1 ;# $Id: Storable.pm,v 1.0.1.13 2001/12/01 13:34:49 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 1.0.1.13 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
13 ;# Revision 1.0.1.12 2001/08/28 21:51:51 ram
14 ;# patch13: fixed truncation race with lock_retrieve() in lock_store()
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
20 ;# Revision 1.0.1.10 2001/03/15 00:20:25 ram
21 ;# patch11: updated version number
23 ;# Revision 1.0.1.9 2001/02/17 12:37:32 ram
24 ;# patch10: forgot to increase version number at previous patch
26 ;# Revision 1.0.1.8 2001/02/17 12:24:37 ram
27 ;# patch8: fixed incorrect error message
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
32 ;# Revision 1.0.1.6 2000/11/05 17:20:25 ram
33 ;# patch6: increased version number
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
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
43 ;# Revision 1.0.1.3 2000/09/29 19:49:01 ram
44 ;# patch3: updated version number
46 ;# Revision 1.0.1.2 2000/09/28 21:42:51 ram
47 ;# patch2: added lock_store lock_nstore lock_retrieve
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
53 ;# Revision 1.0 2000/09/01 19:40:41 ram
54 ;# Baseline for first official release.
59 package Storable; @ISA = qw(Exporter DynaLoader);
61 @EXPORT = qw(store retrieve);
63 nstore store_fd nstore_fd fd_retrieve
67 lock_store lock_nstore lock_retrieve
71 use vars qw($forgive_me $VERSION);
74 *AUTOLOAD = \&AutoLoader::AUTOLOAD; # Grrr...
77 # Use of Log::Agent is optional
80 eval "use Log::Agent";
82 unless (defined @Log::Agent::EXPORT) {
96 # They might miss :flock in Fcntl
100 if (eval { require Fcntl; 1 } && exists $Fcntl::EXPORT_TAGS{'flock'}) {
101 Fcntl->import(':flock');
113 sub retrieve_fd { &fd_retrieve } # Backward compatibility
116 # Determine whether locking is possible, but only when needed.
122 return $CAN_FLOCK if defined $CAN_FLOCK;
123 require Config; import Config;
125 $Config{'d_flock'} ||
126 $Config{'d_fcntl_can_lock'} ||
137 # Store target object hierarchy, identified by a reference to its root.
138 # The stored object tree may later be retrieved to memory via retrieve.
139 # Returns undef if an I/O error occurred, in which case the file is
143 return _store(\&pstore, @_, 0);
149 # Same as store, but in network order.
152 return _store(\&net_pstore, @_, 0);
158 # Same as store, but flock the file first (advisory locking).
161 return _store(\&pstore, @_, 1);
167 # Same as nstore, but flock the file first (advisory locking).
170 return _store(\&net_pstore, @_, 1);
173 # Internal store to file routine
177 my ($file, $use_locking) = @_;
178 logcroak "not a reference" unless ref($self);
179 logcroak "wrong argument number" unless @_ == 2; # No @foo in arglist
182 open(FILE, ">>$file") || logcroak "can't write into $file: $!";
183 unless (&CAN_FLOCK) {
184 logcarp "Storable::lock_store: fcntl/flock emulation broken on $^O";
187 flock(FILE, LOCK_EX) ||
188 logcroak "can't get exclusive lock on $file: $!";
190 # Unlocking will happen when FILE is closed
192 open(FILE, ">$file") || logcroak "can't create $file: $!";
194 binmode FILE; # Archaic systems...
195 my $da = $@; # Don't mess if called from exception handler
197 # Call C routine nstore or pstore, depending on network order
198 eval { $ret = &$xsptr(*FILE, $self) };
199 close(FILE) or $ret = undef;
200 unlink($file) or warn "Can't unlink $file: $!\n" if $@ || !defined $ret;
201 logcroak $@ if $@ =~ s/\.?\n$/,/;
203 return $ret ? $ret : undef;
209 # Same as store, but perform on an already opened file descriptor instead.
