1 ;# $Id: Storable.pm,v 1.0.1.12 2001/08/28 21:51:51 ram Exp $
3 ;# Copyright (c) 1995-2000, Raphael Manfredi
5 ;# You may redistribute only under the same terms as Perl 5, as specified
6 ;# in the README file that comes with the distribution.
8 ;# $Log: Storable.pm,v $
9 ;# Revision 1.0.1.12 2001/08/28 21:51:51 ram
10 ;# patch13: fixed truncation race with lock_retrieve() in lock_store()
12 ;# Revision 1.0.1.11 2001/07/01 11:22:14 ram
13 ;# patch12: systematically use "=over 4" for POD linters
14 ;# patch12: updated version number
16 ;# Revision 1.0.1.10 2001/03/15 00:20:25 ram
17 ;# patch11: updated version number
19 ;# Revision 1.0.1.9 2001/02/17 12:37:32 ram
20 ;# patch10: forgot to increase version number at previous patch
22 ;# Revision 1.0.1.8 2001/02/17 12:24:37 ram
23 ;# patch8: fixed incorrect error message
25 ;# Revision 1.0.1.7 2001/01/03 09:39:02 ram
26 ;# patch7: added CAN_FLOCK to determine whether we can flock() or not
28 ;# Revision 1.0.1.6 2000/11/05 17:20:25 ram
29 ;# patch6: increased version number
31 ;# Revision 1.0.1.5 2000/10/26 17:10:18 ram
32 ;# patch5: documented that store() and retrieve() can return undef
33 ;# patch5: added paragraph explaining the auto require for thaw hooks
35 ;# Revision 1.0.1.4 2000/10/23 18:02:57 ram
36 ;# patch4: protected calls to flock() for dos platform
37 ;# patch4: added logcarp emulation if they don't have Log::Agent
39 ;# Revision 1.0.1.3 2000/09/29 19:49:01 ram
40 ;# patch3: updated version number
42 ;# Revision 1.0.1.2 2000/09/28 21:42:51 ram
43 ;# patch2: added lock_store lock_nstore lock_retrieve
45 ;# Revision 1.0.1.1 2000/09/17 16:46:21 ram
46 ;# patch1: documented that doubles are stringified by nstore()
47 ;# patch1: added Salvador Ortiz Garcia in CREDITS section
49 ;# Revision 1.0 2000/09/01 19:40:41 ram
50 ;# Baseline for first official release.
55 package Storable; @ISA = qw(Exporter DynaLoader);
57 @EXPORT = qw(store retrieve);
59 nstore store_fd nstore_fd fd_retrieve
63 lock_store lock_nstore lock_retrieve
67 use vars qw($forgive_me $VERSION);
70 *AUTOLOAD = \&AutoLoader::AUTOLOAD; # Grrr...
73 # Use of Log::Agent is optional
76 eval "use Log::Agent";
78 unless (defined @Log::Agent::EXPORT) {
92 # They might miss :flock in Fcntl
97 if (exists $Fcntl::EXPORT_TAGS{'flock'}) {
98 Fcntl->import(':flock');
110 sub retrieve_fd { &fd_retrieve } # Backward compatibility
113 # Determine whether locking is possible, but only when needed.
119 return $CAN_FLOCK if defined $CAN_FLOCK;
120 require Config; import Config;
122 $Config{'d_flock'} ||
123 $Config{'d_fcntl_can_lock'} ||
134 # Store target object hierarchy, identified by a reference to its root.
135 # The stored object tree may later be retrieved to memory via retrieve.
136 # Returns undef if an I/O error occurred, in which case the file is
140 return _store(\&pstore, @_, 0);
146 # Same as store, but in network order.
149 return _store(\&net_pstore, @_, 0);
155 # Same as store, but flock the file first (advisory locking).
158 return _store(\&pstore, @_, 1);
164 # Same as nstore, but flock the file first (advisory locking).
