrieve(*STDIN); |
7a6a85bf |
480 | |
481 | You can also store data in network order to allow easy sharing across |
482 | multiple platforms, or when storing on a socket known to be remotely |
483 | connected. The routines to call have an initial C<n> prefix for I<network>, |
484 | as in C<nstore> and C<nstore_fd>. At retrieval time, your data will be |
485 | correctly restored so you don't have to know whether you're restoring |
dd19458b |
486 | from native or network ordered data. Double values are stored stringified |
487 | to ensure portability as well, at the slight risk of loosing some precision |
488 | in the last decimals. |
7a6a85bf |
489 | |
9e21b3d0 |
490 | When using C<fd_retrieve>, objects are retrieved in sequence, one |
7a6a85bf |
491 | object (i.e. one recursive tree) per associated C<store_fd>. |
492 | |
493 | If you're more from the object-oriented camp, you can inherit from |
494 | Storable and directly store your objects by invoking C<store> as |
495 | a method. The fact that the root of the to-be-stored tree is a |
496 | blessed reference (i.e. an object) is special-cased so that the |
497 | retrieve does not provide a reference to that object but rather the |
498 | blessed object reference itself. (Otherwise, you'd get a reference |
499 | to that blessed object). |
500 | |
501 | =head1 MEMORY STORE |
502 | |
503 | The Storable engine can also store data into a Perl scalar instead, to |
504 | later retrieve them. This is mainly used to freeze a complex structure in |
505 | some safe compact memory place (where it can possibly be sent to another |
506 | process via some IPC, since freezing the structure also serializes it in |
507 | effect). Later on, and maybe somewhere else, you can thaw the Perl scalar |
508 | out and recreate the original complex structure in memory. |
509 | |
510 | Surprisingly, the routines to be called are named C<freeze> and C<thaw>. |
511 | If you wish to send out the frozen scalar to another machine, use |
512 | C<nfreeze> instead to get a portable image. |
513 | |
514 | Note that freezing an object structure and immediately thawing it |
515 | actually achieves a deep cloning of that structure: |
516 | |
517 | dclone(.) = thaw(freeze(.)) |
518 | |
519 | Storable provides you with a C<dclone> interface which does not create |
520 | that intermediary scalar but instead freezes the structure in some |
c261f00e |
521 | internal memory space and then immediately thaws it out. |
7a6a85bf |
522 | |
dd19458b |
523 | =head1 ADVISORY LOCKING |
524 | |
525 | The C<lock_store> and C<lock_nstore> routine are equivalent to C<store> |
526 | and C<nstore>, only they get an exclusive lock on the file before |
527 | writing. Likewise, C<lock_retrieve> performs as C<retrieve>, but also |
528 | gets a shared lock on the file before reading. |
529 | |
530 | Like with any advisory locking scheme, the protection only works if |
531 | you systematically use C<lock_store> and C<lock_retrieve>. If one |
532 | side of your application uses C<store> whilst the other uses C<lock_retrieve>, |
533 | you will get no protection at all. |
534 | |
535 | The internal advisory locking is implemented using Perl's flock() routine. |
536 | If your system does not support any form of flock(), or if you share |
537 | your files across NFS, you might wish to use other forms of locking by |
538 | using modules like LockFile::Simple which lock a file using a filesystem |
539 | entry, instead of locking the file descriptor. |
540 | |
7a6a85bf |
541 | =head1 SPEED |
542 | |
543 | The heart of Storable is written in C for decent speed. Extra low-level |
4d3295e3 |
544 | optimizations have been made when manipulating perl internals, to |
545 | sacrifice encapsulation for the benefit of greater speed. |
7a6a85bf |
546 | |
547 | =head1 CANONICAL REPRESENTATION |
548 | |
549 | Normally Storable stores elements of hashes in the order they are |
550 | stored internally by Perl, i.e. pseudo-randomly. If you set |
551 | C<$Storable::canonical> to some C<TRUE> value, Storable will store |
552 | hashes with the elements sorted by their key. This allows you to |
553 | compare data structures by comparing their frozen representations (or |
554 | even the compressed frozen representations), which can be useful for |
555 | creating lookup tables for complicated queries. |
556 | |
557 | Canonical order does not imply network order, those are two orthogonal |
558 | settings. |
559 | |
c261f00e |
560 | =head1 FORWARD COMPATIBILITY |
561 | |
562 | This release of Storable can be used on a newer version of Perl to |
563 | serialize data which is not supported by earlier Perls. By default |