210 # Returns undef if an I/O error occurred.
213 return _store_fd(\&pstore, @_);
219 # Same as store_fd, but in network order.
222 my ($self, $file) = @_;
223 return _store_fd(\&net_pstore, @_);
226 # Internal store routine on opened file descriptor
231 logcroak "not a reference" unless ref($self);
232 logcroak "too many arguments" unless @_ == 1; # No @foo in arglist
233 my $fd = fileno($file);
234 logcroak "not a valid file descriptor" unless defined $fd;
235 my $da = $@; # Don't mess if called from exception handler
237 # Call C routine nstore or pstore, depending on network order
238 eval { $ret = &$xsptr($file, $self) };
239 logcroak $@ if $@ =~ s/\.?\n$/,/;
240 local $\; print $file ''; # Autoflush the file if wanted
242 return $ret ? $ret : undef;
248 # Store oject and its hierarchy in memory and return a scalar
249 # containing the result.
252 _freeze(\&mstore, @_);
258 # Same as freeze but in network order.
261 _freeze(\&net_mstore, @_);
264 # Internal freeze routine
268 logcroak "not a reference" unless ref($self);
269 logcroak "too many arguments" unless @_ == 0; # No @foo in arglist
270 my $da = $@; # Don't mess if called from exception handler
272 # Call C routine mstore or net_mstore, depending on network order
273 eval { $ret = &$xsptr($self) };
274 logcroak $@ if $@ =~ s/\.?\n$/,/;
276 return $ret ? $ret : undef;
282 # Retrieve object hierarchy from disk, returning a reference to the root
283 # object of that tree.
292 # Same as retrieve, but with advisory locking.
298 # Internal retrieve routine
300 my ($file, $use_locking) = @_;
302 open(FILE, $file) || logcroak "can't open $file: $!";
303 binmode FILE; # Archaic systems...
305 my $da = $@; # Could be from exception handler
307 unless (&CAN_FLOCK) {
308 logcarp "Storable::lock_store: fcntl/flock emulation broken on $^O";
311 flock(FILE, LOCK_SH) || logcroak "can't get shared lock on $file: $!";
312 # Unlocking will happen when FILE is closed
314 eval { $self = pretrieve(*FILE) }; # Call C routine
316 logcroak $@ if $@ =~ s/\.?\n$/,/;
324 # Same as retrieve, but perform from an already opened file descriptor instead.
328 my $fd = fileno($file);
329 logcroak "not a valid file descriptor" unless defined $fd;
331 my $da = $@; # Could be from exception handler
332 eval { $self = pretrieve($file) }; # Call C routine
333 logcroak $@ if $@ =~ s/\.?\n$/,/;
341 # Recreate objects in memory from an existing frozen image created
342 # by freeze. If the frozen image passed is undef, return undef.
346 return undef unless defined $frozen;
348 my $da = $@; # Could be from exception handler
349 eval { $self = mretrieve($frozen) }; # Call C routine
350 logcroak $@ if $@ =~ s/\.?\n$/,/;
357 Storable - persistency for perl data structures
362 store \%table, 'file';
363 $hashref = retrieve('file');
365 use Storable qw(nstore store_fd nstore_fd freeze thaw dclone);
368 nstore \%table, 'file';
369 $hashref = retrieve('file'); # There is NO nretrieve()
371 # Storing to and retrieving from an already opened file
372 store_fd \@array, \*STDOUT;
373 nstore_fd \%table, \*STDOUT;
374 $aryref = fd_retrieve(\*SOCKET);
375 $hashref = fd_retrieve(\*SOCKET);
377 # Serializing to memory
378 $serialized = freeze \%table;
379 %table_clone = %{ thaw($serialized) };
381 # Deep (recursive) cloning
382 $cloneref = dclone($ref);
385 use Storable qw(lock_store lock_nstore lock_retrieve)
386 lock_store \%table, 'file';
387 lock_nstore \%table, 'file';
388 $hashref = lock_retrieve('file');
392 The Storable package brings persistency to your perl data structures
393 containing SCALAR, ARRAY, HASH or REF objects, i.e. anything that can be
394 convenientely stored to disk and retrieved at a later time.