167 return _store(\&net_pstore, @_, 1);
170 # Internal store to file routine
174 my ($file, $use_locking) = @_;
175 logcroak "not a reference" unless ref($self);
176 logcroak "wrong argument number" unless @_ == 2; # No @foo in arglist
179 open(FILE, ">>$file") || logcroak "can't write into $file: $!";
180 unless (&CAN_FLOCK) {
181 logcarp "Storable::lock_store: fcntl/flock emulation broken on $^O";
184 flock(FILE, LOCK_EX) ||
185 logcroak "can't get exclusive lock on $file: $!";
187 # Unlocking will happen when FILE is closed
189 open(FILE, ">$file") || logcroak "can't create $file: $!";
191 binmode FILE; # Archaic systems...
192 my $da = $@; # Don't mess if called from exception handler
194 # Call C routine nstore or pstore, depending on network order
195 eval { $ret = &$xsptr(*FILE, $self) };
196 close(FILE) or $ret = undef;
197 unlink($file) or warn "Can't unlink $file: $!\n" if $@ || !defined $ret;
198 logcroak $@ if $@ =~ s/\.?\n$/,/;
200 return $ret ? $ret : undef;
206 # Same as store, but perform on an already opened file descriptor instead.
207 # Returns undef if an I/O error occurred.
210 return _store_fd(\&pstore, @_);
216 # Same as store_fd, but in network order.
219 my ($self, $file) = @_;
220 return _store_fd(\&net_pstore, @_);
223 # Internal store routine on opened file descriptor
228 logcroak "not a reference" unless ref($self);
229 logcroak "too many arguments" unless @_ == 1; # No @foo in arglist
230 my $fd = fileno($file);
231 logcroak "not a valid file descriptor" unless defined $fd;
232 my $da = $@; # Don't mess if called from exception handler
234 # Call C routine nstore or pstore, depending on network order
235 eval { $ret = &$xsptr($file, $self) };
236 logcroak $@ if $@ =~ s/\.?\n$/,/;
238 return $ret ? $ret : undef;
244 # Store oject and its hierarchy in memory and return a scalar
245 # containing the result.
248 _freeze(\&mstore, @_);
254 # Same as freeze but in network order.
257 _freeze(\&net_mstore, @_);
260 # Internal freeze routine
264 logcroak "not a reference" unless ref($self);
265 logcroak "too many arguments" unless @_ == 0; # No @foo in arglist
266 my $da = $@; # Don't mess if called from exception handler
268 # Call C routine mstore or net_mstore, depending on network order
269 eval { $ret = &$xsptr($self) };
270 logcroak $@ if $@ =~ s/\.?\n$/,/;
272 return $ret ? $ret : undef;
278 # Retrieve object hierarchy from disk, returning a reference to the root
279 # object of that tree.
288 # Same as retrieve, but with advisory locking.
294 # Internal retrieve routine
296 my ($file, $use_locking) = @_;
298 open(FILE, $file) || logcroak "can't open $file: $!";
299 binmode FILE; # Archaic systems...
301 my $da = $@; # Could be from exception handler
303 unless (&CAN_FLOCK) {
304 logcarp "Storable::lock_store: fcntl/flock emulation broken on $^O";
307 flock(FILE, LOCK_SH) || logcroak "can't get shared lock on $file: $!";
308 # Unlocking will happen when FILE is closed
310 eval { $self = pretrieve(*FILE) }; # Call C routine
312 logcroak $@ if $@ =~ s/\.?\n$/,/;
320 # Same as retrieve, but perform from an already opened file descriptor instead.
324 my $fd = fileno($file);
325 logcroak "not a valid file descriptor" unless defined $fd;
327 my $da = $@; # Could be from exception handler
328 eval { $self = pretrieve($file) }; # Call C routine
329 logcroak $@ if $@ =~ s/\.?\n$/,/;
337 # Recreate objects in memory from an existing frozen image created
338 # by freeze. If the frozen image passed is undef, return undef.