564 | Storable will attempt to do the right thing, by C<croak()>ing if it |
565 | encounters data that it cannot deserialize. However, the defaults can be |
566 | changed as follows |
567 | |
568 | =over 4 |
569 | |
570 | =item utf8 data |
571 | |
572 | Perl 5.6 added support for Unicode characters with code points > 255, |
573 | and Perl 5.8 has full support for Unicode characters in hash keys. |
574 | Perl internally encodes strings with these characters using utf8, and |
575 | Storable serializes them as utf8. By default, if an older version of |
576 | Perl encounters a utf8 value it cannot represent, it will C<croak()>. |
577 | To change this behaviour so that Storable deserializes utf8 encoded |
578 | values as the string of bytes (effectively dropping the I<is_utf8> flag) |
579 | set C<$Storable::drop_utf8> to some C<TRUE> value. This is a form of |
580 | data loss, because with C<$drop_utf8> true, it becomes impossible to tell |
581 | whether the original data was the Unicode string, or a series of bytes |
582 | that happen to be valid utf8. |
583 | |
584 | =item restricted hashes |
585 | |
586 | Perl 5.8 adds support for restricted hashes, which have keys restricted to |
587 | a given set, and can have values locked to be read only. By default |
588 | when Storable encounters a restricted hash on a perl that doesn't support |
589 | them, it will deserialize it as a normal hash, silently discarding any |
590 | placeholder keys and leaving the keys and all values unlocked. To make |
591 | Storable C<croak()> instead, set C<$Storable::downgrade_restricted> to |
592 | a false value. To restore the default set it back to some C<TRUE> value. |
593 | |
e8189732 |
594 | =item files from future versions of Storable |
595 | |
596 | Earlier versions of Storable would immediately croak if they encountered |
597 | a file with a higher internal version number than the reading Storable |
598 | knew about. Internal version numbers are increased each time new data |
599 | types (such as restricted hashes) are added to the vocabulary of the file |
600 | format. This meant that a newer Storable module had no way of writing a |
601 | file readable by an older Storable, even if writer didn't store newer |
602 | data types. |
603 | |
604 | This version of Storable will defer croaking until it encounters a data |
605 | type in the file that it does not recognize. This means that it will |
606 | continue to read files generated by newer Storable modules which are careful |
607 | in what they write out, making it easier to upgrade Storable modules in a |
608 | mixed environment. |
609 | |
610 | The old behaviour of immediate croaking can be re-instated by setting |
611 | C<$Storable::accept_future_minor> to false. |
612 | |
c261f00e |
613 | =back |
614 | |
615 | Both these variables have no effect on a newer Perl which supports the |
616 | relevant feature. |
617 | |
7a6a85bf |
618 | =head1 ERROR REPORTING |
619 | |
620 | Storable uses the "exception" paradigm, in that it does not try to workaround |
621 | failures: if something bad happens, an exception is generated from the |
622 | caller's perspective (see L<Carp> and C<croak()>). Use eval {} to trap |
623 | those exceptions. |
624 | |
625 | When Storable croaks, it tries to report the error via the C<logcroak()> |
626 | routine from the C<Log::Agent> package, if it is available. |
627 | |
212e9bde |
628 | Normal errors are reported by having store() or retrieve() return C<undef>. |
629 | Such errors are usually I/O errors (or truncated stream errors at retrieval). |
630 | |
7a6a85bf |
631 | =head1 WIZARDS ONLY |
632 | |
633 | =head2 Hooks |
634 | |
635 | Any class may define hooks that will be called during the serialization |
636 | and deserialization process on objects that are instances of that class. |
637 | Those hooks can redefine the way serialization is performed (and therefore, |
c261f00e |
638 | how the symmetrical deserialization should be conducted). |
7a6a85bf |
639 | |
640 | Since we said earlier: |
641 | |
642 | dclone(.) = thaw(freeze(.)) |
643 | |
644 | everything we say about hooks should also hold for deep cloning. However, |
645 | hooks get to know whether the operation is a mere serialization, or a cloning. |
646 | |
647 | Therefore, when serializing hooks are involved, |
648 | |
649 | dclone(.) <> thaw(freeze(.)) |
650 | |
651 | Well, you could keep them in sync, but there's no guarantee it will always |
652 | hold on classes somebody else wrote. Besides, there is little to gain in |
653 | doing so: a serializing hook could only keep one attribute of an object, |