396 It can be used in the regular procedural way by calling C<store> with
397 a reference to the object to be stored, along with the file name where
398 the image should be written.
399 The routine returns C<undef> for I/O problems or other internal error,
400 a true value otherwise. Serious errors are propagated as a C<die> exception.
402 To retrieve data stored to disk, use C<retrieve> with a file name,
403 and the objects stored into that file are recreated into memory for you,
404 a I<reference> to the root object being returned. In case an I/O error
405 occurs while reading, C<undef> is returned instead. Other serious
406 errors are propagated via C<die>.
408 Since storage is performed recursively, you might want to stuff references
409 to objects that share a lot of common data into a single array or hash
410 table, and then store that object. That way, when you retrieve back the
411 whole thing, the objects will continue to share what they originally shared.
413 At the cost of a slight header overhead, you may store to an already
414 opened file descriptor using the C<store_fd> routine, and retrieve
415 from a file via C<fd_retrieve>. Those names aren't imported by default,
416 so you will have to do that explicitely if you need those routines.
417 The file descriptor you supply must be already opened, for read
418 if you're going to retrieve and for write if you wish to store.
420 store_fd(\%table, *STDOUT) || die "can't store to stdout\n";
421 $hashref = fd_retrieve(*STDIN);
423 You can also store data in network order to allow easy sharing across
424 multiple platforms, or when storing on a socket known to be remotely
425 connected. The routines to call have an initial C<n> prefix for I<network>,
426 as in C<nstore> and C<nstore_fd>. At retrieval time, your data will be
427 correctly restored so you don't have to know whether you're restoring
428 from native or network ordered data. Double values are stored stringified
429 to ensure portability as well, at the slight risk of loosing some precision
430 in the last decimals.
432 When using C<fd_retrieve>, objects are retrieved in sequence, one
433 object (i.e. one recursive tree) per associated C<store_fd>.
435 If you're more from the object-oriented camp, you can inherit from
436 Storable and directly store your objects by invoking C<store> as
437 a method. The fact that the root of the to-be-stored tree is a
438 blessed reference (i.e. an object) is special-cased so that the
439 retrieve does not provide a reference to that object but rather the
440 blessed object reference itself. (Otherwise, you'd get a reference
441 to that blessed object).
445 The Storable engine can also store data into a Perl scalar instead, to
446 later retrieve them. This is mainly used to freeze a complex structure in
447 some safe compact memory place (where it can possibly be sent to another
448 process via some IPC, since freezing the structure also serializes it in
449 effect). Later on, and maybe somewhere else, you can thaw the Perl scalar
450 out and recreate the original complex structure in memory.
452 Surprisingly, the routines to be called are named C<freeze> and C<thaw>.
453 If you wish to send out the frozen scalar to another machine, use
454 C<nfreeze> instead to get a portable image.
456 Note that freezing an object structure and immediately thawing it
457 actually achieves a deep cloning of that structure:
459 dclone(.) = thaw(freeze(.))
461 Storable provides you with a C<dclone> interface which does not create
462 that intermediary scalar but instead freezes the structure in some
463 internal memory space and then immediatly thaws it out.
465 =head1 ADVISORY LOCKING
467 The C<lock_store> and C<lock_nstore> routine are equivalent to C<store>
468 and C<nstore>, only they get an exclusive lock on the file before
469 writing. Likewise, C<lock_retrieve> performs as C<retrieve>, but also
470 gets a shared lock on the file before reading.
472 Like with any advisory locking scheme, the protection only works if
473 you systematically use C<lock_store> and C<lock_retrieve>. If one
474 side of your application uses C<store> whilst the other uses C<lock_retrieve>,
475 you will get no protection at all.