342 return undef unless defined $frozen;
344 my $da = $@; # Could be from exception handler
345 eval { $self = mretrieve($frozen) }; # Call C routine
346 logcroak $@ if $@ =~ s/\.?\n$/,/;
353 Storable - persistency for perl data structures
358 store \%table, 'file';
359 $hashref = retrieve('file');
361 use Storable qw(nstore store_fd nstore_fd freeze thaw dclone);
364 nstore \%table, 'file';
365 $hashref = retrieve('file'); # There is NO nretrieve()
367 # Storing to and retrieving from an already opened file
368 store_fd \@array, \*STDOUT;
369 nstore_fd \%table, \*STDOUT;
370 $aryref = fd_retrieve(\*SOCKET);
371 $hashref = fd_retrieve(\*SOCKET);
373 # Serializing to memory
374 $serialized = freeze \%table;
375 %table_clone = %{ thaw($serialized) };
377 # Deep (recursive) cloning
378 $cloneref = dclone($ref);
381 use Storable qw(lock_store lock_nstore lock_retrieve)
382 lock_store \%table, 'file';
383 lock_nstore \%table, 'file';
384 $hashref = lock_retrieve('file');
388 The Storable package brings persistency to your perl data structures
389 containing SCALAR, ARRAY, HASH or REF objects, i.e. anything that can be
390 convenientely stored to disk and retrieved at a later time.
392 It can be used in the regular procedural way by calling C<store> with
393 a reference to the object to be stored, along with the file name where
394 the image should be written.
395 The routine returns C<undef> for I/O problems or other internal error,
396 a true value otherwise. Serious errors are propagated as a C<die> exception.
398 To retrieve data stored to disk, use C<retrieve> with a file name,
399 and the objects stored into that file are recreated into memory for you,
400 a I<reference> to the root object being returned. In case an I/O error
401 occurs while reading, C<undef> is returned instead. Other serious
402 errors are propagated via C<die>.
404 Since storage is performed recursively, you might want to stuff references
405 to objects that share a lot of common data into a single array or hash
406 table, and then store that object. That way, when you retrieve back the
407 whole thing, the objects will continue to share what they originally shared.
409 At the cost of a slight header overhead, you may store to an already
410 opened file descriptor using the C<store_fd> routine, and retrieve
411 from a file via C<fd_retrieve>. Those names aren't imported by default,
412 so you will have to do that explicitely if you need those routines.
413 The file descriptor you supply must be already opened, for read
414 if you're going to retrieve and for write if you wish to store.
416 store_fd(\%table, *STDOUT) || die "can't store to stdout\n";
417 $hashref = fd_retrieve(*STDIN);
419 You can also store data in network order to allow easy sharing across
420 multiple platforms, or when storing on a socket known to be remotely
421 connected. The routines to call have an initial C<n> prefix for I<network>,
422 as in C<nstore> and C<nstore_fd>. At retrieval time, your data will be
423 correctly restored so you don't have to know whether you're restoring
424 from native or network ordered data. Double values are stored stringified
425 to ensure portability as well, at the slight risk of loosing some precision
426 in the last decimals.
428 When using C<fd_retrieve>, objects are retrieved in sequence, one
429 object (i.e. one recursive tree) per associated C<store_fd>.
431 If you're more from the object-oriented camp, you can inherit from
432 Storable and directly store your objects by invoking C<store> as
433 a method. The fact that the root of the to-be-stored tree is a
434 blessed reference (i.e. an object) is special-cased so that the
435 retrieve does not provide a reference to that object but rather the
436 blessed object reference itself. (Otherwise, you'd get a reference
437 to that blessed object).
441 The Storable engine can also store data into a Perl scalar instead, to
442 later retrieve them. This is mainly used to freeze a complex structure in
443 some safe compact memory place (where it can possibly be sent to another
444 process via some IPC, since freezing the structure also serializes it in
445 effect). Later on, and maybe somewhere else, you can thaw the Perl scalar
446 out and recreate the original complex structure in memory.
448 Surprisingly, the routines to be called are named C<freeze> and C<thaw>.
449 If you wish to send out the frozen scalar to another machine, use
450 C<nfreeze> instead to get a portable image.
452 Note that freezing an object structure and immediately thawing it
453 actually achieves a deep cloning of that structure:
455 dclone(.) = thaw(freeze(.))