654 | which is probably not what should happen during a deep cloning of that |
655 | same object. |
656 | |
657 | Here is the hooking interface: |
658 | |
bbc7dcd2 |
659 | =over 4 |
7a6a85bf |
660 | |
661 | =item C<STORABLE_freeze> I<obj>, I<cloning> |
662 | |
663 | The serializing hook, called on the object during serialization. It can be |
664 | inherited, or defined in the class itself, like any other method. |
665 | |
666 | Arguments: I<obj> is the object to serialize, I<cloning> is a flag indicating |
667 | whether we're in a dclone() or a regular serialization via store() or freeze(). |
668 | |
669 | Returned value: A LIST C<($serialized, $ref1, $ref2, ...)> where $serialized |
670 | is the serialized form to be used, and the optional $ref1, $ref2, etc... are |
671 | extra references that you wish to let the Storable engine serialize. |
672 | |
673 | At deserialization time, you will be given back the same LIST, but all the |
674 | extra references will be pointing into the deserialized structure. |
675 | |
676 | The B<first time> the hook is hit in a serialization flow, you may have it |
677 | return an empty list. That will signal the Storable engine to further |
678 | discard that hook for this class and to therefore revert to the default |
679 | serialization of the underlying Perl data. The hook will again be normally |
680 | processed in the next serialization. |
681 | |
682 | Unless you know better, serializing hook should always say: |
683 | |
684 | sub STORABLE_freeze { |
685 | my ($self, $cloning) = @_; |
686 | return if $cloning; # Regular default serialization |
687 | .... |
688 | } |
689 | |
690 | in order to keep reasonable dclone() semantics. |
691 | |
692 | =item C<STORABLE_thaw> I<obj>, I<cloning>, I<serialized>, ... |
693 | |
694 | The deserializing hook called on the object during deserialization. |
695 | But wait. If we're deserializing, there's no object yet... right? |
696 | |
697 | Wrong: the Storable engine creates an empty one for you. If you know Eiffel, |
698 | you can view C<STORABLE_thaw> as an alternate creation routine. |
699 | |
700 | This means the hook can be inherited like any other method, and that |
701 | I<obj> is your blessed reference for this particular instance. |
702 | |
703 | The other arguments should look familiar if you know C<STORABLE_freeze>: |
704 | I<cloning> is true when we're part of a deep clone operation, I<serialized> |
705 | is the serialized string you returned to the engine in C<STORABLE_freeze>, |
706 | and there may be an optional list of references, in the same order you gave |
707 | them at serialization time, pointing to the deserialized objects (which |
708 | have been processed courtesy of the Storable engine). |
709 | |
212e9bde |
710 | When the Storable engine does not find any C<STORABLE_thaw> hook routine, |
711 | it tries to load the class by requiring the package dynamically (using |
712 | the blessed package name), and then re-attempts the lookup. If at that |
713 | time the hook cannot be located, the engine croaks. Note that this mechanism |
c261f00e |
714 | will fail if you define several classes in the same file, but L<perlmod> |
212e9bde |
715 | warned you. |
716 | |
7a6a85bf |
717 | It is up to you to use these information to populate I<obj> the way you want. |
718 | |
719 | Returned value: none. |
720 | |
721 | =back |
722 | |
723 | =head2 Predicates |
724 | |
c261f00e |
725 | Predicates are not exportable. They must be called by explicitly prefixing |
7a6a85bf |
726 | them with the Storable package name. |
727 | |
bbc7dcd2 |
728 | =over 4 |
7a6a85bf |
729 | |
730 | =item C<Storable::last_op_in_netorder> |
731 | |
732 | The C<Storable::last_op_in_netorder()> predicate will tell you whether |
733 | network order was used in the last store or retrieve operation. If you |
734 | don't know how to use this, just forget about it. |
735 | |
736 | =item C<Storable::is_storing> |
737 | |
738 | Returns true if within a store operation (via STORABLE_freeze hook). |
739 | |
740 | =item C<Storable::is_retrieving> |
741 | |
742 | Returns true if within a retrieve operation, (via STORABLE_thaw hook). |
743 | |
744 | =back |
745 | |
746 | =head2 Recursion |
747 | |
748 | With hooks comes the ability to recurse back to the Storable engine. Indeed, |
749 | hooks are regular Perl code, and Storable is convenient when it comes to |
750 | serialize and deserialize things, so why not use it to handle the |
751 | serialization string? |
752 | |
753 | There are a few things you need to know however: |
754 | |
bbc7dcd2 |
755 | =over 4 |
7a6a85bf |
756 | |
757 | =item * |
758 | |