477 The internal advisory locking is implemented using Perl's flock() routine.
478 If your system does not support any form of flock(), or if you share
479 your files across NFS, you might wish to use other forms of locking by
480 using modules like LockFile::Simple which lock a file using a filesystem
481 entry, instead of locking the file descriptor.
485 The heart of Storable is written in C for decent speed. Extra low-level
486 optimization have been made when manipulating perl internals, to
487 sacrifice encapsulation for the benefit of a greater speed.
489 =head1 CANONICAL REPRESENTATION
491 Normally Storable stores elements of hashes in the order they are
492 stored internally by Perl, i.e. pseudo-randomly. If you set
493 C<$Storable::canonical> to some C<TRUE> value, Storable will store
494 hashes with the elements sorted by their key. This allows you to
495 compare data structures by comparing their frozen representations (or
496 even the compressed frozen representations), which can be useful for
497 creating lookup tables for complicated queries.
499 Canonical order does not imply network order, those are two orthogonal
502 =head1 ERROR REPORTING
504 Storable uses the "exception" paradigm, in that it does not try to workaround
505 failures: if something bad happens, an exception is generated from the
506 caller's perspective (see L<Carp> and C<croak()>). Use eval {} to trap
509 When Storable croaks, it tries to report the error via the C<logcroak()>
510 routine from the C<Log::Agent> package, if it is available.
512 Normal errors are reported by having store() or retrieve() return C<undef>.
513 Such errors are usually I/O errors (or truncated stream errors at retrieval).
519 Any class may define hooks that will be called during the serialization
520 and deserialization process on objects that are instances of that class.
521 Those hooks can redefine the way serialization is performed (and therefore,
522 how the symetrical deserialization should be conducted).
524 Since we said earlier:
526 dclone(.) = thaw(freeze(.))
528 everything we say about hooks should also hold for deep cloning. However,
529 hooks get to know whether the operation is a mere serialization, or a cloning.
531 Therefore, when serializing hooks are involved,
533 dclone(.) <> thaw(freeze(.))
535 Well, you could keep them in sync, but there's no guarantee it will always
536 hold on classes somebody else wrote. Besides, there is little to gain in
537 doing so: a serializing hook could only keep one attribute of an object,
538 which is probably not what should happen during a deep cloning of that
541 Here is the hooking interface:
545 =item C<STORABLE_freeze> I<obj>, I<cloning>
547 The serializing hook, called on the object during serialization. It can be
548 inherited, or defined in the class itself, like any other method.
550 Arguments: I<obj> is the object to serialize, I<cloning> is a flag indicating
551 whether we're in a dclone() or a regular serialization via store() or freeze().
553 Returned value: A LIST C<($serialized, $ref1, $ref2, ...)> where $serialized
554 is the serialized form to be used, and the optional $ref1, $ref2, etc... are
555 extra references that you wish to let the Storable engine serialize.
557 At deserialization time, you will be given back the same LIST, but all the
558 extra references will be pointing into the deserialized structure.
560 The B<first time> the hook is hit in a serialization flow, you may have it
561 return an empty list. That will signal the Storable engine to further
562 discard that hook for this class and to therefore revert to the default
563 serialization of the underlying Perl data. The hook will again be normally
564 processed in the next serialization.
566 Unless you know better, serializing hook should always say:
568 sub STORABLE_freeze {
569 my ($self, $cloning) = @_;
570 return if $cloning; # Regular default serialization
574 in order to keep reasonable dclone() semantics.
576 =item C<STORABLE_thaw> I<obj>, I<cloning>, I<serialized>, ...
578 The deserializing hook called on the object during deserialization.
579 But wait. If we're deserializing, there's no object yet... right?
581 Wrong: the Storable engine creates an empty one for you. If you know Eiffel,
582 you can view C<STORABLE_thaw> as an alternate creation routine.