457 Storable provides you with a C<dclone> interface which does not create
458 that intermediary scalar but instead freezes the structure in some
459 internal memory space and then immediatly thaws it out.
461 =head1 ADVISORY LOCKING
463 The C<lock_store> and C<lock_nstore> routine are equivalent to C<store>
464 and C<nstore>, only they get an exclusive lock on the file before
465 writing. Likewise, C<lock_retrieve> performs as C<retrieve>, but also
466 gets a shared lock on the file before reading.
468 Like with any advisory locking scheme, the protection only works if
469 you systematically use C<lock_store> and C<lock_retrieve>. If one
470 side of your application uses C<store> whilst the other uses C<lock_retrieve>,
471 you will get no protection at all.
473 The internal advisory locking is implemented using Perl's flock() routine.
474 If your system does not support any form of flock(), or if you share
475 your files across NFS, you might wish to use other forms of locking by
476 using modules like LockFile::Simple which lock a file using a filesystem
477 entry, instead of locking the file descriptor.
481 The heart of Storable is written in C for decent speed. Extra low-level
482 optimization have been made when manipulating perl internals, to
483 sacrifice encapsulation for the benefit of a greater speed.
485 =head1 CANONICAL REPRESENTATION
487 Normally Storable stores elements of hashes in the order they are
488 stored internally by Perl, i.e. pseudo-randomly. If you set
489 C<$Storable::canonical> to some C<TRUE> value, Storable will store
490 hashes with the elements sorted by their key. This allows you to
491 compare data structures by comparing their frozen representations (or
492 even the compressed frozen representations), which can be useful for
493 creating lookup tables for complicated queries.
495 Canonical order does not imply network order, those are two orthogonal
498 =head1 ERROR REPORTING
500 Storable uses the "exception" paradigm, in that it does not try to workaround
501 failures: if something bad happens, an exception is generated from the
502 caller's perspective (see L<Carp> and C<croak()>). Use eval {} to trap
505 When Storable croaks, it tries to report the error via the C<logcroak()>
506 routine from the C<Log::Agent> package, if it is available.
508 Normal errors are reported by having store() or retrieve() return C<undef>.
509 Such errors are usually I/O errors (or truncated stream errors at retrieval).
515 Any class may define hooks that will be called during the serialization
516 and deserialization process on objects that are instances of that class.
517 Those hooks can redefine the way serialization is performed (and therefore,
518 how the symetrical deserialization should be conducted).
520 Since we said earlier:
522 dclone(.) = thaw(freeze(.))
524 everything we say about hooks should also hold for deep cloning. However,
525 hooks get to know whether the operation is a mere serialization, or a cloning.
527 Therefore, when serializing hooks are involved,
529 dclone(.) <> thaw(freeze(.))
531 Well, you could keep them in sync, but there's no guarantee it will always
532 hold on classes somebody else wrote. Besides, there is little to gain in
533 doing so: a serializing hook could only keep one attribute of an object,
534 which is probably not what should happen during a deep cloning of that
537 Here is the hooking interface:
541 =item C<STORABLE_freeze> I<obj>, I<cloning>
543 The serializing hook, called on the object during serialization. It can be
544 inherited, or defined in the class itself, like any other method.
546 Arguments: I<obj> is the object to serialize, I<cloning> is a flag indicating
547 whether we're in a dclone() or a regular serialization via store() or freeze().
549 Returned value: A LIST C<($serialized, $ref1, $ref2, ...)> where $serialized
550 is the serialized form to be used, and the optional $ref1, $ref2, etc... are
551 extra references that you wish to let the Storable engine serialize.
553 At deserialization time, you will be given back the same LIST, but all the
554 extra references will be pointing into the deserialized structure.
556 The B<first time> the hook is hit in a serialization flow, you may have it
557 return an empty list. That will signal the Storable engine to further
558 discard that hook for this class and to therefore revert to the default
559 serialization of the underlying Perl data. The hook will again be normally
560 processed in the next serialization.
562 Unless you know better, serializing hook should always say:
564 sub STORABLE_freeze {
565 my ($self, $cloning) = @_;
566 return if $cloning; # Regular default serialization
570 in order to keep reasonable dclone() semantics.