759 | You can create endless loops if the things you serialize via freeze() |
760 | (for instance) point back to the object we're trying to serialize in the hook. |
761 | |
762 | =item * |
763 | |
764 | Shared references among objects will not stay shared: if we're serializing |
765 | the list of object [A, C] where both object A and C refer to the SAME object |
766 | B, and if there is a serializing hook in A that says freeze(B), then when |
767 | deserializing, we'll get [A', C'] where A' refers to B', but C' refers to D, |
768 | a deep clone of B'. The topology was not preserved. |
769 | |
770 | =back |
771 | |
772 | That's why C<STORABLE_freeze> lets you provide a list of references |
773 | to serialize. The engine guarantees that those will be serialized in the |
774 | same context as the other objects, and therefore that shared objects will |
775 | stay shared. |
776 | |
777 | In the above [A, C] example, the C<STORABLE_freeze> hook could return: |
778 | |
779 | ("something", $self->{B}) |
780 | |
781 | and the B part would be serialized by the engine. In C<STORABLE_thaw>, you |
782 | would get back the reference to the B' object, deserialized for you. |
783 | |
784 | Therefore, recursion should normally be avoided, but is nonetheless supported. |
785 | |
786 | =head2 Deep Cloning |
787 | |
788 | There is a new Clone module available on CPAN which implements deep cloning |
789 | natively, i.e. without freezing to memory and thawing the result. It is |
790 | aimed to replace Storable's dclone() some day. However, it does not currently |
791 | support Storable hooks to redefine the way deep cloning is performed. |
792 | |
0a0da639 |
793 | =head1 Storable magic |
794 | |
795 | Yes, there's a lot of that :-) But more precisely, in UNIX systems |
796 | there's a utility called C<file>, which recognizes data files based on |
797 | their contents (usually their first few bytes). For this to work, |
8b793558 |
798 | a certain file called F<magic> needs to taught about the I<signature> |
0a0da639 |
799 | of the data. Where that configuration file lives depends on the UNIX |
800 | flavour, often it's something like F</usr/share/misc/magic> or |
8b793558 |
801 | F</etc/magic>. Your system administrator needs to do the updating of |
802 | the F<magic> file. The necessary signature information is output to |
c261f00e |
803 | STDOUT by invoking Storable::show_file_magic(). Note that the open |
8b793558 |
804 | source implementation of the C<file> utility 3.38 (or later) |
805 | is expected to contain the support for recognising Storable files, |
806 | in addition to other kinds of Perl files. |
0a0da639 |
807 | |
7a6a85bf |
808 | =head1 EXAMPLES |
809 | |
810 | Here are some code samples showing a possible usage of Storable: |
811 | |
812 | use Storable qw(store retrieve freeze thaw dclone); |
813 | |
814 | %color = ('Blue' => 0.1, 'Red' => 0.8, 'Black' => 0, 'White' => 1); |
815 | |
816 | store(\%color, '/tmp/colors') or die "Can't store %a in /tmp/colors!\n"; |
817 | |
818 | $colref = retrieve('/tmp/colors'); |
819 | die "Unable to retrieve from /tmp/colors!\n" unless defined $colref; |
820 | printf "Blue is still %lf\n", $colref->{'Blue'}; |
821 | |
822 | $colref2 = dclone(\%color); |
823 | |
824 | $str = freeze(\%color); |
825 | printf "Serialization of %%color is %d bytes long.\n", length($str); |
826 | $colref3 = thaw($str); |
827 | |
828 | which prints (on my machine): |
829 | |
830 | Blue is still 0.100000 |
831 | Serialization of %color is 102 bytes long. |
832 | |
833 | =head1 WARNING |
834 | |
835 | If you're using references as keys within your hash tables, you're bound |
c261f00e |
836 | to disappointment when retrieving your data. Indeed, Perl stringifies |
7a6a85bf |
837 | references used as hash table keys. If you later wish to access the |
838 | items via another reference stringification (i.e. using the same |
839 | reference that was used for the key originally to record the value into |
840 | the hash table), it will work because both references stringify to the |
841 | same string. |
842 | |
843 | It won't work across a C<store> and C<retrieve> operations however, because |
844 | the addresses in the retrieved objects, which are part of the stringified |
845 | references, will probably differ from the original addresses. The |
846 | topology of your structure is preserved, but not hidden semantics |
847 | like those. |
848 | |
849 | On platforms where it matters, be sure to call C<binmode()> on the |
850 | descriptors that you pass to Storable functions. |
851 | |
852 | Storing data canonically that contains large hashes can be |
853 | significantly slower than storing the same data normally, as |