584 This means the hook can be inherited like any other method, and that
585 I<obj> is your blessed reference for this particular instance.
587 The other arguments should look familiar if you know C<STORABLE_freeze>:
588 I<cloning> is true when we're part of a deep clone operation, I<serialized>
589 is the serialized string you returned to the engine in C<STORABLE_freeze>,
590 and there may be an optional list of references, in the same order you gave
591 them at serialization time, pointing to the deserialized objects (which
592 have been processed courtesy of the Storable engine).
594 When the Storable engine does not find any C<STORABLE_thaw> hook routine,
595 it tries to load the class by requiring the package dynamically (using
596 the blessed package name), and then re-attempts the lookup. If at that
597 time the hook cannot be located, the engine croaks. Note that this mechanism
598 will fail if you define several classes in the same file, but perlmod(1)
601 It is up to you to use these information to populate I<obj> the way you want.
603 Returned value: none.
609 Predicates are not exportable. They must be called by explicitely prefixing
610 them with the Storable package name.
614 =item C<Storable::last_op_in_netorder>
616 The C<Storable::last_op_in_netorder()> predicate will tell you whether
617 network order was used in the last store or retrieve operation. If you
618 don't know how to use this, just forget about it.
620 =item C<Storable::is_storing>
622 Returns true if within a store operation (via STORABLE_freeze hook).
624 =item C<Storable::is_retrieving>
626 Returns true if within a retrieve operation, (via STORABLE_thaw hook).
632 With hooks comes the ability to recurse back to the Storable engine. Indeed,
633 hooks are regular Perl code, and Storable is convenient when it comes to
634 serialize and deserialize things, so why not use it to handle the
635 serialization string?
637 There are a few things you need to know however:
643 You can create endless loops if the things you serialize via freeze()
644 (for instance) point back to the object we're trying to serialize in the hook.
648 Shared references among objects will not stay shared: if we're serializing
649 the list of object [A, C] where both object A and C refer to the SAME object
650 B, and if there is a serializing hook in A that says freeze(B), then when
651 deserializing, we'll get [A', C'] where A' refers to B', but C' refers to D,
652 a deep clone of B'. The topology was not preserved.
656 That's why C<STORABLE_freeze> lets you provide a list of references
657 to serialize. The engine guarantees that those will be serialized in the
658 same context as the other objects, and therefore that shared objects will
661 In the above [A, C] example, the C<STORABLE_freeze> hook could return:
663 ("something", $self->{B})
665 and the B part would be serialized by the engine. In C<STORABLE_thaw>, you
666 would get back the reference to the B' object, deserialized for you.
668 Therefore, recursion should normally be avoided, but is nonetheless supported.
672 There is a new Clone module available on CPAN which implements deep cloning
673 natively, i.e. without freezing to memory and thawing the result. It is
674 aimed to replace Storable's dclone() some day. However, it does not currently
675 support Storable hooks to redefine the way deep cloning is performed.
679 Here are some code samples showing a possible usage of Storable:
681 use Storable qw(store retrieve freeze thaw dclone);
683 %color = ('Blue' => 0.1, 'Red' => 0.8, 'Black' => 0, 'White' => 1);
685 store(\%color, '/tmp/colors') or die "Can't store %a in /tmp/colors!\n";
687 $colref = retrieve('/tmp/colors');
688 die "Unable to retrieve from /tmp/colors!\n" unless defined $colref;
689 printf "Blue is still %lf\n", $colref->{'Blue'};
691 $colref2 = dclone(\%color);
693 $str = freeze(\%color);
694 printf "Serialization of %%color is %d bytes long.\n", length($str);
695 $colref3 = thaw($str);
697 which prints (on my machine):
699 Blue is still 0.100000
700 Serialization of %color is 102 bytes long.
704 If you're using references as keys within your hash tables, you're bound
705 to disapointment when retrieving your data. Indeed, Perl stringifies
706 references used as hash table keys. If you later wish to access the
707 items via another reference stringification (i.e. using the same
708 reference that was used for the key originally to record the value into
709 the hash table), it will work because both references stringify to the
712 It won't work across a C<store> and C<retrieve> operations however, because
713 the addresses in the retrieved objects, which are part of the stringified
714 references, will probably differ from the original addresses. The
715 topology of your structure is preserved, but not hidden semantics
718 On platforms where it matters, be sure to call C<binmode()> on the
719 descriptors that you pass to Storable functions.