572 =item C<STORABLE_thaw> I<obj>, I<cloning>, I<serialized>, ...
574 The deserializing hook called on the object during deserialization.
575 But wait. If we're deserializing, there's no object yet... right?
577 Wrong: the Storable engine creates an empty one for you. If you know Eiffel,
578 you can view C<STORABLE_thaw> as an alternate creation routine.
580 This means the hook can be inherited like any other method, and that
581 I<obj> is your blessed reference for this particular instance.
583 The other arguments should look familiar if you know C<STORABLE_freeze>:
584 I<cloning> is true when we're part of a deep clone operation, I<serialized>
585 is the serialized string you returned to the engine in C<STORABLE_freeze>,
586 and there may be an optional list of references, in the same order you gave
587 them at serialization time, pointing to the deserialized objects (which
588 have been processed courtesy of the Storable engine).
590 When the Storable engine does not find any C<STORABLE_thaw> hook routine,
591 it tries to load the class by requiring the package dynamically (using
592 the blessed package name), and then re-attempts the lookup. If at that
593 time the hook cannot be located, the engine croaks. Note that this mechanism
594 will fail if you define several classes in the same file, but perlmod(1)
597 It is up to you to use these information to populate I<obj> the way you want.
599 Returned value: none.
605 Predicates are not exportable. They must be called by explicitely prefixing
606 them with the Storable package name.
610 =item C<Storable::last_op_in_netorder>
612 The C<Storable::last_op_in_netorder()> predicate will tell you whether
613 network order was used in the last store or retrieve operation. If you
614 don't know how to use this, just forget about it.
616 =item C<Storable::is_storing>
618 Returns true if within a store operation (via STORABLE_freeze hook).
620 =item C<Storable::is_retrieving>
622 Returns true if within a retrieve operation, (via STORABLE_thaw hook).
628 With hooks comes the ability to recurse back to the Storable engine. Indeed,
629 hooks are regular Perl code, and Storable is convenient when it comes to
630 serialize and deserialize things, so why not use it to handle the
631 serialization string?
633 There are a few things you need to know however:
639 You can create endless loops if the things you serialize via freeze()
640 (for instance) point back to the object we're trying to serialize in the hook.
644 Shared references among objects will not stay shared: if we're serializing
645 the list of object [A, C] where both object A and C refer to the SAME object
646 B, and if there is a serializing hook in A that says freeze(B), then when
647 deserializing, we'll get [A', C'] where A' refers to B', but C' refers to D,
648 a deep clone of B'. The topology was not preserved.
652 That's why C<STORABLE_freeze> lets you provide a list of references
653 to serialize. The engine guarantees that those will be serialized in the
654 same context as the other objects, and therefore that shared objects will
657 In the above [A, C] example, the C<STORABLE_freeze> hook could return:
659 ("something", $self->{B})
661 and the B part would be serialized by the engine. In C<STORABLE_thaw>, you
662 would get back the reference to the B' object, deserialized for you.
664 Therefore, recursion should normally be avoided, but is nonetheless supported.
668 There is a new Clone module available on CPAN which implements deep cloning
669 natively, i.e. without freezing to memory and thawing the result. It is
670 aimed to replace Storable's dclone() some day. However, it does not currently
671 support Storable hooks to redefine the way deep cloning is performed.
675 Here are some code samples showing a possible usage of Storable:
677 use Storable qw(store retrieve freeze thaw dclone);
679 %color = ('Blue' => 0.1, 'Red' => 0.8, 'Black' => 0, 'White' => 1);
681 store(\%color, '/tmp/colors') or die "Can't store %a in /tmp/colors!\n";
683 $colref = retrieve('/tmp/colors');
684 die "Unable to retrieve from /tmp/colors!\n" unless defined $colref;
685 printf "Blue is still %lf\n", $colref->{'Blue'};
687 $colref2 = dclone(\%color);
689 $str = freeze(\%color);
690 printf "Serialization of %%color is %d bytes long.\n", length($str);
691 $colref3 = thaw($str);
693 which prints (on my machine):