c261f00e |
854 | temporary arrays to hold the keys for each hash have to be allocated, |
7a6a85bf |
855 | populated, sorted and freed. Some tests have shown a halving of the |
856 | speed of storing -- the exact penalty will depend on the complexity of |
857 | your data. There is no slowdown on retrieval. |
858 | |
859 | =head1 BUGS |
860 | |
861 | You can't store GLOB, CODE, FORMLINE, etc... If you can define |
862 | semantics for those operations, feel free to enhance Storable so that |
863 | it can deal with them. |
864 | |
865 | The store functions will C<croak> if they run into such references |
866 | unless you set C<$Storable::forgive_me> to some C<TRUE> value. In that |
867 | case, the fatal message is turned in a warning and some |
868 | meaningless string is stored instead. |
869 | |
870 | Setting C<$Storable::canonical> may not yield frozen strings that |
871 | compare equal due to possible stringification of numbers. When the |
872 | string version of a scalar exists, it is the form stored, therefore |
873 | if you happen to use your numbers as strings between two freezing |
874 | operations on the same data structures, you will get different |
875 | results. |
876 | |
dd19458b |
877 | When storing doubles in network order, their value is stored as text. |
878 | However, you should also not expect non-numeric floating-point values |
879 | such as infinity and "not a number" to pass successfully through a |
880 | nstore()/retrieve() pair. |
881 | |
882 | As Storable neither knows nor cares about character sets (although it |
883 | does know that characters may be more than eight bits wide), any difference |
884 | in the interpretation of character codes between a host and a target |
885 | system is your problem. In particular, if host and target use different |
886 | code points to represent the characters used in the text representation |
887 | of floating-point numbers, you will not be able be able to exchange |
888 | floating-point data, even with nstore(). |
889 | |
c261f00e |
890 | C<Storable::drop_utf8> is a blunt tool. There is no facility either to |
891 | return B<all> strings as utf8 sequences, or to attempt to convert utf8 |
892 | data back to 8 bit and C<croak()> if the conversion fails. |
893 | |
7a6a85bf |
894 | =head1 CREDITS |
895 | |
896 | Thank you to (in chronological order): |
897 | |
898 | Jarkko Hietaniemi <jhi@iki.fi> |
899 | Ulrich Pfeifer <pfeifer@charly.informatik.uni-dortmund.de> |
900 | Benjamin A. Holzman <bah@ecnvantage.com> |
901 | Andrew Ford <A.Ford@ford-mason.co.uk> |
902 | Gisle Aas <gisle@aas.no> |
903 | Jeff Gresham <gresham_jeffrey@jpmorgan.com> |
904 | Murray Nesbitt <murray@activestate.com> |
905 | Marc Lehmann <pcg@opengroup.org> |
9e21b3d0 |
906 | Justin Banks <justinb@wamnet.com> |
907 | Jarkko Hietaniemi <jhi@iki.fi> (AGAIN, as perl 5.7.0 Pumpkin!) |
dd19458b |
908 | Salvador Ortiz Garcia <sog@msg.com.mx> |
909 | Dominic Dunlop <domo@computer.org> |
910 | Erik Haugan <erik@solbors.no> |
7a6a85bf |
911 | |
912 | for their bug reports, suggestions and contributions. |
913 | |
914 | Benjamin Holzman contributed the tied variable support, Andrew Ford |
915 | contributed the canonical order for hashes, and Gisle Aas fixed |
916 | a few misunderstandings of mine regarding the Perl internals, |
917 | and optimized the emission of "tags" in the output streams by |
918 | simply counting the objects instead of tagging them (leading to |
919 | a binary incompatibility for the Storable image starting at version |
920 | 0.6--older images are of course still properly understood). |
921 | Murray Nesbitt made Storable thread-safe. Marc Lehmann added overloading |
922 | and reference to tied items support. |
923 | |
924 | =head1 TRANSLATIONS |
925 | |
926 | There is a Japanese translation of this man page available at |
927 | http://member.nifty.ne.jp/hippo2000/perltips/storable.htm , |
928 | courtesy of Kawai, Takanori <kawai@nippon-rad.co.jp>. |
929 | |
930 | =head1 AUTHOR |
931 | |
0ba8809e |
932 | Storable was written by Raphael Manfredi F<E<lt>Raphael_Manfredi@pobox.comE<gt>> |
933 | Maitainance is now done by the perl5-porters F<E<lt>perl5-porters@perl.orgE<gt>> |
934 | |
935 | Please e-mail us with problems, bug fixes, comments and complaints, |
936 | although if you have complements you should send them to Raphael. |
937 | Please don't e-mail Raphael with problems, as he no longer works on |
938 | Storable, and your message will be delayed while he forwards it to us. |
7a6a85bf |
939 | |
940 | =head1 SEE ALSO |
941 | |
c261f00e |
942 | L<Clone>. |
7a6a85bf |
943 | |
944 | =cut |
945 | |