721 Storing data canonically that contains large hashes can be
722 significantly slower than storing the same data normally, as
723 temprorary arrays to hold the keys for each hash have to be allocated,
724 populated, sorted and freed. Some tests have shown a halving of the
725 speed of storing -- the exact penalty will depend on the complexity of
726 your data. There is no slowdown on retrieval.
730 You can't store GLOB, CODE, FORMLINE, etc... If you can define
731 semantics for those operations, feel free to enhance Storable so that
732 it can deal with them.
734 The store functions will C<croak> if they run into such references
735 unless you set C<$Storable::forgive_me> to some C<TRUE> value. In that
736 case, the fatal message is turned in a warning and some
737 meaningless string is stored instead.
739 Setting C<$Storable::canonical> may not yield frozen strings that
740 compare equal due to possible stringification of numbers. When the
741 string version of a scalar exists, it is the form stored, therefore
742 if you happen to use your numbers as strings between two freezing
743 operations on the same data structures, you will get different
746 When storing doubles in network order, their value is stored as text.
747 However, you should also not expect non-numeric floating-point values
748 such as infinity and "not a number" to pass successfully through a
749 nstore()/retrieve() pair.
751 As Storable neither knows nor cares about character sets (although it
752 does know that characters may be more than eight bits wide), any difference
753 in the interpretation of character codes between a host and a target
754 system is your problem. In particular, if host and target use different
755 code points to represent the characters used in the text representation
756 of floating-point numbers, you will not be able be able to exchange
757 floating-point data, even with nstore().
761 Thank you to (in chronological order):
763 Jarkko Hietaniemi <jhi@iki.fi>
764 Ulrich Pfeifer <pfeifer@charly.informatik.uni-dortmund.de>
765 Benjamin A. Holzman <bah@ecnvantage.com>
766 Andrew Ford <A.Ford@ford-mason.co.uk>
767 Gisle Aas <gisle@aas.no>
768 Jeff Gresham <gresham_jeffrey@jpmorgan.com>
769 Murray Nesbitt <murray@activestate.com>
770 Marc Lehmann <pcg@opengroup.org>
771 Justin Banks <justinb@wamnet.com>
772 Jarkko Hietaniemi <jhi@iki.fi> (AGAIN, as perl 5.7.0 Pumpkin!)
773 Salvador Ortiz Garcia <sog@msg.com.mx>
774 Dominic Dunlop <domo@computer.org>
775 Erik Haugan <erik@solbors.no>
777 for their bug reports, suggestions and contributions.
779 Benjamin Holzman contributed the tied variable support, Andrew Ford
780 contributed the canonical order for hashes, and Gisle Aas fixed
781 a few misunderstandings of mine regarding the Perl internals,
782 and optimized the emission of "tags" in the output streams by
783 simply counting the objects instead of tagging them (leading to
784 a binary incompatibility for the Storable image starting at version
785 0.6--older images are of course still properly understood).
786 Murray Nesbitt made Storable thread-safe. Marc Lehmann added overloading
787 and reference to tied items support.
791 There is a Japanese translation of this man page available at
792 http://member.nifty.ne.jp/hippo2000/perltips/storable.htm ,
793 courtesy of Kawai, Takanori <kawai@nippon-rad.co.jp>.
797 Raphael Manfredi F<E<lt>Raphael_Manfredi@pobox.comE<gt>>