695 Blue is still 0.100000
696 Serialization of %color is 102 bytes long.
700 If you're using references as keys within your hash tables, you're bound
701 to disapointment when retrieving your data. Indeed, Perl stringifies
702 references used as hash table keys. If you later wish to access the
703 items via another reference stringification (i.e. using the same
704 reference that was used for the key originally to record the value into
705 the hash table), it will work because both references stringify to the
708 It won't work across a C<store> and C<retrieve> operations however, because
709 the addresses in the retrieved objects, which are part of the stringified
710 references, will probably differ from the original addresses. The
711 topology of your structure is preserved, but not hidden semantics
714 On platforms where it matters, be sure to call C<binmode()> on the
715 descriptors that you pass to Storable functions.
717 Storing data canonically that contains large hashes can be
718 significantly slower than storing the same data normally, as
719 temprorary arrays to hold the keys for each hash have to be allocated,
720 populated, sorted and freed. Some tests have shown a halving of the
721 speed of storing -- the exact penalty will depend on the complexity of
722 your data. There is no slowdown on retrieval.
726 You can't store GLOB, CODE, FORMLINE, etc... If you can define
727 semantics for those operations, feel free to enhance Storable so that
728 it can deal with them.
730 The store functions will C<croak> if they run into such references
731 unless you set C<$Storable::forgive_me> to some C<TRUE> value. In that
732 case, the fatal message is turned in a warning and some
733 meaningless string is stored instead.
735 Setting C<$Storable::canonical> may not yield frozen strings that
736 compare equal due to possible stringification of numbers. When the
737 string version of a scalar exists, it is the form stored, therefore
738 if you happen to use your numbers as strings between two freezing
739 operations on the same data structures, you will get different
742 When storing doubles in network order, their value is stored as text.
743 However, you should also not expect non-numeric floating-point values
744 such as infinity and "not a number" to pass successfully through a
745 nstore()/retrieve() pair.
747 As Storable neither knows nor cares about character sets (although it
748 does know that characters may be more than eight bits wide), any difference
749 in the interpretation of character codes between a host and a target
750 system is your problem. In particular, if host and target use different
751 code points to represent the characters used in the text representation
752 of floating-point numbers, you will not be able be able to exchange
753 floating-point data, even with nstore().
757 Thank you to (in chronological order):
759 Jarkko Hietaniemi <jhi@iki.fi>
760 Ulrich Pfeifer <pfeifer@charly.informatik.uni-dortmund.de>
761 Benjamin A. Holzman <bah@ecnvantage.com>
762 Andrew Ford <A.Ford@ford-mason.co.uk>
763 Gisle Aas <gisle@aas.no>
764 Jeff Gresham <gresham_jeffrey@jpmorgan.com>
765 Murray Nesbitt <murray@activestate.com>
766 Marc Lehmann <pcg@opengroup.org>
767 Justin Banks <justinb@wamnet.com>
768 Jarkko Hietaniemi <jhi@iki.fi> (AGAIN, as perl 5.7.0 Pumpkin!)
769 Salvador Ortiz Garcia <sog@msg.com.mx>
770 Dominic Dunlop <domo@computer.org>
771 Erik Haugan <erik@solbors.no>
773 for their bug reports, suggestions and contributions.
775 Benjamin Holzman contributed the tied variable support, Andrew Ford
776 contributed the canonical order for hashes, and Gisle Aas fixed
777 a few misunderstandings of mine regarding the Perl internals,
778 and optimized the emission of "tags" in the output streams by
779 simply counting the objects instead of tagging them (leading to
780 a binary incompatibility for the Storable image starting at version
781 0.6--older images are of course still properly understood).
782 Murray Nesbitt made Storable thread-safe. Marc Lehmann added overloading
783 and reference to tied items support.
787 There is a Japanese translation of this man page available at
788 http://member.nifty.ne.jp/hippo2000/perltips/storable.htm ,
789 courtesy of Kawai, Takanori <kawai@nippon-rad.co.jp>.
793 Raphael Manfredi F<E<lt>Raphael_Manfredi@pobox.comE<gt>>